基于GCP的大规模无人机影像检索方法  

Large-scale UAV Image Retrieval Method Based on GCP

作　　者：周宁 杨元维 王萍 高贤君[1] 方军 GCP ZHOU Ning;YANG Yuanwei;WANG Ping;GAO Xianjun;FANG Jun(School of Geosciences,Yangtze University,Wuhan 430100,China;Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China;Key Laboratory of Earth Observation of Hainan Province,Sanya,Hainan 572029,China;National-local Joint Engineering Laboratory of Geo-spatial Information Technology,Hunan University of Science and Technology,Xiangtan,Hunan 411201,China;Hunan Provincial Key Laboratory of Geo-information Engineering in Surveying,Mapping and Remote Sensing,Hunan University of Science and Technology,Xiangtan,Hunan 411201,China)

机构地区：[1]长江大学地球科学学院,武汉430100 [2]中国科学院空天信息创新研究院,北京100094 [3]海南省地球观测重点实验室,海南三亚572029 [4]湖南科技大学地理空间信息技术国家地方联合工程实验室,湖南湘潭411201 [5]湖南科技大学测绘遥感信息工程湖南省重点实验室,湖南湘潭411201

出　　处：《遥感信息》2021年第2期136-144,共9页Remote Sensing Information

基　　金：国家重点研发计划项目(2018YFB0505500、2018YFB0505504);自然资源部地理国情监测重点实验室开放基金项目(2020NGCM07);武汉大学测绘遥感信息工程国家重点实验室开放基金项目(18R04);海南省地球观测重点实验室开放基金项目(2020LDE001);国家自然科学基金项目(41701537);湖北省教育厅科学研究计划资助项目(Q20181317);长江大学大学生创新创业基金项目(2019022);湖南省重点研发计划项目(2018GK2015)。

摘　　要：鉴于利用无人机影像数据进行高精度测绘及地表三维信息提取时,从具有定位定向信息、数据规模大、影像分布密集等特点的数据集中检索包含地面控制点(ground control point,GCP)的目标影像存在检索速度较慢的问题,提出了基于GCP的影像检索方法。首先,对无人机影像点集构建最小外包矩形与基准坐标系;其次,利用GCP数据构建合适大小的索引矩形(ground control point rectangle,GCP-Rec);最后,依据GCP-Rec对影像数据批量建立索引,最终生成树形索引。效率对比实验表明,该方法能快速完成大规模无人机影像的空间索引构建,提升影像检索效率,且其检索性能优于传统的R树与四叉树等空间索引。影响因素对照实验表明,影像检索效率与GCP数量成反比关系,设定GCP-Rec边长为2.0~2.4倍航带间距时检索精度最优。In the process of high-precision mapping and extracting 3D surface information using unmanned aerial vehicle(UAV)imaging data,it is difficult to find the target image with ground control points(GCP)quickly and accurately from the data set with the positioning and orientation information,large data scale,and dense images distribution.Therefore,this paper proposes a large-scale UAV image retrieval method based on GCPs.Firstly,the minimum area bounding rectangle and two-dimensional rectangular coordinate system are constructed from the image point.Next,GCPs are used to establish an appropriate size index rectangle(GCP-Rec).Finally,according to GCP-Rec,the index is batched to generate a tree index.Comparison experiments show that this method can quickly construct a large-scale UAV image index and improve image retrieval efficiency.Its performance is better than the traditional R-tree and quad-tree spatial indexes.Comparison experiments indicate that the retrieval efficiency is inversely proportional to the number of GCPs,and the retrieval accuracy is optimal when the side length of GCP-Rec is set to 2-2.4 times the belt spacing.

关 键 词：地面控制点 空间索引 影像检索 无人机 最小面积外包矩形 树形索引结构 

分 类 号：P231[天文地球—摄影测量与遥感]