基于卷积自编码和哈希算法的图像检索研究  

Image Retrival Research Based on Convolutional Auto⁃encoder and Hash Algorithm

作　　者：周纤 邱奕敏[1] 吴振宇 ZHOU Qian;QIU Yi-min;WU Zhen-yu(School of Information Science and Engineering,Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]武汉科技大学信息科学与工程学院,湖北武汉430081

出　　处：《仪表技术与传感器》2020年第11期105-110,共6页Instrument Technique and Sensor

基　　金：国家自然科学基金面上项目(61373109);湖北省自然科学基金资助项目(2018CFB346,2019CFB138)。

摘　　要：针对目前图像检索的精度和速度不高等问题,提出一种基于卷积自编码神经网络和哈希编码策略的图像检索方法。该方法使用在MNIST和Fashion-MNIST数据集上预训练的卷积自编码网络模型对实验图像数据集提取编码器特征来得到图像特征的深层表示。所提取的特征具有自编码和卷积神经网络两者的优点。然后使用哈希策略将这些深层表示进行编码,从而获得图像的二进制码,通过计算汉明距离的相似度对图像进行检索。在2个大型数据集MNIST和Fashion-MNIST上进行大量实验,与现有技术相比,该算法具有更高的查准率、查全率和平均检索率。Aiming at the problems of low precision and speed of image retrieval,an image retrieval method based on convolu⁃tional autoencoder neural network and hash coding strategy was proposed.The method extracted the encoder features from the ex⁃perimental image dataset using a pre⁃trained convolutional autoencoder network model pre⁃trained on the MNIST and Fashion-MNIST data sets to obtain a deep representation of the image features.The extracted features had the advantage of both autoencod⁃er and convolutional neural networks.These deep representations were then encoded using a hashing strategy to obtain the binary code of the image,which was retrieved by calculating the similarity of the Hamming distance.A large number of experiments were carried out on two large data sets,MNIST and Fashion-MNIST.Compared with the prior art,the algorithm has higher precision,re⁃call and average retrieval rate.

关 键 词：图像检索 卷积自编码神经网络 哈希编码 卷积神经网络 汉明距离 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]