In this paper,an efficient approach for urban buildings′ extraction regarding adjacency relationship between buildings and their shadows is presented.First,based on the results of multi-scale segmentation with a smaller scale,the initial classification of the images is preformed,and then buildings′ shadows are extracted,employing their geometric and spectral properties.Second,some primitive objects of buildings are extracted from objects of bigger scales by strict thresholds,and adjacency relationship between shadows and other objects is computed.Finally,from undefined objects,the missing buildings′ objects are recovered with the help of the adjacency,generating complete region of buildings.The experimental results demonstrate that extraction accuracy of the proposed approach is higher,and that spatial adjacency is a better selection of features for the information extraction from high resolution images.

The proposed approach improves urban buildings extraction accuracy by using spatial adjacency between buildings and their shadows in high-resolution images.

Shadow-Assisted Object-Oriented Extraction of Urban Buildings

Key Takeaway: The proposed approach improves urban buildings extraction accuracy by using spatial adjacency between buildings and their shadows in high-resolution images.