Developing a Camera-Based Perspective Transformation Method for Quantifying Driver Direct Visibility for Passenger Vehicles

This study outlines a camera-based perspective transformation method for measuring driver direct visibility, which produces 360-degree view maps of the nearest visible ground points. This method is ideal for field data collection due to its portability and minimal space requirements. Compared with ground truth assessments using a physical grid, this method was found to have a high level of accuracy, with all points in the vehicle front varying less than 0.30 m and varying less than 0.6 m for the A- and B-pillars. Points out of the rear window varied up to 2.4 m and were highly sensitive to differences in the chosen pixel due to their greater distance from the camera. Repeatability through trials of multiple measurements per vehicle and reproducibility through measures from multiple data collectors produced highly similar results, with the greatest variations ranging from 0.19 to 1.38 m. Additionally, three different camera lenses were evaluated, resulting in comparable results within 0.12–0.99 m. A parametric study looking at seat position and eye height suggests that assessing different eye heights may be the most insightful method for identifying the range of possible driver visibility. The visibility maps created using this method will allow researchers to assess driver blind zones and how they change with vehicle structural trends over time, assess the role of blind zones in crash scenarios with vulnerable road users, and provide consumers with information about comparative visibility for potential vehicle purchases.