Volume XLII-4/W18
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-4/W18, 271–277, 2019
https://doi.org/10.5194/isprs-archives-XLII-4-W18-271-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-4/W18, 271–277, 2019
https://doi.org/10.5194/isprs-archives-XLII-4-W18-271-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

  18 Oct 2019

18 Oct 2019

3D THERMAL MAPPING OF BUILDING ROOFS BASED ON FUSION OF THERMAL AND VISIBLE POINT CLOUDS IN UAV IMAGERY

M. Dahaghin, F. Samadzadegan, and F. Dadras Javan M. Dahaghin et al.
  • School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran

Keywords: Thermal Infrared Imaging, Unmanned Aerial Vehicles, Building Roof, Point Cloud Generation, Fusion

Abstract. Thermography is a robust method for detecting thermal irregularities on the roof of the buildings as one of the main energy dissipation parts. Recently, UAVs are presented to be useful in gathering 3D thermal data of the building roofs. In this topic, the low spatial resolution of thermal imagery is a challenge which leads to a sparse resolution in point clouds. This paper suggests the fusion of visible and thermal point clouds to generate a high-resolution thermal point cloud of the building roofs. For the purpose, camera calibration is performed to obtain internal orientation parameters, and then thermal point clouds and visible point clouds are generated. In the next step, both two point clouds are geo-referenced by control points. To extract building roofs from the visible point cloud, CSF ground filtering is applied, and the vegetation layer is removed by RGBVI index. Afterward, a predefined threshold is applied to the normal vectors in the z-direction in order to separate facets of roofs from the walls. Finally, the visible point cloud of the building roofs and registered thermal point cloud are combined and generate a fused dense point cloud. Results show mean re-projection error of 0.31 pixels for thermal camera calibration and mean absolute distance of 0.2 m for point clouds registration. The final product is a fused point cloud, which its density improves up to twice of the initial thermal point cloud density and it has the spatial accuracy of visible point cloud along with thermal information of the building roofs.