摘要
景点图像识别是当前图像识别领域和智慧旅游领域的一项基本任务。景点图像识别属于大规模图像检索,哈希检索算法是检索中一种常用的方法。针对传统哈希算法以及深度哈希算法存在的问题,改进现有的特征提取策略,提出一种改进的深度学习哈希检索方法。使用特定的领域来划分景点图像,通过领域区分来提取具有更好表达能力的景点特征,利用深度学习训练哈希函数以进一步优化网络性能。实验结果表明,该方法能够有效识别景点图像,取得了查准率95.69%、查全率93.36%、F1测度值94.51%的良好效果。
Image recognition of scenic spot is a basic task in the field of image recognition and intelligent tourism.Image recognition of scenic spot belongs to large-scale image retrieval.Hash retrieval algorithm is a common method in large-scale image retrieval.In order to solve the problems of traditional Hash algorithm and deep Hash algorithm,existing feature extraction strategies are improved,and a improved Hash retrieval method based on deep learning is proposed.The scenic spot image is classified by specific fields.Scenic spot features with better expression ability are extracted through domain differentiation,and Hash function is trained through deep learning to further optimize network performance.The experimental results show that the method can effectively identify scenic spot,and precision is 95.69%,recall is 93.36%,the F1 value is 94.51%.
作者
单慧琳
洪智毅
张银胜
王兴涛
SHAN Huilin;HONG Zhiyi;ZHANG Yinsheng;WANG Xingtao(College of Electronic and Information Engineering,Wuxi University,Wuxi 214105,Jiangsu,China;College of Electronic and Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China)
出处
《实验室研究与探索》
CAS
北大核心
2022年第5期12-17,共6页
Research and Exploration In Laboratory
基金
国家自然科学基金项目(61372128)
南京信息工程大学自然科学研究项目(2020yng003)。
关键词
景点图像识别
深度学习
哈希检索
卷积神经网络
特征提取
image recognition of scenic spot
deep learning
Hash retrieval
convolutional neural network
feature extraction
