Performance evaluation of low-resolution monocular vision-based velocity estimation technique for moving obstacle detection and tracking
R. Rajesh, P. V. Manivannan · International Journal of Vehicle Autonomous Systems · 2023
Moving obstacle detection and its velocity estimation is crucial in an autonomous vehicle for better manoeuvring in real-road conditions. A modified spatial calibration approach proposed in this paper leverages the monocular vision system to estimate the velocity of detected moving obstacle by solving the Scaling Factor (SF) problem. Further, the Least Absolute Residual (LAR)/bisquare method-based SF smoothing is proposed to estimate the velocity much more accurately in day and night conditions. Moreover, the effect of camera resolution in velocity estimation has been studied by testing the developed moving obstacle detection and velocity estimation algorithm with data obtained from a Low-Resolution (LR) and High-Resolution (HR) monocular vision camera and also compared with the estimated velocity from the stereo vision system. The comparison concludes that using the proposed approach, the LR monocular vision provides a cost-effective, computationally inexpensive solution with reduced error and better detection range.