Research On Human Body Detection And Tracking Technology Based On UAV Vision

In many areas,human detection technology has a universal application,and it is the focus of research in the field of computer vision.The contents of human detection include the extraction of human body,the detection of human body,the identification of human body,the tracking of human body and so on.There are some difficulties and challenges in human body testing.The reason is that the complexity of the application of human detection technology and the environment is different.In addition,the detection of the movement of the human body has a random,the diversity of attitude and clothing differences,and the human body may also be blocked when detected.In real life,the commonly used body detection system is a fixed camera monitoring system.This system detects and tracks the human body entering the target area.However,when the human body leaves the target area,the monitoring system can not continue to track the target human body.In view of the limitation of the above-mentioned human detection application scene,this paper puts forward the research on the detection and tracking of the target human body with the small four-rotor unmanned aerial vehicle under the air moving platform.In this paper,the main innovations are as follows:1.In the human body detection,In this paper,we adopt the methods on human HOG features as well as support vector machines SVM.In this paper,the PCA algorithm is used to reduce the HOG dimension,when the real-time requirements can not be met,due to too many HOG feature dimensions.In the process of human detection based on HOG characteristics,this paper adopts the HOG feature combined with the template matching algorithm to improve the accuracy of human detection,in order to solve the problem of missed and false detection.In the course of pedestrian movement,because the head and shoulders of the pedestrian has a good invariance,the template of this paper adopts the head and shoulder template of the human body.2.In the aspect of human tracking,this paper uses CamShift algorithm to track the human body.However,if the body was completely blocked,CamShift algorithm is prone to the loss of the target situation in the tracking process.So in view of the above situation,this paper uses CamShift algorithm combined with particle filter algorithm to improve the tracking effect of the algorithm.3.In the practical application of human detection technology,when we manually select the frame to track the human body,problems often exist that the frame is not accurate or the size of the box is not appropriate.Therefore,this paper adopts the method of automatically detecting the target human body.This method can realize the automatic tracking of the human body without manual frame.Through the experiment,it shows that the human body detection and tracking algorithm in this paper has a good effect,and this effect can meet the real-time requirements.