Person Target Tracking System Of UAV

With the continuous development of UAV technology,the tracking and photographing of person targets,which base on a good image platform for computer vision with the good vision of UAV in the air,has gradually become a famous direction in the development of UAV technology.However,when persons enter a complex environment,due to the shielding of the environment,the change of their own state,and the change of illumination,UAV-based person target long term tracking facing some problems such as target loss,tracking failure,target detection failure,etc.Therefore,this paper analyzes the difficulties in the UAV person target tracking task,research on the person target tracking method of rotorcraft UAV in complex environment.Based on deep learning,this method builds a centralized target tracking system around the UAV flight platform,which can be more effectively used in multimedia,military and other fields.This paper takes the UAV person target tracking in a complex environment as the background.Therefore,the complex environment with a large number of trees,roads and vehicles is selected as the test environment to conduct a systematic study of this paper.The specific contents are as follows:First,in view of the traditional target detection algorithm embodies the detection precision of the problem by researching the basic structure of neural network based on the depth of the airborne platform is constructed to finish the ground target real-time detection and precise positioning method,improve the target detection and target feature fusion network,in a complex environment based on flight platform,this paper realizes the real-time target detection of human body.Secondly,in-depth analyzes and experiment contrast the deficiency of the traditional target tracking algorithm,aiming at the object shielding in traditional algorithm and the state change of the lack of response,this paper research on the Re ID method,a method based on the measurement study,based on Re ID method,this paper build a real-time target tracking method in flight platform.Then,this paper studies the monocular camera image target motion state analysis method based on the target detection and target tracking results,analyzes and constructs the rotorcraft UAV target tracking simulation system,and verifies the rotorcraft UAV person target tracking ability in the simulation system.Finally,this paper deployed the rotorcraft UAV target tracking system on the rotorcraft UAV platform,and verified the person target tracking ability in a complex environment through quantitative experiments,and verified the robustness and real-time performance of the system against person occlusion and target motion state changes in the real environment.The system finally realizes the real-time and stable person target tracking in complex environment such as occlusion,target state change and approximate feature interference.