Moving object detection based on image sequences

According to the problems of numerous noises and incomplete detection of moving object in dynamic environments, a method for moving target detection based on pyramid-style multi-resolution model was proposed. The areas of target were obtained in low resolution and the details were in high resolution. For complex environment, an image difference operation scheme using high/low double-threshold substituted for traditional single threshold was proposed, which could obtain thresholds automatically. The scale transformation of the current and background frames was carried out first and a group of images in different resolutions were obtained. Then the high/low threshold differential images in different scale spaces were got. At last, an image fusion process was done from the high layer to the low layer, and an integrated object image with few noises was obtained. Experimental results show that the proposed method achieves a high accuracy for the moving object extraction, it has obvious advantages especially in complex dynamic environment. The precision is 0.802 for single object and 0.615 for multi object.