Browse Topic: Photogrammetry

Items (126)
The International Roughness Index (IRI) is a key indicator for evaluating the performance of road surfaces. However, traditional measurement methods only focus on the evaluation data of a single longitudinal section and do not consider the lateral difference between the actual contact area between the tire and the road surface, which may lead to inaccurate evaluation results. In recent years, with the advancement of 3D laser scanning and digital photogrammetry technology, full-section data acquisition has brought new possibilities for roughness evaluation. However, how to find a balance between data fineness and computing efficiency has become a core problem that needs to be solved. Based on the principle of interaction between vehicles and road surfaces, this paper proposes to include only the pavement height data within the tire width range into IRI analysis, and establishes an evaluation framework based on standard tire-ground contact width. This method not only retains the key information of horizontal unevenness, but also increases the computing speed by reducing the amount of data. In the experiment, a vehicle-mounted three-dimensional measurement system was used to measure the height of the actual roads, and the variation patterns of IRI values under different contact widths were analyzed. The study found that the tire-ground contact width will affect the IRI calculation results, but the degree of impact varies according to the wheel position. The IRI value of the left wheel area is not sensitive to the change in contact width, while the IRI value will fluctuate significantly due to the lateral slope of the road surface or structural asymmetry in the right wheel area. We recommend that 200 mm be used as the recommended value for the tire-ground contact width, which can not only ensure the evaluation accuracy, but also improve detection efficiency.
An, HuazhenWang, RuiHan, XiaokunLuo, Yingchao
Photogrammetry is a commonly used type of analysis in accident reconstruction. It allows the location of physical evidence, as shown in photographs and video, and the position and orientation of vehicles, other road users, and objects to be quantified. Lens distortion is an important consideration when using photogrammetry. Failure to account for lens distortion can result in inaccurate spatial measurements, particularly when elements of interest are located toward the edges and corners of images. Depending on whether the camera properties are known or unknown, various methods for removing lens distortion are commonly used in photogrammetric analysis. However, many of these methods assume that lens distortion is the result of a spherical lens or, more rarely, is solely due to distortion caused by other known lens types and has not been altered algorithmically by the camera. Today, several cameras on the market algorithmically alter images before saving them. These camera systems use proprietary distortion correction algorithms to change photographs and videos before visually presenting them to the user. Some cameras have been found to produce images where the spherical lens distortion has been partially mitigated, resulting in vertical objects such as power poles appearing straight while horizontal features such as rooflines remain curved. A video or photograph produced by these camera systems can display unexpected distortion patterns that cannot be corrected using spherical lens distortion correction techniques. This paper presents methods for correcting digital camera images that display algorithmically altered distortion mitigation. The methodology is demonstrated using two common dashcams, each implementing a proprietary distortion correction algorithm. The results for each camera are given. The methodology may be applied to any camera system to correct its unique lens distortion.
Pittman, KathleenMockensturm, EricBuckman, TaylorWhite, Kirsten
Camera matching photogrammetry is widely used in the field of accident reconstruction for mapping accident scenes, modeling vehicle damage from post collision photographs, analyzing sight lines, and video tracking. A critical aspect of camera matching photogrammetry is determining the focal length and Field of View (FOV) of the photograph being analyzed. The intent of this research is to analyze the accuracy of the metadata reported focal length and FOV. The FOV from photographs captured by over 20 different cameras of various makes, models, sensor sizes, and focal lengths will be measured using a controlled and repeatable testing methodology. The difference in measured FOV versus reported FOV will be presented and analyzed. This research will provide analysts with a dataset showing the possible error in metadata reported FOV. Analysts should consider the metadata reported FOV as a starting point for photogrammetric analysis and understand that the FOV calculated from the image metadata will likely not match exactly the final solved FOV in a virtual camera match.
Smith, Connor A.Erickson, MichaelHashemian, Alireza
Shadow positions can be useful in determining the time of day that a photograph was taken and determining the position, size, and orientation of an object casting a shadow in a scene. Astronomical equations can predict the location of the sun relative to the earth, and therefore the position of shadows cast by objects, based on the location’s latitude and longitude as well as the date and time. 3D computer software includes these calculations as a part of their built-in sun systems. In this paper, the authors examine the sun system in the 3D modeling software 3ds Max to determine its accuracy for use in accident reconstruction. A parking lot was scanned using a FARO LiDAR scanner to create a point cloud of the environment. A camera was then set up on a tripod at the environment, and photographs were taken at various times throughout the day from the same location. This environment was 3D modeled in 3ds Max based on the point cloud, and the sun system in 3ds Max was configured using the date and time of the photographs. Photogrammetry techniques were used to align undistorted photographs to the 3D environment, and rendered images were compared to the photographs. The results show that the 3ds Max sun system recreates the sun's position adequately, making it a reliable tool for accurately determining shadow locations and dimensions.
Barreiro, EvanErickson, MichaelSmith, ConnorCarter, NealHashemian, Alireza
This paper introduces a method to solve the instantaneous speed and acceleration of a vehicle from one or more sources of video evidence by using optimization to determine the best fit speed profile that tracks the measured path of a vehicle through a scene. Mathematical optimization is the process of seeking the variables that drive an objective function to some optimal value, usually a minimum, subject to constraints on the variables. In the video analysis problem, the analyst is seeking a speed profile that tracks measured vehicle positions over time. Measured positions and observations in the video constrain the vehicle’s motion and can be used to determine the vehicle’s instantaneous speed and acceleration. The variables are the vehicle’s initial speed and an unknown number of periods of approximately constant acceleration. Optimization can be used to determine the speed profile that minimizes the total error between the vehicle’s calculated distance traveled at each measured position, subject to various constraints. A test was designed to demonstrate the proposed method, using two synchronized video cameras and an instrumented test vehicle coming to a hard stop on a controlled roadway. The cameras were positioned to capture the vehicle approach in the initial field of view, and the final point of rest in the secondary field of view, with an area of indeterminate vehicle motion occurring in the region between the two camera setups. The test vehicle was driven at an unknown speed and initiated a hard brake in the uncovered region between camera systems. Using camera match photogrammetry, the position of the vehicle was measured at discrete times as it traversed the scene. Using the proposed method, optimization was shown to accurately determine the hard braking point and the speed of the vehicle as it traveled between camera systems.
Snyder, SeanCallahan, MichaelWilhelm, ChristopherJohnk, ChrisLowi, AlvinBretting, Gerald
Tesla Model 3 and Model Y vehicles come equipped with a standard dashcam feature with the ability to record video in multiple directions. Front, side, and rear views were readily available via direct USB download. Additional types of front and side views were indirectly available via privacy requests with Tesla. Prior research neither fully explored the four most readily available camera views across multiple vehicles nor field camera calibration techniques particularly useful for future software and hardware changes. Moving GPS instrumented vehicles were captured traveling approximately 7.2 kph to 20.4 kph across the front, side, and rear views available via direct USB download. Reverse project photogrammetry projects and video timing data successfully measured vehicle speeds with an average error of 2.45% across 25 tests. Previously researched front and rear camera calibration parameters were reaffirmed despite software changes, and additional parameters for the side cameras calculated.
Jorgensen, MichaelSwinford, ScottImada, KevinFarhat, Ali
Testing aircraft antennas is challenging since optimal tests are made after antenna installation. Aircraft are often taken to anechoic antenna test facilities which create long lead times, transportation hassle, and very high costs. Portable alternatives exist but often have compromised testing fidelity. Innovators at the NASA Glenn Research Center have developed the PLGRM system, which allows an installed antenna to be characterized in an aircraft hangar. All PLGRM components can be packed onto pallets, shipped, and easily operated.
