Surface Quality Inspection Of Industrial Sewing Machine Upper Knives Based On Machine Vision

As a consumable in the garment and textile industry with great demand,the knife on the sewing machine is more inclined to use visual detection technology in the production process.In order to meet the testing requirements of enterprises,a cutter defect detection system for KR23 upper knives is developed to realize the detection of cutter surface defects and edge defects.In addition,the following researches are carried out in aspects of lighting scheme,detection platform,algorithm design and software system:1.According to the analysis of the distribution characteristics and causes of various cutter defects,combined with the comparison of the advantages and disadvantages of common acquisition equipment,the hardware equipment model is selected.And the lighting scheme is determined according to the imaging effect and experimental test:coaxial lighting scheme is used for the cutter surface,and dark field lighting scheme is used for the blade part under skewed light.2.In order to reduce the uneven light caused by Lambert effect and solve the problem of imaging field of vision,the floating ROI splicing technology based on small area contour matching was proposed.The template matching based on the minimum rotation rectangle was used to screen cutter types using the spliced image contour.After comparing the least square fitting effect of ellipse and circle,the size and position of fixed hole were calculated using the fitting data of the latter.3.A two-dimensional Otsu threshold segmentation algorithm based on gradient is designed for images with indistinguishable background and foreground,to realize background separation of tungsten steel cutter head and part of cutter body defects.The shape-based template matching method was improved to increase the mask template,and the contours involved in matching were screened to improve the matching rate of images with large contour differences.Pyramid and parallel techniques were used to accelerate the matching speed.4.For the character recognition and unground defects that are not suitable for threshold segmentation and template matching,the former extracts the defects by matching the corresponding template difference after accurate segmentation based on gray projection,and the latter extracts the defects which are confused with the gray level of the knives body by using the improved region growth method on the color model that can highlight the defects.5.For the defects of upper cutter blade which need to be detected by industrial microscope,article designs the quadratic fitting method of skeletonization,interpolation reconstruction method and analyzes the detection effect deeply.For the bottom cutter blade image with three-peak feature in gray histogram,the iterative histogram double-peak method was used to extract the defect.Finally,the cutter testing software system and experimental testing platform were designed according to the testing process,and experiments were designed to verify the detection performance of the algorithm.Experimental results show that this detection algorithm can meet the requirements of industrial detection,and realize efficient and accurate detection of defects.