Font Size: a A A

Moving Object Detection Based On Gaussion Mixture Model

Posted on:2013-10-24Degree:MasterType:Thesis
Country:ChinaCandidate:M N YinFull Text:PDF
GTID:2248330377959184Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
As the spread of the multimedia, video images have been affected by the attention ofpeople. The moving target detection extracted prospects change area from the backgroundimage. The detect technology is based on moving region corresponding pixels, so theeffective extract moving region for target classification, target tracking and behaviorunderstanding and so on work is of vital significance.The method of moving video target detection is the foundation of picture processingtechnology, but how to detect the moving targets efficiently and accurately is a hot issue.Moving targets detection can be divided into two levels: moving target detection methods instatic background and moving object detection methods in the movement background. Instatic background, though the traditional method can detect the moving targets, because ofthe light, the shadow, background noise causes the range of detected moving target is toobig and the internal cavity. In movement background, how could be effectively estimate thecamera movement component and compensate the camera movement situation, that is thekey point of target detection.According to the problems, for the static background, the paper presents a movingobject detection method combine SUSAN edge detection operators and mixture Gaussianmodel background subtraction. This method uses the Gaussian mixture model backgroundsubtraction extract prospect goal, then by morphology processing rough extract the outlineof moving target.; At the same time, to this frame image after SUSAN edge detectionprocessing. The two methods adopt "parallel" way, then the results after "and", extract themovement of more accurate target. For the moving background, the paper presents TDS fastblock matching search method to estimate the video image, after estimate the component,by motion compensation methods remove the camera movement component, that istransform from "move" into "static", then use the target detection methods of video imagesin the static background to detect the moving object.Proved by the experiment, according to the static background, this paper presents theresearch methods compared with traditional method can not only effectively detect movingtargets, but also can effectively overcome the effects of shadow, light and noise; For the moving background, this paper presents method can effectively improve the estimation ofthe camera movement and accurately detect the moving target.
Keywords/Search Tags:gaussian mixture model, background subtraction, SUSAN edge detection, movement target detection
PDF Full Text Request
Related items