摘要
对于织物缺陷的检测 ,可以使用多种不同的图像处理技术 .而具有多分辨特性的小波变换是一种分析图像的新方法 ,它的变尺度特性与人类视觉中的空间频率多通道相吻合 .使用小波分析的方法对 3种织物缺陷进行检测分类 .首先将织物图像进行 3层小波分解 ,然后把小波分解后的图像灰度值作为特征参数输入到 BP神经网络进行检测识别 ,实验结果表明 ,用这种方法识别织物缺陷识别率可达到 98% .
To fabric defects, there are a lot of image-based inspection techniques. However, wavelet (transform) is a new kind of multiresolution algorithm, and its multiresolution character corresponds to (time-frequency) multiresolution of human vision. So wavelet transform is used to inspect and classify three kinds of fabric defects. First, decomposing fabric image to three level. Then, identifying defects to use gray value of the third level as character parameters of BP neural network, The result of experiment shows that the fabric defect's identifying rate can attain 98%.
出处
《纺织高校基础科学学报》
CAS
2004年第4期364-367,共4页
Basic Sciences Journal of Textile Universities
基金
陕西省自然科学基金项目 ( 99C1 8)
关键词
小波分析
BP神经网络
织物缺陷
wavelet transform
BP neural network
fabric defects