顾及地理空间信息的无人机影像匹配像对提取方法  

A method for extracting match pairs of UAV images considering geospatial information

作　　者：任超锋 蒲禹池 张福强 REN Chaofeng;PU Yuchi;ZHANG Fuqiang(College of Geological Engineering and Geomatics,Chang’an University,Xi’an 710054,China)

机构地区：[1]长安大学地质工程与测绘学院,西安710054

出　　处：《自然资源遥感》2022年第1期85-92,共8页Remote Sensing for Natural Resources

基　　金：国家自然科学基金项目“非常规摄影测量模式下无人机影像匹配像对的提取方法研究”(编号:41801383);国家重点研发计划项目“岩溶山区特大滑坡早期识别与监测预警”(编号:2018YFC1504805);中央高校基本科研业务费“海量无人机影像三维自动重建关键技术研究”(编号:300102269206);“三维模型智能重建”(编号:300102269304)共同资助。

摘　　要：针对无人机影像三维重建中匹配像对提取适应性差、效率低、需准确的先验知识等问题,提出一种顾及无人机影像地理空间信息的匹配像对提取方法。首先利用主成分分析(principal component analysis,PCA)方法将高维度的特征降至低维特征,提高检索词典构建效率;其次,通过计算查询影像之间的反距离权重因子,构建综合检索因子,提高相似影像之间的可区分性;最后,通过计算检索阈值,舍弃阈值后的无效匹配像对,进一步提高了影像查准率。实验结果表明,与传统脚印图法和128维特征检索方法相比,该方法获得了更高的处理效率及更全面的稀疏重建结果,尤其针对海量无人机数据更具优势。To overcome the shortcomings such as poor adaptability,low efficiency,and the demand for prior knowledge in the 3D reconstruction using UAV images,this study proposed a method for extracting match pairs of UAV images considering geospatial information.The steps of this method are stated as follows.Firstly,reduce high-dimensional features of the images to low-dimensional features using the principal component analysis(PCA)method to improve the construction efficiency of the retrieval vocabulary.Secondly,construct a comprehensive retrieval factor by calculating the inverse distance weighting factor between query images to improve the distinguishability between similar images.Finally,discard invalid match pairs by calculating the retrieval threshold to improve the query precision of images.The experimental results show that,compared to the traditional footprint map method and 128-dimensional feature retrieval method,this method enjoys higher processing efficiency and more comprehensive sparse reconstruction results,especially for the massive UAV data.

关 键 词：匹配像对 视觉词袋 无人机 三维重建 影像检索 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]