Research On The Technology Of Target Detection Based On Visual Attention And Unit_Linking PCNN Hole-filter

Based on camera motion or not during collecting the video information, moving targetdetection can be divided into the static background detection or the dynamic scenedetection. It is one of the important contents and key technology of machine vision. With thedevelopment and application of machine vision, moving object detection technology hasbeen applied widely in many fields.This paper based on the video image sequence studies on the target detection techniquein dynamic scenes. It was proposed to combine Itti visual attention and Unit_Linking PCNNhole-filter algorithm together to achieve the detection of moving target in dynamic scene.HSV color space for color description method is more in line with the way of humanvisual system when understanding and descripting the target color. Therefore, in thispaper, improving the Itti visual attention model in the HSV color space, making it moreclose to the habit of eye observing and describing the target, and shorten the running time ofthe algorithm.The traditional Itti visual attention model does not account for the movement status ofthe target in a scene, the paper by adding motion vector feature channels in it, makesthe moving target in the notice, and is able to detect the moving object in the dynamic scene.The target boundary is vague through the Itti model extracting, and the targetintermediate sometimes appeared empty, so that the target object is not connected. Or thedetection area is too small to be surrounded by the detected moving targets. So it must be tointroduce hole-filter to make the target detection out. In this paper, use Unit_Linking PCNNto make it be hole-filter, then get the area of moving target. Through the experimental analysisand detection efficiency is satisfactory, which proves the feasibility and effectiveness of themethod.