Abstract: Traditional manual method of metal zipper defects detection has the features of inefficiency, poor stability and high false positive rate. To these problems, this paper presents a new detection method based on Pulse Coupled Neural Network（PCNN）and gray mutation detection. Depending on the characteristics of metal zipper image, the processing speed of image segmentation will be increased by improving the traditional PCNN, which can extract the feature images of zipper scoops by combining with morphology. A new proposed method can automatically distinguish the metal zipper scoops defects through the region pixel statistics and gray mutation detection. A detection system is designed for this experimental study. Experiment results show the detection method of this paper is quick, accurate and feasible.

Key words: Pulse Coupled Neural Networks（PCNN）, scoops defect detection, image preprocessing, gray level mutation

摘要： 针对传统金属拉链缺陷人工检测方法效率低、稳定性差、误检率高等缺点，提出一种基于脉冲耦合神经网络（Pulse Coupled Neural Networks，PCNN）和灰度跃变检测的金属拉链缺陷检测方法。针对拉链图像的特点，通过对传统PCNN进行改进以提高金属拉链图像二值分割处理速度；将传统PCNN和形态学理论相结合，提取链齿特征图像；采用区域像素统计与灰度跃变检测的方法实现金属拉链缺陷自动检测；完成检测系统的设计并进行实验研究。实验结果表明提出的检测方法快速、准确、可行。

关键词: 脉冲耦合神经网络, 链齿缺陷检测, 图像处理, 灰度跃变