Moving Objects Detection Based on Gaussian Models with HSV

Abstract

An important step in video tracking and identification is moving target detection. The new method combined Gaussian background models with HSV remove space shadow, better adapted to the situation of background changed and updated, avoid segmentation errors caused by shadow. First, the Gaussian model was constructed, and a new way was adopted to update background. Then, a coarse foreground image could be obtained by background subtraction. Next, moving shadows was eliminated by a shadow detection algorithm using hue, saturation and brightness information. Finally, we performed post processing, including morphology filtering, connected component analysis and seeded region growing. The experiments prove using the detection results and mean shift algorithm can get a better tracking.