摘要
为有效提取中药材图像的特征,提高中药材图像分类准确率并提升检索性能,对中药材图像的梯度方向直方图形状特征和局部二元模式纹理特征进行研究,对2种特征进行维数改进,提出一种基于形状特征和纹理特征的中药材检索方法。使用改进的图像梯度方向直方图和分块局部二元模式进行形状及纹理的特征提取;对提取得到的特征向量进行线性组合;采用一对一方式构造多分类器,使用支持向量机进行分类检索。实验结果表明,组合降维特征提取算法能在中药材图像数据集中取得较好的识别效果。
To extract features of Chinese herbal medicine images more effectively as well as to improve the classification accuracy and the retrieval performance of Chinese herbal medicine images, the shape features of the histogram of oriented gradient and the texture features of the local binary pattern in Chinese herbal medicine images were studied, and both feature dimensions were im- proved, a Chinese herbal medicine retrieval method based on shape features and texture features was proposed. Firstly, the IHOG (improved histogram of oriented gradient) algorithm was developed to extract shape features of images and the PLBP (partitioned local binary pattern) method was applied to extract texture features of images. Secondly, through the linear combi- nation of the two kinds of features mentioned above, a feature vector was gotten. Lastly, the one-vs-one SVM (support vector machine) classification was constructed to retrieval images from the Chinese herbal medicine dataset. The experimental results show that algorithms with the linear combination of these two kinds of reduced features can obtain good recognition rate in Chi- nese herbal medicine image datasets.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第11期3903-3907,共5页
Computer Engineering and Design
基金
四川省应用基础计划基金项目(2013JY0086)
四川省科技创新苗子工程基金项目(2012ZZ047
20132033)
关键词
分块局部二元模式
梯度方向直方图
支持向量机
中药材检索
特征降维
partitioned local binary pattern
histogram of oriented gradient
support vector maehine
Chinese herbal medicineimage retrieval
feature dimension reduction