Research On Surface Defect Detection Technology Of Automobile Hub Based On Machine Vision

In the manufacturing process of automobile hub,it is usually necessary to machine the outer surface of the hub,which will inevitably produce scratches,scratches and other defects.After the hub processing is completed,it is usually produced by workers to check whether the hub has defects.The traditional manual inspection method has the disadvantages of low efficiency,time-consuming and low detection accuracy,which is difficult to meet the needs of industrial automation.Because of the rapid development of machine vision in recent years,convolutional neural network has achieved better application in image recognition and image classification.It will be an inevitable trend to use machine vision instead of manual detection.This paper proposes to use machine vision technology to detect surface defects of automobile hub.In order to realize the automatic detection of automobile hub surface defects,this paper mainly focuses on the key technologies of automatic detection,including detection scheme,hub location technology and defect detection algorithm.A detection and recognition algorithm based on image processing and convolution neural network for automobile hub surface defects is proposed.According to the requirement of automobile hub surface defect detection,the scheme of wheel hub surface defect detection is designed.According to the complex structure of wheel hub,the vision system is installed at the end of robot actuator to collect the image of wheel hub surface.The selection of industrial camera,industrial lens,lighting source and lighting mode is studied.Different lighting modes are compared.According to the characteristics of automobile hub surface,the coaxial illumination mode was chosen.Finally,the actual hardware system was built using the selected visual components,which laid the foundation for the subsequent defect detection.In order to realize the automatic detection of hub defects,hub positioning is indispensable.This paper proposes the use of machine vision technology to realize hub positioning,and carries out camera calibration and hand-eye calibration.Through image processing technology,the center coordinates of hub center hole and installation hole are extracted to realize hub positioning,and the accuracy of the positioning algorithm is verified by experiments.。The image preprocessing algorithm is studied.The main purpose of the image preprocessing algorithm is to highlight the defect features of the image and facilitate the recognition of the image defect.The image preprocessing algorithms in this paper include image filtering,image gray transformation and image sharpening.The median filter,logarithmic gray transformation and Laplacian image sharpening are selected to process the image through experimental comparative analysis,which effectively highlights the defective features of the hub surface.The main structure and characteristics of convolutional neural network are introduced.Based on the platform of MATLAB,a convolution neural network is built.After the training of the neural network and the debugging of the parameters,the convergent neural network model is obtained.By comparing the test results of the algorithm with that of the convolution neural network algorithm without image processing,it is proved that the algorithm in this paper effectively improves the accuracy of defect detection.Finally,the hardware system experiment proves that the algorithm can realize the surface defect detection of automobile hub.