3D Mesh Generation from DSM Images

This project showcases the process of converting Digital Surface Model (DSM) data into high-quality 3D mesh and point cloud outputs. The DSM image used for this process, along with the generated outputs, are displayed below.

Updates

• 1.0 VersionInitial Release '24.08.09
• 1.1 Version Zero Division Error Fixed, Empty Information Area is now filled with lowest Z level. '24.08.10
• 1.2 Version Now You can convert greyscale image, or RGB image to DSM '24.08.12

Target DSM Image

This is the target DSM image that was used to create the 3D data:

Outputs

Below are the results of the 3D data creation process:

Point Cloud Created

Mesh Created

Object Created (OBJ)

Features

• DSM Image Processing: Converts DSM images into point clouds.
• 3D Mesh Generation: Creates 3D meshes using Poisson surface reconstruction.
• Multi-format Export: Outputs results in both .ply and .obj formats.
• Height Adjustment: Adjusts building heights to better match reality.

You can adjust Created 3D Object's height for purpose(Verical images etc..)

Usage

Prerequisites

Ensure you have Python 3.x installed and the following libraries:

pip install numpy pillow open3d tqdm

Running the Script

1. Set Input and Output Paths: Edit the script to set the correct file paths for your DSM image input and desired output locations.

2. Run the Script:

(if you need)

   pip install requirements.txt

python main.py

3. View the Results: The script will generate point cloud and 3D mesh files in both .ply and .obj formats, which you can open with 3D visualization software like MeshLab, Blender, or Open3D.

Example Command

python main.py --input /path/to/your/DSM_image.tif --output_ply /path/to/save/output.ply --output_obj /path/to/save/output.obj

Adjustable Parameters

The script includes several adjustable parameters that can be tuned according to your specific needs:

1. Height Scaling Factor

• Variable: height_scaling_factor

• Description: Adjusts the height (z-values) of the DSM data to better reflect the real-world heights of buildings.

• Default: 1.5

• Example: height_scaling_factor = 2.0 will double the height of buildings.

2. Poisson Reconstruction Depth

• Variable: depth
• Description: Controls the depth of the Poisson surface reconstruction. Higher values increase the level of detail.
• Default: 12
• Example: depth=10 for less detail, depth=15 for more detail.

3. Normal Estimation Parameters

• Variables: radius, max_nn
• Description: Defines the search radius and the maximum number of nearest neighbors for normal estimation.
• Defaults: radius=1.0, max_nn=30
• Example: radius=2.0, max_nn=50 for wider and more comprehensive normal estimation.

4. Vertex Density Threshold

• Variable: density_threshold
• Description: Controls the threshold for removing low-density vertices from the mesh.
• Default: 0.01 (1% quantile)
• Example: density_threshold = 0.05 will remove more vertices, creating a cleaner but potentially less detailed mesh.

Contact

If you have any questions, suggestions, or issues, please feel free to reach out.

• Email: [email protected]

This project was created and maintained by Jaden Jang, a 4th-year undergraduate student in Civil & Environmental Engineering at Yonsei University.

License

This project is licensed under the MIT License - see the LICENSE file for details.