Target Detection And Tracking System Of An Unmanned Surface Vehicle

The demand for USV is rising,as a result of increasing danger,dirtiness and dullness of marine operations.Meanwhile,new technology,including high quality sensors,high bandwidth communication and intelligent controller provides a possibility for USV.Obstacle detection and target tracking is a basic task for USV,and is can be widely applied to marine military operation,marine research and marine rescue.In this paper,a complete system based on GNC structure is developed to achieve target detection and tracking,and its validity is verified by a trimaran model test.The software system contains seven modules,namely,data preprocessing,data segmentation,feature extraction,data association,state estimation,mapping and target tracking control.Data preprocessing transforms original data into 2D matrixes,which decreases the computation quantity of the whole system.Data segmentation generates separated clusters with DBSCAN method.Features of each cluster is extracted with PCA.Data association establishes relations between known objects and clusters using nearest neighbor algorithm.State estimation is accomplished by Kalman Filter,which predicts and estimates moving state of objects.Mapping module stores increasing information of static obstacles.Target tracking control module uses heading controller,which is based on pure pursuit navigation,to achieve target tracking navigation.An autonomous target detection and tracking system for Unmanned Surface Vehicles(USVs)is developed in this study.The system uses scanning data of Lidar to detect obstacles in the environment,extract their features and autonomously track a target with certain features.