摘要
为提高肝脏B超图像的诊断准确率,研究了将粒子群算法(Particle Swarm Optimization,PSO)和支持向量机(Support Vec-tor Machine,SVM)相结合进行肝脏B超图像识别的方法;该方法首先提取肝脏B超图像的空域和频域的纹理特征,然后运用SVM对108幅肝脏B超图像进行分类,利用PSO算法优化SVM的模型参数,最后将该方法与基于网格搜索法优化的SVM和基于BP神经网络的分类方法进行了对比;实验结果表明,在PSO-SVM算法下,所提取的两种纹理特征相结合能够有效地描述肝脏B超图像,基于粒子群优化算法的支持向量机模型具有较高的识别精度,平均分类准确率达94.44%,这就表明PSO-SVM算法适用于对肝脏B超图像的识别。
To improve the diagnostic accuracy of liver B ultrasound images, an approach was proposed based on Particle Swarm Optimi- zation （PSO） and Support Vector Machine （SVM）. First, texture features of spatial and frequency domains were extracted from liver B ul trasound images. Then, SVM was used to conduct the classification of 108 liver B ultrasound images. And PSO algorithm was used to opti- mize the model parameters. Last, the proposed method was compared with that SVM based on grid search method and BP neural network. Experimental results show, in the PSO-SVM algorithm, the combination of two texture features that extracted can effectively describe liver B ultrasound images. And SVM model based on PSO has higher recognition accuracy with its average accuracy up to 94.44 %, which con- firmed PSO-SVM algorithm is suitable for the recognition of liver B ultrasound images.
出处
《计算机测量与控制》
CSCD
北大核心
2012年第9期2491-2493,2500,共4页
Computer Measurement &Control
关键词
支持向量机
粒子群优化算法
灰度共生矩阵
小波变换
support vector machine
particle swarm optimization
gray level concurrence matrix
wavelet transform