Video of an event recorded from a moving camera contains information not only useful for reconstructing the locations and timing of an event, but also the velocity of the camera attached to the moving object or vehicle. Determining the velocity of a video camera recording from a moving vehicle is useful for determining the vehicle’s velocity and can be compared with speeds calculated through other reconstruction methods, or to data from vehicle speed monitoring devices. After tracking the video, the positions and speeds of other objects within the video can also be determined. Video tracking analysis traditionally has required a site inspection to map the three-dimensional environment. In instances where there have been significant site changes, where there is limited or no site access, and where budgeting and timing constraints exist, a three-dimensional environment can be created using publicly available aerial imagery and aerial LiDAR. This paper presents a methodology for creating a three-dimensional environment and performing video tracking analysis without a site visit. To validate the methodology, a blind study was conducted where three different videos were tracked. Each video presented a different traffic scenario including oncoming traffic, cross traffic, and passing traffic. The speed of the vehicle from which the video was recorded was determined through the video tracking process, and speed was also determined for a second vehicle visible within the videos. The speed of both vehicles in each video was then compared to vehicle speeds measured with vehicle instrumentation using Harry’s LapTimer to evaluate the accuracy of vehicle speeds determined through video and object tracking.
Terpstra, TobyMcDonough, SeanHelms, EthanBeier, StevenHessell, David
Shadow positions can be useful in determining the time of day that a photograph was taken and determining the position, size, and orientation of an object casting a shadow in a scene. Astronomical equations can predict the location of the sun relative to the earth, and therefore the position of shadows cast by objects, based on the location’s latitude and longitude as well as the date and time. 3D computer software have begun to include these calculations as a part of their built-in sun systems. In this paper, the authors examine the sun system in the 3D modeling software Blender to determine its accuracy for use in accident reconstruction. A parking lot was scanned using Faro LiDAR scanner to create a point cloud of the environment. A camera was then set up on a tripod at the environment and photographs were taken at various times throughout the day from the same location in the environment. This environment was then 3D modeled in Blender based on the point cloud, and the sun system in Blender was set up using the date and time of the photographs. The photographs from the environment were then undistorted and aligned to the 3D environment using photogrammetry techniques, and images were rendered in the same positions to compare the shadows in Blender’s Cycles render engine to the photographs. Through this process, the authors determined that Blender’s sun system recreates the sun position adequately, and can be used to accurately determine the location and dimensions of shadows cast by objects of known dimensions and location.
Barreiro, EvanCarter, NealHashemian, Alireza
Optical Image Stabilization (OIS) is a technology used in cameras and camcorders to reduce blur and shaky images or videos caused by unintentional camera movements. The primary goal of OIS is to counteract motion and maintain the stability of the image being captured, resulting in clearer, sharper, and more stable photos and videos. PhotoModeler, a photogrammetry software, advises users to turn off OIS on their cameras. Since the iPhone 7, OIS has become standard on all iPhones and cannot be deactivated. When calibrating an iPhone camera for photogrammetry, the OIS affects the calibration project's marking residual. In photogrammetry and 3D modeling terminology, "marking residual" typically refers to the difference between the observed image points and the corresponding points predicted by the photogrammetric process and refers to pixels. In other words, it represents the error between the actual image measurements and the values calculated by the photogrammetric algorithm. Because of OIS, the marking residual for the calibration project for an OIS-equipped camera is often outside the range recommended by PhotoModeler. As camera phones are now ubiquitous, this study aims to understand the effect of the OIS in modern camera phones on the accuracy of a PhotoModeler project. PhotoModeler projects were done using photographs taken with iPhone 7, 8, XS, 11, 12, 13, and 14 Pro models, all equipped with OIS. The results of this study demonstrate that for OIS-equipped cameras, approximately 95 percent of points measured via a calibrated camera project were within 1 to 11 mm (0.04 to 0.42 inches) of their true position, approximately 95 percent of points measured via an exemplar camera project were within 1 to 12 mm (0.03 to 0.48 inches) of their true position, and approximately 95 percent of points measured via a targetless exemplar camera project were within 1 to 14 mm (0.04 to 0.54 inches) of their true position.
Neal, JosephLeipold, TaraPetroskey, Karla
Creating a 3-dimensional environment using imagery from small unmanned aerial systems (sUAS, or unmanned aerial vehicles -UAVs, or colloquially, drones) has grown in popularity recently in accident reconstruction. In this process, ground control points are placed at an accident scene and an sUAS is flown over an accident site and a series of overlapping, high resolution images are taken of the site. Those images and ground control points are then loaded onto a computer and processed using photogrammetric software to create a 3-dimensional point cloud or mesh of the site, which then can be used as a tool for recreating an accident scene. Many software packages have been created to perform these tasks, and in this paper, the authors examine RealityCapture, a newer photogrammetric software, to evaluate its accuracy for the use in accident reconstruction. It is the authors’ experience that RealityCapture may at times produce point clouds with less noise that other software packages. To do so, Propeller Aeropoints were placed along a stretch of road to provide ground control points for the photogrammetric software and an sUAS was used to take a series of overlapping aerial images of the road. The pictures and ground control points were then processed using Pix4Dmapper, a validated and widely used photogrammetric software package, and a point cloud was created. The same process was then performed in RealityCapture. The two resulting point clouds were then compared using Cloud Compare software. Through this process, the authors determined that RealityCapture’s algorithm for creating a point cloud based on overlapping images is comparable to Pix4Dmapper’s algorithm, and is adequate for use in accident reconstruction.
Barreiro, EvanCarter, Neal
The 3D crush model can be obtained by any suitable photogrammetry method using this image set and is intended to graphically represent in photographs the shape and orientation of the damaged surface(s) relative to the undamaged, or least damaged, portion of the vehicle. The procedure is intended to provide an image set sufficient to determine, with the use of photogrammetric methodologies, the 3D location of points on the crushed surface of the damaged vehicle. Measurement of the exterior damaged surface(s) on a vehicle is a necessary step in quantifying the deformation caused by a collision and the energy dissipated by the deformation process. The energy analysis is sometimes called a crush analysis. Evaluation of the energy dissipated is useful in reconstructing the change in the velocity of the vehicles (delta-V) involved in a collision. This guideline is intended for use by investigators who do not have photogrammetry expertise, special equipment or training and may be constrained to a short time period for photography. The person who captures the image set is not expected to perform the photogrammetry or to necessarily understand how it is accomplished. It may be used by law enforcement personnel, safety officials, insurance adjusters, and other interested parties. This guideline has a written body which explains the foundation for the procedure and an appendix which specifies the procedure with accompanying diagrams. Appendix A is intended to be used as a separable field guide. This procedure recommends capturing more images than are absolutely necessary rather than specifying the criteria for selecting the absolute minimum set.
Crash Data Collection and Analysis Standards Committee
A new spatial calibration procedure has been introduced for infrared optical systems developed for cases where camera systems are required to be focused at distances beyond 100 meters. Army Combat Capabilities Development Command Armaments Center, Picatinny Arsenal, NJ All commercially available camera systems have lenses (and internal geometries) that cannot perfectly refract light waves and refocus them onto a two-dimensional (2D) image sensor. This means that all digital images contain elements of distortion and thus are not a true representation of the real world. Expensive high-fidelity lenses may have little measurable distortion, but if sufficient distortion is present, it will adversely affect photogrammetric measurements made from the images produced by these systems. This is true regardless of the type of camera system, whether it be a daylight camera, infrared (IR) camera, or camera sensitive to another part of the electromagnetic spectrum. The most common examples of large-scale lens distortions are known as barrel and pincushion effects, which are illustrated in figure 1. If these images were a truly planar reproduction of the real world, the curved lines in the images would appear as straight lines. Essentially, this can be thought of as the focal length (conversion from pixel distance to real-world distance) not being uniform throughout the image. Spatial calibration aims to build a transform to correct for large-scale distortion effects and effectively flatten an image so that the focal lengths (x and y) are uniform throughout the field of view.
