Panoramic tree inventory
First I would like to thank Brad for setting up my account so I can post my ideas here.
I own 80 acres of woodlands in Southern Oregon. I would love to be able to inventory every tree on it. Arial photos the county has of my property are not quite detailed enough, and they show the crown of the tree but not the size of the trunk. Seedlings are completely hidden.
Using a video camera I could do video panoramas at various spots on the property. To obtain depth would require either dual video cameras for parallax or a laser mounted a foot or so off of the side of the video camera. Dual video cameras would be out of phase with each other, and that would need to be accounted for in creating the depth information. Would a level's laser be powerful enough to see at 100' off of bark? If it were then the position of the laser spot on the video image would be an indication of the depth of the object.
Or maybe mount the camera on a sliding track. Leave the tripod in the same place, but the first pass has the camera in the center of tripod rotation, the second pass moves it a foot away from that center. Having a marker, like a stick with a reflector on it, at a fixed distance from the tripod (using a string) would help with calibrating and converging the images. Also by mounting the camera sideways I could get a little bit more vertical information since that would make the picture higher than it is wide.
I would print out a map beforehand and mark the spots roughly where I captured the panorama. I might even leave a stake in the ground for next year's inventory. My GPS doesn't work well under dense canopy, so I wouldn't rely on having it for this project. It might make it easier to process if I had a directional indicator, like always starting the panorama from magnetic north according to a compass. Leaving a colored stake in the ground and being sure to capture it on my next panorama would help align the panoramas as well.
From there a map should be able to be constructed. Each panorama would be turned into a disc on the map, with depth information showing how far the tree is from the center of the disc. Ideally the discs would overlap enough to have redundant information for some trees and the stakes, which would help align the rest of the trees on the map. The map should be able to show the trees and also areas of the property with "unknown" information, from which I could figure out what other spots would be good to take more panoramas from.
The panoramas themselves would be useful to see how the forest is changing over time. They could possibly be aligned and shown one above the other. In the ten years I've been here seedlings have grown taller than myself. I've thinned some areas, removing dense areas of small trees to allow just a few of the biggest trees to thrive, on the theory they will be getting more nutrients that the other seedlings are no longer "stealing."
This would also be a useful tool for monitoring timber sales to see the before & after of a harvest. There are many different ways to sustainably manage a forest. I would have a much better mental picture of the effects of various practices if I could really see the before and after of the work I do, in six or 12 month intervals. Forestry inventories are typically done by sampling a fixed area, counting stems and measuring their diameter at breast height. A tool like this could automate the capture of that kind of data, and help people get a good picture of what is going on in the forest.
Comments
Moz
Sun, 2005-06-26 19:41
Permalink
would love to see something like this
and not just for forest management - there's a bit of a gap in the market for 2D->3D modelling right now. I just fear the computational cost might be horrendouse. It's bad enough now stitching panoramas from styill cameras. Although video is pathetic resolution which would help (since it's going to be shown on a screen that's 2MP at most).
Add new comment