面向内容的众源影像聚合检索与智能筛选  

Content-based crowdsourcing imagery retrieval and intelligent screening

作　　者：崔萌 谢曹东 单杰[1,2] CUI Meng;XIE Caodong;SHAN Jie(School of Remote Sensing and InformationEngineering,Wuhan University,Wuhan 430079, China;School of Civil Engineering,PurdueUniversity,Indiana-IN 47907,USA)

机构地区：[1]武汉大学遥感信息工程学院,武汉430079 [2]普渡大学土木工程学院,美国印第安纳州47907

出　　处：《测绘科学》2019年第3期165-171,共7页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41271431)

摘　　要：众源数据存在基数大、来源复杂、相关性低等特点,使得检索和筛选成为研究的热点,针对当前研究未能有效利用众源地理数据源的问题,该文提出了一种基于聚合平台的网络图片动态检索流程,以语义关键字和地理坐标为索引进行多图片库的聚合搜索。利用深度神经网络进行影像进行筛选的方法,解决了传统筛选方法筛选精度不高的缺点,具有高识别率和高效的特点。实验表明,本文提出的方案能够有效搜集高质量的众源影像,对于众源地理信息处理有潜在的价值。Crowdsourcing geographic data is faced with problems such as large amount,complex source and low correlation,which makes the retrieval and screening become the hotspot of research. In this paper, an aggregation platform of network pictures and a dynamic searching process on that platform using semantic keywords and geographic coordinates were proposed. Using deep neural network,the proposed process screen images with higher recognition rate and efficiency compared with traditional methods. Experiments showed that the proposed process could effectively collect network pictures,and had potential value for processing geographic information of various sources.

关 键 词：众源影像 深度学习 影像检索 服务聚合 智能筛选 

分 类 号：P208[天文地球—地图制图学与地理信息工程]