融合LiDAR点云的大型建筑场景多时序遥感变化分析OA

Due to the complex shape and significant occlusion phenomenon of large high-density building clusters,a multi-temporal remote sensing change analysis method for large-scale building scenes was proposed to solve the problem of reduced geometric and attribute accuracy in multi-temporal remote sensing change detection work in this scene,which integrated light detection and ranging(LiDAR)point cloud.Its characteristic lies in performing band-level fusion between the normalized digital surface model derived from LiDAR point clouds and remote sensing imagery,thereby enhancing the synergistic repre-sentation of the three-dimensional geometry and spectral texture of buildings.Based on the fused data,a linear iterative clus-tering and segmentation method extracted features including multispectral histograms,directional gradient histograms,and combined local binary pattern histograms.Cosine similarity measures were then applied to calculate change intensity informa-tion,enabling coordinated matching of building space and attribute data.Building recognition was achieved using a dual mul-tiscale manifold ordering network,which constructed a graph cut optimization model to generate candidate areas for changed buildings.Feature point matching and fully connected conditional random field optimization refined fuzzy change boundaries to yield change detection results.Tests on case studies show that the proposed method achieves an overall boundary matching score(BMS)higher than 0.85 and a spatial-attribute collaborative matching index(SACI)higher than 0.96 during the six-month monitoring period,effectively supporting high-precision change detection in complex building scenes.This method provides a reliable reference for the dynamic monitoring,planning,and supervision of urban buildings and holds practical value for enhancing the operational application of remote sensing technology in building change analysis.