基于SVM、TSVM与ELM的图像检索算法对比研究

Study of Image Retrieval Algorithm based on SVM,TSVM and ELM

分类算法应用于图像检索中,可有效解决图像检索中的分类问题,缩小低层特征与高层特征之间的鸿沟,提高检索精度。以图像颜色与纹理特征并结合图像分块特征作为低层综合特征,借鉴词袋(Bag of Words)模型,利用K均值(K-means)聚类算法,分别采用支持向量机(SVM)、直推式支持向量机(TSVM)以及极限学习机(ELM)三种学习机制,对corel图像库进行分类检索。实验表明,ELM分类器的识别准确率高于SVM和TSVM分类器,且检索速度快。

Classification algorithm is applied to image retrieval,which can effectively solve the problem of classification of image retrieval,narrowing the gap between low- level features and high- level features,and improve the retrieval precision. In image color,texture features combined with the feature of image content features as low comprehensive,draw lessons from the word Bag（ Bag of Words） model,using K- means（ K- means algorithm）,and support vector machine（ SVM）,straight push support vector machines（ TSVM） with extreme learning machine（ ELM） three learning mechanism are respectively used,therefore classifying corel image database retrieval. Experiments show that the ELM classifier recognition accuracy is higher than the SVM and TSVM classifier,and the retrieval speed is also rapid.