The scene geometry file (or .P file) contains the locations of the camera for each image in addition to it's calibration properties. If you are a researcher in the Computer Vision field you may find this information useful for your own reconstruction algorithms. Or if you are a graphic artist then it may help you in creating augmented overlays or re-texturing the 3D model.
The format of the file is as follows, the first line corresponds to the total number of cameras described in the file. After this first line each of the individual camera entries has a total of 5 lines. The first of these lines (i.e. 0:7) describes the 'camera number':'image number', note that the camera number will always be 0. The next line describes the approximate calibration for the image. 3dsee uses the plumb-bob model for the lense of the camera (see
here) .
3dsee represents the projection of a homogenous 3D point X to a homogenous 2D image point x according to :
x = K[R | t]X.
Where K is as follows :
| f x sx f x skew cx|
| 0 f cy|
| 0 0 1|.
- f - focal length (in pixels)
- sx - aspect ratio
- skew - CCD array skew
- cx - x-coordinate principal point (in pixels)
- cy - y-coordinate principal point (in pixels)
The radial distortion coefficients (k1, k2, k3) and tangential distortion coefficients (p1, p2) may or may not be present but operate on the undistorted camera coordinates according to the plumb-bob model (see
here)
In summary the information given on the second line is as follows :
f sx skew cx cy (| k1 k2 k3 p1 p2).The next 3 lines of the camera entry are the rotation (R) and translation (t) of the camera in the scene written as 3x4 matrix :
R t.The coordinates of the vertices in the .ply (polygon) files are also in this coordinate system.