Machine Vision Technology And Application Of Defect Detection

For high quality products, on the surface of the product characteristics also have strict requirements. But due to improperly subsequent processing of raw materials, it will make the product surface defects, which not only affect the external appearance, and may even reduce its quality. The stand or fall of badminton products depends largely on the quality of the feather piece, and the defects of the leaves of feather is an important factor affecting the quality of piece of feathers. In actual production, feather leaf surface defect detection is an essential link. Therefore, this paper with feathers as the research object, combined with machine vision technology to realize the feather leaf surface defect detection and classification recognition, has the important theoretical practical application value and economic benefits.Defects on the Feather leaf are mainly stain, bug eat by insects, local yellow. Accurate segmentation of these feather leaf defects will largely improve the quality of the feather piece of sorting.This paper focuses on the enhancement methods for uneven illumination feather piece image caused by collecting device,which on the basis of feather image automatic acquisition platform, preprocesses and segment feather image, and finally completes the classification of feature extraction and recognition of defects. The main work and innovations are as follows:1. Analysis the feather light environment and establish the feather light mathematical model when image acquisition to get the feather irradiation intensity distribution curve in theory,and with this propose the feather image correction method and calculation expression.2. Analysis feather image brightness distribution characteristics,propose a segmentation method based on the improved mean shift. manifestate the differences between feather leaf and feather stem, highlight the color features of feather stem,and combined with the regional mergers and threshold segmentation to finally realize the feather stem’s segmentation.3. Improve the traditional CV model level set segmentation algorithm, put forward a quickly CV level set algorithm which integrated into the global gradient information and prior knowledge of the target. this algorithm can convergence quickly and mark off the target area from complex background image grayscale contains multiple hierarchy contour and has good adaptability. Experiments show that this method is used for feather defect segmentation results are accurate and fast.4. Extraction the defect geometrical characteristics parameters, and add color features to constitute the feature vector for defect classification and decision. Combined with the feature of the two kinds of description in order to improve the recognition accuracy.This paper realize the algorithms and methods by Matlab and VC6.0. And for each method is analyzed and experimented in the simulation,which has the good effect. Finally do experiment for defect image classification decision, and the experiment obtains a good recognition effect, the algorithms and methods have certain practical value.