Research On Pose Detection For Loading System With Complex Conditions Based On Computer Vision

Mobile Harbor(MH)is a new type of marine cargo transfer system proposed by KAIST in recent years.The system can transfer goods at a certain level of wave and windy sea.Currently,the container has been used to transfer heavy goods at the field of sea transportation.However,under this special condition,it is especially important to compensate for the movement of some degrees of freedom caused by the waves in order to achieve precise docking of specific goods with specific locations,which has great research significance for the cargo handover of the sea and the circular target of the hole.First of all,to accurately map the target in real-world coordinates with the coordinates in the industrial camera,the camera needs to be calibrated.This paper compares the checkerboard calibration method,the circular calibration method and the application of MATLAB through a simple monocular ranging experiment.Calibrate the calibration accuracy of the toolbox,select the optimal calibration method,and improve it to complete the calibration of the camera.Secondly,as the identified target is a circle,and the space circle appears in the form of ellipse in the imaging plane.The voting algorithm based on Hough transform and the optimal algorithm based on least squ ares method are studied.Through the simulation,the advantages and disadvantages of the two types of algorithms are understood.Combined with their respective advantages and disadvantages,an improved ellipse fitting algorithm is proposed.The center of the ellipse is determined by voting method and the ellipse is fitted by least squares method.The algorithm is proven to be effective.Thirdly,based on the successful fitting of the ellipse,the projection model of the circular feature in space is studied.What is more,the relationship between the position of the circle in the space and the ellipse of the imaging plane is found,and the projection cone model of the space circle is established.The ambiguity problem of the circle is proposed to eliminate the ambiguity of the circle by the interframing method.At the same time,for the dynamic processing,due to the time difference generated during the data transmission,the EKF,UKF and PF,three kinds of prediction filtering for nonlinear motion,are compared.And the tracking algorithm most suitable for the scene of the subject is selected.Finally,based on theoretical analysis and simulation we have already worked on,the feasibility of the algorithm and the accuracy of the measurement are verified through experiments.