Our research presents an image-based rooftop measurement system used in an unmanned aerial vehicle controlled by Python code. The background of our measurements was based on spatial data reconstruction, using multiple images taken by an unmanned aerial vehicle, while taking into consideration the known positions of the aircraft. Through the Scale-Invariant Feature Transform method, matching parts in two different images were found, taking into consideration the surrounding points, while disturbing points were removed. The rotation and translation matrices were calculated and triangulation to reconstruct the position of the observed points in a 3D coordinate system was used. Finally, the points located at the very edge of the roof were selected and by that the distance between them was determined. The final task resulted in roof measurements that are automatically determined by the software using Python code, with deviations from ground-truth averaging no more than half a metre.
|
