Research On Vision Positioning Technology Of Gantry Robot Based On Semi-physical Simulation

In order to achieve high-precision and rapid positioning of gantry robots for woodworking furniture manufacturing,this paper studies the visual positioning method of gantry robots based on semi-physical simulation technology.The main contents are as follows:First,design and build a gantry robot vision measurement experiment system based on semi-physical simulation.The system uses Unity software to realize the simulation of the gantry robot's wooden door processing scene,and uses the industrial camera to collect the virtual simulation scene projected by the projector,and realizes the solution of the relative pose of the gantry robot's end in the virtual scene in the interactive environment.Secondly,the Center Net circle detection algorithm enhanced based on the graph convolution global inference module is proposed.The algorithm uses the center point-based detection algorithm to detect circles in markers,and introduces the global inference module based on the graph neural network to learn the mutual relationship between the targets.The experimental results show that the proposed method has good performance and can improve the overall detection accuracy of the circle in the marker.Then,in view of the low detection efficiency and accuracy of the existing circle detection algorithm on the circular marker image obtained by the high-resolution camera,a new type of fast circle detection algorithm is proposed.The algorithm first uses the Yolov4-tiny algorithm to quickly locate the markers in the image,and uses the proposed completeness ranking method to obtain candidate circles from the arcs.Experimental results show that the detection time of this algorithm in high-resolution marker images is only 11% of the existing circle detection algorithm,which greatly improves the speed of circle detection in the task of high-precision visual positioning.Finally,a semi-physical simulation visual positioning experiment of the gantry robot is carried out.The circle detection algorithm based on completeness sorting is used to detect the feature points,and then the pose of the manipulator is calculated according to the information of the feature points,and the manipulator positioning of the gantry robot is realized.In the accuracy test experiment,the experimental errors of different feature extraction algorithms and algorithms under different working distances are compared,which proves that the proposed semi-physical vision measurement system can achieve millimeter-level measurement accuracy.