摘要
研究皮肤图像特征提取问题,在皮肤图像症状识别过程中,针对选择出对皮肤症状分类能力强、准确识别图像诊断方法问题,为提高识别率,提出采用遗传算法和小样本、非线性的支持向量机结合起来。通过遗传算法优化对皮肤症状特征空间进行搜索的同时,采用支持向量机对提取的皮肤显微图像的特征参数进行优化组合。在对5类典型皮肤症状进行仿真,使皮肤图像症状的特征通过组合的诊断识别率由87.24%提高到98.15%。实验结果表明,所采用的遗传算法与支持向量机结合的方法对皮肤症状图像识别率的提高是十分有效的,有利于皮肤病症的临床诊断研究。
In the research of automatic recognition of skin microimage,feature selection is a key process in pattern recognition and affects the design and performance of the classifier.Based on support vector machine technique and genetic algorithm,a new method which utilizes genetic algorithm to search feature space without the hard condition of conventional statistics,is put forward to select rational feature from the feature group of the skin microimage.By this new method,the cross validation accuracy of 98.15% has been obtained with the selected features by the classifier system for 5-class skin symptom classification,versus the accuracy of 87.24 with all the features.So the experimental results are satisfactory.
出处
《计算机仿真》
CSCD
北大核心
2010年第11期267-269,362,共4页
Computer Simulation
关键词
遗传算法
支持向量机
皮肤症状
模式识别
Genetic algorithms
Support vector machines（SVM）
Skin microimage
Pattern recognition
