1- Shahid Beheshti University
2- Shahid Beheshti University ,a_milan@sbu.ac.ir
3- K. N. Toosi University of Technology
2- Shahid Beheshti University ,
3- K. N. Toosi University of Technology
Abstract: (54 Views)
3D Reconstruction is one of the key topics in the field of surveying, particularly photogrammetry, used to extract accurate geometric information from objects and environments. Conventional 3D reconstruction methods typically rely on multi-view images and positional information of imaging stations, which can be challenging in certain operational scenarios. In this study, a novel method for generating 3D models using the MiDaS deep learning model is presented. This method uses only a single 2D image to produce a depth map and then generates a 3D model using the Poisson surface reconstruction technique. In this approach, there is no need for positional information or imaging station angles. To evaluate the accuracy of the proposed model, it was compared with a baseline 3D model created using photogrammetric methods. The results showed that the proposed model performs well with a Root Mean Square Error (RMSE) of 0.775 cm, indicating a small relative discrepancy with the baseline model. This study demonstrates that deep learning models like MiDaS can be effectively used for 3D reconstruction in scenarios where multi-view imaging is not feasible. Additionally, using more advanced versions of this model could improve reconstruction accuracy and expand its applications in surveying, photogrammetry, and 3D modeling.
Keywords: Depth Estimation, 3D Reconstruction, Single Image, Deep Learning, Photogrammetry, Machine Vision
Type of Study: Original Research |
Subject:
Remote sensing and GIS
Received: 2025/03/25 | Accepted: 2025/08/16
Received: 2025/03/25 | Accepted: 2025/08/16
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |