Page 507 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

Evaluating the effect of key point filtering on the accuracy of the 3D model generated by UAV
images

Seyed Arya Fakhri * 1 , Seyed Arya Fakhri 2 , Masood Varshosaz 3 , Vahid Mousavi 4

1 Department of Surveying and Geomatics Engineering, University College of Engineering, University of Tehran
aryafakhri@ut.ac.ir
2 Faculty of Geodesy and Geomatics, K. N. Toosi University of Technology
arvin.fakhri@email.kntu.ac.ir
3 Faculty of Geodesy and Geomatics, K. N. Toosi University of Technology
varshosazm@kntu.ac.ir
4 Faculty of Geodesy and Geomatics, K. N. Toosi University of Technology
vmoosavy@mail.kntu.ac.ir

Abstract: In recent years, the use of UAV-based photogrammetry has been very popular due to its many benefits
and low cost and it has been used in various applications. Low-strength lenses are commonly used in the construction
of cameras used in drones, which causes various errors in the imaging process. Errors related to the camera geometry
affect the extracted points and reduce the accuracy of the bundle adjustment and 3D model. The points extracted
from the edges of images are mostly affected by these errors more than other. Therefore, in this study, the effect of
filtering these areas in the generation of 3D models has been investigated. For this purpose, different areas of the
images taken by the drone at two heights of 60 and 120 meters have been masked. Then the key points are extracted
from the masked images and 3D models are produced. Before the flight operation, the control points in the study area
were read using GPS and entered into modeling as check and control points so that the different models could be
compared quantitatively. The results of different models showed that masking the image in the best case increases the
surface accuracy of the 3D model by 40%.

Keywords: UAV photogrammetry, Key points filtering, Image matching, Image inhancement.

16