Skip to content

Latest commit

 

History

History
193 lines (146 loc) · 6.58 KB

Plan.md

File metadata and controls

193 lines (146 loc) · 6.58 KB
Raw

Plan for Lunar PSR Image Enhancement and Interactive Deployment

Project Overview

Objective

  • Develop a robust software system to enhance low-light images from Permanently Shadowed Regions (PSRs) of lunar craters using Chandrayaan-2 datasets.
  • Improve Signal-to-Noise Ratio (SNR) and resolution to create high-quality, annotated maps for:
    • Lunar landing site selection.
    • Geomorphological studies.
  • Deploy an interactive platform that integrates a scrollable, spherical Moon visualization (like Google Earth) with layered overlays for enhanced usability.

Phase 1: Data Acquisition and Preprocessing

1.1 Datasets

This project leverages the following Chandrayaan-2 datasets:

  1. OHRC (Orbiter High-Resolution Camera):

    • Purpose: Primary dataset for lunar surface imagery.
    • Data: Calibrated and derived images in GeoTIFF format (L1 and L2).
    • Key Use: Source for low-light enhancement and resolution improvement.

  2. TMC2 (Terrain Mapping Camera-2):

    • Purpose: Provides Digital Elevation Models (DEM) and ortho-images.
    • Data: Calibrated (L1) and derived (L2) DEMs in GeoTIFF.
    • Key Use: Geometric corrections, terrain context, and validation.

  3. DFSAR (Dual-Frequency Synthetic Aperture Radar):

    • Purpose: Offers subsurface insights and water ice detection.
    • Data: Calibrated and derived radar data in GeoTIFF format.
    • Key Use: Validate surface features and subsurface anomalies.

  4. IIRS (Imaging Infrared Spectrometer):

    • Purpose: Provides hyperspectral data for surface composition analysis.
    • Data: Spectral cubes in binary format with XML metadata.
    • Key Use: Validate material composition and refine surface detection.

1.2 Workflow for Data Acquisition

1.2.1 Data Access

  1. Download Data:
  2. Data Format:
    • OHRC, TMC2, DFSAR: GeoTIFF (calibrated and derived levels).
    • IIRS: Binary spectral cubes with accompanying metadata.

1.2.2 Organization

Standardize the directory structure:

/LunarEnhancement
├── raw_data
│   ├── ohrc
│   ├── tmc2
│   ├── dfsar
│   ├── iirs
├── preprocessed_data
├── models
├── outputs
├── overlays
└── documentation

1.3 Preprocessing Steps

1.3.1 OHRC (High-Resolution Images)

  • Normalization:
    • Scale pixel values between 0–1 for uniformity in model training.
  • Geometric Alignment:
    • Use gdalwarp to align OHRC images with TMC2 DEM.
  • Noise Reduction:
    • Apply median filtering or wavelet-based denoising to remove speckle noise.

1.3.2 TMC2 (DEM and Ortho-images)

  • Extract and preprocess DEM using GDAL.
  • Align ortho-images to OHRC data using mutual metadata.

1.3.3 DFSAR (Radar Data)

  • Denoise subsurface radar data.
  • Isolate radar signatures relevant to PSR regions.

1.3.4 IIRS (Spectral Data)

  • Apply spectral filtering to remove low-confidence bands.
  • Normalize spectral bands for material classification.

1.4 Integration of Preprocessed Data

  1. Align all datasets to a consistent CRS (e.g., Moon 2000/IAU CRS).
  2. Resample to a uniform spatial resolution (e.g., 10m/pixel).
  3. Merge datasets for analysis and visualization.

Phase 2: Model Development

2.1 Goals

  • Enhance OHRC imagery with deep learning.
  • Generate high-resolution outputs validated by auxiliary datasets.

2.2 Deep Learning Framework

2.2.1 DCGAN (Deep Convolutional GAN)

  • Purpose: Enhance low-light images by improving SNR.
  • Implementation:
    • Generator: Uses convolutional layers to brighten low-light regions.
    • Discriminator: Validates the naturalness of generated enhancements.
    • Training Data: Illuminated OHRC images as ground truth.

2.2.2 ESRGAN (Enhanced Super-Resolution GAN)

  • Purpose: Upscale enhanced images while retaining texture and edge details.
  • Enhancements:
    • Incorporate TMC2 DEM to constrain elevation-induced distortions.

2.3 Training Details

  1. Use NVIDIA GPUs for training.
  2. Augment training data with rotation, cropping, and simulated noise.
  3. Implement perceptual loss for visual quality and pixel-based loss for fidelity.

Phase 3: Integration, Deployment, and Visualization

3.1 Data Integration

  • Overlay DFSAR data on enhanced OHRC images to highlight subsurface structures.
  • Use IIRS data for material validation by mapping spectral composition over enhanced regions.

3.2 Google Earth-Type Deployment

3.2.1 Features

  1. Spherical Moon Interface:
    • Scrollable, zoomable 3D model of the Moon.
    • Overlays stitched OHRC images for a complete lunar surface map.
  2. Layer Options:
    • Base Layer: OHRC images (raw and enhanced).
    • Altitude: TMC2 DEM for elevation data.
    • PSR Filter: Highlight PSRs for easy identification.
    • Enhanced Images: Overlay enhanced OHRC images for detailed inspection.

3.2.2 Implementation

  1. Frameworks:
    • CesiumJS for 3D visualization.
    • Three.js for custom enhancements.
  2. Backend:
    • Flask/FastAPI to serve datasets dynamically.
  3. Data Storage:
    • Use AWS S3 for hosting large GeoTIFFs.
  4. Frontend Deployment:
    • Integrate CesiumJS with a React or plain HTML/CSS interface.

3.3 Workflow Automation

  • Use Airflow or Prefect to automate preprocessing, enhancement, and visualization pipelines.
  • Enable real-time updates as new datasets become available.

3.4 Validation

  1. Quantitative:
    • Evaluate SNR improvement (>30%) and resolution enhancement (>25%).
  2. Qualitative:
    • Compare results with known lunar features for accuracy.

Phase 4: Documentation and User Support

4.1 Documentation

  1. Technical Guide:
    • Details preprocessing, model architectures, and deployment pipelines.
  2. User Guide:
    • Instructions for navigating the Moon visualization interface.

4.2 Training and Outreach

  • Provide video tutorials for researchers.
  • Conduct workshops for lunar mission planners.

Conclusion

This refined plan integrates advanced image processing with a state-of-the-art visualization platform. The final deliverable not only enhances scientific understanding of lunar PSRs but also provides a practical tool for exploration planning. This scalable, interactive system lays the foundation for future lunar data analysis and dissemination.