基于计算机图像识别技术的坯布疵点检测研究

Research for Defect Detection of Fabric Based on Computer Image Recognition Technology

提出了一种新的坯布疵点检测方法:利用布匹图像的自相关函数所提取织物的纹理单元作为分割窗口,建立正常纹理与异常纹理2类样本;应用Fisher分类器,通过对布匹图像以纹理单元为单位逐次扫描进行2类判别,经过图像运算将正常类纹理单元消除,得到疵点的待测图像;再采用一些图像运算处理技术检出疵点并进行评分。进行了相应的软件编程,实验结果显示疵点识别率可达94%。

At present, defect detection during the manufacturing is a manual work; there are a lot of weaknesses, such as low detection efficiency, high miss rate. These affect the production quality and restrict the improvement of production efficiency. A new method was presented. The basic design thought is that using the fabric texture unit extracted by the autocorrelation function of fabric images as separate windows, establishing two samples of normal and abnormal texture, applying Fisher classifier to discriminate the two kinds through successive scanning with the texture unit of fabric image as unit, after image operations eliminating normal category of texture unit. The defect images to be detected can be obtained. And then detect defects and score by some image operation process techniques. The software programming is provided and test results show that the defect identification rate is 94%.