Backround: Aerial photography and remote sensing can be used to measure current land cover and recent land use change. However, these technologies are fairly recent and therefore time series are limited when using them. In contrast, historical landscape photos exist for many locations from much further back in time. Repeat-photography studies return to the locations of historical photographs a take identical photographs to examine changes. However, in general, these studies are limited to qualitative measures of change inferred manually because these obliquely taken photographs can't easily be georeferenced.

Question: I'm helping with an repeat-photography study in Alaska, we have historical (up to 100 years old) photos of the landscape (with lat/long) and we'll be retaking photos this year at the same locations. We also have access to a high resolution DEM and aerial imagery. The goal is to quantify forest cover change between the photos. I'm looking for a method to georeference these obliquely taken photographs, drape them over the landscape, and create a map to be compared with aerial photography. Essentially I want to transform an oblique photo to an orthophoto.

  • @PolyGeo I've made an effort to refine the question, thought it's still a bit vague because I really have almost no idea where to start – Matt SM Mar 1 '16 at 20:02
  • 1
    I do not understand why landscape features must be persistent. If you start with an accurately georeferenced photo and end with another accurately georeferenced photo, you will have no problem in comparing them. Perhaps the most salient issue concerns constructing a sufficiently accurate map of the century-old data. You say these old photos have "lat-long," but how accurately have those coordinates been established and what amount of distortion is present within the photos? – whuber Mar 1 '16 at 20:24
  • @whuber Perhaps I haven't done the best job of explaining the problem. These are just normal photos of the landscape, i.e. not from an aerial view. I know the location of the photographer and the direction they were facing. What we're trying to do is essentially infer the aerial view of the landscape (to measure forest cover) from these normal horizontally taken photos. I'm increasingly thinking that this tasks is probably not possible though... – Matt SM Mar 1 '16 at 23:51
  • I was gonna say your question reminded me of this: nature.com/nature/journal/v411/n6837/abs/411546a0.html, then i realized you are the author of that paper :) – TonyC May 30 '16 at 15:02
up vote 4 down vote accepted

I've finally found a tool that does this. This task is referred to as mono-photogrammetry or monoplotting and involves referencing a single oblique and unrectified photo to a DEM to produce georeferenced data for use in a GIS. This is similar to photogrammetry, except you only have a single image.

The WSL Monoplotting-Tool is specifically designed for this task. My understanding is that you georeference the image by connecting control points (i.e. topographical features) in the photo and the DEM. The tool then orthorectifies the image to create polygons, which can be compared to traditional orthophotos or other spatial data.

This recent paper (pdf here) outlines the use of the tool for repeat-photography studies. They also perform ground truthing to test accuracy and found points were within 15m of their actual 3D location.

Will update this answer further once I've used the software.

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.