What question did this study set out to answer?

The aim is to improve the accuracy of 3D building reconstruction and contour measurement using UAV-acquired point clouds affected by non-uniform density.

March 17, 2026Open Access

Density-Constrained Contour Analysis for Unmanned Aerial Vehicle-Based 3D Building Reconstruction and Measurement

Puntos clave

The aim is to improve the accuracy of 3D building reconstruction and contour measurement using UAV-acquired point clouds affected by non-uniform density.
Segmented UAV building point clouds to isolate different roofs.
Constructed a point density function model for contour lines using a multi-factor function.
Employed a vector clustering algorithm to determine main vectors.
Established principal points and contours based on calculated density and cut-off distance.
The method showed good reconstruction effects despite varying Gaussian noise.
It outperformed existing methods like Polyfit and Chen's in precision.
Demonstrated competitive results compared to deep learning approaches while needing fewer faces and without training data.

Resumen

3D reconstruction and contour measurement of building point cloud is the key to realize 3D digital city modeling. However, the sparse and non-uniform distribution of UAV-acquired building point clouds often leads to significant challenges for previous methods in achieving accurate 3D reconstruction and reliable contour measurement. In view of this, this paper proposes a 3D reconstruction and contour measurement method of building point cloud based on peak point density of main contour. Firstly, the UAV building point cloud is segmented and dfferent roofs are obtained. Secondly, the point density function model of contour line based on multi-factor function is constructed, and the main vectors are determined by vector clustering algorithm. Thirdly, the cut-off distance is determined based on density, and constraint parameters for determining principal points, and the principal vectors are established according to the cut-off distance and density. Finally, the principal contour fitting lines of the building were determined based on principal points and vectors. The building corner points were derived through an adjacent contour line intersection-merging strategy, thereby enabling 3D reconstruction and contour measurement of the building. Experimental results show that the proposed method has a good reconstruction effect on UAV building point cloud under different Gaussian noise. Moreover, the proposed method achieves a higher level of refinement compared to Polyfit and Chen’s methods, while also demonstrating competitive performance against the recent deep learning-based PolyGNN approach on orthogonal building structures, yet with significantly fewer faces and without requiring training data. Additionally,the proposed method can obtain high precision reconstruction and contour measurement results with fewer Faces. The source code will be available at: https://github.com/xijiangyue04/Reconstruction-of-UAV-PCD-based-DGM .

Leer artículo completoexternamente

Me gusta

Guardar

Ver artículo completo