Moving object detection algorithm based on template matching in wavelet transform and color segmentation

One of the problems to be solved in moving object tracking is how to detect moving objects in dynamic background. In order to track moving objects accurately and to reduce computation time, a new algorithm based on template matching in wavelet transformation and color segmentation is proposed. Firstly, the serial frames are transformed using wavelet, then the local region containing object can be obtained in low-low frequency sub-band by global optimal searching method using template matching, which reduces the searching time greatly; next the. color information of original object is distilled, and the local area of each frame is transformed into HSV color space for color segmentation; finally the color feature multi-channel fusion method is carried for the above segmentation results which can detect and track the moving object more accurately. The international standard testing videos and practical videos are used in our experimentation to test the algorithm proposed here. Experimental results indicate that this algorithm is highly effective in detecting and tracking moving objects in dynamic scene.