期刊文献+

基于显著性区域和蚁群算法的图像检索研究

Image retrieval based on saliency region and ant colony algorithm
原文传递
导出
摘要 针对提取到的图像特征受背景信息干扰,不能有针对性地提取到所需要的图像信息影响检索精度。为了解决这一问题,本文提出一种基于改进VGGNet(visual geometry group network)和蚁群算法的图像显著性区域检索算法。首先,利用类激活映射(class activation mapping,CMA)算法对图像显著性区域进行提取,剔除图像背景信息;然后使用训练好的RS-VGG16模型提取图像显著性区域特征来表征图像。引入主成分分析(principal component analysis,PCA)算法,对高维特征进行降维的同时减少特征信息的损失。最后,引入蚁群算法对检索结果进行优化。在corel;000数据集上,选取基于VGG16网络的图像全局特征检索算法以及传统的BOF(bag of features)图像检索算法进行对比试验。本文提出算法相较于基于VGG16网络的图像检索算法,平均查准率(mean average precision,MAP)值平均提升约4.36%,相较于传统的BOF算法,MAP值平均提升约16.99%。实验结果表明本文提出算法能够很好地去除图像背景信息的干扰,具有更优的检索性能。 Since the extracted image features are interfered by background information,the retrieval accuracy is affected by the failure to extract the required image information.In order to solve this problem,this paper proposes an image saliency region retrieval algorithm based on improved visual geometry group network(VGGnet)and ant colony algorithm.Firstly,class activation mapping(CMA)algorithm is used to extract the salient region of the image and remove the background information.Then the trained RS-VGG16 model is used to extract the salient regional features of the image to represent the image.Principal component analysis(PCA)algorithm is introduced to reduce the dimensionality of high-dimensional features while reducing the loss of feature information.Finally,the ant colony algorithm was introduced to optimize the retrieval results.In the corel;000 data set,the image global feature retrieval algorithm based on VGG16 network and the traditional bag of features(BOF)image retrieval algorithm are selected for comparative experiment.Compared with the image retrieval algorithm based on VGG16 network,the MAP value of the proposed algorithm is improved by about 4.36%on average,and compared with the traditional BOF algorithm,the mean average precision(MAP)value is improved by about 16.99%on average.Experimental results show that the algorithm proposed in this paper can remove the interference of image background information well and has better retrieval performance.
作者 夏思珂 雷志勇 XIA Sike;LEI Zhiyong(Key Electronic Information Engineering College,Xizan Technology University,Xifan,Shaanxi 710000,China)
出处 《光电子.激光》 CAS CSCD 北大核心 2021年第12期1300-1306,共7页 Journal of Optoelectronics·Laser
关键词 图像检索 神经网络 注意力机制 主成分分析(principal component analysis PCA) 蚁群算法 image retrieval neural network attentional mechanism principal component analysis(PCA) ant colony algorithm
  • 相关文献

参考文献2

二级参考文献8

共引文献9

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部