To address the issue of motion perception during ship mooring,a binocular camera-based ship mooring motion perception algorithm was proposed,achieving pose detection during ship mooring.The algorithm adopted a perception strategy of recognition followed by measurement.First,the Mask-R CNN neural network was used to obtain the ship′s mask;then,based on the ship′s mask,binocular disparity map,and the internal and external parameters of the binocular camera,the ship′s point cloud was calculated.Finally,the NDT-ICP algorithm was applied to register the ship′s point cloud,yielding the pose transformation matrix,which was then used to compute the ship′s pose during mooring.Compared with the measured values,the algorithm achieves a minimum displacement RPD of 2.21 mm,a minimum rotation RPD of 2.93°,a maximum displacement RMSE of 1.135 mm,and a maximum rotation RMSE of 3.809°,with a detection video frame rate of 8 FPS.Therefore,the algorithm is approved to demonstrate good detection capability and hold significant value for ship mooring activities.
维普
万方数据