Lunar Surface 3D Procedural Terrain Generation

Table of Contents

• Project Overview
• Key Features
• Detailed Components
• Current Progress
• Data Sources
• Technologies Used
• Contributing
• Acknowledgments
• License
• Contact

Project Overview

This project aims to generate high-resolution 3D procedural terrain for lunar surfaces using advanced machine learning techniques. By combining Deep Convolutional Generative Adversarial Networks (DCGAN), Pix2Pix models, and GAN-based Digital Elevation Model (DEM) upscaling, we create detailed and accurate representations of lunar terrain based on data from the Chandrayaan 2 Orbiter's Terrain Mapping Camera.

The project is inspired by and builds upon the work presented in the paper:

Zhang, F., Wu, B., Di, K., Liu, Z., Liu, Z., Liu, Y., & Ye, M. (2022). A Generative Adversarial Network for Pixel-Scale Lunar DEM Generation from High-Resolution Monocular Imagery and Low-Resolution DEM. Remote Sensing, 14(15), 3684. https://doi.org/10.3390/rs14153684

Key Features

• DCGAN-based heightmap generation
• Progressive growing for high-resolution (512x512) heightmaps
• Pix2Pix GAN for high-resolution terrain map generation from heightmaps
• DEM upscaling using GAN-based methods
• Integrated pipeline for end-to-end lunar terrain generation
• Evaluation metrics for generated terrain accuracy

Detailed Components

1. DCGAN for Heightmap Generation

The DCGAN (Deep Convolutional Generative Adversarial Network) is used to generate lunar terrain heightmaps.

Architecture:

• Generator: Transposed convolutional layers with batch normalization and ReLU activation
• Discriminator: Convolutional layers with batch normalization and LeakyReLU activation

Key Features:

• Progressive growing implementation for high-resolution output
• Spectral normalization for improved training stability
• Custom loss function incorporating gradient penalty

2. Pix2Pix GAN for Terrain Map Generation

The Pix2Pix GAN transforms heightmaps into detailed, photorealistic lunar terrain images.

Architecture:

• Generator: U-Net architecture with skip connections
• Discriminator: PatchGAN for local and global feature assessment

Key Features:

• Custom data augmentation pipeline for lunar terrain
• Perceptual loss using pre-trained VGG network
• Multi-scale discriminator for improved global coherence

3. DEM Upscaling

GAN-based upscaling technique to generate high-resolution DEMs from low-resolution input and high-resolution imagery.

Architecture:

• Based on the approach described in Zhang et al. (2022)
• Modified U-Net generator with residual blocks
• PatchGAN discriminator similar to Pix2Pix

Key Features:

• Integration of low-resolution DEM and high-resolution image data
• Custom loss function combining adversarial, L1, and gradient difference losses
• Evaluation metrics specific to DEM quality (RMSE, slope accuracy)

4. Integrated Pipeline

A comprehensive system that combines all components for end-to-end lunar terrain generation.

Key Features:

• Automated workflow from input data to final terrain model
• User-friendly interface for parameter adjustment
• Visualization tools for generated terrains
• Export options for various 3D modeling formats

Current Progress

• Implemented DCGAN for 64x64 heightmap generation
• Implemented DCGAN for 128x128 heightmap generation
• Progressive growing implementation (in progress)
• Pix2Pix GAN implementation
• DEM upscaling module
• Integrated pipeline development

## Data Sources

- Heightmaps extracted from DTM (Digital Terrain Model) of the Terrain Mapping Camera onboard the Chandrayaan 2 Orbiter
- High-resolution terrain maps/images from the Lunar Reconnaissance Orbiter Camera (LROC)
- 5m resolution DEMs (Digital Elevation Models) from LROC
- 0.2m resolution images of lunar terrain from LROC Narrow Angle Camera (NAC)

## Technologies Used

- Python 3.8+
- PyTorch 1.9+
- NumPy
- SciPy
- GDAL for geospatial data processing
- Matplotlib and Seaborn for visualization
- OpenCV for image processing
- MLflow for experiment tracking

## Contributing

We welcome contributions to this project. Please follow these steps to contribute:

1. Fork the repository
2. Create a new branch (`git checkout -b feature/AmazingFeature`)
3. Make your changes
4. Commit your changes (`git commit -m 'Add some AmazingFeature'`)
5. Push to the branch (`git push origin feature/AmazingFeature`)
6. Open a Pull Request

Please read [CONTRIBUTING.md](CONTRIBUTING.md) for details on our code of conduct and the process for submitting pull requests.

## Acknowledgments

- Chandrayaan 2 mission and the Indian Space Research Organisation (ISRO) for providing valuable lunar data
- NASA's Lunar Reconnaissance Orbiter mission for high-resolution lunar imagery and DEMs
- The authors of "A Generative Adversarial Network for Pixel-Scale Lunar DEM Generation from High-Resolution Monocular Imagery and Low-Resolution DEM" for their innovative approach to DEM upscaling