一种新的基于小波分析与神经网络的织物疵点检测与识别方法

A New Method to Inspect and Recognize Fabric Defects Based on Wavelet Analysis and Neural Network

提出了一种新的基于小波分析与神经网络的织物疵点检测与识别方法,根据疵点图像形态的先验知识,在织物图像小波分解高频分量中运用数学形态学中的开运算,结合小波高频各子带反映不同边缘细节的特点,去除由棉籽壳与背景光从经、纬纱之间的空隙透射而成的荧光点在织物图像上形成的噪声,提取特征参数,利用神经网络BP算法,有效地检测与识别了缺纬、断经、油污、破洞等常见疵点,并具有识别正确率高、检测速度快等优点。

A new method to inspect and recognize fabric defects based on wavelet analysis and neural network is presented. The method, based on prior knowledge of characteristics of defect image, using open transform of mathematical morphology on high\|frequency components of wavelet decomposition of fabric image, combining with the characteristics that different wavelet subbands show different edge details, eliminating noises that generated on the fabric image by cotton shell and fluorescent dots formed by the back face light coming through the intersections of warps and wefts, extracting the feature parameters, utilizing the BP algorithm of neural network, can efficiently inspect and recognize four common fabric defects—weft-lacking, warp-lacking, oil stains and holes, and have advantages with high identification correctness and high inspection speed.