Font Size: a A A

Background Extraction And Moving Target Detection In Surveillance Video Based On SIFT Algorithm

Posted on:2019-02-26Degree:MasterType:Thesis
Country:ChinaCandidate:L CaoFull Text:PDF
GTID:2428330545483683Subject:Control Engineering
Abstract/Summary:PDF Full Text Request
Surveillance video technology is widely used.The detection and tracking of moving targets in video is a key and difficult point.The moving targets in surveillance video can be divided into two parts:background extraction and foreground detection.The foreground detection is the detection of moving targets.At present,there are still many deficiencies in the background extraction and foreground detection of surveillance video.Taking into account that SIFT feature points have good robustness to environmental changes,such as brightness changes,illumination changes,rotations,noise,and so on,this paper uses SIFT algorithm to detect and track the moving target in monitoring video.For the background extraction problem,this paper extracts the background SIFT feature points,describes the "background" in the form of feature points.To some extent,eliminating the impact of environmental changes.For the moving target detection,the background feature points of the background frame and the moving target image frame are matched to remove the background feature points in the moving target image frame,and the moving target feature points are obtained,and then the moving target is detected.In order to detect the moving target in complex background,this paper appropriately increases the number and density of the feature points of the image frame,and matches them by blocks.Then the position information of the SIFT feature points is used to remove the background that the small matching cannot be removed.Experimental results show that under the complex environment such as illumination and shadow,this method is better than the traditional algorithm such as background difference method,frame difference method and so on.Finally,based on the detection of moving targets,this paper makes use of the nature of SIFT feature points and the position parameters in SIFT to track the target through SIFT feature point matching.Compared with other traditional target tracking algorithms such as MeanShift tracking algorithm,the advantages of this method are that the target tracking can be carried out directly on the basis of target detection without pre given moving target or manually selecting target area.Secondly,the algorithm has low complexity,high efficiency and moving target.The velocity of the target is insensitive.
Keywords/Search Tags:Surveillance video, SIFT algorithm, Background extraction, Target Detection, Target Tracking
PDF Full Text Request
Related items