Research On Position And Attitude Recognition Algorithm Of Unmanned Surface Vehicle Based On Binocular Vision

As the product of the combination of unmanned technology and traditional ship,USV(Unmanned Surface Vehicle)has the characteristics of strong maneuverability,high level of informatization and intelligentization.USV are usually deployed and recovered by medium and large ships or shore-based facilities,especially for the short-range recovery of USV,which requires higher accuracy and stability of positioning and orientation of navigation system.In this paper,an actual camera model is constructed based on the pinhole camera model and the principle of lens imaging,and the calibration method of a single camera is deduced.The mapping matrix is calculated according to the internal parameters of the camera obtained by calibration and used for camera imaging distortion correction.Furthermore,the binocular camera model is constructed according to the single camera model and the principle of epipolar geometry,and the binocular correction steps are deduced.Based on YOLOV5 target detection algorithm,a target detection system of USV is established in this paper,and the image is collected and enhanced for model training.Among these models,the one with m AP@0.5 being 0.993 and m AP@0.5:0.95 being 0.762 is selected for testing.The results show that the model selected can accurately identify USV with different poses within the range of 40 m.On the basis of the above research work,this paper proposes an USV position and attutiude recognition algorithm based on SURF feature point detection and stereo matching.The algorithm detects the feature points of the USV area and performs filtering,clustering and horizontal pole line matching,then realizes the identification of the USV’s position and attitude based on the principle of stereo imaging.A virtual simulation model is established to verify the algorithm.The results show that theoretically the average attitude angle error of the algorithm is 0.902° and the average position error is 0.010 m in a range of 5m,which verify the feasibility of the algorithm.Finally,an experimental platform for USV position and attutiude recognition is built,calibrated and tested.The USV’s position and attutiude obtained by the algorithm in this paper and RTK device during recovery are compared and the error analysis is made.According to the analysis results,the average position error is 0.59 m and the average attitude angle error is 4.40°.The research work in this paper provides a reference for the research and development of the guidance system for short-range recovery of USV based on vision.