词袋特征支持下的无人机影像滑坡解译模型  

An Automatic Landslide Interpretation Model of UAV Imagery Based on BoW

作　　者：李政[1] 李永树[1] 郭加伟[1] 张帅毅[1] 刘锟铭 

机构地区：[1]西南交通大学地球科学与环境工程学院,成都610031

出　　处：《遥感信息》2016年第5期24-29,共6页Remote Sensing Information

基　　金：国家"十二五"科技支撑计划课题(2014BAL01B04);中央高校基本科研业务费专项资金(2682013ZT26);国土资源部地学空间信息技术重点实验室开放基金(KLGSIT2015-03)

摘　　要：川西地区滑坡灾害频发,给当地人民生命、财产安全带来了极大威胁。如何利用无人机高分影像构建滑坡灾害解译模型是实现快速自动解译的关键。针对该问题,首先,对比了多种影像特征提取方法,然后将词袋特征和支持向量机引入到构建滑坡灾害解译模型中,提出了一种基于词袋特征的无人机影像滑坡灾害解译模型。最后,选取512汶川地震及420芦山地震系列无人机影像构建了滑坡灾害样本库,基于该样本库进行了实验,分类准确度达到76%,相比于传统目视解译在效率上有明显优势。分析表明,所提出方法可用于高分影像滑坡自动解译。此外,该模型可用于大面积高分影像快速山地滑坡灾害定位,可实现滑坡灾害的快速检测,对于灾后应急具有一定的指导作用。West Sichuan area frequently experiences landslides,which brings great threat to the local people＇s life and property safety.How to use UAV high spatial resolution images to construct an interpretation model of landslide disaster is the key to achieve fast and automatic interpretation.For this problem,we firstly compare multiple methods of extracting image feature,then introduce the bag-of-words（BoW）and support vector machine（SVM）into the interpretation model,and finally propose an automatic landslide interpretation model of UAV images based on BoW.And,we select a series of UAV images of 5·12 Wenchuan earthquake and 4·20Lushan earthquake to build sample library of landslide disaster.Experiments based on this sample library show that the comprehensive classification accuracy（ACC）proposed in this paper is 76%,which has better effiency than traditional visual interpretation,which proves that this method can be available for the automatic interpretation.Meanwhile,this method can be used to rapidly locate the landslide disaster within a large area of UAV images,which has guidance significance for emergency response after the landslide disater.

关 键 词：山地滑坡 词袋特征 无人机影像 机器学习 遥感图像信息提取 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]