All commercially available camera systems have lenses (and internal geometries) that cannot perfectly refract light waves and refocus them onto a two-dimensional (2D) image sensor. This means that all digital images contain elements of distortion and thus are not a true representation of the real world. Expensive high-fidelity lenses may have little measurable distortion, but if sufficient distortion is present, it will adversely affect photogrammetric measurements made from the images produced by these systems. This is true regardless of the type of camera system, whether it be a daylight camera, infrared (IR) camera, or camera sensitive to another part of the electromagnetic spectrum.
In the scope of development or certification processes for the flight under known icing conditions, aircraft have to be tested in icing wind tunnels under relevant conditions. The documentation of these tests has to be performed at a high level of detail. The generated data is used to prove the functionality of the systems, to develop new systems and for scientific purposes, for example the development or validation of numerical tools for ice accretion simulation. One way of documenting the resulting ice geometry is the application of an optical 3D scanning or reconstruction method. This work investigates and reviews optical methods for three-dimensional reconstructions of objects and the application of these methods in ice accretion documentation with respect to their potential of time resolved measurement. Laboratory tests are performed for time-of flight reconstruction of ice geometries and the application of optical photogrammetry with and without multi-light approach. The results of the pre-tests and the review of existing methods are evaluated with respect to scaling of the methods for application in a large icing wind tunnel. As a result of this process, multi-view photogrammetry is used for 3D reconstruction of ice accretion on a common research model wing tip installed in the icing wind tunnel of Rail Tec Arsenal. The results are compared with 3D laser scans of the final ice geometry. The presented approach allows a time resolved quantitative documentation of an icing process without interrupting the experimental ice accretion process.
Neubauer, ThomasKozomara, DavidPuffing, ReinhardTeufl, Luca
Aircraft icing is an important subject for investigation due to its critical effects on flight performance. Ice accretion analysis is commonly carried out using computational tools, from which parameters such as the mean ice shape and roughness characteristics can be obtained, as these parameters have a strong effect on the physics of aerodynamics and ice accretion. Hence, the accurate digitization of a generated ice shape through ice measurement techniques is of crucial importance. This study aimed to validate the use of photogrammetry for measurement of ice geometries and roughness on UAV airfoils, by comparing it with the cast-and-mold method. Two test cases, one mixed and second rime ice, were analyzed, each case with three subcases varying in the number of photographs used. For test case 1, mixed ice, photogrammetry method resulted in an underestimation of mean ice height by 0.5 mm in the smooth zone and overestimation by 0.2 mm and 0.6 mm on the pressure and suction sides, respectively, in the rough zone with feathers compared to the 3D-scan of the mold. The absolute surface roughness error amongst the 3 datasets was ±0.1 mm. Results indicated that the subcase with the most photographs had the least amount of ice geometry errors, but the impact of number of images on the surface roughness was negligible. In test case 2, rime ice, the results showed that even with a smaller number of photos, surface roughness was captured well, given the base geometrical noise roughness to be 0.03 mm. These findings indicate that good predictions of surface roughness can be made with a small number of photos for rime ice surfaces. Possible sources of error in capturing ice geometry include lighting, shadows, camera angles, and software reconstruction errors. The addition of global control points on the stagnation line can provide a better reference for the reconstruction software and reduce error. Accuracy of surface roughness can be improved by reducing the base geometrical noise, which can be achieved with a painting technique using a smaller droplet distribution. In conclusion, photogrammetry is a viable alternative for measurement of surface roughness on UAV airfoils.
Baghel, Anadika PaulSotomayor-Zakharov, DenisKnop, InkenOrtwein, Hans-Peter
Recent Tesla models contain four integrated onboard cameras that serve the Autopilot and Self-Driving Capabilities of the vehicle and act as a dashcam by recording footage to a local USB drive. The purpose of this study is to analyze the footage recorded by the integrated cameras and determine its suitability for speed determinations of both the host vehicle and surrounding vehicles through photogrammetry analyses. The front and rear cameras of the test vehicle (2019 Tesla Model 3) were calibrated for focal length and lens distortion characteristics. Two types of tests were performed to determine host vehicle speed: constant-speed and acceleration. Several frames from each test were analyzed. The distance between camera locations was used to gather vehicle speed through a time distance analysis. These speeds were compared to those gathered via the onboard GPS instrumentation. Two additional types of tests were performed to determine surrounding vehicle speeds: a vehicle approaching from the rear and an offset vehicle approaching from the front. For both tests, the Tesla was stationary. Several frames from each test were analyzed via reverse projection, using a point cloud of the approaching vehicles. The speeds obtained through photogrammetry were compared to GPS instrumentation onboard the approaching vehicle. The mean difference between photogrammetry and GPS instrumentation ranged between 0.38 and 0.72 mph across all tests.
Molnar, Benjamin T.Peck, Louis R.
Centrifugal Pendulum Vibration Absorber (CPVA for short) is used to absorb torsional vibrations caused by the shifting motion of the engine. It is increasingly used in modern powertrains. In the research of the dynamic characteristics of the CPVA, it is necessary to obtain the real motion of the pendulum to compensate the fitting performance of mathematical model. The usual method is to install an angle sensor to measure the movement of the pendulum. On the one hand, the installation of the sensor will affect its movement to a certain extent, so that the measurement results do not match the actual motion. On the other hand, the motion of the pendulum is not only the rotational motion around the rotational axis of the CPVA rotor, but also has translation relative to it. As a result, it is difficult to obtain accurate motion only by the angle sensor. We proposed a non-contact centrifugal pendulum motion measurement method. A high-speed camera is used to photograph the motion of the CPVA. For two adjacent images recording its motion, we use the ORB descriptor with rotation invariance to obtain the key points of the images. The Brute-Force Matcher(BF Mathcer) is used to match these points. Then, we calculate the rotational motion of the absorber rotor and the translation of the centrifugal pendulum relative to the rotor after peeling off the rotational motion around it. The method can obtain the motion of the centrifugal pendulum without affecting the motion of the CPVA. It is a feasible method without installation limitation of sensors to study the dynamic characteristics of the CPVA.
Li, WeijunWu, GuangqiangZhang, Yi
NASA researchers have developed a compact, cost-effective imaging system using a co-linear, high-intensity LED illumination unit to minimize window reflections for background-oriented schlieren (BOS) and machine vision measurements. The imaging system tested in NASA wind tunnels can reduce or eliminate shadows that occur when using many existing BOS and photogrammetric measurement systems; these shadows occur in existing systems for a variety of reasons, including the severe back-reflections from wind tunnel viewing port windows and variations in the refractive index of the imaged volume.
The paper focuses on the analyses of photogrammetry measurement results on the two-blade, hingeless MERIT rotor with diameter 1.8 m at the two rotor speeds 900 and 1800 RPM and collective pitch angles 0-12 in hover and their comparison to simulation results calculated in CAMRAD II. Blade tip displacements in flap, lead-lag, axial, and torsional direction are shown as a function of collective pitch and rotor speed. Radial displacements in flap and lag direction depict the influence of the pitch bearing play and blade attachment. The structural blade model is validated by using static DIC deformation measurements and shows very good agreement. Calculated and measured thrust polars match very well with the use of a free wake model in the simulation. The combination of measured flapping angle and calculated flapping moment gives a stiffness estimation for a virtual flap hinge. The influence of hinge offset and stiffness are shown by parameter adjustment. Flap deformations of the rotating blades leave a tip flap offset of less than 1 mm in average, which corresponds to the statically measured effect of the bearing play at the blade tip. The study shows that photogrammetry is a valuable tool to identify and tune parameters of a numerical model.
