Structure from Motion (SfM) Photogrammetric Processing of Historical Aerial Photographs

Submitted by atripp on

Ecologists and natural resource managers have used historical aerial photography to analyze long-term vegetation changes in a wide range of ecosystems. In national park units like Lassen Volcanic National Park, historical aerial photographs can provide baseline information for a wide range of geophysical and ecological studies, such as geomorphology changes, post-fire regeneration, and tree/shrub encroachment. Historical aerial images are challenging to work with if they are not accurately georeferenced and orthorectified, and images taken prior to 1973 lack camera calibration reports that provide interior orientation parameters (principal point, fiducial measurements, lens focal length) that are required for traditional photogrammetric processing. Structure from Motion photogrammetry (SfM), a technique that uses multiple overlapping photographs to create 3D point clouds, overcomes this limitation by automatically estimating internal camera geometry, position, and orientation based on image data alone.

USGS has made high-resolution (800 dpi) scans of ~1,250 hard copy aerial photos collected between 1941 and 2004 over Lassen Volcanic National Park, California. The photos include partial or complete coverage of the park for the years 1941, 1952–1956, 1966, 1973–1974, 1988, 1998, and 2004. SfM algorithms were used to automate the process of photogrammetric mosaicking and orthorectification. SfM image matching algorithms do not require detailed flight paths or orientation files and require only a small number of Ground Control Points (GCPs) to locate and match the raw, unstructured images. Within a 430 km2 area of Lassen National Park, 300 GCPs were identified from surface features identifiable in both historical and current images. Camera calibration reports were obtained from the USDA Forest Service RSAC Camera Report Database when available. For each photo year researchers are generating a georeferenced dense point cloud and polygonal mesh model to build orthomosaics with high spatial accuracy.

The resulting images may be used to quantify changes in forest density and fuel loads in Lassen National Park, and to estimate rates of conifer encroachment into wetlands and meadows. The geodatabase of high-resolution and high-quality historical orthomosaics produced through this project will provide baseline geospatial data that can help address many other geoscience and ecological investigations and support park decision making for years to come.

A) Example of a Structure from Motion (SfM) orthomosaic of Lassen National Park generated from 39 aerial photographs taken in 1966. Original black and white photographs were taken with a camera focal length of 208.27 mm at 1:15840 scale. The red box indicates approximate location of Dersch Meadows. B) Increasing conifer encroachment into Dersch Meadows based on imagery from 1941, 1952, 1966, 1973, 1998, and 2020.

 

Sensor
Platform
Author Name
Miguel Villarreal; Olivia Duane
Author Email
mvillarreal@usgs.gov; oduane@contractor.usgs.gov