Estimation of structure and motion parameters for a roaming robot that scans the space

Recursive estimation of motion and a scene model with a two-camera system of divergent view

“…A multi-camera system of single-viewpoint structure was used in [20] to obtain planar motion and structure parameters by using observations in three different positions and non-linear optimization. The single-viewpoint structure means that all optical centers of the cameras coincide at one point and can be constructed by using a multi-mirror and multi-camera so that virtual optical centers coincide.…”

“…The shortcomings of this system are a cumbersome mechanism and a finite time required to capture an image with a wide angular range. Finally, we can adopt a system having multiple cameras [19][20][21][22][23][24][25]. The system can acquire a large field of view and alleviate the ambiguity of confusion between translation and rotation, keeping uniform and high image resolution.…”

Planar motion and three‐dimensional structure recovery from image sequences based on a fish‐eye projection model

“…(4) and (5); substituting these into Eqs. (8) and (9) (13) upward to zero, we can obtain the 2F-by-3 matrix [Ũ], the 3-by-3 diagonal matrix [W ], and the 3-by-P matrix [Ṽ]:…”

“…Previous work [8,9] has proposed methods for simultaneously estimating the camera location and three-dimensional structure from omnidirectional images. These methods have used geometric features of a scene such as segments.…”

Extrinsic Camera Parameter Recovery from Multiple Image Sequences Captured by an Omni-Directional Multi-camera System

“…They are called the shape-from-motion (SFM). Although these techniques are also attempted for omni-directional camera systems [15][16][17], they cannot deal with a large number of images because such methods are sensitive to feature tracking errors and these errors must be accumulated. The problem of nonunified center of projection in an OMS has not also been well discussed.…”