Heuschneider, VerenaBerghammer, FlorianAbdelmoula, AmineHajek, ManfredSirohi, Jayant
Rising electric scooter popularity has seen a surge in electric scooter crashes. Crash reconstructionists increasingly have access to global positioning system (GPS) data for micromobile vehicle trips, and GPS devices can produce a wealth of data about cyclists’, scooterists’, and other riders’ road paths and route usage. However, prior research has demonstrated that GPS positional accuracy is less reliable for more nuanced roadway positioning, such as which lane a vehicle occupies, as well as within-lane movements, such as acceleration and deceleration⁠. This limitation presents a challenge for crash reconstructionists that may have access to GPS data and require second-by-second positional accuracy to determine such nuanced maneuvers and vehicle positioning in their analysis. The purpose of this study was to explore the positional accuracy of five GPS units for a micromobile vehicle during three different ride conditions: acceleration, deceleration, and constant speed. The same devices were also tested for stationary accuracy and power cut-off scenarios. To obtain precise data from GPS units, tests were performed with an electric scooter ridden in rural landscapes with clear skies. Location data from the portable GPS devices were compared to reference data obtained from photogrammetry methods based on video recorded by DJI Mavic 2 drones. It was found that the overall average positional deviation from baseline across the devices and three ride conditions was 6.68 ft. The five devices also showed inconsistencies for which of the three ride stages had the greatest and least positional error. These findings can help investigators and crash reconstructionists quantify these devices’ GPS positional accuracy when using such data in their forensic analysis.
Engleman, KrystinaVega, HenrySuway, JeffreyDesai, Elvis
Photoscanning photogrammetry is a method for obtaining and preserving three-dimensional site data from photographs. This photogrammetric method is commonly associated with small Unmanned Aircraft Systems (sUAS) and is particularly beneficial for large area site documentation. The resulting data is comprised of millions of three-dimensional data points commonly referred to as a point cloud. The accuracy and reliability of these point clouds is dependent on hardware, hardware settings, field documentation methods, software, software settings, and processing methods. Ground control points (GCPs) are commonly used in aerial photoscanning to achieve reliable results. This research examines multiple GCP types, flight patterns, software, hardware, and a ground based real-time kinematic (RTK) system. Multiple documentation and processing methods are examined and accuracies of each are compared for an understanding of how capturing methods will optimize site documentation.
Mckelvey, NathanKing, CharlesTerpstra, TobyHashemian, AlirezaMitchell, Steven
Aerial photoscanning is a software-based photogrammetry method for obtaining three-dimensional site data. Ground Control Points (GCPs) are commonly used as part of this process. These control points are traditionally placed within the site and then captured in aerial photographs from a drone. They are used to establish scale and orientation throughout the resulting point cloud. There are different types of GCPs, and their positions are established or documented using different technologies. Some systems include satellite-based Global Positioning System (GPS) sensors which record the position of the control points at the scene. Other methods include mapping in the control point locations using LiDAR based technology such as a total station or a laser scanner. This paper presents a methodology for utilizing publicly available LiDAR data from the United States Geological Survey (USGS) in combination with high-resolution aerial imagery to establish GCPs based on preexisting site landmarks. This method is tested and compared to accuracies achieved with traditional control point systems.
Terpstra, TobyMckelvey, NathanKing, EricHashemian, AlirezaKing, Charles
Accident scene data obtained from photographs and videos are vital to the analysis performed during accident reconstruction. They allow forensic analysts to precisely determine the orientation and location of evidence. For that reason, the term digital evidence was adopted and is commonly used by forensic analysts in conjunction with retro-projection methods as an aid to reconstruct the events leading up to the incident in question. Photogrammetry is a retro-projection method commonly used by analysts to match scene photographs with calibrated control points obtained from three-dimensional point cloud data collected at the subject accident site and visualized on the accident scene images. From this merger, the photographs allow for the determination of the point at which the camera was positioned that took the subject image. In general, the point cloud data allows for increased accuracy during the photogrammetry process. Video footage obtained from the scenes can be exported as frames and then used in the same manner as digital images in photogrammetry. 360-degree cameras are a relatively newer type of camera that can capture their entire surroundings by having multiple wide-angle lenses. The research and testing presented in this study intend to analyze images from 360-degree cameras and justify their usage in photogrammetry. The present study considered three models of 360-degree cameras of various specifications, performance, and cost. All cameras were positioned in the same location and the same orientation as each other. Point cloud data of the test site was obtained using a Leica RTC360 Laser Scanner. The same point cloud data was used in the photogrammetric analysis for each camera. Testing was conducted outdoors in natural light, as it would be during typical forensic investigative procedures. The test scenario was conducted with 360-degree cameras. Distance between designated test objects was physically measured and used as a benchmark to the distances determined by point cloud photogrammetric analysis. The distances between the designated test objects were large enough to be considered appropriate for typical accident reconstruction analyses. A distance comparison was made against the physically measured benchmark distance for each 360-degree camera, as well as against the point cloud data.
Morales, Roberto C.Farias, Edgar
The world is going through the fourth industrial revolution, where digital transformation is one of the global market trends. To maintain competitive advantages and sustainable businesses, an increasing number of companies and organizations are embracing digital transformation processes. These organizations are changing their business and processes and creating new business models with the help of digital technologies. Taking all industries and business models to unprecedented heights and in a certain way consolidating globalization. For such digital transformation, technologies like IoT (Internet of Things), artificial intelligence, machine learning, neural networks, and others are increasingly common. This paper seeks to define what technical aspects are involved to implement digitalization in the process of vehicle collision data analysis. In this sense, insurance companies are aware of the changes and are trying to follow the trends and updating themselves to provide better services and with better quality. The applied methodology is divided into an analysis of the existing 3D metrology techniques and a pilot project in order to evaluate the application of the selected technique and analyze results and gains in the incident inspection process. The set of these methodologies allowed the proposed application to be validated.
Stano, Pedro Henrique Silva
Traffic cameras, dash-cameras, surveillance cameras, and other video sources increasingly capture critical evidence used in the accident reconstruction process. The iNPUT-ACE Camera Match Overlay tool can utilize photogrammetry to project a two-dimensional video onto three-dimensional point cloud software to enable measurements to be directly taken from the video. Those measurements are commonly used, and critical for, the determination of vehicle speed in accident reconstruction. The accuracy of the Camera Match Overlay tool has not yet been thoroughly examined. To validate the use of the tool to measure vehicle speed for accident reconstruction, data were collected from a series of tests involving three traffic cameras, a stationary and moving dash-camera, a stationary and moving cell-phone camera, and a doorbell surveillance camera. Each camera provided unique specifications of quality and focal length to ensure the tool would be tested in a variety of scenarios. Vehicles drove past the various cameras at a variety of distances and angles with vehicle speeds that ranged from approximately 15.5 kph (9.6 mph) to 90.2 kph (56.1 mph) as recorded by a RACELOGIC VBOX GPS system. A bright flash was utilized to synchronize timing between VBOX and video data. The resulting comparison between VBOX data and the tool’s distance and timing data revealed the iNPUT-ACE Camera Match Overlay tool was an effective method for analyzing vehicle speed across a variety of video sources.
Jorgensen, MichaelSwinford, ScottJones, Brian
This paper introduces a method for calculating vehicle speed and uncertainty range in speed from video footage. The method considers uncertainty in two areas; the uncertainty in locating the vehicle’s position and the uncertainty in time interval between them. An abacus style timing light was built to determine the frame time and uncertainty of time between frames of three different cameras. The first camera had a constant frame rate, the second camera had minor frame rate variability and the third had more significant frame rate variability. Video of an instrumented vehicle traveling at different, but known, speeds was recorded by all three cameras. Photogrammetry was conducted to determine a best fit for the vehicle positions. Deviation from that best fit position that still produced an acceptable range was also explored. Video metadata reported by iNPUT-ACE and Mediainfo was incorporated into the study. When photogrammetry was used to determine a vehicle’s position and speed from video recorded by a constant frame rate camera, the results closely matched the speeds reported by the instrumented vehicle being measured. This low uncertainty resulted from the constant frame rate eliminating error in time, and from low error in the vehicle’s position through photogrammetry. For the variable frame rate camera, uncertainty in speed was dependent on the time between frames analyzed as well as any uncertainty in position. Quantification of this uncertainty has value for the reconstructionist. Determining speed of the vehicle in the variable frame rate video could be improved by incorporating frame timing reported by iNPUT-ACE or through other video analysis techniques and software that measure precise time differences between each frame.
