Three-dimensional reconstruction of plants and a single building from UAV images

Abstract

Abstract: The three-dimensional (3D) models of buildings and plants from UAV images become increasingly popular for city construction. However, whether the previous 3D modeling precision of large-scale buildings can be further enhanced and whether that of plants from UAV images is acceptable still remain to be investigated. For these ends, this research studied the 3D modeling precision of a basketball hall and a row of Euonymus japonicas based on images from a DJI Inspire-1 UAV system. The data were processed with Pix4D to calculate the camera parameters, which were then processed with ContextCapture and Photoscan to generate the 3D models. The displayed height, width and crown breadth in the 3D models with the actual measured data were compared. The results showed that the errors of the 3D models in each method were within tolerance. The ContextCapture displayed a higher accuracy while the Photoscan a higher reconstruction efficiency. The r.m.s. of the respective percentage errors for the basketball hall with Photoscan and ContextCapture were4.9 cm and2.3 cm while those for the Euonymus japonicas were1.2 cm and0.7 cm. The results reveal two implications: the large-scale modeling precision in theory can be improved; the plants modeling from UAV images can be a better alternative because of its satisfying precision as well as its own much lower cost and less redundant data. Keywords: UAV photogrammetry, 3D reconstruction, precision, single building, plants DOI: 10.33440/j.ijpaa.20200304.141 Citation: Ma Y, Chen Y W, Zhang L F, Zhang Z T, Wang J R, Yu W H, Xing Z, Hu Y H, Liu Z J. Three-dimensional reconstruction of plants and a single building from UAV images. Int J Precis Agric Aviat, 2020; 3(4): 80–88.