Library to pre-process scanned historical images for Structure from Motion (SfM) surface reconstruction and photogrammetric analysis.
- Download imagery from historical image archives
- Supported archives:
- North American Glacier Photography (NAGAP)
- USGS Earth Explorer
- Built-in application to create fiducial marker templates
- OpenCV template matching to detect fiducial marker coordinates
- Fiducial markers are detected at sub-pixel precision
- Can detect 4 midside and/or 4 corner fiducials
- Replaces poor matches with np.nan based on threshold
- Computes estimated principal point
- Quality Control
- Outputs window image around detected fiducial marker for visual verification
- Creates qc plots for fiducial coordinates and intersection angles before and after affine transformation
- Routines to detect proxy for midside fiducial markers, when actual fiducial markers are cropped out of image frame
- Computes affine transform between calibrated (true) fiducial marker coordinates and detected coordinates
- Affine transforms images
- Requires minimum of 3 successfully detected fiducial markers to perform restitution
- Crops images about principal point to standard size
- Contrast Limited Adaptive Historgram Equalization (CLAHE) to improve match point detection during SfM processing
See notebooks for processing examples.
$ git clone https://github.com/friedrichknuth/hipp.git
$ cd ./hipp
$ conda env create -f environment.yml
$ conda activate hipp
$ pip install -e .
Historical Structure from Motion (HSfM)
Knuth, F., Shean, D., Bhushan, S., Schwat, E., Alexandrov, O., McNeil, C., Dehecq, A., Florentine, C. and O’Neel, S., 2023. Historical Structure from Motion (HSfM): Automated processing of historical aerial photographs for long-term topographic change analysis. Remote Sensing of Environment, 285, p.113379. https://doi.org/10.1016/j.rse.2022.113379
Bradski, G. (2000). "The OpenCV Library". Dr. Dobb's Journal of Software Tools.