The Detection And Tracking Of Moving Targets With Unsteady Image For Unmanned Surface Vehicle

The surface of the earth is composed of several continents and oceans surrounding the mainland.The sea area accounts for about 360 million square kilometers,accounting for about 71%of the earth's surface area.It can be said that most of the earth is covered by sea water.The ocean is not only the main path of world trade,but also contains abundant natural resources.China and western developed countries have always attached importance to the development of marine equipment and enhanced control and development of the ocean.An unmanned surface vehicle(usually referred to as USV)is an unmanned surface ship that has been developed in recent years.Unmanned surface vehicle can be used for a variety of civil and military missions,such as marine environmental monitoring,personnel search,anti-mine mines;island map mapping,offshore facility maintenance and hydrographic surveys;submarine anti-submarine,anti-special operations and port patrols,combating the sea Smuggling,anti-piracy,etc.Therefore,the research on unmanned surface vehicle has important theoretical significance and practical application value.This thesis relies on "WAM-V-USV" as the carrier research object,and mainly studies the depth detection and tracking method of unmanned surface vehicle moving target under unsteady image conditions.The research goal of this thesis is that the proposed method can be applied to the surface target detection and tracking problem through the practical application of the unmanned surface vehicle.Through the surface mage processing,the information of the moving target or obstacle is obtained,which is a necessary condition for the autonomous operation of the unmanned surface vehicle,which helps to improve the intelligence of the unmanned surface vehicle.According to the research content,this thesis is mainly divided into the following five parts:Firstly,the research status of unmanned surface vehicle at home and abroad and the development of surface unconscious water surface sensing technology are reviewed.The research status of unmanned surface vehicle based on optical visual perception and laser radar sensing technology is reviewed,including water surface target detection technology and Surface target tracking technology.Secondly,the characteristics of the water surface image are summarized.The evaluation indexes PNSR and SSIM for evaluating the quality of the water surface image are studied.The water surface image is verified by the two methods.On this basis,a method for evaluating picture quality based on the three-channel distance difference of color images is proposed,which verifies the practicability and accuracy of the method.The water antenna detection of canny and Hough junction and the water antenna detection combined with SVM and Roberts are realized,and the advantages and disadvantages of the two methods are verified respectively.Finally,the problem of reflection on the water surface target is studied,and a method of eliminating reflection by morphological method is proposed,which solves the problem that the target detection and tracking performance is degraded due to the reflection of the water surface target.Then,the optical image information collected by the camera and the point cloud information collected by the laser radar are used for information fusion,and the optical image is processed by the deep neural network to obtain the target type,the detection confidence and the pixel position.At the same time,the point cloud information of the lidar image is used to obtain the distance and orientation information of the target.Finally,the information is combined to obtain the type of surface target or obstacle,the confidence of the detection,and the distance and bearing information from the unmanned surface vehicle.Next,the tracking method of the water surface target is studied.The improved YOLO-V3 framework,Dlib target tracking and improved optical flow field method are used to track the surface position information of the water surface,and the method is improved.The data set is tracked by the data set.The method was validated to verify the performance and effectiveness of the method used to track targets on the surface.Finally,the architecture of the "WAM-V-USV" test platform is introduced,and then the environment-aware system is designed based on the hardware system and software system of the test platform.The field test is carried out to verify the information fusion target detection algorithm and target tracking algorithm based on deep learning,and the test data is processed and analyzed to verify the accuracy and practicability of the target detection and tracking methods.