摘要
为构建在线生物文献核磁共振图像库,通过分析在线医学文献图像的特点,用塔式梯度方向直方图进行图像特征提取,结合图像对应的文本标注,采用基于高斯过程的分类方法设计实现了一个在线生物文献MRI图像识别系统。实验结果表明,该系统比基于单一特征的系统识别率更高,同时比基于标准的SVM和KNN的识别方法性能更好。表明该系统的设计是可行、可靠和有效的。
In order to construct an MRI image database from online biologic literature,the features of images in online biologic literature are analyzed and a method is proposed using the pyramid histogram of words to represent image features,together with word frequency features of the text,and then target images are identitied by the classification method based on the Gaussian process（GP）,and an MRI image recognition system based on online biological literature is designed and realized.Experimental result shows the system has a higher recognition accuracy than systems based on single feature,and also has a better performance than standard recognition methods such as SVM and KNN.The result shows the design of this system is feasible,reliable and efficient.
出处
《计算机工程与设计》
CSCD
北大核心
2012年第5期1886-1889,共4页
Computer Engineering and Design
基金
浙江省自然科学基金项目(NY1101359)
