基于局部投影统计法的螺纹缺陷检测

为解决传统接触式螺纹测量方法费时且程序冗长的缺陷，提出一种基于机器视觉技术的螺纹缺陷检测算法。对捕获的螺纹图像进行中值滤波、迭代法二值化与Canny边缘提取处理|通过分析螺栓图像中螺纹缺陷断口位置灰度值的变化，提出一种基于DOG模型的螺纹自动检测方法。为验证该算法性能，用基于形状的模板匹配算法作为对照进行实验。结果表明，局部投影统计算法能有效提取螺纹缺陷图像的缺陷信息，螺纹缺陷图像识别率在95%以上。该方法可快速有效地降低噪声，准确迅速地定位缺陷点，提高生产线螺栓可替换性。

Thread Defect Detection on Packaging Production Line Based on Local Projection Statistics

The work aims to deal with problems that current methods of measuring screw threads either suffer from time consuming or procedure tedious. In this paper， median filtering， iterative binarization and Canny edge extraction are carried out for the captured screw images.Then this paper presents an automatic screw thread inspection method based on the DOG model by analyzing the change of the gray value of the position of the screw thread defect in the bolt image. In addition a shape-based template matching algorithm as a controlled experiment is used to compare with the proposed algorithm. The experimental results showed that local projection statistical algorithm has higher recognition accuracy for screw thread defect image. The recognition rate is above 95%. The method can quickly and effectively reduce noise， which comprehensively extract defect information of the screw thread image. And this algorithm can accurately and quickly locate defect points for improving the replaceability of the bolts.