Research On Surface Crack Detection Of Parts Based On Digital Image Processing

The superficial cracks and stress concentration of Mechanical parts will suddenlycause component fatigue fracture in the process of using, which directly affects their lifeand safety, and along with great harm, thus it has very important significance to detectsurface micro cracks of mechanical parts timely. This paper carries out a research ondetecting surface crack defect of mechanical parts wit detection technology based ondigital image processing. An experimental software system has been designed with thecorresponding digital image processing algorithms by VS2008, which complete theprocessing for static image of surface crack of parts and feature extraction of crack, andprovides an effective evaluation system on detecting surface crack defect of parts.Firstly, this paper describe the content, characteristics and operation methods ofdigital image processing and make a brief introduction on the structure of detectionsystem based on digital image processing, as well as its general working procedure.Secondly, improved processing has been made to crack image for the gray blur,noise interference, etc., which is caused by the external environment or the transmissionprocess when in the field to get the image. Based on analyzing the gray histogramwhich is drawn to tell the distribution of the grey level, this paper takes a subsectionlinear stretch for the image, compress the noise in the both ends of the gray value area incrack image, while highlighting the needed details. Then, this paper has analyzed thesource and types of the noise in image, and it has compared the filtering effect ofneighborhood averaging and median filtering method. Under a certain extent, medianfilter is a kind of nonlinear filters which can overcome the details fuzzy of the imagecaused by linear filters, such as mean filter, and it is ideal for crack image filterprocessing. This paper determined the effectiveness of the median filtering, and furtherincreased the improved median filtering processing. In order to extract the crack area,this paper analyzed the threshold segmentation methods according to the distribution ofthe grey level of the crack image, and highlighted the maximum variance and iterationthreshold segmentation methods, whereby the image is obtained with the crack andbackground area separated. On this basis, Because of the result of crack image aftersegmentation exists defects such as burrs and scatter noise which distributed outside thetarget area from a certain distance, so finally getting the crack binary image with the boundary are preserved better by using appropriate structural elements formorphological operations, and connecting with the isolated noise removal method basedon threshold length.In the paper, finally analyzing the characteristics, such as length, perimeter andarea of the crack from the surface crack image of parts by using scanning label,contour tracing, and the method is proposed based on morphological thinning, and thendraw circularity parameters. So the detection and judgment on surface crack defect ofthe part is achieved.