Monocular vision based distance estimation algorithm for pedestrian collision avoidance systems

Abstract

this paper presents a novel and effective distance estimation algorithm based on monocular vision for pedestrian collision avoidance systems (PCAS). Vehicle -camera geometry and intrinsic parameters of camera are used to estimate the distances of pedestrians detected in front of the ego vehicle. An innovative method to estimate the pitch angle for the camera is provided in the proposed solution. The advantage of the intrinsic parameters based method is that the algorithm is computationally efficient to meet the real time requirements of the system. In addition to this, it takes into account the system requirements, in the sense, that instead of estimating point to point distance between the camera and the target pedestrian, the vertical distance is estimated with the help of vertical and pitch angles. Practical issues like pixel density per distance value shrinkage at higher distances, lens distortion and jitters are addressed and solutions to them are provided. In multiple experimental tests under real world driving conditions, it has been confirmed that the proposed solution works effectively with a minimal percentage mean error rate.

Vision-based Approach for Automated Social Distance Violators Detection

“…However, most of the applications of distance estimation and motion predictions are in the advanced driver assistance systems (ADAS). Some of these applications are proposed in [10], [12].…”

The Correlation between Vehicle Vertical Dynamics and Deep Learning-Based Visual Target State Estimation: A Sensitivity Study

“…To cope with varying pitch angles a virtual horizon can additionally be computed and used as a reference line [43]. Another method is measuring the top or bottom light ray hitting the camera’s imaging sensor to compensate sensor movement [44]. With a Lane Departure Warning system on board, the acquired lane information can be paired with vehicle classifiers to generate a solid ground-plane and design a more robust distance estimator [38,45].…”