Reserch On Seam Geometry Detection And Defect Recognition Classification Based On Machine Vision

The seam quality inspection process in the garment production process is a key part of garment production.At present,the seam quality inspection line of garment enterprises basically relies on manual visual inspection,and there will be cases of false detection and missed inspection,which are costly and inefficient.Therefore,there is an urgent need for a scam quality inspection method that can be accurately,quickly,and can be operated for a long time.The main contents of the thesis arc as follows:(1)Through the on-the-spot investigation and detection target analysis of the seam inspection line of garment enterprises,an experimental platform suitable for the geometric quantity detection and defect identification of the seam is constructed,and the position correction model of the scam geometry detection is constructed.Due to the complex and diverse background of the stitching and the presence of noise interference,the stitch image is bilaterally filtered to achieve the "edge"feature of the stitch.According to the characteristics that the stitch points of the stitch are arc-shaped,the Sobel operator template in eight directions is selected,and the sub-pixel edge extraction of the stitch is performed by using the polynomial difference method in the gray-gradient direction on the edge of the stitch to ensure the stitch point.The robustness of the extraction.(2)Using the support vector machine(SVM)algorithm based on HOG feature,the classified stitch defect samples are extracted by HOG feature to obtain the seam defect feature data set.The SVM classifier suitable for seam defect classification is trained by using the feature data set to identify and classify the seam defect categories.(3)The experimental tests were completed on the seam samples,and the effectiveness of the proposed algorithm was verified.