In view of the existing problems of single data source, data delay and data interruption in sensing the surrounding navigation targets during the navigation of the existing USV, a USV maritime navigation target sensing data fusion method based on radar and GPS data source is proposed. Based on the minimum error method, a radar original image data analysis algorithm is proposed, and the target data is preprocessed through data elimination and time-space unification methods. The track association algorithm model based on Euclidean distance and Mahalanobis distance and the fusion data weight distribution model based on analytic hierarchy process and expert evaluation method are constructed, The overall fusion method is verified by USV measured navigation data. The results show that the target raw data preprocessing method is reasonable and reliable, and the fusion algorithm is stable and reliable, which can provide technical and theoretical support for USV maritime navigation target perception, safe navigation and rapid collision avoidance.