The Method Research Of Workpiece Surface Defect Detecion Based On Machine Vision

Since the 1980s,with the rapid development of digital image processing and higher requirement of product quality,conventional manual work can't achieve satisfying detection result for many reasons. so that people pay more attention to detection technology based on machine vision .This paper studies some key technologies on surface detection of workpiece, implement the correlative arithmetics based on VC6.0, the result of actual running is deeply analyzed, and it demonstrates that these methods are valid.Firstly,object location and correction is an important technology.This paper using analytical method of connectivity to locate object.after that, mainly studied the Least Squares fitting,the principal axis of inertia law,the Hough transform,and there has made improve to these method,resolved object location and correction problem of some workpieces.Secondly,surface defect detction on untextured image and textured image.This paper has researched several convetional detection arithmetic,these method are Template Matching,Statistical Method,Dynamic Threshold Arithmetic.This paper put forward an improved Dynamic Threshold Arithmetic,By combining the gray feature of whole and local,the outcome indicating that this method get good detection result.And then,this paper has researched several convetional detection arithmetics of textured image, these methods are Fourier Transformation,Gabor Filtering Methods,Wavelet Transformation.This paper put forward an improved Wavelet Transformation to wipe off the texture part of image, and then using Dynamic Threshold Method to detect defection.Finally, this paper extract feature and classify those detected defect.We ex- tract shape and projection feature and classify them for the trait of defection in the surface of copper plating.