基于单像近景摄影测量的滑坡裂缝探测方法  

Landslide Crack Detection Method Based on Single Image Close-range Photogrammetry

作　　者：刘志奇[1] 李天子[1] 刘昌华[1] 张慧慧[2] Liu Zhiqi;Li Tianzi;Liu Changhua;Zhang Huihui(School of Surveying and Land Information Engineering,Henan Polytechnic University,Jiaozuo 454000,China;College of Resources and Civil Engineering,Northeastern University,Shenyang 110819,China)

机构地区：[1]河南理工大学测绘与国土信息工程学院,河南焦作454000 [2]东北大学资源与土木工程学院,辽宁沈阳110819

出　　处：《金属矿山》2018年第8期108-113,共6页Metal Mine

基　　金：煤炭联合基金项目(编号:U1261206)

摘　　要：滑坡监测能够为预测可能发生滑坡灾害的时间和范围、制定减灾方案提供可靠依据。针对目前基于近景摄影测量的滑坡监测需要建立立体模型、数据处理复杂、数据量大、难以实现实时监测预警的不足,设计了一种适合实时、远程传输的单像滑坡监测方案。该方案主要采用滑移前后相邻影像匹配的方式,用同名点的位移场计算滑移矢量场。在确定用于该方案的最佳影像匹配算法时,对比分析了SIFT、SURF和ASIFT 3种常用算法在旋转、缩放、光照变化和仿射变换4种情况下的匹配性能,并根据该监测方案需要确保滑坡区域匹配点的密度和滑坡发生初期仿射变换较小的特点,认为SIFT算法最适合于该监测方案。在此基础上,提出了基于SIFT特征点的二维场位移矢量模型和相邻影像对应像素灰度值跳跃检验模型2种滑坡范围圈定模型,根据2种模型圈定的滑坡范围在理论上的差异,将2种模型进行耦合,能够有效探测出滑坡裂缝,为滑坡预警提供可靠依据。Landslide monitoring provides a reliable basis for forecasting the time and spot of landslide hazards and formulating disaster reduction plans.According to the conventional close-range photogrammetry that a three-dimensional model is needed,which data processing flow is complicated and the data volume is large,it is difficult to realize the real-time monitoring and early warning.A single image landslide monitoring scheme for real-time and remote transmission is proposed.The scheme mainly adopts the method of matching adjacent images before and after sliding,and the sliding vector field is calculated by the displacement field of the same feature point.In order to determine the best image matching algorithm for the scheme,the matching properties of SIFT（Scale Invariant Feature Transform）,SURF（Speeded Up Robust Features）and ASIFT（Affine-SIFT）algorithm are compared in the four situations of rotation,zooming,illumination and affine transformation.According to the monitoring scheme,it is necessary to guarantee the density of the matching point of the landslide region and the initial affine transformation of the landslide is small,then the study finds that the SIFT algorithm is best suited to the monitoring scheme.On that basis,a two-dimensional field displacement vector model based on SURF feature points and a corresponding gray scale jumping test model for adjacent images are proposed.Based on the theoretical difference between the two models,the two types of monitoring models are coupled with the ability to detect cracks in the landslide.Due to the landslide cracks is the basis of landslide precursor detection,therefore,the scheme proposed in this paper could provide an important support for landslide warning.

关 键 词：近景摄影测量 滑坡监测 SIFT 滑坡裂缝 滑坡预警 仿射变换 

分 类 号：P234.1[天文地球—摄影测量与遥感] P642.22[天文地球—测绘科学与技术]