Research On Pedestrian Gait Identification And Tracking System In UAV Scene

With the vigorous development of UAV technology and artificial intelligence technology,UAVs carrying cameras and other visual sensors to complete machine vision tasks show excellent performance.In recent years,with the development of deep learning,target recognition and tracking technology has developed rapidly.UAV platform equipped with target recognition and tracking algorithm is widely used in human-computer interaction,intelligent transportation,military and other fields,making people’s life more intelligent.However,in the process of target recognition and tracking,UAV has some problems,such as long target distance,low resolution and scale change.In order to solve these problems and enable UAV to accurately identify and track targets,this topic studies the pedestrian gait identification and tracking system in UAV scene,including gait identification algorithm,target tracking algorithm and the construction and experiment of UAV target locking system,The specific research contents of this paper are as follows:Firstly,based on the characteristics of UAV far away from pedestrian targets and low image resolution,gait features are used as long-distance identity recognition features.Aiming at the problem that the recognition accuracy of gait recognition algorithm is not ideal when carrying items and clothes are changed,a sequential attention convolution module is improved,and this module is used to build a recognition network and test it on CASIA standard data set.The accuracy of the algorithm is in the ordinary The accuracy of carrying items and changing clothes was improved by 4.9%,4.9% and 4.2% respectively.At the same time,the test of the algorithm on the self built data set further verified the effectiveness of the algorithm in the UAV application scenario.Then,in order to realize the active approach of UAV to the target and improve the robustness of target tracking algorithm to target scale change,this paper improves the scale adaptive correlation filter target tracking algorithm and tests it on the uav123 target tracking standard data set.The accuracy and success rate of the algorithm under the challenge of scale change are improved by 5.0% and 17.2% respectively,which can realize stable tracking in UAV scene.The experiments show that the UAV can complete the task of spatial tracking and image processing,and the UAV can complete the task of target recognition.