中文摘要：

针对目前各种缝纫线迹质量检测全部依赖人工目视检测和逐针检查，导致工作量巨大，且容易由于视觉的疲劳造成漏检或误检的问题，以降落伞缝纫线迹为例，提出一种基于立体视觉的线迹质量自动检测方法．首先对立体摄像机进行标定，获取立体摄像机的内参数和相对位置姿态；然后对立体影像进行图像分类、线迹目标稳健提取、同名特征匹配等一系列处理：最后结合立体摄像机标定结果计算缝纫线迹上针脚点的三维空间坐标，实现缝纫线迹属性参数的计算和质量评定．利用多组降落伞缝纫线迹数据进行实验的结果表明，采用文中方法检测的线迹密度误差在±0．2针以内，均匀度基本相同，完全能够满足缝纫线迹检测精度要求，从而验证了该方法的可行性和有效性．

英文摘要：

At present, quality inspection of all kinds of sewing stitches relies on manual visual inspection needle by needle. This method may lead to undetection or false detection easily due to the visual fatigue, and which workload is very huge. A quality auto-inspection method of parachute sewing stitch based on stereo vision is proposed in this paper. First, calibrate the stereo cameras to obtain intrinsic parameters and the relative position and posture of stereo cameras. Then, series algorithms of stereo images processing were implemented including image classification, target extracting and feature matching. Finally, combining cali- bration result of stereo cameras, three-dimensional coordinates of sewing stitch were calculated to obtain the attribute parameters and assess the quality of parachute sewing stitch. Experiment results of multiple groups of parachute sewing stitch data show that the stitch density error is within ± 0.2 pin and the degree of uniformity is almost equal compared with manual detection results. This can fully meet the accuracy requirement of sewing stitch detection, and demonstrate the feasibility and effectiveness of this method.