Moving Target Detection Algorithm Research Background Subtraction And Frame Difference Method

The detection of moving targets is an important topic in the ifeld of computer vision, and hasa wide range of applications in the aerospace, video surveillance, intelligent transportation andother ifelds. Studying moving target detection algorithm of video image sequence has importanttheoretical value and practical signiifcance. This paper focuses on the analysis of several commonmoving target detection algorithms which based on background subtraction and frame difference.The main work of this paper is as follows:(1)Analyzed the background of moving object detection, the current situation and videoimage processing theory of target detection. Studied three moving target detection algorithms,which are optical lfow method, the frame difference method and background subtraction. Theprinciples of these algorithms and improved algorithms are analyzed. At last, this paper studiedthe modeling methods of the background subtraction.(2)The ifxed learning rate is adopted by traditional Gaussian mixture model，in other words,the learning rate can not change along with the background, which reduces the efifciency ofbackground updating. An improved moving target detection algorithm is proposed based on thebackground modeling algorithms. The learning rate is divided into two stages by the improvedalgorithms, which are the early stage of background modeling and the stage after the formation ofbackground. The adaptive learning rate is adopted by both of the stages, so that makes thebackground updating immediately, eliminates the blur of a moving target, and improves thedetection accuracy. The experimental results show that the improved algorithm can detectmoving targets more accurately, eliminate shadows, and with a better adaptability and robustness.(3)To solve the problem of incomplete contour when the color of the target and backgroundare similar, this paper has carried on the improvement based on analyzing the traditional framedifference method for target detection. First the improved frame difference method is adopted onthe target for the moving object detection, then we carry the Canny operator on the result aboveto extract the edge information, and ifnally the outcome from Canny operator and the result fromframe difference are done with the OR operation. This method can solve the problem of thesimilar color between target and background better and can get more accurate moving targets.Experimental results show that, the proposed algorithm which based on edge detection algorithmand frame difference algorithm can be applied to complex environment. Also the proposedalgorithm has slight color limitations and better applicability.