Beauchamp, GrayPentecost, DavidKoch, DanielHashemian, AlirezaMarr, JamesCordero, Rheana
Forensic disciplines are called upon to locate evidence from a single camera or static video camera, and both the angle of incidence and resolution can limit the accuracy of single image photogrammetry. This research compares a baseline of known 3D data points representing evidence locations to evidence locations determined through single image photogrammetry and evaluates the effect that object resolution (measured in pixels), and angle of incidence has on accuracy. Solutions achieved using an automated process where a camera match alignment is calculated from common points in the 2D imagery and the 3D environment, were compared to solutions achieved in a more manual method by iteratively adjusting the camera’s position, orientation, and field-of-view until an alignment is achieved. This research independently utilizes both methods to achieve photogrammetry solutions and to locate objects within a 3D environment. Results are compared for a greater understanding of the accuracies that can be achieved using camera matching photogrammetry for evidence placement when only a single image is available.
Terpstra, TobyHashemian, AlirezaGillihan, RobertKing, EricMiller, SethNeale, William
Photogrammetry is a commonly used and accepted technique within the field of accident reconstruction for taking measurements from photographs. Previous work has shown the accuracy of optimized close-range photogrammetry techniques to be within 2 mm compared to other high accuracy measurement techniques when using a known calibrated camera. This research focuses on the use of inverse camera close-range photogrammetry, where photographs from an unknown camera are used to model a vehicle. Photogrammetry is a measurement technique that utilizes triangulation to take measurements from photographs. The measurements are dependent on the geometry of the camera, such as the sensor size, focal length, lens type, etc. Three types of cameras were tested for accuracy; a high-end commercial camera, a point and shoot camera, and a cell phone camera. This study indicates that in a properly conducted inverse photogrammetry project, an analyst can be 95% confident the true position of a point will be within 6.25 mm (0.25 inches) of the location of the point obtained via photogrammetry.
Neal, JosephFunk, CharlesSproule, David
A new measurement capability was created by combining photogrammetry and metrology techniques to accurately measure one half of the XV-15 Tilt Rotor Research Aircraft at the Smithsonian’s Udvar-Hazy museum. The challenges imposed by the fuselage and surrounding environment at Udvar-Hazy were overcome by careful application of photogrammetry and metrology techniques. Data analyses and processing included the use of multiple reverse engineering programs to accurately generate a complete 3-dimensional water-tight geometry of the aircraft and rotor blade. This paper describes the photogrammetry and metrology measurement systems, technology and hardware set-up, data analysis and processing methods, future work, and lessons learned. In addition, selected measurement results of the fuselage and rotor blade are presented.
Cummings, HaleyDominguez, MichelleSolis, EduardoSilva, ChristopherBowman, BelenBurek, Shirley
Feasibility in Manufacturing of autonomous unmanned aerial vehicles at low cost allows the UAV developers to bring it out with numerous applications for society. Civil domain is a widely developing platform which initiated the development of UAV for civilian applications like bridge inspection, building monitoring, life or strength estimation of historical places and also outdoor and indoor mapping of buildings. These autonomous UAVs with high resolution camera fly over and around the construction sites, buildings, mines and captures images of various locations and point clouds in all sides of the building and creates a 3D map by using photogrammetry techniques. The software auto generates the report and updates it to the cloud which can be accessed online. Autonomous operations are quite difficult in new environments which requires SLAM (simultaneous localization and mapping) to operate the UAV between open spaces. This paper describes the technique of mapping a construction site using a quadcopter and determine the completion of such constructions using image processing and machine learning techniques. Obstacle avoidance during the autonomous flight using ultrasonic sensors provide greater flexibility of the vehicle to move around the buildings.
V, HariprasadMS, YaswanthV, SathishN, YamunaV, Karthick SreenivasanK, Sivakumar
The aerodynamic effects of Cold Soaked Fuel Frost have become increasingly significant as airworthiness authorities have been asked to allow it during aircraft take-off. The Federal Aviation Administration and the Finnish Transport Safety Agency signed a Research Agreement in aircraft icing research in 2015 and started a research co-operation in frost formation studies, computational fluid dynamics for ground de/anti-icing fluids, and de/anti-icing fluids aerodynamic characteristics. The main effort has been so far on the formation and aerodynamic effects of CSFF. To investigate the effects, a generic high-lift common research wind tunnel model and DLR-F15 airfoil, representing the wing of a modern jet aircraft, was built including a wing tank cooling system. Real frost was generated on the wing in a wind tunnel test section and the frost thickness was measured with an Elcometer gauge. Frost surface geometry was measured with laser scanning and photogrammetry. The aerodynamic effect of the frost was studied in a simulated aircraft take-off sequence, in which the speed was accelerated to a typical rotation speed and the wing model was then rotated to an angle of attack used at initial climb. Time histories of the lift coefficient were measured with a force balance. The experiments showed that depending on the ambient temperature the frost may evaporate/melt during the take-off sequence. Lift losses after rotation with CSFF contamination at ambient temperatures of 4° to 7°C above freezing point were measured to be 4 to 5 % for roughness values, k/c, below 10-3. For comparison, lift loss tests with typical anti-icing fluids were performed resulting to roughly equal lift losses. This paper gives an overview of the performed activities.
Koivisto, PekkaSoinne, ErkkiBroeren, AndyBond, Thomas
Surface Contamination Effects on CRM Wing Section Model2019-01-19766/10/2019
The aerodynamic effects of Cold Soaked Fuel Frost have become increasingly significant as aircraft manufacturers have applied for to allow it during aircraft take-off. The Federal Aviation Administration and the Finnish Transport Safety Agency signed a Research Agreement in aircraft icing research in 2015 and started a research co-operation in frost formation studies, computational fluid dynamics for ground de/anti-icing fluids, and de/anti-icing fluids aerodynamic characteristics. The main effort has been so far on the formation and effects of CSFF. To investigate the effects a HL-CRM wing wind tunnel model, representing the wing of a modern jet aircraft, was built including a wing tank cooling system. Real frost was generated on the wing in a wind tunnel test section and the frost thickness was measured with an Elcometer gauge. Frost surface geometry was measured with laser scanning and photogrammetry. The aerodynamic effect of the frost was studied in a simulated aircraft take-off sequence, in which the speed was accelerated to a typical rotation speed and the wing model was then rotated to an angle of attack used at initial climb. Time histories of the lift coefficient were measured with a force balance. Time histories of the upper surface boundary layer displacement thickness were measured with a boundary layer rake. For comparison the effects of typical anti-icing fluids, sandpaper and smooth PVC plastic sheet were also measured. The lift losses correlated with average surface contamination roughness height and the boundary layer displacement thickness increment.
