基于无人机测绘图像的大面积地形变化特征提取方法  

A method for extracting features of large-area terrain changes based on unmanned aerial vehicle surveying and mapping images

作　　者：邱龙 QIU Long(Zhuhai Institute of Surveying and Mapping,Zhuhai,Guangdong 519000,China)

机构地区：[1]珠海市测绘院,广东珠海519000

出　　处：《北京测绘》2024年第6期930-935,共6页Beijing Surveying and Mapping

基　　金：国家自然科学基金(41431178)。

摘　　要：为了解决大面积地形变化特征提取效果较差的问题,本文提出基于无人机测绘图像的大面积地形变化特征提取方法。利用信息重构法收集无人机拍摄的图像信息,计算地形边缘尺度;利用图像滤波算法预处理大面积图像。测量和计算地形特征参数,利用加权全极化分解地形变化特征,通过特征选择因子与特征判断因子加强特征的区分效果,利用合成孔径雷达(SAR)特征分类方法划分地形特征;通过随机森林和分类器,处理最佳加权SAR特征数据集;根据决策树的投票结果,完成地形变化特征提取。实验证明,应用本文所提方法,大面积地形特征的尺寸与实际值最为匹配,且提取山地与平地区域特征的用时更短;整体特征提取效果好,便于快速调整测绘路径和参数,从而实现地形变化的实时监测。In order to solve the problem of poor feature extraction performance for large-scale terrain changes,a feature extraction method for large-scale terrain changes based on unmanned aerial vehicle(UAV) surveying and mapping images was proposed.Image information captured by UAVs was collected using the information reconstruction method,and the terrain edge scale was calculated.The large-area images were preprocessed using image filtering algorithms.Terrain feature parameters were measured and calculated,and terrain change features were decomposed using weighted full polarization.The feature differentiation effect was enhanced through feature selection factors and feature judgment factors,and terrain features were classified using synthetic aperture radar(SAR) feature classification methods.The optimal weighted SAR feature dataset was processed through random forests and classifiers.Terrain change feature extraction was completed based on the voting results of the decision tree.Experiments have shown that after the application of the proposed method,the size of large-scale terrain features best matches the actual values,and the time required to extract features from mountainous and flat areas is shorter.The overall feature extraction effect is good,making it easy to quickly adjust the surveying and mapping path and parameters and thus achieving real-time monitoring of terrain changes.

关 键 词：特征提取 数字高程模型 图像滤波算法 地形曲率 无人机测绘 

分 类 号：P237[天文地球—摄影测量与遥感]