Research of obstacle recognition method for USV based on laser radar

In order to settle the problem of obstacle recognition for Unmanned Surface Vessel (USV), a novel method is presented to distinguish between vessels and coastlines based on point cloud data (PCD) from ship borne laser radar. Primarily, echo data of inland river targets and noise is analyzed and filtered. Furthermore, targets are extracted from PCD based on distance. Subsequently, three robust target features are presented for targets classification. Consequently, based on the three features, the k-Nearest Neighbors (KNN) classification method is used to identify the obstacle. Experimental results show that the method has good performance on obstacle recognition for USV.