Soinne, ErkkiRosnell, Tomi
UAV Icing: Ice Accretion Experiments and Validation2019-01-20376/10/2019
Atmospheric icing is a key challenge to the operational envelope of medium-sized fixed-wing UAVs. Today, several numeric icing codes exist, that all have been developed for general aviation applications. UAVs with wingspans of several meters typically operate at Reynolds numbers an order of magnitude lower than commercial and military aircraft. Therefore, the question arises to what extent the existing codes can be applied for low-Reynolds UAV applications to predict ice accretion. This paper describes an experimental campaign at the Cranfield icing wind tunnel on a RG-15 and a NREL S826 airfoil at low velocities (25-40m/s). Three meteorological icing conditions have been selected to represent the main ice typologies: rime, glaze, and mixed ice. Each case has been run at least twice in order to assess the repeatability of the experiments. Manual ice shape tracings have been taken at three spanwise locations for each icing case. The liquid water content calibration was performed according to ARP5905 using the icing blade method. The tests have initially shown significantly higher water contents than anticipated, which could be traced to dimensional differences of the blade at Cranfield, as well as low flow velocities. This systematic error was resolved by simulating the droplet collection coefficients on the off-specification blade. In addition to manual tracings, photogrammetry and a handheld laser-scanner were used to capture the ice shapes. The results indicate that manual tracings are still the most efficient method, although there is potential in exploring the alternatives further. Additionally, numerical simulations with two icing codes, LEWICE and FENSAP-ICE, were performed on a rime and a glaze case. For rime, the simulations show a good agreement with the experiment, whereas the glaze case exhibits significant differences.
Hann, Richard
A Non-Contact Technique for Vibration Measurement of Automotive Structures2019-01-15036/5/2019
The automotive and aerospace industries are increasingly using the light-weight material to improve the vehicle performance. However, using light-weight material can increase the airborne and structure-borne noise. A special attention needs to be paid in designing the structures and measuring their dynamics. Conventionally, the structure is excited using an impulse hammer or a mechanical shaker and the response is measured using uniaxial or multi-axial accelerometers to obtain the dynamics of the structure. However, using contact-based transducers can mass load the structure and provide data at a few discrete points. Hence, obtaining the true dynamics of the structure conventionally can be challenging. In the past few years, stereo-photogrammetry and three-dimensional digital image correlation have received special attention in collecting operating data for structural analysis. These non-contact optical techniques provide a wealth of distributed data over the entire structure. However, the stereo camera system is limited by its field of view of the cameras and can only measure the response on the parts of the structure that cameras have the line of sight. Therefore, a single pair of Digital Image Correlation (DIC) cameras may not be able to provide deformation data for the entire structure. In current work, a multi-view 3D DIC approach is used to predict the vibrational characteristics of a full vehicle. A pair of DIC cameras is roved over the entire vehicle to capture the deformation data of each field of view. The measured data includes the geometry and displacement data which is mapped into the global coordinate system using 3D transformation matrices. The obtained data in the time domain for each field of view is transformed to the frequency domain using the Fast Fourier Transformation (FFT) to extract the operational deflection shapes and resonant frequencies for each field of view. The obtained deflection shapes are scaled and stitched in the frequency domain to extract the operating deflection shapes of full vehicle.
Srivastava, VanshajBaqersad, Javad
Small unmanned aerial systems have gained prominence in their use as tools for mapping the 3-dimensional characteristics of accident sites. Typically, the process of mapping an accident site involves taking a series of overlapping, high resolution photographs of the site, and using photogrammetric software to create a point cloud or mesh of the site. This process, known as image-based scanning, is explored and analyzed in this paper. A mock accident site was created that included a stopped vehicle, a bicycle, and a ladder. These objects represent items commonly found at accident sites. The accident site was then documented with several different unmanned aerial vehicles at differing altitudes, with differing flight patterns, and with different flight control software. The photographs taken with the unmanned aerial vehicles were then processed with photogrammetry software using different methods to scale and align the point clouds. The point cloud data produced with different vehicle / flight pattern / altitude combinations was then quantitatively compared to terrestrial LiDAR scan data. The results are presented here, as well as recommendations based on equipment and desired output.
Carter, NealHashemian, AlirezaMckelvey, Nathan
The Application of Augmented Reality to Reverse Camera Projection2019-01-04244/2/2019
In 1980, research by Thebert introduced the use of photography equipment and transparencies for onsite reverse camera projection photogrammetry [1]. This method involved taking a film photograph through the development process and creating a reduced size transparency to insert into the cameras viewfinder. The photographer was then able to see both the image contained on the transparency, as well as the actual scene directly through the cameras viewfinder. By properly matching the physical orientation and positioning of the camera it was possible to visually align the image on the image on the transparency to the physical world as viewed through the camera. The result was a solution for where the original camera would have been located when the photograph was taken. With the original camera reverse-located, any evidence in the transparency that is no longer present at the site could then be replaced to match the evidences location in the transparency. Reverse camera projection is useful for both determining the location of historical evidence, where it is no longer physically in existence, as well as for directing the investigator to evidence still at the site that may otherwise have been overlooked during a site inspection. With the advent of augmented reality, an entirely digital process of this technique is now possible. This paper both presents a digital methodology and provides reference to a publicly available, augmented reality application developed specifically for this process by the authors. The accuracy of the application and methodology is then demonstrated through field studies with reported results.
Terpstra, TobyBeier, StevenNeale, William
Accuracy Assessment of Three-Dimensional Site Features Generated with Aid of Photogrammetric Epipolar Lines in PhotoModeler and Using Minimal sUAS Imagery2019-01-04104/2/2019
Photogrammetry is widely used in the accident reconstruction community to extract three-dimensional information from photographs. This article extends a prior study conducted by the authors, whereby model accuracy was assessed for a technique that exploited vehicle edges and epipolar line projections to construct 3D vehicle models, by examining 3D roadway and site features. To do so, artificial images were generated using an ideal computer-generated camera within a computer-assisted drawing environment to allow for a known reference model to compare with results produced using photogrammetry. A systematic study was undertaken by modeling the curvature, elevation, and super-elevation of a roadway and associated markings, sidewalks, and buildings, either by relying on discrete points or utilizing epipolar lines. The models were assessed for accuracy, and the sensitivity of the accuracy to camera elevation was considered. Subsequently, the photogrammetric procedures were applied to actual sites, and the results were compared with 3D total station and scanner measurements. A further goal of this study was to evaluate modeling accuracy for cases in which a minimal number of sUAS images were included in the photogrammetry project. The findings of the current study corroborated the prior effort when scaled for the size of the objects modeled. It was demonstrated in this study that using photographs taken with a calibrated digital camera, and taking advantage of epipolar lines when modeling, allowed analysts to construct wireframe models of a real-world site, straight and curved edges included, that exhibited an average residual error of 4.1 cm (SD = 2.6 cm) when compared to scanned measurements, resulting in nominal dimensions of the object within 0.2% of the measured dimensions. The modeled site grades fell within ±0.4% of the measured grades.
Long, AndreaNoll, Scott Allen
Reconstruction of 3D Accident Sites Using USGS LiDAR, Aerial Images, and Photogrammetry2019-01-04234/2/2019
The accident reconstruction community has previously relied upon photographs and site visits to recreate a scene. This method is difficult in instances where the site has changed or is not accessible. In 2017 the United States Geological Survey (USGS) released historical 3D point clouds (LiDAR) allowing for access to digital 3D data without visiting the site. This offers many unique benefits to the reconstruction community including: safety, budget, time, and historical preservation. This paper presents a methodology for collecting this data and using it in conjunction with aerial imagery, and camera matching photogrammetry to create 3D computer models of the scene without a site visit. To determine accuracies achievable using this method, evidence locations solved for using only USGS LiDAR, aerial images and scene photographs (representative of emergency personnel photographs) were compared with known locations documented using total station survey equipment and ground-based 3D laser scanning. The data collected from three different site locations was analyzed, and camera matching photogrammetry was performed independently by 5 different individuals to locate evidence. On average, the resulting evidence for all three test sites was found to be within 3.0 inches (8cm) of known evidence locations with a standard deviation of 1.7 inches (4cm). To further evaluate the quality of the USGS LiDAR, a comparative point cloud analysis of the roadway surfaces was performed. On average, 85% of the USGS LiDAR points were found to be within .5 inches of the ground-based 3D scanning points.
