Research Of Surface Error Detection Based On Digital Image Processing

In order to continuously improve product quality and production efficiency,surface mount electronic components and metal parts online automatic detection ofsurface defects in the production process is becoming increasingly important. Forsurface mount electronic components and a variety of defects in the metal surface, Idesigned a set of machine vision based on surface defects can be achieved in realtime online, no damage automatic detection system. The system uses a CCD arrayand multi-channel image acquisition card as part of image acquisition, improvedspeed detection system and reduces the performance requirements of the CCD, thesystem under the current conditions are relatively easy to achieve real-time onlinedetection; using automatically selected image segmentation threshold based on thethreshold value of the actual application of the work information extracted from theimage out and scanned workpiece image information to achieve the automaticmeasurement system; the workpiece information scanned to remove the apertureedge of the workpiece, using the automatically selected threshold metal surface ofthe image binarization segmentation, in order to achieve the automatic extractionand recognition of various defects.With the continuous improvement of the level of technology and electronicsindustries become more stringent quality requirements in the automated productionprocess, surface mount electronic components and small metal componentsautomatic surface defects detection technology has been more and more attention.Due to the small surface-mount electronic components and metal components, thereare different types of surface defects, this thesis of various common defects wereclassified, and then the paper were the most common small circular metalcomponents and stickers chip capacitors designed for the carrier and itscorresponding surface defect detection system, which mainly use the current populardigital image processing technology, so the detection of high precision process, fast.Low cost, non-invasive, accurate and fast, so that the machine vision is ofgreat significance in the field of defect detection. This article specifically study thedefect detection algorithm based on digital image processing and MATLABsimulation, in order to better apply to reality. The more in-depth study completed insurface defects recognition algorithm made.Components on the surface defect detection, this paper can be used for defectdetection for each basic image processing algorithms, such as image filtering,image enhancement, morphological image processing, image segmentation, werestudied. Components on the surface defect detection, this thesis can be used for defectdetection for each basic image processing algorithms, such as image filtering,image enhancement, morphological image processing, image segmentation, werestudied. For easy detection of scratches and marks, we use a method to extractimage edge, the edge of the image points are analyzed. In this paper, the firstalgorithm is used to enhance the use of Canny operator weighted multi-scaledetection algorithm. The method uses a multi-scale advanced ideas using the Gaussfilter in three different scales, a gradient of the image processing and filtering,respectively, under a separate scale formation then weighted gradient map to obtainthe total gradient of FIG.Due to the enhanced use of Canny operator to calculate the weightedmulti-scale quantum detection algorithm is very large, the actual production isdifficult to meet the requirements, it is also using a differential edge detectionalgorithms, but the differential edge detection algorithm to extract the outer edge ofthe image is not accurate enough, and higher requirements for image filtering, sonot very suitable. Also proposed using edge detection method based onmathematical morphology. After a large number of experimental studies, thismethod can more precisely identify the scratches and marks defects and faster.This paper presents a method to detect the use of, the main use of Matlabsimulation, a large number of components of the samples were tested, although noton the components positioned on the bad parts detection accuracy rate of98%.