基于多证据的血液白细胞自动分类  

Automatic Classification of Leukocytes in Blood Based on Multi-evidence

作　　者：庄杨凯[1] 周平[1] 

机构地区：[1]浙江理工大学信息学院,杭州310018

出　　处：《浙江理工大学学报（自然科学版）》2013年第3期367-371,382,共6页Journal of Zhejiang Sci-Tech University(Natural Sciences)

摘　　要：对白细胞分类方法提出了一种基于形状特征学习的分类算法。该算法首先从细胞核中提取基于欧氏距离变换的不变矩特征和形态学特征;然后,所提取的特征可以实现对单核细胞、淋巴细胞、嗜碱性粒细胞的分类;最后,提取灰度共生矩阵作为纹理特征,通过支持矢量机实现对剩余白细胞类别的分类。实验结果表明,该方法具有很好的分类准确率及较短的处理时间。This paper puts forward a classification algorithm based on shape feature study for leukocytes classification method. This algorithm first extracts invariant moment feature and morphological feature based on Euclidean distance transform from nucleus; then, features extracted can realize the classification of mononuclear cell, lymphocyte and basophilic granulocyte; gray-level co-occurrence matrix is extracted as texture feature and the classification of residual leukocyte categories is realized through support vector machine. The experimental result shows that this method has a very good classification accuracy rate and short processing time.

关 键 词：白细胞分类 欧氏距离变换 不变矩 形态学特征 纹理特征 支持矢量机 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]