Click to enlarge the photos above.
I have been learning a new workflow for the 3000 kite mapping photos I captured in Alaska last month (for context, see http://fastie.net/?p=1678). These photos cannot be aligned properly in MapKnitter because the available base maps for this remote part of Alaska are low resolution. The amount that you can zoom in at MapKnitter is a function of base map resolution, and at maximum zoom for the best base map, most of my scaled kite photos are so tiny that I can’t see enough detail to align them (e.g., see https://mapknitter.org/maps/monahan-flats). I tried Microsoft ICE and Hugin, but standard stitching algorithms don’t seem to align these photos well, and that kind of stitching does not maintain georeferencing. I think the program I need is the professional version of Agisoft Photoscan, but the educational license costs $549.
So I grabbed a few screen captures of the appropriate area from Bing Maps (their Alaska Range coverage is currently far superior to Google’s) and aligned them manually in Photoshop. Then I increased the resolution of that Photoshop project by a factor of 50 so the kite photos I aligned on top would display at that higher resolution. The Bing map showed individual trees which helped align four high altitude (~400-500 feet) kite photos of the area of one study plot. Those kite photos had enough detail to align 16 lower altitude (50-100 feet) kite photos covering just the study plot.
The resulting orthophoto mosaic allowed me to locate 51 spruce seedlings that had been decorated in the field with pink flagging. Using a field sketch map, I identified each seedling by number (except for three that I failed to add to the field map, but they have numbers now).
When all the image alignment was done and no more scaling or warping was needed, I rasterized each image so I could adjust brightness and blend the edges where needed. I think that rasterization happens at the resolution of the Photoshop project, so I made sure it was about as high as the lowest altitude photos. Photoshop was running rather slowly with all those layers of full resolution images, so the process was a little tedious.
The image gallery above includes some photos captured by the KAP rig during launch or landing showing the type of vegetation in this plot. The tundra varies from thin cover of reindeer lichen and moss to low shrubs (low blueberry and shrub birch). Below are three full resolution images embedded from gigapan.com illustrating the process of constructing the seedling map. Click the Gigapan logo in the bottom right corner to see them at gigapan.com.
Update -- the gigapans might not display anymore, so here they are at gigapan.com:
Now I have just three more plots to do. And the other study site. And there is an infrared version of each of the regular photos allowing NRG and NDVI versions of everything. Yikes.