A methodology of vehicle speed estimation based on optical flow

It has great significance to acquire vehicle speed for active safety system. This paper presents a methodology for identifying vehicle speed by obtaining a sparse optical flow from image sequences. Distinct corners can be detected by Harris corner detector after image enhancement. Then, Lucas-Kanade method for optical flow calculation is utilized to match the sparse feature set of one frame on the consecutive frame. In order to improve the accuracy of optical flow, RANSAC algorithm is introduced to optimize the matched corners. Finally, the vehicle speed can be determined by averaging all the speeds estimated by every optimized matched corner. The results of field test indicated that the computation time of the developed method to execute for one time was 59ms, and the mean error of speed estimation relative to the measurement of GPS was 0.121 m/s. The developed method can achieve satisfying performance, such as accuracy and output frequency.