一种基于SVM多特征综合的图像分类方法

分别利用多通道Gabor滤波器和HSV颜色模型对图像进行特征提取,得到两种特征空间。用顺序向前浮点法搜索,以J-M距离(Jeffreys-Matusitas distance)为评价指标进行特征选择,最后利用综合后的特征数据在SVM基础上实现图像的监督分类。上述方法提高了彩色纹理图像和遥感图像的分类正确率。实验表明,多特征融合的分类效果比单一特征要好。

A multi-feature image classification method based on SVM

This paper proposed a feature extraction algorithm using multi-channel Gabor filter and HSV color model respectively to generate two feature spaces.This feature selecting method employs Sequential Forward Float Selection (SFFS) with Jeffreys - Matusitas distances as evaluation index.Then it implemented the supervised classification of images with fusion feature data based on SVM.The proposed method improved the classification accuracy in both color texture images and remote sensing images,and experimental results indicated that the classifying quality of the multi-feature fusion is better than that of single feature.