基于循环神经网络的图像检索技术
Image Retrieval Technology Based on a Recurrent Neural Network
周春良1
作者信息
- 1. 郑州西亚斯学院电信与智能制造学院,河南郑州 450000
- 折叠
摘要
传统图像检索方法只关注图像的浅层特征,难以深入地捕捉图像的整体结构,导致提取的特征往往无法全面反映图像内容,影响了图像检索的准确性和效率.为此,文章提出一种基于循环神经网络的图像检索技术.采集图像数据并构建图像数据库;对数据库中的图像进行去均值和归一化处理;利用循环神经网络抽取处理后的图像数据特征,得到图像的表征信息;将处理后的不同表征信息进行融合,作为输入数据进行图像编码处理;计算编码后的图像之间的相似度,从图像数据库中检索出与查询图像最相似的图像,从而完成图像检索任务.经实验测试,该技术具有较高的查准率,能够准确识别并检索与查询图像相似的图像,且检索效率显著提升.
Abstract
Traditional image retrieval methods only pay attention to the shallow features of the image,and it is difficult to capture the overall structure of the image in depth,resulting in the extracted features often unable to fully reflect the image content,which affects the accuracy and efficiency of image retrieval.Therefore,the paper proposes an image retrieval technology based on cyclic neural network.The image data is collected and an image database is constructed.The images in the database are de-averaged and normalized,and the characteristics of the processed image data are extracted by using the cyclic neural network to obtain the representation information of the image.The processed different representation information is fused and used as input data for image coding.Calculate the similarity between the encoded images,and retrieve the image most similar to the query image from the image database,thus completing the image retrieval task.The experimental result shows,the technology has a high precision.It can accurately identify and retrieve images similar to the query image,and the retrieval efficiency is significantly improved.
关键词
循环神经网络/图像检索/检索技术/图像表征信息/相似度Key words
recurrent neural network/image retrieval/retrieval technology/image representation information/similarity引用本文复制引用
出版年
2024
NETL
NSTL
万方数据