Moving Human Detection And Tracking Based On Regions Of Interest

With the development of computer technology and image processing technology in recent decades,target detection and tracking system have been widely applied in various fields.Based on the in-depth researches about moving target detection and tracking of video sequences,People proposed a variety of effective methods.Especially,the detection and tracking of moving human have become one of the hot research topics for application prospects and potential economic value in the field of intelligence surveillance.First of all,the paper summarizes the research background and significance of the topic,and briefly introduces some detection and tracking systems of video-based,then analyzes and compares the proposed detection and tracking methods.It pays attention to the research how to select the ROI(regions of interest) of the human body as a target region and how to do the subsequent tracking.In terms of target detection,the automatic detection of ROI in moving human is introduced further,according to the background subtraction to obtain the complete region of human body.This method successfully simulates the process of human visual system to form the fixation point of moving human using Saliency-map,and automatically detect the ROI in body by region of joint saliency based on multiple fixation point.The results show that it can be regarded as a stable and reliable automatic detection method of ROI in human body,the detected region accord with the attention location while eyes observe the moving target,which can be applied in subsequent moving human tracking based on region color-feature.In terms of target tracking,proposes a SSD-MS combined tracker based on ROI,which combining respective advantages of the SSD tracker and Mean-Shift tracker.Aiming at the quick movement of human,scale changes and occlusion etc,the Kalman filter is used to estimate the state of target.Especially,when the object undergoes partial occlusion,the multiple small regions with Mean-Shift by using the points of joint saliency generated in the process of automatic detection,which can effectively solve the tracking in the case of partial occlusion.A variety of experiment tests of video sequence with moving human proves that combined tracker with the ROI by automatic detected can effectively accomplish the tracking of moving human,in case of the quick movement,scale changes and occlusion etc.