Unmanned Surface Vehicle Target Detectionbased On 3D Lidar

The Unmanned Surface Vehicle(USV)is a multi-functional surface artificial intelligence platform.The USV platform’s demand for environmental awareness is multi-faceted.In the process of executing missions at low speeds,the important point is target detection,and on the basis of target detection,targets can be further identified and classified for collision avoidance operations.Or task change.Among current sensors,three-dimensional laser radar is used as the main sensor of the target detection system because of its advantages of high ranging accuracy and large detection range.Therefore,the focus of this study is based on the real-time target detection of 3D laser radar installed on the unmanned ship platform,and further expand the target recognition.This paper will mainly implement the detection of targets in the surrounding environment information.In order to further analyze the target detection results,it will be identified based on the results of the target detection.The target detection aspect first introduced the Velodyne VLP-16 3D laser radar sensor,and analyzed the internal reference model and external parameter model of coordinate calibration,and preprocessed the data of the laser radar.Then,based on the point cloud processing and target detection and recognition method based on grid map,the original 3D laser radar point cloud is rasterized,and further attribute judgments are made.In order to reduce the interference noise,the hanging point and single point filtering in the grid map are separately performed.Because the grid map detection is prone to undersegmentation,the grid-based gradient segmentation method is used to supplement the grid map method,and the gradient points of the laser beam echo points are segmented to provide an effective basis for the grid map clustering.The outline feature extraction of the target.For target identification and classification,this article divides the targets in the USV environment into three categories: ships,small water targets,and other non-water targets.The main idea of classification is to distinguish between surface targets and non-surface targets first,and then to distinguish the ships based on geometric features in the surface targets.For the problems that the ship’s geometry is complex and difficult to identify,three specific feature descriptions are proposed: echo intensity characteristics,height distribution characteristics,and attitude properties.Finally,based on the characteristics of the target’s grid map,real-time target detection based on support vector machine classifier is realized.A simulation experiment was carried out on the vehicle platform to verify the validity of the target detection feature in this paper,and the overall system can have better target detection accuracy and real-time performance.