Reconstruction of 3D Building Edges to Improve DSM Generated from UAV Multi-Images

Articles in Press

Document Type : Original Article

Authors

Department of Geodesy and Surveying Engineering, Tafresh University, 79611-39518, Tafresh, Iran

10.22059/eoge.2025.390296.1168

Abstract

Digital surface model (DSM) generation of urban scenes encounters some challenges especially for man-made features boundaries with high altitude. It leads to the saw-tooth effects at the features edges in the orthophoto images. To mitigate these artifacts, this study proposes a novel methodology employing three-dimensional (3D) line segment extraction, enabling precise reconstruction of building edges in three-dimensional space. Firstly, the relevant 2D line from multi-images are extracted then, these segments reconstruct the 3D line segments by intersecting multi-planes, estimating the optimum intersection parameters of line based on least square methodology, and applying restrictions at the end point of the line. Finally, the 3D line segments are divided into discrete 3D points to be included in the 3D point clouds. After the classification of points on both sides of the line and masking points around the line between the inner and outer edges as the parapet wall of the roof, the height of the nearest line point is assigned to all points within the mask. Regarding the numerical results, 99% improvement in height point cloud consistency attained through advanced masking techniques. As a result, a more complete and accurate TIN can be developed to provide important essential data for products such as DSM and orthophoto. Experimental results of this approach show that saw-tooth effects at the edges of the building can be greatly improved; so, the quality of DSM and orthophoto increases significantly.

Keywords

Main Subjects

Earth Observation and Geomatics Engineering

Articles in Press, Accepted Manuscript
Available Online from 06 September 2025

Files

Share

How to cite

Statistics