基于无人机影像的露天矿工程量监测分析方法  

Method of Engineering Volume Monitoring and Calculation for Open-Pit Mine from UAV Images

作　　者：许志华[1] 吴立新[2,3] 陈绍杰[4] 王植[2] 

机构地区：[1]北京师范大学民政部/教育部减灾与应急管理研究院,北京100875 [2]东北大学资源与土木工程学院,辽宁沈阳110819 [3]中国矿业大学环境与测绘学院,江苏徐州221116 [4]龙岩学院资源工程学院,福建龙岩364012

出　　处：《东北大学学报（自然科学版）》2016年第1期84-88,共5页Journal of Northeastern University(Natural Science)

基　　金：国家重点基础研究发展计划项目(2011CB707102);中央高校基本科研业务费专项资金资助项目(105565GK)

摘　　要：提出了一种基于无人机影像序列的露天矿工程量(采剥量、堆放量等)计算方法.该方法利用旋翼无人机搭载低成本便携式数码摄像机获取露天矿山不同时间的视频帧或影像序列.基于运动恢复结构(Sf M)和多目立体视觉(PMVS)算法,自动生成矿山完整、致密的三维点云.研究设计了一种基于形态不变区的点云配准方法进行两期点云空间配准,并采用DTM三角网差值法计算矿山工程量.矿堆体积变化无人机监测实验结果表明,该方法重建点云模型的点间相对误差小于±1%,堆放体积变化监测精度接近92%,基本达到地面Li DAR扫描的堆放体积变化监测精度.The image sequences from an unmanned aerial vehicle （UAV） are used to calculate the engineering volume （overburden amount, stacking amount, etc. ） of open-pit mine. Firstly, two sets of video frames or optical images of the open-pit mine are collected with a time interval using a portable digital camera installed on an octocopter. Next, two groups of the point clouds are automatically generated by implementing structure from motion （SfM） and patch-based multi- view stereo （PMVS） algorithms. And then, the two point clouds are fine registered with a constant region-based registration method. Finally, the engineering volume is computed with a differential method for digital terrain model triangulated irregular network （DTM-TIN）. It shows that the relative error of the point cloud model is lower than ＋ 1% in the experiment for change detection of a stacking stockpile with UAV images. Moreover, the accuracy for monitoring the volume change is up to 92% , which is comparable to that of a terrestrial laser scanning.

关 键 词：无人机 露天矿 运动恢复结构(SfM) 多目立体视觉(PMVS) 工程量 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]