I'm newbie, but learned a lot just reading this and many others foruns. But, I still have doubt about the operation of UAVs for imagery.
Well, I do not understand how the camera of UAVs (as SenseFly and Gatewing X100 and others) they do photos with the coordinates of each picture taken, for mount the orthomosaicos with them.
The GPs of these UAVs are connected to the camera?
I watched several videos on NET, but I can't see, The cameras aren't connected to GPS. Should be modified only with firmware to take the pictures in sequence, a specific time interval, I think.
The software I know to create orthomosaic, would need a file with the geographic coordinates of each image.
Thanks for any help... and sorry my poor english.
The way surveyors do this is :
They demarcate the area to be photographed on the ground station, and create a flight plan that covers the area, maybe in a 'ladder' pattern. The flight plan must ensure the correct side to side and forward overlap of images, as required by the Ortho Software, and as limited by the camera field of view, height above ground, etc.
They also place ACCURATE targets on ground, typically along the periphery of the are - these targets are Differential GPS located to within a cm or so. These targets must also be clearly visible on the photos.
The A/C then flies the ladder, taking photos - for example ( see my blog on SurVoyeur) for a 100hectare area with 2cm ground resolution you need a 20Mpixel image, using a 45mm lens and flying approx 140m above ground. This will give about 800 photos with around 30% side overlap ( a little low - 50% is muchbetter = another 300photos...) and around 65% forward overlap. Camera MUST be fixed focus lense, at infinity - auto focus slows down the camera, and a 20m/s flight speed you need rapid shots - ie, one photo every 1.4 seconds with the above lens/height above ground, etc.
Most software does NOT need the camera position or A/C attitudes, but having that does help the Ortho processing.
The Ortho software then processes the image and you give the locations of your targets and then the image is 'fitted' to those control points.
800 photos of 20Mpixel ( as above) takes approx 200hours to process on a quad core PC with 16GB ram....
Our SurVoyer system and ground control station software automates the whole process, except for the actual processing of images. It all works very well. We achieve a 2cm ground resolution with a sub 4cm accuracy in horizontal over a 2.5km x 2.5km area. Accuracy in height is sub 3cm over the same area. We have some interesting processed images and ortho data from one of the other commercial systems ( mentioned in this forum subject...) showing up to 3m errors horizontally..