中文摘要：

提出了一种基于增量式拉普拉斯嵌入和支持向量机的图像识别方法,该方法首先利用增量式拉普拉斯特征映射对数据点进行维数约减和特征提取;再应用以统计学习理论为基础的支持向量机对图像进行分类识别。在降维过程中,该方法能够最优保持原始空间数据点的局部信息,克服了PCA降维算法从全局考虑而丢失局部信息的缺点,并且对测试集的嵌入坐标增量式计算的特点很好地减少了运算量。实验证明,该方法的图像识别率明显高于传统的PCA线性降维方法,具有可行性。

英文摘要：

This paper presents an image recognition method based on incremental Laplacian Eigenmap and SVM （Support Vector Machine）. The incremental Laplacian Eigenmap is used to reduce the dimension and extract the feature, and its incremental property can reduce calculation. Then SVM based on the statistical learning theory is used for image recognition. Laplacian Eigenmap has locality preserving property, so it can circumvent the weakness of PCA that it always reduces dimension in global perspective and would lose local information. Simulations show that the SVM performance with incremental Laplacian Eigenmap in image recognition is better than that with PCA.