Key Technologies On Visual Object Tracking And Guidance For Manipulator Machining Of Large Components

Large spacecraft cabins are high-end equipment for tasks such as space technology research and deep space exploration.Their manufacturing and processing focus on the most cutting-edge theories and technologies in today’s industry.In recent years,the intelligent mobile robot processing system has great potential in the processing of large components due to its intelligence,high efficiency,flexibility,economy and other characteristics.However,the number of local features to be processed that are constrained by associations on the surface of the cabin are numerous and irregularly distributed.The dimensions,shapes and postures of various types of features are different.The surface of the cabin has strict cleanliness requirements.Difficulty positioning.In the face of such a complex measurement environment,the measurement mode based on cooperative objects is limited,the robotic arm processing/detection unit is difficult to guide to the feature area to be processed,and the flexibility of robotic arm processing is limited.In recent years,computer vision perception technology based on pattern recognition has developed rapidly,and the recognition accuracy of features has been continuously improved,which provides new ideas for solving this problem.In addition,the accuracy of the traditional static open-loop guidance form is affected by the performance of the vision sensor and the robotic arm,and the online perception and realtime operation capabilities are poor,and the guidance process is uncontrollable.In contrast,the closed-loop visual servo guidance has global/local progressive stability,which enables the robotic arm to quickly reach the desired pose and overcome the uncertainties in the model(including the robot,vision system,and environment).Improve visual positioning accuracy.As a hot spot of machine vision,online visual object tracking technology has the characteristics of strong real-time performance and good continuity,and meets the requirements of continuous feedback of robotic arm visual servoing.In view of this,this paper has carried out the research on the key technology of visual object tracking and guidance in the processing of large-scale component manipulators.In order to solve feature occlusion and environmental interference,a color saliency-aware object tracking method is proposed,which uses color statistics as a model of image boundary connectivity cues,which greatly improves the robustness of object tracking.In order to solve the in-plane affine and transmission distortion of the feature image to be processed caused by the change of the camera angle of view,this paper proposes a decoupled Fourier Merlin transform based on the independence of the scale change in the log-polar coordinate,which can effectively estimate the object rotation and scale.information to adapt the aspect ratio changes of the feature to be processed in the image plane.Comparative experiments on multiple authoritative datasets verify the advancement of the proposed object tracking method.Through the proposed robust visual object tracking method,the pose of the feature to be processed in the image is obtained in real time,and combined with the distance measurement of the depth camera,the stable guidance of the robotic arm based on visual servoing is realized.In addition,in order to shorten the time for large-scale guidance,on the basis of accurately identifying the dynamic parameters of the robot arm,the optimal trajectory of time is planned with the dynamics as the constraint,so as to improve the execution efficiency of the robot arm guidance.Finally,using ROS as a platform,a robotic arm guidance system is built for the surface features of the cabin support to be processed.The two types of typical supports of the cabin are used as the object parts to verify the reliability and accuracy of the actual operation of the method in this paper.It realizes large-scale rapid guidance and local accurate visual servo guidance,which meets the guidance requirements of the robotic arm processing/detection unit to be processed.