Motion human detection based on difference background image
Yubin Wu · Guangdian gongcheng · 2007
Based on difference background images, an improved real-time background update algorithm was proposed. With the background learning process, the mean square error of grey value of each pixel was estimated by sequence images. According to the 3σ rule of normal distribution, the influence caused by the variety of background was avoided. A threshold segmentation algorithm based on the average grey value of the image was applied, which made the algorithm adapt to different illumination. A shadow removing algorithm based on the vertical projection was proposed. It was shown that there were different vertical projection characteristics between human body and shadow. It was used to eliminate the shadow of the motion human efficiently. The experimental results demonstrate that the algorithm can extract the motion human effectively and completely under different cases of illumination.