Terpstra, TobyDickinson, JordanHashemian, AlirezaFenton, Stephen
In this paper will be explained how photogrammetry and tracking technologies are a highly accurate alternative to accelerometers instrumented sensors related to distances calculations between objects or vehicle interior parts and the dummies. Photogrammetry is used to calculate the real-world point’s position on an image. The tracking system uses algorithms to follow points and keep the same center point at each movie frame. A software application combines these two elements to provide position, velocity, acceleration and angles of every point on the movie for the 3-dimensional axis. The tracking technology can be applied for on dummy’s analysis head impact criterion (HIC) against internal structure and objects as the pole. The use of internal sensors for this kind of analysis, only offers a yes/no response and yet tracking provides the exact distance between head and the interior components. Using tracking technology the distance between the dummy’s head and any other structural part of the vehicle can be known and even the distance between the head and the steering wheel can be calculated. Therefore, knowing the distances and the resultant forces from the sensors, improvements at the restraint systems can be done in order to decrease the severity implemented to the dummy from the restraint systems itself. In this paper several crash test had been analyzed to determine if the implemented methodology is accurate. Results are shown and conclusions about the benefits of using the tracking technology are explained.
Molina, David Company
The accident reconstruction community relies on photogrammetry for taking measurements from photographs. Camera matching, a close-range photogrammetry method, is a particularly useful tool for locating accident scene evidence after time has passed and the evidence is no longer physically visible. In this method, objects within the accident scene that have remained unchanged are used as a reference for locating evidence that is no longer physically available at the scene such as tire marks, gouge marks, and vehicle points of rest. Roadway lines, edges of pavement, sidewalks, signs, posts, buildings, and other structures are recognizable scene features that if unchanged between the time of accident and time of analysis are beneficial to the photogrammetric process. In instances where these scene features are limited or do not exist, achieving accurate photogrammetric solutions can be challenging. Off-road incidents, snow-covered roadways, rural areas, and unpaved roadways are examples where available scene features may be limited. Other factors like the number of photographs, the specific vantage of the photographs, and occlusion of recognizable features within these photographs can also limit the number of common features available for use in camera matching. In these instances, camera matching solutions can be improved by extending the 3D environment to include objects visible in the distance such as mountains, valleys, and other notable landmarks that are typically outside of the scope of 3D scene mapping. This article demonstrates a method for obtaining and using this elevation data in combination with 3D scene mapping for camera matching photogrammetry. Photogrammetric solutions with limited scene features are compared to photogrammetric solutions based on the same limited scene features with the addition of digital elevation models. Solution accuracies from both scenarios are then individually evaluated to demonstrate improvements through the use of elevation models. In this study, the incorporation of digital elevation modeling at a site with limited scene features demonstrates a 74% improvement for evidence located through camera matching photogrammetry. For further evaluation, the camera match solutions were compared in combined solutions, where information obtained from one camera match was used to inform the next. This was done for both the scenario with digital elevation models and the scenario without. The results demonstrate how the number of available photos can influence the overall accuracy of photogrammetry solutions.
Terpstra, TobyDickinson, JordanHashemian, Alireza
Accident reconstructionists will typically document scenes, evidence, vehicles or objects of interest by using 3-dimensional laser scanners. These techniques are well documented, utilized and can be extremely accurate. However, when the subject of documentation involves surfaces that include intricate, highly reflective, and/or complex geometry (motorcycles, wheelchairs, stairs, etc.) the commercially available laser scanners can produce obscuring dense stray and scattered points which results in point clouds that could require tedious manual registration and/or optimization. This paper compares a FARO Focus laser scanner, Pix4DMapper, and Agisoft’s Photoscan point cloud data to FARO ARM measurements of vehicles, other transportation devices and architectural features. It was shown that the Pix4DMapper and Agisoft’s Photoscan point cloud data resulted in detailed and accurate point cloud data compared to the FARO ARM measurements. Additionally, the input data for Pix4DMapper and Agisoft’s Photoscan is easy to capture and required minimal processing and did not require extensive, time consuming, optimization of individual scans. This paper demonstrates the use of contemporary photogrammetry softwares, Pix4DMapper and Agisoft’s Photoscan, as accurate, time and cost effective alternatives to laser scanners.
Grimes, ClareRoescher, ToddSuway, Jeffrey AaronWelcher, Judson
In an accident reconstruction, vehicle speeds and positions are always of interest. When provided with scene photographs or fixed-location video surveillance footage of the crash itself, close-range photogrammetry methods can be useful in locating physical evidence and determining vehicle speeds and locations. Available 3D modeling software can be used to virtually match photographs or fixed-location video surveillance footage. Dash- or vehicle-mounted camera systems are increasingly being used in light vehicles, commercial vehicles and locomotives. Suppose video footage from a dash camera mounted to one of the vehicles involved in the accident is provided for an accident reconstruction but EDR data is unavailable for either of the vehicles involved. The literature to date describes using still photos to locate fixed objects, using video taken from stationary camera locations to determine the speed of moving objects or using video taken from a moving vehicle to locate fixed objects. However, techniques to evaluate the position, speed and acceleration of moving objects seen in video taken from moving locations have not been evaluated. To address the increasing prevalence of dash cams and other in-vehicle video and the value in using such video in vehicle crash reconstruction, this paper describes techniques for determining the position and speed of a moving object from digital video taken from a moving vehicle. Evaluations of the accuracy of those techniques were done when provided three different levels of information about the environment: 1 Aerial Photography (USGS) 2 Survey Data (Total Station) 3 3D Scan Data (of both the environment and vehicles)
Manuel, Emmanuel JayMink, RichardKruger, Daniel
Photogrammetry is widely used in the automotive and accident reconstruction communities to extract three-dimensional information from photographs. Prior studies in the literature have demonstrated the accuracy of such methods when photographs contain easily-identifiable, distinct points; however, it is often desirable to determine measurements for locations where a seam, edge, or contour line is available. To exploit such details, an analyst can control the direction that the epipolar line is projected onto the camera plane by strategic selection of photographs. This process constrains the search for the corresponding 3D point to a straight line that can be projected perpendicular to the seam, edge, or contour line. Thus, the goal of this study was to evaluate the modeling accuracy for cases in which an analyst uses epipolar lines in a workflow. To do so, artificial images were created using a computer-generated camera within a computer-assisted drawing environment to allow for a known reference model to compare with results produced using photogrammetry. A systematic study was undertaken by modeling two-dimensional curves on a plane, three-dimensional curves on a curved surface, and then curved edges on a vehicle model. Each model was assessed for accuracy, and the sensitivity of the accuracy to camera placement was carefully examined and explained. Finally, the procedures were applied to an actual vehicle, for which the results were compared to a 3D laser scan of the vehicle. In conclusion, the average residual error between a photogrammetry model created with the aid of epipolar lines and 3D scanned points for a three-dimensional vehicle edge feature was 1.69 mm (SD = 0.55 mm).
Long, AndreaNoll, Scott Allen
Photogrammetry and the accuracy of a photogrammetric solution is reliant on the quality of photographs and the accuracy of pixel location within the photographs. A photograph with lens distortion can create inaccuracies within a photogrammetric solution. Due to the curved nature of a camera’s lens(s), the light coming through the lens and onto the image sensor can have varying degrees of distortion. There are commercially available software titles that rely on a library of known cameras, lenses, and configurations for removing lens distortion. However, to use these software titles the camera manufacturer, model, lens and focal length must be known. This paper presents two methodologies for removing lens distortion when camera and lens specific information is not available. The first methodology uses linear objects within the photograph to determine the amount of lens distortion present. This method will be referred to as the straight-line method. The second methodology utilizes photogrammetry principles and 3D point cloud data to solve for and remove lens distortion. This method will be referred to as the point cloud method. Using cameras with known distortion parameters, both methodologies are presented and individually evaluated against publically available, library-based, distortion removal solutions. Based on the results of lens distortion removal from cameras with known lens distortion, the straight-line method was found to improve pixel location within a photograph by an average of 82 percent and by as much as 99 percent. The point cloud method was found to improve pixel location by an average of 40 percent and by as much as 66 percent.
