The Research And Application Of Surface Defect Image Detection Technology

With the improvement of living standards, people have higher requirements for the quality of the wide range of products, which also make higher demands for quality controlling of industrial production line. Defective products must be rejected before they can be used. Badminton is a sport many people likes, badminton quality is good or bad depends largely on the quality of its main raw materials. The effect of feather-chip defects ranges from the appearance to the quality of the badminton, therefore, combined with machine vision technology, the subject points out that feathers-chip defects can be automatically detected by using image processing technology, and it has great significance in application.Based on the platform of automatic acquisition for feathers image, the paper focus on pre-processing and segmentation of the image-defect information, and then the extraction and recognition classifier of this defect features. In this paper, main work is as follows:1. Study the human visual characteristics and color space, and select the appropriate color space in different stages of image processing. Select the HSV color space, combined with the human visual characteristics in the image enhancement stage, and CIEL*a*b*color space is selected according to the characteristics of defects in the defect segmentation and recognition stage.2. Research the enhancement algorithm in the Retinex, such as SSR, MSR. MSRCR and so on, propose the new Retinex algorithm based on the adaptive brightness of visual characteristics, of that eliminates the uneven illumination and maintains the color and border of image, to enhance the image quality.3. On the basis of the level set, claim the Fast Level Set segmentation algorithm based on the CIEL*a*b*color space, the algorithm select CIEL*a*b*color space, according to feather leaf defects of its own characteristics, so that stains, partly yellowish can be characterized in contrast more significantly, then optimize re-initialization of the level set function, the selection of the velocity field, the structure of the signed distance function and the selection of the iteration time, and the experiments show that the method has accurate segmentation results and shorter split time4. Extracting the characteristic parameters of defects, is conducive to classification and decision. Study the extracted defects characteristic parameters ranging from the geometric characteristics, morphological characteristics color characteristics and so on, then, formulate specific classification method to classify these defects, which improves the recognition accuracy.Through the Matlab and VC6.0software, the paper fulfills the above algorithms and methods, and in the final experimental verification for detection of the image surface defects, the results are very good.