Moving Object Selection Using Background Suppression and PCNN

A novel object detection model based on Unit-linking PCNN, optical flow and topological property is proposed. In this paper, according to velocity differences between the moving object and its background, the approach calculates the velocity amplitude of each pixel of the image, through which the flow velocity amplitude are quantized using as threshold to decide whether a pixel is a part of an object or the background. Besides, using Unit linking PCNN whole filter expresses the connectivity and integrity, which are significant topology properties, in moving object attention selection. Experimental results on four databases indicate that the proposed method outperforms other methods both in detection accuracy and background suppression effect under the same conditions.