Moving vehicle detection and tracking in image seuqnces

摘要

This paper presents an algorithm that tracks moving vehicles in an outdoor environment. The algorithm was developed based on four heuristis deived from the properties of moving vehicles, maxium velocity, small velocity changes, coherent motion, and continuous motion. The algotihm was developed based on four heuristics on moving vheicles, namely maximum velocity, sall velocity change, choeretn motion and continuous motion. the algotihm has three major computational components, compute pixel motion vectors using a search window and a matching whindow, eliminate irrelevant moving poxels using a proximilty flter and directional filters, and track the oving vheicles through multiple image frames using a proximity filter. The algorithm has been implemented and tested on image sequences taking in outdoor environments. The test results are presented in the paper.