摘要
基于合成孔径雷达干涉测量(InSAR)技术,结合热点分析和机器学习算法,进行区域高形变斜坡自动提取和潜在滑坡人工智能识别研究,以提升潜在滑坡识别效率和准确性,解决传统人工调查和目视解译无法有效识别位于高位、隐蔽性较强的潜在滑坡问题.结果表明,基于热点分析自动提取20处高形变区域,提取正确率、错分率和漏分率分别为74.31%、25.69%和11.80%,证实热点分析方法能够有效应用于InSAR高形变区自动识别和提取.基于识别的高形变区,结合历史滑坡灾害发育特征,利用机器学习算法建立潜在滑坡预测模型,采用表现最佳的自适应提升模型对自动提取区域进行预测,预测召回率和准确率分别为81%和65%,能够实现潜在滑坡的有效识别.
We extract regional high deformation slope automatically and artificial intelligence identification of potential landslides based on interferometric synthetic aperture radar(InSAR) surface deformation monitoring technology combined with hotspot analysis and machine learning,to improve the efficiency and accuracy of potential landslide identification and solve the problem that the traditional manual investigation and visual interpretation can not effectively identify the high and hidden potential landslide.The results indicated that:based on the hotspot analysis technology,a total of 20 severe deformation areas are extracted.The extraction accuracy rate,the error rate and the missing rate are 74.31%,25.69% and 11.80% respectively,which proves that hotspot analysis can be effectively applied to the automatic recognition and extraction of InSAR high deformation region.Based on the identified high deformation area,combined with the development characteristics of historical landslide disasters,the potential landslide prediction model is built using machine learning.The best AdaBoost prediction model is used to conduct prediction in the areas of extract automatically,and the recall and the accuracy rate are 81% and 65% respectively,which can prove the effectiveness of identifying potential landslides.
作者
梁懿文
张毅
苏晓军
刘旺财
李媛茜
王爱杰
孟兴民
LIANG Yi-wen;ZHANG Yi;SU Xiao-jun;LIU Wang-cai;LI Yuan-xi;WANG Ai-jie;MENG Xing-min(School of Earth Sciences,Lanzhou University,Lanzhou 730000,China;Technology&Innovation Center for Environment Geology and Geohazards Prevention of Gansu Province,Lanzhou 730000,China;College of Earth and Environmental Sciences,Lanzhou University,Lanzhou 730000,China)
出处
《兰州大学学报(自然科学版)》
CAS
CSCD
北大核心
2022年第2期212-221,共10页
Journal of Lanzhou University(Natural Sciences)
基金
国家重点研发计划项目(2018YFC1504704)
甘肃省科技重大专项项目(19ZD2FA002)
国家自然科学基金青年基金项目(42007232)
甘肃省青年科技基金项目(20JR5RA223)
甘肃省科技计划项目(18YF1WA114)
自然资源部中国地质调查局地质调查项目(DD20189270)。
关键词
潜在滑坡
基于合成孔径雷达干涉测量技术
机器学习
人工智能
potential landslide
interferometric synthetic aperture radar
machine learning
artificial intelligence