초록 ▼

Determining the self-motion in space of an imaging device (e.g., a television camera) by analyzing image sequences obtained through the device. Three-dimensional self-motion is expressed as a combination of rotations about the horizontal and vertical camera axes and the direction of camera translati

Determining the self-motion in space of an imaging device (e.g., a television camera) by analyzing image sequences obtained through the device. Three-dimensional self-motion is expressed as a combination of rotations about the horizontal and vertical camera axes and the direction of camera translation. The invention computes the rotational and translational components of the camera self-motion exclusively from visual information. Robust performance is achieved by determining the direction of heading (i.e., the focus of expansion) as a connected region instead of a single location on the image plane. The method can be used to determine when the direction of heading is outside the current field of view and when there is zero or very small camera translation.

대표청구항 ▼

An imaging device capable of determining its own self-motion from its imaging, comprising: imaging means for detecting a sequence of images in a field of view; token tracking means, connected to said imaging means, for determining two-dimensional displacement vectors of selected tokens for each cons

An imaging device capable of determining its own self-motion from its imaging, comprising: imaging means for detecting a sequence of images in a field of view; token tracking means, connected to said imaging means, for determining two-dimensional displacement vectors of selected tokens for each consecutive pair of images, and wherein said token tracking means tracks stationary parts of a visual environment in the images, using two-dimensional, references such as corner points, contour segments, region boundaries as references, from image to image, wherein the two-dimensional displacement vectors are the result of camera motion; seeker means, connected to said token tracking means, for selecting candidate locations for focus of expansion under forward translation of said imaging means and for focus of contraction under backward translation of said imaging device, for forming a connected image region of the candidate locations and a range of corresponding rotations, and for outputting focus of expansion or contraction and image rotation estimates; and optimal rotation means, connected to said token tracking means and to said seeker means, for determining optimal three-dimensional rotation angles plus an error value for a selected candidate location for focus of expansion or contraction.