Research On Detection And Tracking Technology Of Moving Objects Under Video Surveillance

Moving object detection and tracking is the focus of research and application in the field of computer vision.As the key technology in intelligent video surveillance,it is mainly used in object detection,intelligent monitoring and other fields,and the effect of detection has a great influence on subsequent processing such as object positioning and tracking.Although people have done a lot of research on these two technologies,there are still some problems that need to be solved.Based on the previous research,this study takes intelligent transportation as the research background,and uses the key technologies in intelligent video surveillance to conduct more in-depth research on the detection and tracking technology of moving object under surveillance video,and realize the detection and tracking of moving object in the surveillance video area.This paper mainly includes the following aspects of the study:(1)In the aspect of moving object detection,this paper first introduces common detection methods:optical flow method,inter-frame difference method and background subtraction method,and analyzes the advantages and disadvantages of these three methods.In the process of moving object detection,there are some challenges such as the disturbance of chaotic background,the influence of illumination,noise and shadow.In this study,a background subtraction method based on wavelet blocks is proposed for the challenge in object detection.In the background modeling stage,a Gaussian background modeling method with less running time is provided.The background is reconstructed based on Gaussian mixture model(GMM)of the mean images of image blocks,aiming to simplify the calculations so as to improve the speed of the corresponding operations.In the foreground detection stage,a wavelet-based de-noising method with the semi-soft threshold function is applied to de-noise the object images of the foreground.In the background maintenance stage,an adaptive background maintenance algorithm is provided to dynamically update the background.Experimental results show that the computational complexity is reduced,while the accuracy and adaptability are improved by using this method.(2)In the aspect of moving object tracking,this paper introduces the basic theory research on moving object tracking:random finite sets(RFS),probability hypothesis density(PHD)filtering algorithm,cardinalized probability hypothesis density(CPHD)filtering algorithm and optimal sub-pattern assignment(OSPA)distance.Aiming at the problem of parameter estimation under unknown conditions of detection and clutter,a tracking algorithm for Gaussian mixture PHD filter based on time-varying filtering under extended state space model is proposed.In the update stage,time-varying filtering is introduced to update the measurement values,which can effectively reduce missed detection and false alarm rate in object number estimation,and improve the performance of the tracking algorithm.Aiming at the confusion between missed detection target and clutter in measurement and clutter unknown CPHD filtering algorithms,a tracking algorithm for Gaussian mixture CPHD filter based on adaptive parameter learning mechanism is proposed.In the update stage,the object multiple estimation process based on the measure likelihood function is introduced,and its mixture cardinality distribution is modified in a targeted way,so as to effectively solve the confusion problem and improve the tracking effect of the target tracking,and improves the tracking effect of the target tracking.Experimental results show that the accuracy of the proposed algorithm for tracking targets is high,and stable tracking of moving targets is achieved.