Abstract
In the road video shooting of unmanned aerial vehicle (UAV), the UAV is in an unconstrained state, and the video picture is jitter and unstable. When tracking multiple vehicles in uav video, identity switching of tracking targets is easy to occur. Based on this, an improved DeepSort vehicle target tracking algorithm based on Yolov5s is proposed. When Yolov5s trains the vehicle detection model, DIOU-NMS is used instead of NMS to remove the redundant target position prediction box. Aiming at the fact that DeepSort pre-trained appearance extraction model did not contain the appearance model of the vehicle, the vehicle appearance model was obtained by using the lightweight ShuffleNet V2 network for vehicle rerecognition training on VeRi data. The improved algorithm is used to evaluate the MOT of UAVDT data sets. The experimental results show that compared with the original algorithm, the tracking identity switching (IDSW) of this experimental method is reduced by 49.1%, MOTA is improved by 3.2% and MOTP is improved by 0.3%.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
