Potholes are a common road hazard that significantly compromise road safety. Water filled potholes can be particularly dangerous. These hidden hazards may cause vehicles to hydroplane [1], leading to a loss of control and potential collisions. At night or in low visibility conditions, such potholes can appear deceptively shallow, increasing the risk of severe suspension damage or tire blowouts. Additionally, deep water intrusion can affect critical components such as the exhaust system, air intake, or electrical wiring, potentially leading to engine stalling or short circuits.
This research proposes a novel approach for identifying and determining the depth of potholes, especially those that are filled with water. By integrating YOLO, cutting edge computer vision methods like stereo imaging and Lidar. We hope to create a system that can precisely detect and evaluate potholes' severity, reducing the risks connected to these road hazards.
A structured 2k factorial Design of Experiment (DOE) methodology is used to extensively evaluate the performance, accuracy, and limitations of both approaches under a variety of surface and environmental conditions. The stereo imaging approach uses polarization to enhance contrast and reduce glare, which significantly increases depth estimate accuracy in difficult lighting conditions. On the other hand, the LIDAR technique produces 3D point cloud data with excellent resolution, allowing for accurate depth assessment even in low visibility situations. To ensure a reliable comparison between the two systems, the DOE framework carefully specifies the water content, lighting, measurement distance, and important experimental factors. Furthermore, interaction analysis is used to look at how system and environmental factors work together to affect each method's accuracy and dependability.