Research On Vehicle Target Detection And Tracking Technology Based On UAV

With the rapid development of UAV in recent years,advances in hovering,endurance,and gimbal camera technology have provided new directions of the development for many industries.At the same time,with the development of computer vision and deep learning,the ability to detect and track targets is becoming stronger and stronger.There is a great value of combining the two for vehicle detection and tracking.This research fits the current research content of intelligent transportation systems,and it illustrates the development trend of future road traffic research.The difficulties of the research lie in not only collecting vehicle images on UAV,but also algorithmically implementing vehicle target detection and tracking in complex backgrounds.At the same time,it is necessary to overcome the problem of image processing capabilities for airborne computing centers compared with traditional computers.The entire UAV system has higher requirements in miniaturization and rapid response.Based on UAV platform,this dissertation studies the related technology of vehicle target detection and tracking.The main research of the dissertation includes:(1)System deployment,including UAV image acquisition units and airborne computing centers.The whole process of vehicle image collection,processing,and calculation is integrated in the platform.Compared with the traditional camera system,it has the characteristics of low cost and high efficiency.(2)Vehicle detection algorithm,the optical flow method,frame difference method,background difference method,and deep learning-based algorithms are compared and analyzed.Finally,the YOLO algorithm in deep learning is used to detect vehicle targets,and PVcar vehicles are made separately for the vehicle targets.The data set is used as the basis for subsequent vehicle tracking.(3)Vehicle tracking algorithm,a tracking algorithm is proposed by combining YOLO and Camshift.Based on the advantages of YOLO and Camshift algorithms,the first frame uses a method of image enhancement called Retinex to make it easier to detect vehicle targets with YOLO algorithm.Camshift performs the subsequent tracking.In the case when the vehicle target is partially occlusion,the target is updated by the YOLO algorithm to avoid the loss of vehicle target during the tracking process as well as test it on UAV system to verify its effectiveness.(4)Integrated technology,the integration of UAV system together with vehicle detection and tracking algorithms are realized,and the prototype of the system is developed.According to the UAV system,a vehicle tracking software is designed to save and output the position after tracking the vehicle target,and generate a tracking log for searching.