Algorithm of Vehicle Trajectory Extraction using Fusion of Hu Moments and Grayscale Feature

摘要

In order to perform vehicle behavior analysis, the accurate extracting vehicle trajectory becomes more and more important. This study proposes an algorithm for vehicle trajectory extraction based on the fusion of Hu moments and grayscale feature, in which the main task is about image matching. Firstly, SIFT and the pavement image matching algorithm based on the fusion of Hu moments and grayscale feature are used to obtain matching points and the wrong matching points are removed by using a custom classification method. Then according to the position change between matching points, offsets and rotation angles between each two images are calculated. Finally, the vehicle trajectory is obtained by the offsets and rotation angles for all the measuring images. Experiments show that the measuring accuracy of the algorithm meets the requirements of vehicle behavior analysis and the vehicle trajectory can be extracted accurately when the vehicle doesn't receive GPS signals.