Two-step pedestrian searching method based on Lab color space

As urbanization keeps accelerating in China, video surveillance and image processing technology have been applied to many fields. Since policeman often fail to find their suspector in huge video information with high pedestrian volume especially in malls and at subway stations, pedestrian recognition can reduce time cost when we need to find a specific person. We utilize temporal difference and mathematical morphology to extract objects and denoise them in the video material, analyzing color and texture features through Lab color space. Therefore, we can filter most irrelevant objects efficiently only with a photo or main color of the lost person provided.