Abstract: By studying several common knit defects, improved phase angle transform algorithm for knit defects feature extraction and the Mahalanobis distance for knit binarization to achieve defect detection are used in this method. Mathematical principles to achieve knit defect classification are used. This method not only can detect subtle small defects, points of defects, lines of defects, blocks of defects, but also the shape and size of defects is as closer as original kint defects compared with the previous knit defect detection.

Key words: phase angle conversion, binarization, knitted fabric defect detection, geometry

摘要： 基于编织物常见的几种疵点为研究对象，利用改进的相角变换算法对编织物疵点进行特征提取，并在此基础上利用马氏距离对编织物二值化以实现疵点检测，利用几何数学原理实现对编织物疵点的分类。方法实现简单，相对于以往的编织物疵点检测，不仅使检测出的疵点形状大小更接近疵点原样，又能检测出不易察觉的小疵点，并且对于点、线、块状的疵点检测都有良好的效果，使得编织物疵点类型识别更加容易。

关键词: 相角变换, 二值化, 编织物疵点识别, 几何关系