基于双向2DPCA 和 SVM 的胃粘膜肿瘤细胞识别

摘要

We propose a tumour cells recognition algorithm for gastric mucosa,it is based on Bi2DPCA (bidirectional 2 dimensional principal component analysis)and SVM(support vector machine)and aims at the features of high dimensionality,irregularity and complexity the tumour cells ’image has.Bidirectional 2DPCA operates the feature extraction on both row and column direction of the image simultaneously,so greatly reduces the feature dimensions of image.Combining the advantage of statistical theory-based SVMin classification and recognition,the method skilfully solves the nonlinear problem by introducing kernel function,so that quickly and efficiently realises cells classification.Experimental results show that the proposed method can improve the classification rates,and the time of algorithm is decreased significantly as well.%针对胃粘膜肿瘤细胞图像的高维性、不规则性及复杂性特征，提出基于双向2DPCA（二维主成分分析）和 SVM（支持向量机）的肿瘤细胞识别方法。双向2DPCA 同时对图像行、列方向进行特征提取运算，大大降低图像特征维数。结合基于统计理论的SVM在分类识别方面的优势，通过引入核函数巧妙地解决非线性问题，从而快速有效地实现细胞分类。实验表明该方法不但有效提高了识别率，而且算法时间明显减少。