Design Of Automatic On-line Detection System For Surface Defects Of Multi-component Finished Parts Based On Machine Vision

Metal isomer components are widely used in industry and life,and their surface defects will directly affect product sales.As labor costs continue to increase,the market is increasingly demanding fully automated production.In the automatic production,product online quality inspection is one of the key links,and machine vision inspection technology can be applied to the surface defect detection of products,which is an important link in the production quality inspection^[1].The galvanized faucet is one of the isomer components,and its surface defect detection method is still based on manual detection.Increasing labor costs are forcing enterprises to seek automated inspection of galvanized faucets,and machine vision comes into being.This topic is based on the combination of machine vision and robots to study the surface defects of galvanized faucets.The main research contents are as follows:1.Designing the overall mechanism of the machine vision inspection experimental platform and its communication system,and design the appropriate machine vision system hardware according to the field of view required for inspection.And designing specific fixtures to achieve galvanized faucet grabbing.2.The principle of monocular camera calibration and nine-point hand-eye calibration is studied,and a calibration scheme is designed according to an example to correct the camera distortion.On the basis of monocular camera calibration,a two-step correlation method between camera and the robot is proposed.3.According to the difference of defects between inner and edge regions of the galvanized faucet surface,the algorithms of image preprocessing,edge segmentation,threshold and feature extraction are studied.At the same time,according to the different surface defects of galvanized faucet,the edge area and the inner area are treated differently to realize the recognition of surface defects.Finally,an online detection system software for real-time communication between the camera,the host computer and the robot is designed.After testing,a product inspection cycle can be controlled within 35 seconds,which is 20%higher than the manual detection efficiency,meeting the enterprise requirements.