一种使用多特征的鱼类图像检索方法

为解决数量巨大、特征复杂的鱼类图像难以用单个特征准确描述,以至于给鱼类图像检索带来困难的问题,提出了一种基于多特征的鱼类图像检索方法——MFFIR,并针对鱼类图像的特点设计了由颜色直方图、特征矩和Gabor特征构成多维特征向量。在此基础上,采用MFFIR法通过计算多维特征向量的特征相关性,实现了鱼类图像的匹配与检索。为验证此方法的有效性,使用"QUT_fish_data"数据集和"DLOU_fish_data"数据集进行了仿真实验,实验结果表明,MFFIR法具有较好的准确率和较强的鲁棒性。

A fish image retrieval method using multiple features

It is impossible to apply single feature to describe fish image which is characterized by large quantity and complicated features. A novel fish image retrieval method is proposed using multiple features as MFFIR and charac-terized to deal with the problem of the fish image retrieval. On the basis of analysis, a multi-features vector com-prised of color histogram feature, color moment feature and Gabor feature is designed based correlation calculation method to calculate the similarity between images. The fish image retrieval method was described in detail, and some experiments were designed to test the performance of MFFIR on “QUT fish data” dataset and “DLOU fish data” dataset. The findings show that MFFIR has good performance and robust.