基于新核函数的支持向量机在物体分类中的应用

View-based object classification using SVM with new kernel

在基于视的物体识别中,将图像的局部信息引入到图像的相似性度量,提出了一种新的图像距离度量,并把它嵌入到支持向量机的核函数中,得到了一种新的核函数——基于局部卡方距离(Chi-square distance)的核函数。物体分类实验结果表明,新算法优于非线性支持向量机,区别张量一阶分解(DTROD),稀疏网络模型(SNW)等方法。

For view-based object recognition,the authors introduced the local image information into image similarity measurement,presented a new image distance measurement,and imbed it into Kernel function used in support vector machines （SVM）. a new kernel-local Chi-square distance based kernel was constructed. Experiments showed that it outperforms some methods such as non-linear SVM,Discriminant Tensor Rank-One Decomposition （DTROD）,Sparse Network of Winnows（SNW）and so on.