基于遗传算法构造的Gabor滤波器在Lab空间的印花织物疵点分割研究

项目名称： 基于遗传算法构造的Gabor滤波器在Lab空间的印花织物疵点分割研究

项目编号： No.61301276

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 无线电电子学、电信技术

项目作者： 景军锋

作者单位： 西安工程大学

项目金额： 22万元

中文摘要： 印花织物是一类在坯布织物基础上的高附加织物，印花织物图案的美观能提高织物的附加值，现在人们对高品质印花织物的需求更甚以往。但是目前织物图像疵点检测技术仍只适用于坯布。本项目分析印花织物的图案信息的差异和联系，研究和改进Lab空间，从而使印花织物的信息可以完全表达。采用改进的Gabor滤波器提取印花织物有效特征信息，使印花纹理信息与织物瑕疵互相独立且易于分割。利用遗传算法的搜索特性，构建提取最优纹理信息的Gabor滤波器。设置遗传算法的标准函数，搜索标准函数的高适应度参数组，从而达到构建Gabor滤波器的目的。在印花织物特征值与Gabor滤波器卷积响应中，进行统计印花织物周期单元能量的研究，从而来分割印花织物疵点。同时，阐明和改进高斯混合模型的分类和模型构造的规律，对印花织物和疵点进行分类。自动检测印花织物疵点算法的实现将为提高印花织物的品质奠定理论基础。

中文关键词： 织物瑕疵检测；Gabor滤波器；遗传算法；颜色空间；稀疏编码

英文摘要： Printed fabric is a high additional fabric on the basis of grey fabric. The beautiful design of printed fabric can improve the fabric additional value. Now, the need of printed fabric with high quality becomes more popular than ever before for people.But the fabric detection technology is still only applied to grey fabric.The analysis of differences and contact on pattern information of printed fabric would be done in this project.Printed fabric information can be fully expressed by way of studying and improving Lab space.Improved Gabor filter extracts printed fabric effective feature information in which printed texture information and fabric defects are independent and easy to split. According to search characteristics of genetic algorithm, the optimal Gabor filter to extract texture information can be constructed. Setting up standard function of genetic algorithm and searching high fitness parameter group of standard function are to achieve the purpose of constructing Gabor filter.In the convolution response of printed fabric characteristic value and Gabor filter, statistical cycle unit energy research on printed fabric is used for printed fabric defect segmentation.At the same time, the law of classification and model structure would be clarified and improved in gaussian mixture model,thus achieving classif

英文关键词： Fabric defects detection；Gabor filter；Genetic algorithm；color space；sparse coding