# A Method for 3D Building Individualization Integrating SAMPolyBuild and Multiple Spatial-Geometric Features

## TL;DR

This paper introduces a new method for accurately creating 3D building models using zero-shot learning and spatial-geometric features, reducing the need for manual work and training data.

## Contribution

The novel framework combines zero-shot learning with spatial-geometric refinement for efficient 3D building individualization.

## Key findings

- The method achieves an F1-score of approximately 0.83 for buildings with typical spatial structures.
- The joint JS-EMD metric effectively identifies building-ground interfaces by analyzing normal vector angles.
- The proposed approach reduces reliance on manual effort and extensive training data for urban modeling.

## Abstract

What are the main findings?
A novel framework integrating SAMPolyBuild’s zero-shot capability with spatial-geometric feature refinement achieves high-precision 3D building individualization without pre-training.The joint JS-EMD metric effectively identifies building-ground interfaces by quantifying spatial distribution shifts in normal vector angles.

A novel framework integrating SAMPolyBuild’s zero-shot capability with spatial-geometric feature refinement achieves high-precision 3D building individualization without pre-training.

The joint JS-EMD metric effectively identifies building-ground interfaces by quantifying spatial distribution shifts in normal vector angles.

What are the implications of the main findings?
The method provides a lightweight, efficient solution for large-scale urban modeling, reducing reliance on manual effort and extensive training data.It establishes a zero-shot learning paradigm that effectively transfers 2D foundation model segmentation capabilities to 3D spatial analysis.

The method provides a lightweight, efficient solution for large-scale urban modeling, reducing reliance on manual effort and extensive training data.

It establishes a zero-shot learning paradigm that effectively transfers 2D foundation model segmentation capabilities to 3D spatial analysis.

Individualization of buildings is one of the key issues in the establishment of three-dimensional (3D) building models. Most existing individualization methods rely on inefficient manual separation, while deep learning approaches require extensive pre-training and are highly influenced by the spatial structure of the models. To address these issues, this paper proposes a novel method for 3D building individualization that integrates SAMPolyBuild with multiple spatial-geometric features. Leveraging the zero-shot learning capability of SAMPolyBuild, the method first performs coarse extraction of individual buildings, then refines the extraction accuracy using multiple spatial-geometric features. Innovatively, two statistical parameters—Jensen-Shannon Divergence and Earth Mover’s Distance—are introduced into the building identification process. To validate the feasibility and effectiveness of the proposed method, experiments were conducted on the Semantic Urban Meshes (SUM) dataset. The results demonstrate that the method can effectively extract individual building models from urban oblique photogrammetric 3D models, achieving an F1-score of approximately 0.83 for buildings with typical spatial structures.