Authors:
Maximilian Jarofka
;
Stephan Schweig
;
Niko Maas
and
Dieter Schramm
Affiliation:
Chair of Mechatronics, University of Duisburg-Essen, Lotharstraße 1, 47057 Duisburg, Germany
Keyword(s):
Artificial Neural Network, COLMAP, Clustering, Driving Simulator, Metashape, Meshroom, Object Classification, Object Detection, Photogrammetry, Process Chain, Unity, VisualSFM.
Abstract:
This paper presents an automated process chain for the reconstruction of characteristic 3D objects, which can be used in a simulation environment. The process chain can distinguish between recurring objects such as trees and cars and specific objects like buildings. To acquire this, it detects and classifies objects in images from a previously recorded video. In contrast to the specific objects, which are reconstructed during the workflow of the process chain, the recurrent objects are loaded from already existing models and are placed multiple times into the simulation environment. In terms of quality a visual comparison between the two integrated programs for the reconstruction (Metashape and Meshroom) is carried out. Furthermore the accuracy of the positioning of standard objects in the Unity game engine is examined.