A proof-of-concept for detecting potential archaeological anomalies (e.g., buried structures) in raster imagery using statistical and computer vision techniques. This project demonstrates a foundational workflow for identifying rectangular patterns that differ from the surrounding terrain.
For a detailed overview of the proposed framework and its application, see A Proposed Framework for AI-Assisted Archaeological Survey.
The detection pipeline operates on single-band grayscale images and uses a multi-step process:
- Noise Reduction: A Gaussian filter smooths the image.
- Edge & Anomaly Detection: The algorithm combines a Canny edge detector with a statistical anomaly finder that flags pixels deviating significantly from the image's mean intensity.
- Component Analysis: Detected pixels are grouped into components, which are then filtered by size to remove noise and isolate objects of interest.
The following shows the pipeline applied to a synthetic image with four buried structures.
| Input Grayscale Image | Detected Edges | Identified Anomalies | Final Detected Features |
|---|---|---|---|
[IMAGE_PLACEHOLDER_1] |
[IMAGE_PLACEHOLDER_2] |
[IMAGE_PLACEHOLDER_3] |
[IMAGE_PLACEHOLDER_4] |
| Note: Replace placeholders with actual output images. |
# Install dependencies
pip install -r requirements.txt
# Run on a synthetic image and show plot
python -m geo_arch.cli --showTo analyze your own grayscale image (e.g., a GeoTIFF):
python -m geo_arch.cli path/to/your/image.tifThis repository is the first step in a larger vision. The planned development path is:
- Phase 1: Core Algorithm: Develop and validate the fundamental detection logic on synthetic data.
- Phase 2: Real-World Data Validation: Test the algorithm on open-source satellite or drone imagery of known archaeological sites.
- Phase 3: Deep Learning Model: Implement a U-Net or similar semantic segmentation model for improved accuracy and generalization.
- Phase 4: Scalable Deployment: Package the tool for large-scale analysis on cloud infrastructure.