Research On The Pose Recognition Method Of Large-size Cabin Based On Machine Vision

At present,the position and attitude detection methods of large-size cabins are mainly to paste visual marks on the surface of the cabin or install target tooling on the end face of the cabin,and detect the visual marks through monocular or multi-eye cameras.The spatial coordinates of the visual landmarks are solved.Since the visual landmarks and the cabin are rigidly connected,the cabin coordinate system can be established according to the coordinates of the landmarks,and the cabin pose can be solved.Errors will occur in the process of pasting the surface of the cabin section and the manufacturing process of the tooling target.The error will be superimposed during the cabin section pose detection process,which will eventually increase the error of the detection result.Reduced the accuracy of cabin pose detection.This paper proposes a pose detection algorithm for detecting large-sized cabins.A cabin pose detection system based on binocular vision is built,and the camera,lens,and cabin pose adjustment platform are selected.The overall process of cabin pose detection is designed.Take a partial image of the end face of the cabin,and take the bolt hole on the end face of the cabin as the detection object.In the process of bolt hole detection,the existing ellipse detection algorithm based on arc support is improved,so that the algorithm can detect the ellipse equation corresponding to the bolt hole.Calculate the spatial coordinates of the center of the bolt hole on the end face.The coordinate system of the cabin is established according to the center of the bolt hole,and the pose of the cabin is calculated.The algorithm is verified by the built cabin position and attitude detection system,and the experiments are carried out on the experimental cabin in the static state,the position movement and the attitude rotation state respectively.The experimental results show that the absolute error of the cabin pose detection algorithm in the translation direction is 0.0216 mm,and the absolute error in the rotation direction is 0.0194°.In the static experiment,the repeated positioning and translation accuracy of the cabin pose detection algorithm is better than 0.02 mm,and the repeated positioning angle accuracy is better than 0.1°.