【摘要】 主要对平纹、斜纹和缎纹组织丝织物的一些常见疵点,如档疵、缺纬、缺经、重纬、油污等进行了智能化判别。先用SONY数码相机在黑色的背景下对疵点进行了拍照得到了图像数据,然后用一系列图像预处理法,如直方图处理变换增加了织物图像的对比度、用计算得到的阈值对织物进行了二值化处理、滤波方法消除二值化处理后图像噪声等,从织物纹理分离出疵点部分,得到了可以分析的织物疵点图像。用灰度统计法对预处理得到的织物疵点图像进行了分析,得到了织物各疵点基本特征值信息。织物疵点智能化判别是用BP神经网络进行的,首先对BP神经网络进行了训练,然后将灰度统计法得到的疵点特征值信息输入到BP神经网络,对疵点进行了分类。更多还原

【Abstract】 The paper mainly deals with the intelligent recognition of such common defects as loom bars, broken filling, brokenwarps, double weft and oil stains in tabby, twill and satin weave fabrics. First the defect images are taken with SONY digital camerawith a black background. Then by series of methods for image pretreatment such as with histogram transform the image contrastbeing increased, with the threshold obtained through calculation the fabrics treated by two-value method and by wave filteringmethod the image noises eliminated, the defect parts are separated from fabric texture and the defect images for analysis areobtained. Then some basic characteristic information of each defect are acquired by analysis the images by grey level statisticalmeans. The intelligent recognition of fabric defects are done by BP neural network. And the BP network should be trained at first,then the defect characteristic values are put into network to be classified.更多还原