Workflow 3D-KAP taking images
Photogrammetry is based on the difference in images taken from a different position of the same surface. Therefore if a 3d-model is to be made of the ground, it is most usefull to aim the camera straight down. Personally i have a camera rig which automatically takes images every 5 seconds, by walking in rays along a field i capture overlapping images. The amount of overlap between the images should be around 60%. But smaller and larger overlap do work fine as long as the distance of the camera position in the images is sufficient.
These three images worked fine for me. As you can see there is about 30% of overlap between the left pair and about 70% between the right pair. Not so obvious the difference in elevation of the camera when the images were shot, but it was quite significant (about 25%).
It is very important that there is sufficient contrast in the images. The best results are obtained on sunny days in the morning and afternoon when the sun is creating small shadows. Creating a digital elevation model of a evenly coloured surface (snow-covered field) is close to impossible because the software can't recognize common points on the images.
Ground control points
In general you want to be able to geo-reference your images (to know the real coordinates of the images, and sizes of objects on them). This can be done by measuring the location of recognisable objects by using (D-) GPS or a Total Station. But it might be hard to find these locations you measured on your images. Another way is to put marks on the ground before you take the pictures and measure the location of these marks. In each images you should try to have at least have 3 measured points (Ground Control Points(GPC's)).
In the middle-left of the image down here you can see a white piece of paper (letter/A4 sized), with a number written next to it. The position of the upper left corner is measured using DGPS, which has a position accuracy of about 15 cm horizontally and 5 cm vertically.
