适用于瓜果分检的一种纹理分类方法
New Texture Classification Method Based on Rotating Invariant Features and SVM
小波变换在纹理特征提取中得到广泛应用，但是往往提取的特征不是旋转不变的。对此，本文提出了一种提取纹理旋转不变特征的新方法。首先对图像进行同心旋转变换（CRT），然后对变换后的纹理图像分别进行旋转不变纹理特征的提取和小波分解，最后利用小波分解各尺度子图像间关系建立灰度共生矩阵，然后提取更多特征。实验表明这些特征可以在图像旋转过程中基本保持不变。为了提高分类正确率，使用支持向量机（SVM）建立分类器，最后的分类试验取得了比较好的效果。
he wavelet transform has been widely used in extraction of texture features, however, the values of the features are usually greatly depend on the rotating angle of the texture picture. This paper presents a new method to extract rotating invariant features. At first, Concentric Rotation Transform（CRT）is applied to the original texture picture. Then some features are extracted from the transformed texture picture. In addition, the wavelet transform is imposed on the transform picture and more features are extracted. The experiment shows that those features remain constant on the whole when the texture picture is rotated. In order to improve the efficiency of classification, a classifier based on SVM is built up. The last experiment’s result is satisfying.
刘劲柏、徐立鸿
自动化技术、自动化技术设备计算技术、计算机技术
纹理分类特征提取旋转不变小波支持向量机
texture classification feature extraction rotating invariance wavelet SVM
刘劲柏,徐立鸿.适用于瓜果分检的一种纹理分类方法[EB/OL].(2004-05-12)[2025-08-21].http://www.paper.edu.cn/releasepaper/content/200405-28.点此复制
评论