Terpstra, TobyMiller, SethHashemian, Alireza
Improvements in computer image processing and identification capability have led to programs that can rapidly perform calculations and model the three-dimensional spatial characteristics of objects simply from photographs or video frames. This process, known as structure-from-motion or image based scanning, is a photogrammetric technique that analyzes features of photographs or video frames from multiple angles to create dense surface models or point clouds. Concurrently, unmanned aircraft systems have gained widespread popularity due to their reliability, low-cost, and relative ease of use. These aircraft systems allow for the capture of video or still photographic footage of subjects from unique perspectives. This paper explores the efficacy of using a point cloud created from unmanned aerial vehicle video footage with traditional single-image photogrammetry methods to recreate physical evidence at a crash scene. The unique aspects of photographs or video taken with unmanned aircraft systems ease some of the challenges of creating point cloud data with ground level footage. To explore the accuracy of this process, the authors constructed a mock scene with physical evidence that is typical of vehicular crashes. The scene was scanned with a FARO laser scanner and photographed. The evidence was then removed and video was taken of the scene from an unmanned aerial vehicle. That video footage was processed with image-based scanning software to create a point cloud, and the point cloud was used as a means to determine the positions and characteristics of the camera at the time the evidence was photographed. The evidence was then reconstructed with traditional single image photogrammetry techniques, and the position and size of the reconstructed evidence was compared to the actual position as documented by the FARO scanner. Through this process, the authors determined that the use of unmanned aerial vehicle footage and image-based scanning software could be used to accurately reconstruct the location of physical evidence.
Carter, NealHashemian, AlirezaRose, Nathan A.Neale, William T.C.
This paper presents a methodology for determining the position and speed of objects such as vehicles, pedestrians, or cyclists that are visible in video footage captured with only one camera. Objects are tracked in the video footage based on the change in pixels that represent the object moving. Commercially available programs such as PFTracktm and Adobe After Effectstm contain automated pixel tracking features that record the position of the pixel, over time, two dimensionally using the video’s resolution as a Cartesian coordinate system. The coordinate data of the pixel over time can then be transformed to three dimensional data by ray tracing the pixel coordinates onto three dimensional geometry of the same scene that is visible in the video footage background. This paper explains the automated process of first tracking pixels in the video footage, and then remapping the 2D coordinates onto three dimensional geometry using previously published projection mapping and photogrammetry techniques. The results of this process are then compared to VBOX recordings of the objects seen in the video to evaluate the accuracy of the method. Some beneficial aspects of this process include the time reduced in tracking the object, since it is automated, and also that the shape and size of the object being tracked does not need to be known since it is a pixel being tracked, rather than the geometry of the object itself.
Neale, William T.Hessel, DavidKoch, Daniel
Video and photo based photogrammetry software has many applications in the accident reconstruction community including documentation of vehicles and scene evidence. Photogrammetry software has developed in its ease of use, cost, and effectiveness in determining three dimensional data points from two dimensional photographs. Contemporary photogrammetry software packages offer an automated solution capable of generating dense point clouds with millions of 3D data points from multiple images. While alternative modern documentation methods exist, including LiDAR technologies such as 3D scanning, which provide the ability to collect millions of highly accurate points in just a few minutes, the appeal of automated photogrammetry software as a tool for collecting dimensional data is the minimal equipment, equipment costs and ease of use. This paper evaluates the accuracy and capabilities of four automated photogrammetry based software programs to accurately create 3D point clouds, by comparing the results to 3D scanning. Both a damaged and undamaged vehicle were documented with video and photographs and on average the damaged vehicle set returned more data points with higher accuracy than the undamaged vehicle set. Four cameras types were evaluated and more accurate results were achieved when using either a DSLR or a point-and-shoot camera than when using a GoPro, or a cell phone camera. Photogrammetry data from video footage was analyzed and found to be both less accurate and to return less data than photographs. By limiting the number of photographs used, it was found that a photogrammetry solution could be achieved with as few as 16 photographs encircling a vehicle, but better results were reached with a larger number of photographs.
Terpstra, TobyVoitel, TiloHashemian, Alireza
Two Transport Rotorcraft Airframe Crash Testbed (TRACT) full-scale tests were performed at NASA Langley Research Center's Landing and Impact Research Facility in 2013 and 2014. Two CH-46E airframes were impacted at 33-ft/s forward and 25-ft/s vertical combined velocities onto soft soil, which represents a severe, but potentially survivable impact scenario. TRACT 1 provided a baseline set of responses, while TRACT 2 included retrofits with composite subfloors and other crash system improvements based on TRACT 1. For TRACT 2, a total of 18 unique experiments were conducted to evaluate ATD responses, seat and restraint performance, cargo restraint effectiveness, patient litter behavior, and activation of emergency locator transmitters and crash sensors. Combinations of Hybrid II, Hybrid III, and ES-2 Anthropomorphic Test Devices (ATDs) were placed in forward and side facing seats and occupant results were compared against injury criteria. The structural response of the airframe was assessed based on accelerometers located throughout the airframe and using three-dimensional photogrammetric techniques. Analysis of the photogrammetric data indicated regions of maximum deflection and permanent deformation. The response of TRACT 2 was noticeably different in the longitudinal direction due to changes in the cabin configuration and soil surface, with higher acceleration and damage occurring in the cabin. Loads from ATDs in energy absorbing seats and restraints were within injury limits. Severe injury was likely for ATDs in forward facing passenger seats.
Annett, MartinLittell, Justin
In the field of accident reconstruction, a reconstructionist will often inspect a crash scene months or years after a crash has occurred. With this passage of time important evidence is sometimes no longer present at the scene (i.e. the vehicles involved in the crash, debris on the roadway, tire marks, gouges, paint marks, etc.). When a scene has not been totally documented with a survey by MAIT or the investigating officers, the reconstructionist may need to rely on police, fire department, security camera, or witness photographs. These photos can be used to locate missing evidence by employing traditional photogrammetric techniques. However, traditional techniques require planar surfaces, matched discrete points, or camera matching at the scene. Sometimes it is not possible to survey discrete points or perform camera matching at the scene due to lack of access (the tops of power poles, elevated bridge features, or objects at a great distance) or for safety reasons (interstate highways with high traffic conditions or on narrow bridges). Other times important evidence can be located on a hill or depressed median and planar photogrammetric methods are not effective. In recent years three dimensional laser scanners have been utilized by accident investigators allowing the reconstructionist to document a crash scene with millions of points in minutes so that a “point cloud” can be created. The equipment is fast, simple to setup, and is very accurate, thus allowing the reconstructionist the ability to take home a three dimensional model of the scene for in depth analysis. This paper will describe the scanning equipment, photo considerations, and a methodology for utilizing three dimensional laser scan data and camera matching to extract evidence from first responder photographs or videos. The accuracies of this technique are compared to other accepted methods such as planar photogrammetry, Discrete Point Software, and photo rectification over point cloud data. The camera matching method of locating evidence has been described in literature before [1,2,7] but technology advances have allowed the method to be expanded upon. The camera matching method now applied to a three dimensional point cloud with millions of points used for the solution allows easy and accurate extraction of evidence on not only flat areas but complex terrains as well. A simulated crash which is called a “staged collision” is analyzed. This “staged collision” represents two types of evidence. One type of evidence is on a flat surface to compare planar accuracies. The second type of evidence is in an area incorporating large changes in the terrain to compare three dimensional accuracies. Various photogrammetry methods are compared in these two types of terrain and it is shown how combining laser scan data with the “virtual camera matching to point cloud” technique can accurately extract evidence from non-planar areas with large changes in terrain.
Coleman, ClayTandy, DonaldColborn, JasonAult, Nicholas
Items per page:
1 – 50 of 126