摘要
鱼类图像具有丰富的颜色和纹理特征,传统的多特征鱼类图像检索采用多特征向量合并方式将颜色与纹理特征合并后进行鱼类图像检索.这种多特征合并方式对表达多种特征的能力较弱,合并后的多特征向量无法有效地将鱼类的颜色与纹理特征进行融合,从而影响了鱼类图像检索算法的性能.针对以上问题,本文在基于灰度图的SURF特征提取算法的基础上提出了基于HSVG四通道的SURF特征图像检索方法,该算法能够有效融合图像中的纹理信息和颜色信息,按通道匹配特征后加权进行鱼类图像检索还可以对鱼类图像背景信息有一定的抗干扰能力.为了验证本文所提算法的有效性,课题组在澳大利亚昆士兰大学提供的QUT_fish_data数据集上进行了仿真实验,实验结果表明,本文所提算法在较大规模数据集以及鱼类自然场景下都具有较好的鲁棒性和准确率.
Fish images are rich in color and texture features.The traditional multi feature fish image retrieval uses multiple feature vectors combine color and texture features.The characteristics of combining methods is weak to express a variety of characteristics ，multi feature vector after the merger can′t effectively integrate fish color and texture feature which will influence the performance of fish image retrieval algorithm.To solve the above problems，the SURF feature extraction algorithm based on the HSVG four channels is proposed based on gray image retrieval method，the algorithm can effectively fuse the texture information and color information in the image，and then fish image retrieval can be carried out by weight after the channel feature matching，so this algorithm also has a certain antiinterference ability to the fish image background information. In order to evaluate the performance of this method，we have made some experiments in QUT_fish_data which is provided by University of Queensland.These experiments suggest that our method has better accuracy and robustness,in large scale datasets and fish natural scene.
作者
张美玲
吴俊峰
于红
孙建伟
罗强
ZHANG Mei-ling;WU Jun-feng;YU Hong;SUN Jian-wei;LUO Qiang(College of Information Engineering,Dalian Ocean University,Dalian 116023,China;Key Laboratory of Marine Information Technology of Liaoning Province,Dalian 116023,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2018年第9期2085-2089,共5页
Journal of Chinese Computer Systems
基金
大连市科技计划项目(2015A11GX022)资助
