基于自适应LBP和SVM的织物疵点检测算法.pdfVIP

第 30 卷第 6期 计算机应用 Vo.l 30 No. 6 2010年 6月 Journal of ComputerApp lications June 2010 : 100 1- 9081( 20 10) 06- 1597- 05 LBP SVM 1, 2 1 付 蓉 , 石美红 ( 1. , 710048; 2. , 710075) (an elinm ay@ 126. com ) : 为准确提取不同种类织物纹理的特征, 提出 一种新的纹理特征描述方法 自适应 部二值模式 ( ALBP) 该方法为不同纹理结构创建相应的主要概率模式子集, 避免了均匀 部二值模式 ( ULBP) 使用同一模式集 描述不同纹理而导致的描述不准确问题在该算法基础上构建 一种基于支持向量机( SVM ) 的织物疵点检测算法, 将 疵点检测问题转化为分类问题实验结果证明, 该算法不仅保持了传统 部二值模式( LBP)的旋转不变多分辨率等 特点, 而且疵点检测结果在视觉上更加清晰误检率更低适用范围更广, SVM 的优秀分类性能也有效地提高了疵点 检测的准确率 : 部二值模式; 支持向量机; 图像分割; 疵点检测 : TP39 1. 4 : A Fabric d efect detection based on adap tive LBP and SVM 1, 2 1 FU Ron , SH I M eihon ( 1. S chool of C omp uter Science, X ian Poly technic University, X ian Shaanx i 710048, C hina; 2. School of E lectronic Eng ineering, X idian University, X ian Shaanx i 710075, Ch ina ) A bstract: An advanced local b inary patterns m ethod was proposed to describe the main mi a e features. Adaptive Local B inary Patterns (A LBP) method selected the frequently occurrin patterns to construct the m ain pattern set, wh ich avoids us in the sam e pattern set to depict d ifferent texture structures in the traditional uniform local binary patterns. Based on the proposed method, an effective fabric defect detection al orithm of Support Vector M ach ine ( SVM ) w as desi ned. First, the features of the trainin sam ples w ere extracted accord in to the set and were fed to SVM. Then the testin mi a e was equally divided into detection w