首页|顾及无人机影像多特征信息的滑坡识别

顾及无人机影像多特征信息的滑坡识别

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
滑坡灾害具有较强的突发性和破坏力,快速、准确识别滑坡对受灾区的应急救援、灾害评估具有重要意义.无人机(unmanned aerial vehicle,UAV)具备操作灵活、高时效性和高分辨率等优势,为特定区域的滑坡数据获取提供了强大的技术支持.以云南省绿春县为研究区,以UAV影像为数据源,首先构建研究区的数字正射影像(digital orthophoto map,DOM),在此基础上进行多尺度分割;然后综合光谱、形状和纹理等多特征信息,计算样本分离度后优化特征空间;最后基于面向对象的贝叶斯(Bayes)分类方法进行滑坡识别与精度分析.研究结果表明:采用Bayes方法得到的滑坡识别总体精度(overall accuracy,OA)达 92.49%,Kappa系数达 0.888,其中滑坡区域的生产者精度(producer accuracy,PA)和用户精度(user accuracy,UA)分别为 89.84%和 83.17%.此外,与决策树(decision tree,DT)、随机森林(random forest,RF)和支持向量机(support vector machine,SVM)3 种分类方法的滑坡识别结果进行精度对比,Bayes方法的OA较其他分类方法提高 3.26%～5.86%,Kappa系数提高0.048～0.092.该方法对于复杂、破碎山地的滑坡识别精度较高,能够满足基于高分辨率UAV影像的滑坡精细化识别应用需求.
Landslide identification considering multi-feature information of UAV images
Landslide disaster has the characteristics of suddenness and high destructiveness.It is of great significance to identify landslide quickly and accurately for emergency rescue and disaster assessment in affected areas.Unmanned aerial vehicle(UAV)has the advantages of flexible operation,high timeliness and high resolution,which can provide powerful technical support for landslide data acquisition in specific areas.Lvchun County of Yunnan Province was taken as the research area,and UAV visible light image was used as the data source.Firstly,the digital orthophoto map(DOM)of the study area is constructed,and multi-scale segmentation is carried out.Then,the spectrum,shape,texture and other multi-feature information are integrated to calculate the sample separation degree and optimize the feature space.Finally,the landslide identification and accuracy analysis are carried out based on object-oriented Bayes classification method.The results show that the overall accuracy(OA)of landslide identification obtained by Bayes method is 92.49%,and the Kappa coefficient is 0.888.The producer accuracy(PA)and user accuracy(UA)of landslide are 89.84%and 83.17%,respectively.In addition,compared with the landslide identification results of decision tree(DT),random forest(RF)and support vector machine(SVM),the OA of Bayes method is 3.26%～5.86%higher than that of other classification methods,and the Kappa coefficient is 0.048～0.092 higher.This method has high accuracy for landslide identification in complex and fractured mountains,and can meet the application requirements of landslide fine identification based on high-resolution UAV images.

object-orientedBayes classificationlandslideUAV imagesmulti-feature

张钰洁、刘佳佳、孙龙、易邦进、李佳

展开 >

云南师范大学 地理学部,云南 昆明 650500

中国能源建设集团云南省电力设计院有限公司,云南 昆明 650051

云南省设计院集团有限公司,云南 昆明 650103

云南省地质科学研究所,云南 昆明 650051

展开 >

面向对象 贝叶斯分类 滑坡 无人机影像 多特征

国家自然科学基金项目云南省基础研究计划项目兴滇英才支持计划兴滇英才支持计划云南省院士专家工作站云南省基础研究专项重点项目

41961061202301AT070061YNWR-QNBJ-2020-048YNWR-QNBJ-2020-1032017IC063202201AS070024

2024

10.13577/j.jnd.2024.0510

自然灾害学报
中国地震局工程力学所 中国灾害防御协会

自然灾害学报

CSTPCD北大核心
影响因子：0.862
ISSN：1004-4574
年,卷(期)：2024.33(5)
  • 18