I've slogged through hundreds of hours of work creating my own software for my own application. I'm stitching together 200+ drone photos to form an image of my property.
The main difference is the drone photos are quite detailed--really to an arbitrary level of detail. The image above is the composite that's shrunk down to a tiny fraction of the 20,000x30,000 pixel size. The excerpt from the composite, below, shows the level of detail that's available in this particular set of photos. I think that image is from three or four different drone photos. These are from 200 feet above the ground.
You can see individual dandelion heads in these photos. The "image" is really a big database of photos and their relative position. You can browse it with a web application, or dump it out into a giant JPG (or TIFF) file. The photos overlap, too, so the same points on the ground are visible from several different angles.
The algorithm that forms the image composites is the current focus of the work I'm doing right now. The simple version that produced the first composite image works OK, but needs a little more sophistication to make an image that looks good and most importantly, is accurate in scale. So once that looks really good, I'll make a giant print and frame it.
What can you do with these images?
The most obvious applications are forestry and agriculture. I could catalog every tree on my property for example. Other property use and management tasks could be aided with such a tool, also. I'll probably put the tools online for people to use as a service once it's polished enough.
No comments:
Post a Comment