Photogrammetry

Producing 3D models of terrain has traditionally been hard work, but the computer vision algorithms created so that robots could navigate their surroundings are making it a lot easier. Using structure-from-motion tools, where texture-like 'features' are tracked in relation to each other in order to interpolate their position in 3D space, it is now possible to take a photo collection, plug it into a piece of software, and end up with a very high-resolution 3D model of the area. Small UAVs are a natural fit for data collection missions because of their ability to fly arbitrary paths at altitudes too low for normal planes, without using expensive flight hours.

Photogrammetry traditionally referred to complex calculations involving prior knowledge of the precise position of a plane, angle of the camera, and lens characteristics, but the term seems to be the most popular one for this rapidly developing field also known as 'Automated Photogrammetry', 'Computer Vision', or 'Structure from Motion'. Unlike the previous meaning of the term, most algorithms require only the photo, sensor size, and focal length as input variables - the rest is solved automatically.

Links

• 123D by Autodesk - For free photogrammetic processing users should try Autodesk's 123D project. The rendering of complex structures can be created with ease using your own computer or by accessing Autodesk's cloud server
• Microsoft Photosynth: Possible KAP Use? - Notes on Kite Aerial Photography - A good thread to introduce people to the concept
• Photogrammetry Forum
• Photosculpt
• Agisoft Photoscan
• vSFM
• Bundler
• Microsoft Photosynth