Research On Metal Part's Surface Defects Base On Machine Vision

The detection of product quality has become the most significant passageway for products to enter the market with the rapid growth of industrial production lines.The traditional production lines depend on manual detection which is not only inefficient and high cost,but also has subjective fatigue of human vision and other factors.So as the manual detection in industrial production lines is not accurate enough.Applied the machine vision in industrial product quality detection can advance the products quality,improve production efficiency,and reform industrial measurement technology with the rapid growth of machine vision at home and abroad.Currently,advancing the detection efficiency and accuracy is the core issue to be researched in modern intelligent production.This topic takes the metal parts produced in industry as the research background,measuring the dimensions and detecting the faults for the two different models of metal parts,and completing the overall system structure of metal parts detection on the basis of machine vision system.The main research contents of this article are as follows:Established a hardware structure system based on machine vision according to the technical requirements of on-line detection of metal parts.The choice of hardware models of machine vision containing CMOS camera,lens,light source,acquisition card and external trigger.Selected the suitable models of light sources and industrial cameras satisfying the detection accuracy,and designed rational focal length and object distance of camera and so on.Then establish the software system and the system platform of software and hardware under VS2010 environment to realize the image acquisition,camera parameter adjustment,image preprocessing method,comparison and analysis of edge feature extraction,dimension measurement,detection and recognition of faults and so on.The treatment flowsheet of software system includes:acquiring image by camera;image preprocessing includes:acquiring ROI,adaptive median filtering,threshold segmentation is applied to the images after pre-processed.Then extracting sub-pixel edge detection of the images by adopting improved morphological gradient operator and quadratic interpolation,contrasting several different sub-pixel edge extraction,applying Hough transform algorithm to detect the lines and circles in the image,and calculating the distance;Adopted template matching to locate and recognize image defects in the defect detection system.Located and recognized four kinds of defects on metal surface,including scratch,unfilled corner,rag and disrupt,and finally verify the rationality of the algorithm by testing samples.The image software processing system is on account of the algorithm library of OPENCV and HALCON,and establish the MFC interface development system under VS2010 environment to achieve the operation and management of system,and the display of human-computer interaction interface.