A comparison of low-cost monocular vision techniques for pothole distance estimation

Abstract

Consider a single camera mounted on the inside of a vehicle's windscreen used for detecting potholes and other obstacles on the road surface. This paper outlines three approaches to the depth estimation problem of determining the distance to these obstacles in the range of 5 m to 30 m. We provide an empirical evaluation of the accuracy of these approaches under various conditions, and make recommendations for when each approach is most suitable. The approaches are based on the pinhole camera model: the simplest approach is based on the geometry of similar triangles, another employs the cross-ratio of a set of collinear points, and the final approach relies on calibration of the camera matrix. We recommend the use of the cross ratio approach for a fixed camera setup and depth estimation almost directly ahead, and an approach using similar triangles when predicting distances at wide angles or adjusting the camera height may be necessary.