一种基于增强模糊分类器的快速图像分类方法  

A Fast Image Classification Method Based on Enhanced Fuzzy Classifier

作　　者：徐小龙[1] 高仲合[2] 韩丽娟[1] 

机构地区：[1]曲阜师范大学实验教学中心,山东日照276826 [2]曲阜师范大学软件学院,山东曲阜273165

出　　处：《电子技术（上海）》2016年第8期45-49,35,共6页Electronic Technology

摘　　要：基于内容的图像匹配是计算机科学中一个具有挑战性的问题,它由两个相对困难的任务组成:识别图像上的对象和快速搜索已识别对象的集合。文章提出了一种基于简单模糊分类器的图像分类新方法,使用局部图像特征在已知类和其它类之间区分,使用增强元学习发现最有代表性的局部特征。所提出的方案使用目前最新的图像数据集进行测试,并与使用支持向量机分类的Bag-of-Features图像表示模型进行比较。新方法能达到更高的分类准确度,并且学习和测试所需的时间也缩短了30%以上。Content based image matching is a challenging problem in computer science. It consists of two relatively difficult tasks： identify objects on the image and quickly search for a set of objects that have been identified. This paper presents a new method of image classification based on simple fuzzy classifier. It uses local image features to distinguish between known classes and other classes. It uses the enhanced meta-learning to find the most representative local features. The proposed method is tested using the currently updated image data set, and compared with the Bo F-SVM image representation model. The new method can achieve higher classification accuracy, and the time required for learning and testing has been reduced by more than 30%.

关 键 词：图像分类 图像检索 模糊分类器 增强元学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]