基于特征提取的SVM图像分类技术的无人机遥感建筑物震害识别应用研究  

Application of SVM Image Classification Technology Based on Feature Extraction in Seismic Damage Identification of Buildings by UAV Remote Sensing

作　　者：张莹[1] 郭红梅[1] 尹文刚 赵真 鲁长江[1] 肖本夫 ZHANG Ying;GUO Hongmei;YIN Wengang;ZHAO Zhen;LU Changjiang;XIAO Benfu(Sichuan Earthquake Administration,Chengdu 610041,China;College of Armed Police Officer,Chengdu 610213,China)

机构地区：[1]四川省地震局,四川成都610041 [2]武警警官学院,四川成都610213

出　　处：《灾害学》2022年第4期30-36,56,共8页Journal of Catastrophology

基　　金：国家重点研发计划项目(2020YFA0710603-07);国家自然科学基金项目(42061073);四川省重点研发项目(2020YFS0451);四川地震科技创新团队专项(201901)。

摘　　要：在现有的建筑物震害信息获取途径中,相比传统的现场调查法,无人机遥感系统具有机动灵活、快速高效等优点,目前已成为一种重要的震害信息获取手段。而在遥感图像中识别建筑物震害时,常用的人工目视解译及现有的计算机自动识别方法存在效率低下、精度不足等缺陷。结合机器学习最新进展,将基于特征提取的SVM图像分类技术应用到无人机遥感建筑物震害识别中,通过尺度不变特征转换(SIFT)提取图像特征后,再采用视觉词袋构建建筑物震害无人机遥感图像特征向量标签库,作为SVM进行图像分类的基础。并以2021年9月16日发生的四川泸县6.0级地震为例,对方法的可行性加以验证。结果表明:该方法可快速准确地从无人机遥感图像中识别出建筑物震害情况。In the existing ways of obtaining earthquake damage information of buildings,compared with the traditional field investigation method,UAV remote sensing system has the advantages of flexibility,speedy and efficiency.And has become an important means of obtaining earthquake damage information presently.When recognizing earthquake damage of buildings in remote sensing images,the commonly used manual visual interpretation and existing computer automatic recognition methods have some defects,such as low efficiency and insufficient accuracy.Therefore,combined with the latest progress of machine learning,we apply SVM image classification technology based on feature extraction to UAV remote sensing building earthquake damage recognition.After extracting image features through scale invariant feature transformation(SIFT),the visual word bag is used to construct the feature vector label Library of UAV remote sensing image of building earthquake damage,which is the basis of SVM image classification.Taking the Luxian M6.0 earthquake on September 16,2021 as an example,the feasibility of the method is verified.The results show that this method can quickly and accurately identify the earthquake damage of buildings from the UAV remote sensing images,and provide effective information support for the government and industry departments to carry out emergency disposal in time.

关 键 词：尺度不变特征转换(SIFT) 特征向量标签库 支持向量机(SVM) 图像分类技术 无人机遥感 建筑物震害识别 四川泸县6.0级地震 

分 类 号：X43[环境科学与工程—灾害防治] P15.5[天文地球—天文学] P315