Abstract: Because bacterium in microbe image has different shapes, is apt to overlap, and gray levels are very close, a new method of micrograph identification was proposed in this paper, which was based on pattern recognition theory of Support Vector Machine (SVM) and used texture feature extracted by Variation. The classification results were then compared with the results obtained by NN theory. The results of study have proved that the method based on pattern recognition theory of support vector machine to the classification of bacterium micrograph may improve the accuracy of image classification.

Key words: bacterium, micrograph, texture, variation, Support Vector Machine (SVM)

摘要： 根据微生物显微图像中微生物形态各异、目标重叠、灰度接近等特性，提出了一种新的显微图像分类识别方法。该方法利用变差函数对微生物显微图像纹理信息进行特征提取，根据支持向量机模式识别原理建立分类识别模型。将该方法应用于两类微生物分类，并与基于神经网络方法的分类结果进行对比分析，结果表明，该方法具有较高的分类精度。

关键词: 微生物, 显微图像, 纹理, 变差函数, 支持向量机