滑坡颗粒图像识别的投影校正方法与应用研究  

A projection calibration method of landslide particles image recognition and its application

作　　者：杨峥 崔圣华 覃亮 郭宁 许向宁[4] 蒙明辉 向国萍 杨晴雯[2] YANG Zheng;CUI Shenghua;QIN Liang;GUO Ning;XU Xiangning;MENG Minghui;XIANG Guoping;YANG Qingwen(School of Civil and Hydraulic Engineering,Xichang University,Xichang,Sichuan 615000,China;State Key Laboratory of Geohazard Prevention and Geoenvironment Protection,Chengdu University of Technology,Chengdu,Sichuan 610059,China;Sichuan Huadi Construction Engineering Co.Ltd.,Chengdu,Sichuan 610081,China;405 Geological Brigade of Sichuan Bureau of Geology and Mineral Resources,Dujiangyan,Sichuan 611830,China)

机构地区：[1]西昌学院土木与水利工程学院,四川西昌615000 [2]成都理工大学地质灾害防治与地质环境保护国家重点实验室,四川成都610059 [3]四川省华地建设工程有限责任公司,四川成都610081 [4]四川省地质矿产勘查开发局四〇五地质队,四川都江堰611830

出　　处：《岩石力学与工程学报》2022年第2期362-376,共15页Chinese Journal of Rock Mechanics and Engineering

基　　金：国家自然科学基金青年基金资助项目(41907254);四川省教育厅项目(18ZB0546);四川省科技计划重点研发项目(2018GZ0499)。

摘　　要：破碎地层滑坡是我国西南山区典型的灾害类型之一,滑坡物质主要由破碎颗粒构成,具有较强的流动性和致灾能力,对滑坡区颗粒分布研究有助于灾害范围预测和灾害防治。基于无人机摄影和图像识别技术,针对投影平面与地形间夹角带来的颗粒投影变形问题,引入投影面积修正系数Ki,实现非统一平面投影和变形校正,从而更准确获得颗粒统计学特征。以2018年4月8日发生于汶川县龙溪乡的阿尔寨破碎层滑坡为例,利用所提方法获取滑坡颗粒数量、位置参数、投影面积、校正面积,对滑坡颗粒尺寸特征及颗粒空间分布规律,对滑坡纵向运动和横向运动模式,纵横2向运动的颗粒空间分布特征,纵向运动过程颗粒破碎与分选作用,巨石颗粒的势能驱动特征和巨石颗粒的惯性优势进行分析;结果表明,阿尔寨破碎层滑坡堆积区相比滑源区小颗粒占比更高,分选性更好,并随着运动距离增加小颗粒占比增高、分选性趋好;占比2.4%的巨石颗粒运动具有显著的势能趋动特征,运动距离相比小颗粒更远,巨石颗粒同时还具有明显的惯性优势,集中分布于滑坡堆积体的中心区域。Broken strata landslide is one of the typical disaster types in southwest mountainous area of China. The landslide material is mainly composed of broken particles and has strong fluidity and disaster-causing ability. The study of particle distributions in the landslide area is helpful to disaster range prediction and disaster prevention.Based on UAV photography and image recognition technology,this paper introduced the correction coefficient of the projection area,Ki,to realize non-uniform plane projection and deformation correction,aiming at the problem of particle projection deformation caused by the angle between the projection plane and the terrain,and so as to obtain the statistical characteristics of particles more accurately. Taking the Aerzhai landslide in Longxi Township,Wenchuan County on April 8,2018 as an example,the number of landslide particles,location parameters,projection areas and correction areas were obtained by the proposed method. The characteristics of the landslide particle size and the particle spatial distribution,the longitudinal and transverse motion patterns of the landslide,the spatial distribution of particles in vertical and horizontal directions,the crushing and sorting of particles during longitudinal motion,the potential energy driving characteristics and the inertia advantage of huge particles were analyzed. The results show that the accumulation area of Aerzhai landslide has higher proportion of small particles and better sorting than the sliding source area. With increasing the moving distance,the proportion of small particles increases and the sorting becomes better. The motion of huge particles with a ratio of 2.4% has significant potential energy driving characteristics,and the distance of movement is farther than that of small particles. At the same time,huge particles also have obvious inertial advantage,concentrated in the central area of landslide accumulation.

关 键 词：边坡工程 阿尔寨滑坡 颗粒识别 投影校正 堆积特征 

分 类 号：P642[天文地球—工程地质学]