Automatic seam pucker evaluation using multi-view stereo vision

Traditional fabric seam pucker evaluation suffers from low efficiency and color sensitivity. To address this issue in color printed fabrics, a novel method based on multi-view stereo vision is proposed. A multi-view imaging system was developed to capture 16 images per sample and reconstruct three-dimensional (3D) point clouds. These clouds were projected onto a two-dimensional (2D) depth map to locate the seam line. Four morphological features within 80 pixels from the seam were extracted as objective indicators. Correlation analysis identified the number of seam puckers and average curvature as key parameters for SVM classification. Experimental results demonstrate that the proposed method effectively classifies seam pucker grades and shows strong potential to replace conventional approaches.