Autonomous Target Recognition And Tracking Of UAV Based On Vision

With the implementation of China's strategy of "Rejuvenating The Sea With Science And Technology" and "Maritime Silk Road".the security situation of maritime traffic is increasingly severe.and the risk of marine pollution is also gradually increasing.In order to strengthen the monitoring and management of marine environment.the research of marine and air heterogeneous robots is gradually becoming important.the combination of unmanned aerial vehicles and unmanned surface vessels can overcome several flaws existing in the current single cruise robots-unmanned surface vessels have quite long cruise ability but limited perspective.UAV have a broad field of vision but lack of battery life.however.to overcome their weaknesses by acquiring their strong points.we can form a stable monitoring and management system by combining the two.which can not only enable it to have a broader vision,but also strengthen the cruise capacity.ensuring the effective supervision of the state on the marine environment.Among them.the precise positioning and tracking control of UAV for unmanned surface vehicle has become one of the key researchs of the heterogeneous robots.In this paper,the "Blue Letter" unmanned surface vehicle of Dalian Maritime University and DJI M100 aircraft are taken as the research objects.It proposes a visual positioning scheme which takes deep learning as object detection and takes improved TLD algorithm as object tracking.This paper focuses on the visual processing algorithm for the tracking of the unmanned surface vehicle at a certain speed.and then makes the tracking control of the unmanned surface vehicle comes true.The main research work of this paper is as follows:(1)It analyzed the software architecture and principle of TLD and KCF tracking algorithms.and compared their advantages and disadvantages with each other.According to the advantages and disadvantages of the two algorithms.it improved the TLD algorithm.and the improved algorithm is proved to have a better tracking effect through VOT verification.and the improved TLD algorithm is also applied as the visual tracking algorithm of the UAV target tracking system.(2)It choosed deep learning detection on the link of unmanned aerial vehicle autonomous tracking.After analyzing the SSD algorithm and the feature map of the unmanned surface vehicle.the SSD algorithm was pruned and optimized to improve the speed and accuracy of unmanned surface vehicle detection.(3)In order to realize the tracking of unmanned surface craft by UAV.DJI M100 is selected as the hardware experimental platform in this paper.and on this basis.a separate control architecture is built according to the actual needs.In the software system.the control mode of the truss tracking and flight tracking is designed.Finally.the effectiveness of the tracking control of the UAV to the unmanned surface vehicle is verified through experiments.it provided a solution for the tracking control of the UAV to the unmanned surface vehicle.