Automatic 3D reconstruction of buildings under single satellite images
This paper aims to address the application challenges of three-dimensional modeling based on satellite imagery in industries such as construction,urban planning,cultural heritage preservation,and industrial manufacturing. It focuses on exploring efficient and rapid methods for generating architectural models from single satellite images. The paper proposes a solution based on deep learning,utilizing a deep convolutional neural network model to extract building footprints. Simultaneously,it introduces an improved Douglas-Peucker algorithm for regularizing building footprints and enhances the architectural model by estimating building heights and identifying roof styles. Ultimately,by synthesizing the aforementioned parameters,the paper successfully achieves efficient and batch reconstruction of buildings from individual satellite images.
3D reconstructionsingle imageinstance segmentationregularizationmultitasking
万方数据
维普
