【作者】 强俊； 周鸣争； 汪军； 楚宁；

【Author】 Qiang Jun, Zhou Ming-zheng, Wang Jun, Chu Ning(Dept. of Comp. Sci & Engn., Anhui University of Technology and Science, Wuhu 241000,China)

【机构】 安徽工程科技学院计算机科学与工程系；

【摘要】 基于核学习算法的基本原理,给出了一种利用核主分量分析(KPCA)提取图像像素间高阶相关的非线性特征并与优化的支持向量机(SVM)相结合来进行内窥镜图像的识别方法。在对核函数的构成条件以及特性进行分析的基础上,根据推广误差上界估计函数,实现了SVM核参数及惩罚因子的优化选取;实验结果表明,KPCA在高维特征空间具有较强的特征选择能力,优化核参数的SVM的分类识别正确率高于最近邻(NN)分类器;该方法具有较高的分类识别效果。更多还原

【Abstract】 In this paper, A method of Endoscopic Image Recognition based on Kernel learning algorithms is given. The method use the Kernel principal component analysis(KPCA) to extract the high order relation nonlinear feature of image pixels, and selects optimal Support Vector Machines(SVM) to perform the endoscopic image recognition. Based on the analysis of the formation conditions and the characteristics of Kernel functions , and according to the function upper bound of generalization error, this paper realizes the optimal kernel parameters and penalization factor of the SVM. The experiment results indicate that the KPCA technique is a powerful feature selection tool in high-dimensional feature space, and optimal SVM can get results which are significantly better than the nearest neighbor(NN) classifier, and the method has higher classification results.更多还原