Research On Machine Vision Positioning Technol-Ogy Of Rivet Hole For Plane Parts

With the upgrading of equipment in manufacturing enterprises,many enterprises are carrying out automatic transformation of equipment in the context of the Made in China 2025 policy,so as to reduce costs and improve work efficiency.Machine vision,as a major detection component for upgrading and upgrading,plays a very important role.When strengthening the skin of the helicopter isolation cabin panel in conjunction with a certain helicopter manufacturer,it is necessary to rivet the partition frame and skin together to increase strength.The project carried out the work of using machine vision technology to locate rivet hole positions,designed a set of automated equipment for nailing,punching,and riveting,and conducted research and analysis on the identification and positioning technology of rivet hole positions.The main research content of the paper is as follows:(1)A machine vision dual robot automatic riveting system has been constructed to address the issues of multiple processes,large footprint,and unreliable human labor in the production process of enterprise isolation cabin panels.This system includes an assembly station,transfer components,nail supply system,and a drilling and riveting system equipped with a machine vision system,achieving the manufacturing,positioning,and riveting functions of the rivet holes on the isolation cabin board,improving the efficiency of rivet riveting and avoiding dangerous labor operations by workers.(2)In view of the large size of the isolation hatch and the high requirements for industrial cameras when the single view panoramic image of the isolation hatch reaches the ideal accuracy,the image mosaic method is proposed to obtain high pixel images using low pixel cameras.And through image preprocessing and rigid transformation during stitching,the stitching accuracy is controlled within one pixel,achieving the usage conditions..(3)According to the possibility of interference with the partition frame during the riveting action of the robotic arm,deep learning is introduced to determine the direction of the aluminum and aluminum edges at the partition frame corners.And through transfer learning technology,four different training models are used to train 1325 angular aluminum images,and pre trained is comprehensively selected_dl_classifier_The resnet50.hdl model has an accuracy of 99.5%.(4)Analyzed the positioning and measurement scheme of rivet holes.By measuring the diameter of the hole in the target field of view,it is possible to guide the control system in providing matching specification rivet operations.By obtaining hole coordinate information through edge detection,threshold segmentation,and other methods,the riveting operation of the guided riveting robot is achieved.(5)We have built a visual recognition and positioning experimental platform.Unify the coordinate system between the robot and the workpiece through hand eye calibration and camera calibration;We have built a software interaction platform to collect images,process images,output target point information,and angle aluminum direction information;Realized the recognition and positioning function of rivet hole positions.