computer vision - Are Extrinsic Camera Parameters classified as a Rotation Matrix? -
i've been reading prince's book computer vision: models, inference, , learning, aim of understanding camera parameters , pose estimation problem , i'm having trouble extrinsic camera parameters. understand it, extrinsic camera parameters consist of rotation matrix , translation vector. rotation matrix transforms world co-ordinate system camera co-ordinate frame. question whether rotation matrix rotation matrix in strict sense; in it's orthogonal , has determinant 1.
i ask because in subsequent chapter on geometric transformations, describes case camera viewing plane (w/z co-ordinate = 0), , introduces affine , projective transformations represented extrinsic camera matrix. i'm confused because such transformations can't achieved using rotation matrix, or wrong? confused
affine , projective transformations represented projection matrix.
for typical case of pinhole camera, can think of projection matrix product p = k * [r | t] of 3x3 upper triangular matrix k representing camera's intrinsic parameters, , 3x4 roto-translation matrix [r | t], r being 3x3 orthonormal rotation matrix, , t 3x1 translation vector. matrix p transforms 4x1 homogeneous 3d point in world frame 3x1 homogeneous 2d point in image coordinates.
the columns of r ordinately components of x,y,z world frame axes in camera coordinates. vector t displacement from origin of camera frame to world frame.