I'm using the Metashape software to generate orthoimages from historic imagery (the 1980's). These images to not have EXIF information, however they do have a center lat/lon specified - but this is the lat/lon from the flight plan, not how it was actually flown, so there is some error there. Each project was flown in 1 flight using 1 camera and recorded on film. This film was then digitized, however the digitizing process ended up creating images whose sizes differ by 1 to 5 pixels. There is a black border that we mask out. Because the images are different sizes they end up showing as different camera types (sensors?) in the Camera Calibration screen.
The flight plan was 1 north-south flight, with images taken such that each image (frame) overlaps the adjacent frame by 2/3, so there's good coverage. I locate ground control points every 7th image by locating corresponding features from modern aerial images and obtaining the lat/long from them, and then getting the elevation for that lat/lon.
My question: when I perform camera optimization using these control points, how effective is it when the images are different sizes (in terms of number of pixels)? Since each of those would be considered separate cameras will the optimization work across all those types. I know the manual says not to modify the image size, but I'm wondering if in this case I trim the images down to be the same size. As I said each image has a black border that gets masked out, so I'd just be trimming back the masked area. Since I know these are all the same camera type it seems to me that the optimization will work better if there is only a camera/sensor in the chunk.
Right now the resulting orthos overlay the modern imagery about 40% of the time without doing this trimming. Any insight into whether this would help generate better output would be greatly appreciated.
Thanks,
Chris Musial