I got 3 aerial photographs from a mountainous area (~3500 - 4000 m.a.s.l), from 1956, 1965, 1977 (no metadata, no camera parameters, scanned white and black). My study catchment is relatively small (1 sq. km) and I would like to use those photos in order to assess potential vegetation changes. I know both the processes of orthorectification and georeferencing, however, I am not sure about which process should be used in this case. For orthorectifying I will require additional information, particularly the detailed DEM, which is very difficult to get for this area, while for georeferencing I will need some control points, a-priori the georeferencing process is 'easier' however I am not sure how comparable the results can be, with respect to orthorectification, and also in terms of overall accuracy assessment. Since the area of interest, is relatively small (each photography, covers approximately 10x10 km), I would not expect severe deformations, yet the area surface is very irregular. Any reference or previous experience in this regard is welcome.
It's hard to answer this question without seeing the images, but I'll try. For orthorectification, you need to know internal orientation, deduced from the camera, and external orientation parameters. Forget about this variant in your case.
If your mountainous area contains remarkable reference points (such as road crossings, free-standing trees, big detached stones, some human-made objects), you can georeference all three images to these points. Use base maps, such as Google or Bing, as a background. Open-source QGIS, with its default Georeferencer plugin, is suitable for this job. You may need to experiment with various transformation types, before getting a decent result.
What I would do is to:
- Add some web base map to QGIS. You can do it with the help of QuickMapServices plugin, which allows you to select from a big list of data providers. Or, perhaps, you have your topographic maps for it.
- Georeference the first aerial image to the base map. At this stage, you may spend much time selecting an appropriate transformation type. E.g., I would try Thin Plate Spline transformation and a lot of georeferencing points along the borders of the aerial image and inside it.
- Georeference the second aerial image to the first one. As the transformation type is already defined, it should go much faster now.
- Georeference the third image to the first one.
Doing 3 and 4 allows you to check RMSE of your georeferencing points against the first aerophotograph. But be careful: usually, spline transformations give an RMS error of around zero; however, this does not mean that the image is precisely aligned in space.
The planar accuracy of your final georeferenced maps can be estimated only when you have ground control points visible on all aerophotographs, and you know their accurate coordinates.
P.S. Please, check out this QGIS Georeferencer tutorial.
P.P.S. I used QGIS Georeferencer to align scanned black-and-white images from American reconnaissance satellites CORONA with modern satellite images. It worked well, but it required a big amount of reference points, which was hard to achieve, considering the territory change over 50 years.