摘要
研究了一种显微图像中特征识别与分类的方法,通过图像增强和二值化,将细胞从背景中完整分离出来,并提取了9个形态学特征参数、12个色彩特征参数和5个纹理特征参数,并使用支持向量机(SVM)对样本进行了分类,利用MATLAB编程的方法建立了训练模型。实验证明,正常细胞的识别率为91.35%,低度病变细胞的识别率达86.68%,高度病变细胞识别率达89.94%,总体识别率为88.93%。可见,本文的特征识别与分类具有较好的效果。
this paper studies a method which identify and classify the feature in a microscopic image,with the image enhancement and binarization,segment the cell from background.We extract the 9 morphological feature parameters,12 color feature parameters and five texture feature parameters,and use the support vector machine( SVM) to classify samples.The raining model is built by MATLAB Experiments results show that normal cell identification rate is 91. 35%,identification rate of low lesion was 86. 68%,and the high lesion rate was89. 94%,and the overall rate is 88. 93%.To sum up,our method has better effect of the feature of the identification and classification.
出处
《自动化与仪器仪表》
2016年第10期197-199,共3页
Automation & Instrumentation
基金
黑龙江省卫计委科研课题(2014-430)
关键词
图像处理
模式分类
特征识别
支持向量机
image processing
pattern classification
feature identification
SVM