Research On Detection Method Of Sea Surface Obstacles For USV-borne Millimeter Wave Radar

Unmanned surface vehicle(USV)is a new unmanned platform which plays an important role in water quality detection,investigation and maritime search and rescue.It has the advantages of small size,flexibility and autonomy.Accurate detection of obstacles in the surrounding environment of USV is the premise of autonomous obstacle avoidance,so as to ensure navigation safety and complete tasks autonomously.The millimeter wave(MMW)radar has high accuracy in short-range detection,strong anti-clutter interference ability,and will not be interfered by environmental factors such as light.When it is used to detect sea obstacle target in front of USV,it can effectively complement other detection sensors such as navigation radar and binocular vision.However,due to the complex sea environment,there are three major problems in the original detection data obtained by MMW radar,such as false alarm,target plots splitting and target missing detection.Therefore,in order to ensure the navigation safety of the USV,the method of plots-centroid is studied in detail in this thesis,aiming at false alarm and target plots splitting,and on this basis,aiming at false alarm and target missing detection,through in-depth study of multi-target tracking method,further optimize the detection results,provide more accurate and effective obstacle target information for autonomous obstacle avoidance of USV.Firstly,in order to reduce the false alarm rate and solve the problem of target plots splitting,a plots-centroid method based on weighted Euclidean distance is proposed.Combined with prior knowledge such as radar echo intensity and effective detection range,the threshold filtering method is used to remove invalid plots.The weighted Euclidean distance between plots is used to measure the information similarity,and the target plots clustering is realized.The position and section width of the obstacle target are calculated,and the range of the obstacle target is represented by the rectangular dangerous area,so as to realize the accurate detection of the obstacle target in front of the USV.The actual ship experiment show that the results obtained by this method are more accurate than the traditional methods,and more suitable for autonomous obstacle avoidance of USV.Then,because there are still a small number of false targets in the result of plots-centroid,and the target missing detection has not been solved,this thesis proposes a multi-target tracking method in front of USV based on attitude indicator correction and two-step association.The coordinate transformation of the target is carried out with the attitude indicator to solve the problem that the abnormal change of the target position information caused by the real-time change of the radar coordinate system affects the target association.After the first target association is completed according to the initial association gate of each known target,the gate of the known target that has not been successfully associated is expanded to carry out the second target association.The accurate target association is realized,and the information of the lost target is estimated.The life cycle method is used to judge the target state and screen effective obstacle targets.The actual ship experiment results show that this method can effectively filter out false targets,reduce the miss detection rate from 20% to5% and the effect is better than traditional methods.Finally,through the investigation of several well-known MMW radar brands at home and abroad,the equipment selection is carried out,and the MMW radar detection hardware system is built.In the compiled environment of Microsoft Visual Studio 2017,the software of MMW radar sea surface obstacle target detection is developed by using C#,which realized the functions of receiving,analyzing and real-time processing of the original detection data of the MMW radar,as well as display and storage of detection results.Based on this system,several actual ship experiment have been successfully carried out to verify the effectiveness of the data processing method in this thesis,the practical application results show that the system is stable and reliable.