Target Recognition and Tracking in Outdoor Environment

Target recognition and tracking is a tough task in computer vision. How to recognize a target accurately and tracking robustly is still a core problem studied by researchers. And the task is even more challenging in outdoor environment. In this paper, a recognition and tracking algorithm suitable for a natural target in outdoor environment is introduced. The natural target we choose is the overhead street lights. The recognition method is based on a template matching algorithm. The template is defined in advance according to the shape of the lights. The proposed tracking algorithm is based on particle filter. The difference between our method and the traditional particle filter is the image feature used. In the proposed method, the particle filter is based on the area of the target rather than the color histogram. In order to cope with the illumination changes, adaptive threshold adjustment is introduced in this paper. Also, the adjustment of camera exposure time is discussed.