Research And Application On The Contour Extraction Of Moving Object In Video Surveillance

Along with the high development of computer technology and the continuous emergence of intelligent electronic products, computer vision and pattern recognition technology as an important focus in the field of computer cognitive science has received enough attention and extensive study of people. In the popular sense,computer vision technology is a study of how to make the machine to "see",to "think" and to process.As a branch of computer vision technology. This paper focuses on the research on object detection and contour extraction of moving targets in video surveillance.This paper firstly introduces several classic method of moving object detection, mainly with their basic principles and comparative analysis,including the methods of optical flow,background subtraction and inter-frame difference.This paper also introduces several common methods of image contour extraction in the beginning. Which is different from the contour extraction in video surveillance,here we mainly aim at the contour of static object,including edge detection operator,morphological processing,and active contour model.Then this paper presents a new method of contour extraction roughly of moving object which combines the edge detection operator of Sobel with the inter-frame difference.The algorithm mainly uses the common information of moving objects which is the result of two methods processing the frames.Considering the efficiency of the algorithm in the meanwhile,this paper presents a new method of moving object’s contour extraction which adjusting to the speed.This method can choose the image sequence flexibly depending on the speed of moving objects.In addition,the paper summarizes a generic model of contour extraction.Also,this paper put forward a method of contour extraction precisely based on morphological processing under the conditions of the method of moving object detection presented above. This method firstly filters out the noise with erosion,then restores and strengthens the connectivity of contour with twice dilation,lastly extracts the contour accurately with the method of skeleton extraction technology.Extracting the contour precisely with morphology is based on what the paper presents before.At the same time,in order to verify the rationality of the algorithm,this paper provides several images which processed by the experiment,including the output we need,which shows good practical significance.Lastly this paper does a summary and outlook about the work it had done,mainly focus on the suitable conditions,and the worthy improvement by the next step.