Television – Camera – system and detail – Swing driven
Reexamination Certificate
1997-12-11
2004-02-17
Garber, Wendy R. (Department: 2612)
Television
Camera, system and detail
Swing driven
C348S208140
Reexamination Certificate
active
06693666
ABSTRACT:
TECHNICAL FIELD
This invention relates generally to computer input devices, and more particularly to digital image capture devices used to provide ranging and tracking information for a computer system.
BACKGROUND ART
The range of an object, i.e. the distance to the object from an observation site, can be determined by the analysis of two or more spatially separated images (often referred to as “binocular images” when there are two images) that are taken from the observation site. In range computation from simultaneously acquired binocular digital images, the area of processing is limited to the visible region of overlap between the two images. To maintain a reasonable region of overlap usually necessitates redirecting the optical axes of the cameras (i.e. changing their vergence) which introduces other problems including a usual necessity to resample the imagery. As is well known to those skilled in the art, “vergence” means the angle between the optical axes of the lenses.
The processing of binocular or multi-view imagery for range computation is easiest when the optical axes are parallel and the imaging surfaces are coplanar—in what is termed parallel epipolar geometry. Because verging the optical axes to optimize the region of image overlap eliminates image-surface co-planarity, the complexities of calculating range increases significantly with non-parallel viewing. This is further compounded when viewing objects at a variety of azimuths and distances where adjustments in the view direction as well as verging would be necessary to retain sufficient image overlap.
For computational purposes, the frame of reference for scene description is usually tied to image location, so changing the image location through vergence adjustments necessitates reconfiguring the frame of reference. Again, adjusting a system's frame of reference increases the computational and conceptual complexity of its analysis.
A similar situation arises for typical monocular (i.e. single image) computer analysis of tracking and scanning in some space before the camera. With subjects able to operate over a broad region before the camera, continued observation generally involves use of either wide-angle optics or a panning/tilting mechanism to properly direct the camera's view direction. These control mechanisms are relatively complex, must move fairly large pieces of equipment (cameras and their lenses), and alter the underlying geometric frame of reference for scene analysis by rotating the frame of reference with the cameras. In addition, the use of wide angle optics works against high resolution analysis, as only larger scene detail is visible.
One approach to solve these acquisition problems in image-based range and tracking computation would be to employ greatly oversized imagers (e.g. imagers having about 3K by 3K or 9×10
6
elements), and select corresponding standard-sized windows within these for processing. However, such an approach would be prohibitively expensive. For example, a 1K by 1K imager sells for well over a thousand dollars. Higher resolution imagers are available at considerably greater price.
A prior art solution to the apparent dichotomy between simple processing (with parallel epipolar geometry) and broad depth and tracking coverage exists in adaptation of perspective-correcting lens systems as used in “view-camera” or “technical-camera” designs. In such designs, an oversized lens is used to image the scene, and lateral repositioning of the lens or imaging platform can be used to redirect the camera without rotating the imaging surface. For single camera use this enables maintaining lines parallel in the world parallel on the image plane; in ranging camera use it enables parallel epipolar geometry.
For example, in U.S. Pat. Nos. 5,063,441 and 5,142,357 of Lipton et al., devices for use in 3D videography are disclosed. More particularly, Lipton et al. teach devices for capturing binocular images for use in stereo videography (stereo movies), with reduced viewing eyestrain, by using dual and triple camera systems. Briefly stated, Lipton et al. teaches an imager controller for epipolar stereo capture in videography, including stereo lenses mounted fixedly together in a single housing. Stereographics Inc., of San Raphael, Calif., produces a product embodying elements of the Lipton et al. patents.
In the matter of two-dimensional imager control, U.S. Pat. No. 5,049,988, of Sefton et al. teaches a system that provides the display of a video capture window for surveillance applications. Phillips, in U.S. Pat. No. 4,740,839, teaches a TV surveillance system operated by sub-sampling a conventional camera, with a result that resembles the Lipton et al. approach of image capture.
As will be appreciated, image capturing of the prior art uses planar sensors due to the high cost, lack of availability, and complexities involved with the use and manufacture of curved or “spherical” sensors. However, spherical sensors have a number of advantages with respect to field of view, view direction, and use in stereo image capture that designers of prior art digital imaging cameras have apparently failed to consider.
DISCLOSURE OF THE INVENTION
The present invention includes a multi-image camera system for automated stereo ranging applications that provides high-resolution broad field coverage over a wide range of subject distances while enabling retention of the epipolar constraints necessary for efficient stereo matching. At the same time, the multi-image camera of the present invention supports monocular image applications such as object tracking.
In one embodiment of the present invention, an imaging device is preferably placed on a three-degree-of-freedom linear motion positioning mechanism whose position can be controlled by a computer. Broad field coverage is preferably attained through use of a wide-angle large-coverage lens (e.g. a lens designed for a 35 mm camera). High resolution is attained through placing the imaging device under the large-coverage lens, so the imaging device's immediate field of view is considerably narrower than that provided by the lens.
In contrast with traditional view-camera usage, the present invention teaches moving the imaging surface instead of the lens. This allows the apparatus to retain co-planarity of the images and, so long as the displacements can be determined to sub-pixel accuracy, maintains a stable frame of reference for the analysis, all while providing the required view re-directions. Moving the lens alters the projective relationships among observations, whereas moving the imager does not.
Image focus may be attained through the traditional rotating-travel focus adjustment although, for quantitative computational tasks, this is often unsatisfactory as the center of projection may vary with lens rotation. For this situation, our preferred embodiment of the moving imager provides a back-focus capability, as will be described below.
Computer control of the platform enables positioning accuracy in the micron range. Stepper motors, for example, operating with full step positioning, can position a platform along an axis to within a few microns of the desired position. Differential stepper control enables increasing the accuracy of this placement (the number of locations where this precision is attainable) by one to two orders of magnitude. Fine specification of absolute imager location can also be attained through use of interferometers or sonar-based ranging systems, as will be described below in more detail.
The planar imaging embodiment of the present invention enjoys a number of advantages over related devices of the prior art. In the present invention, an oversized lens is used to provide a wide field of view of the scene. Movement of the imaging surface under the lens provides high-resolution view selection, which is the equivalent of view redirection of pan and tilt motions without the need for a complex mechanism for accomplishing such motions. With accurate positioning knowledge, this form of imaging provides a stable frame of ref
Baker Henry H.
Kalayjian Nicholas R.
St. Hilaire Pierre
Woodfill John I.
Garber Wendy R.
Interval Research Corporation
Nguyen Luong
Van Pelt & Yi LLP
LandOfFree
Moving imager camera for track and range capture does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Moving imager camera for track and range capture, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Moving imager camera for track and range capture will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3304964