1. Aerial imagery was downloaded from archived historical aerial photo single frames on USGS Earth Explorer (
https://earthexplorer.usgs.gov/). A total of 31 images were downloaded. The project code in the Earth Explorer database is 68008, but the spatial query tools were used to identify and download the relevant imagery. Online linkage for that data set, per metadata, is
https://lta.cr.usgs.gov/Single_Frame_Records
2.Images from the survey area were first aligned using Hugin by rotating and shifting scanned imagery (affine transformation) so that fiducial marks were at the same pixel locations, and all images were cropped to the same center pixel. Images were then imported into Agisoft Photoscan Professional v. 1.4.2 software using the 'Add photos' tool.
3. The photos were processed through an initial alignment and optimization procedure using the following settings:
Alignment - Accuracy: 'High'; Pair selection: 'Reference, generic'; Key point limit: 0 (unlimited); Tie point limit; 0 (unlimited).
Optimization - Lens-calibration parameters f, cx, cy, k1, k2, k3, p1, and p2 were included; b1, b2, and higher-order parameters k4, p3, and p4 were not.
4. The sparse point cloud (tie points; created as a result of photo alignment and optimization) was edited using an iterative error-reduction procedure to filter the data. This was done in several iterations of a process called "Gradual Selection" to first reduce reconstruction uncertainty (to a unitless value of 10) and then projection accuracy (to a weighted value of 3).
5. Ground control points (GCPs) were identified in the lidar data and the 1967 aerial images, primarily along the highway. Latitude, longitude, and elevation were extracted from lidar and added to the PhotoScan project.
6. Another round of "Gradual Selection" was done to reduce the reprojection error (to a value of 0.4 pixels)
7. A dense point cloud was then created with the parameters set to "High" quality and "moderate" depth filtering.
8. The dense cloud was exported as a .las file and imported to CloudCompare, where it was aligned to the 2010 lidar point cloud using a rigid transformation (rotate, translate) derived from a least-squares adjustment using Horn's least-squares method for rigid-body transformation between two coordinate systems (Horn, B.K., 1987, Closed-form solution of absolute orientation using unit quaternions: Journal of the Optical Society of America A, vol. 4, p.629-642)
The transformation matrix using the 2010 lidar as the source data is:
0.999999523163 -0.000057821395 -0.001067430479 4.551651000977
0.000056641620 0.999999344349 -0.001105255098 -2.024248600006
0.001067493809 0.001105194795 0.999998807907 -122.032318115234
0.000000000000 0.000000000000 0.000000000000 1.000000000000
9. The dense point cloud was exported from cloud compare in LAZ format and clipped to a shapefile of the survey area using lastools with the following command:
lasclip -i [input_file.laz] -poly [clip_shape.shp] -odir clipped -olaz
