Multi-moving Object Detection And Tracking Based On Static Scenario

With the popularity of digital camera, moving object detection and trackinghas gradually become the hot spot of research in the field of computer vision.However, the research of moving object detection and tracking is still facing manychallenging issues because of the complex background, diverse shapes andmovement patterns of the object, mutual occlusion of many objects and so on. Sofurther research is needed. The thesis studies the multi-moving object detection andtracking technology based on the static scene. The main work is as follows:1.About the object detection, the thesis proposes a multiple moving objectdetection algorithm based on VIBE and HOG feature classification. First, find thepossible area of object by VIBE. Then, accurately segment the object by the SVMclassifier based on HOG feature. The experimental results show that the algorithmcan effectively solve the object shadows and objectives adhesion problems whichVIBE algorithm can not solve, and it is better than HOG algorithm in terms oftime-consuming or detection accuracy. The algorithm has good robustness, and canaccurately detect moving objects.2.About the object tracking,multi-moving target tracking problem is convertedinto moving target detection and classification matching, then proposes a multiplemoving objects tracking algorithm based on the detection and classification. Themain innovation consists of the following three points:1) The difficult problems ofobject merger, split and scale change which occurred in the multi-target trackingare solved by detecting.2) Solve the problem of how to integrate the variousobject’s appearance features in the classification by incremental extremely randomforest classifier.3) To avoid the interference of similar appearance object, theobjects are classified by the spatial relationship. Experimental results show that themethod can keep track of multi-target,and has a high tracking accuracy.