Design Of Automatic Inspection System For Screwdriver Head Based On Machine Vision

The production process of an electric screwdriver bit is to cut the material,open the cutter head,chamfer the tail of the knife,and mark the rubber sleeve.Every aspect of the production process is likely to produce quality defects with a small defect area.The quality inspection of the electric screwdriver bit is a full inspection.The traditional manual inspection is caused by the inspectors,and after a long time of work,fatigue occurs,resulting in missed inspection,and there is a problem that the personnel cost is excessively consumed.In order to solve the above problems,this paper carried out the research on the automatic detection technology of the head defect of electric screwdriver based on machine vision under the support of the design of the automatic detection system of the screwdriver head based on machine vision in an electronic tool company of Kunshan City,Jiangsu Province.Based on this,an automatic inspection system for the screwdriver head based on machine vision was designed.This paper mainly carries out the following three research work:Firstly,according to the existing equipment of the factory and related technical indicators,the overall structure of the inspection system is designed,including the mechanical part,the electrical control part,and the machine vision part.Secondly,analyze the characteristics of the cutter head defect and the chamfer defect,design the hough circle based chamfer defect detection method and the template matching based cutter head defect detection method.Thirdly,analyze the defect characteristics of the circumferential rubber sleeve,and design a circumferential rubber sleeve defect detection method based on image processing.The circumferential rubber sleeve defect detection based on image processing includes two parts: image stitching and image processing after stitching.In the image mosaic part,this paper proposes a method of simple splicing from the perspective of the matrix for the requirements of this topic.In the image processing part after splicing,the method of quadratic statistical features of gray gradient co-occurrence matrix is proposed for the requirements of this subject,and the method based on feature point splicing plus template matching and based on simple splicing plus gray gradient co-occurrence matrix The results of the statistical feature method were compared.Aiming at the problem of insufficient precision and speed of circumferential rubber sleeve defect detection method after splicing and post-processing,this paper proposes a method for detecting circumferential rubber sleeve defects based on convolutional neural network.Aiming at the requirements of this topic,a phased learning rate adjustment strategy is proposed;a data enhancement method with horizontal directionflipping is proposed.At the same time,in view of the fact that there are few data sets in this subject,the defect detection of the circumferential rubber is proposed by using Fine-tuning technology.Finally,the detection results of the two detection methods are given and analyzed.