Research And Implementation Of Pedestrian Detection And Tracking Algorithm In Intelligent Monitoring

Video surveillance intelligence is an emerging research direction based on computer vision technology.It has received extensive attention and research.Pedestrian detection and tracking is an important part of intelligent monitoring.Due to the influence of factors such as posture changes and mutual occlusion between pedestrians in the monitoring scene,it has always been a hot spot and a difficult point in the field of computer vision.In order to solve the problem of accurate pedestrian detection and tracking under the monitoring scene,the existing pedestrian detection and tracking methods are deeply researched and analyzed.According to the target occlusion in the pedestrian detection process,as well as the target scale change and non-rigid deformation in the pedestrian target tracking,the related algorithms are optimized and improved,and an accurate pedestrian detection and tracking algorithm framework is designed.Test verification was carried out in the substation monitoring scenario.The main work of the thesis is as follows:1.The existing pedestrian detection algorithm is deeply studied,and the reason for the poor detection effect of the adjacent target occlusion is analyzed.Based on the Aggregate Channel Features(ACF)pedestrian detection algorithm,a non-maximum Suppresion(NMS)algorithm is designed to reduce the missed detection rate when adjacent targets occlude each other.The effectiveness of the improved algorithm is compared and verified by experiments.2.The existing single target tracking algorithm is researched and analyzed,and the problem of inaccurate scale estimation and target non-rigid deformation leading to tracking failure is discussed.A multi-scale estimation mechanism is designed to accurately estimate the scale information of the tracking target.At the same time,a method based on gradient direction histogram(HOG)feature and color histogram feature fusion is designed to solve the target non-rigid deformation.The problem that caused the tracking to fail.The experimental comparison and analysis are carried out to verify that the improved algorithm can effectively solve the problem of scale estimation and target non-rigid deformation.3.The design and implementation of the pedestrian detection and tracking algorithm framework in intelligent monitoring is completed.By integrating and integrating the above improved detection and tracking algorithms,and adding a peak sidelobe ratio(PSR)based tracking state detection module and tracking information recording module,a complete automatic pedestrian detection and tracking algorithm framework is formed..The pedestrian target in the actual monitoring scene of the substation can be effectively detected and tracked,and the tracking state is automatically determined by the tracking state detecting module and relevant information is recorded.The algorithm framework runs on the experimental platform in line with real-time requirements and has strong practical value.The research results show that the improvement of the non-maximum suppression module in the detection algorithm can effectively solve the problem of detector miss detection when adjacent targets are occluded.The multi-scale estimation mechanism is introduced to effectively improve the accuracy of the single target tracking algorithm.Sexuality;through the fusion of HOG features and color histogram features,it can effectively solve the non-rigid deformation problem of pedestrian targets during tracking.