Optics: measuring and testing – Shape or surface configuration – Triangulation
Reexamination Certificate
1999-10-15
2002-07-16
Pham, Hoa Q. (Department: 2877)
Optics: measuring and testing
Shape or surface configuration
Triangulation
C356S607000, C356S003090
Reexamination Certificate
active
06421132
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method and apparatus for detecting range images indicative of the shape of objects in a visual scene. More specifically, the present invention is related to a combination of a triangulation-based image formation arrangement and electronic circuit which senses and process radiation received thereon, thereby enabling a rapid, accurate and high resolution collection of range data.
2. Description of the Prior Art
High-speed gathering of high-accuracy 3D information of visual scenes is important in many applications. As employed herein “visual scenes” or“scenes” means visible surfaces of objects in the environment falling within a range sensor's field of view. In computer-aided design (CAD) and computer graphics, for example, it is useful to digitize the 3D shape of these objects as a starting description of solid shapes as to ease and enable further manipulation and analysis of these shapes by the computer. Similarly, in industrial applications, such as object inspection, it is useful to acquire shape of industrial parts and analyze them by a computer. Since robots, as humans, can greatly benefit from the knowledge of 3D information about their environment, range images are extremely useful for robotics. Robotic applications benefiting from range images include automated assembly, obstacle detection, navigation and motion planing, among others.
The importance and usefulness of obtaining range images have been known by those skilled in the art. In a paper entitled, Range Imaging Sensors, published by General Motors Research Labs., Warren, Mich., Research Publication GMR-6090, March 1988, P. J. Besi describes various range imaging sensors. The paper concludes that triangulation-based light stripe methods are the most practical and quite robust in many applications.
As well known to many skilled in the art, a conventional triangulation range imaging method projects a slit ray of light onto a scene. A sensor array, usually a CCD camera, images the scene from an oblique angle. In such arrangement, the intersection of the surface of the object and the slit ray produces a contour in the image indicative of the local object shape. In order to ease the detection of the slit ray in the image, the illumination conditions are usually adjusted so that the projected slit ray generates a prominently bright features in the scene.
Conventional triangulation methods collect range maps one slice at a time. The slit ray illuminating a scene is fixed at a particular position. The scene is projected onto an image plane through a system of lenses. Ordinarily the scene is imaged with a one- or two-dimensional array of photodetectors, such as a CCD image sensor, whose row or rows are substantially perpendicular to the axis of rotation of the slit ray. The sensed image is collected and each row examined by a computer to find the location of the light ray projection in each row. Using this location and the geometric parameters of the triangulation imaging arrangement, the range or distance to the point on the object can be computed.
By continuing this process for each row, one slice of range image is obtained. Then, the laser stripe is repositioned and another slice of the range image is collected. One problem with this known process is that it is too slow as each slice requires at least one camera frame time.
High-speed triangulation approaches have been proposed in prior art in which the slit ray continuously sweeps across the scene. This approach is sometimes called “dynamic triangulation”.
U.S. Pat. No. 4,794,262 to Y. Sato et al. discloses a triangulation setup with a continuously sweeping slit ray across the scene. The scene is viewed with the array of mutually independent photosensors. Each photosensor in the array has its own line of sight and “sees” the slit ray only once as it sweeps by, assuming there are no interreflections among surfaces in the scene. The time t when a particular detector at a particular location sees the laser is recorded. Then using a computer, the position that the slit ray assumed at the instant t is determined. Again, using the location of the particular detector together with the geometric parameters of the slit ray and the triangulation setup, the range along the line of site for the particular detector is computed. In this disclosure, the time is recorded in a memory array whose cells have one-to-one correspondence to the detector array cells.
U.S. Pat. No. 5,107,103 to Gruss et al. discloses a very-large-scale-integration (VLSI) chip method. Each cell in the sensor array has a photodiode, a comparator for thresholding the sensory signal to detect when the slit ray shines across the photosensor and an analog memory for storing the timestamp in each cell. By hard-wiring a memory cell in each cell this method also records time in a memory array whose tells have one-to-one correspondence to the detector array cells. One deficiency of this method is the fact that the thresholding is not a reliable method for detecting the passage of the slit ray. The sensory signal may be unable to reach the preset threshold due to varying reflectivity of the object and circuitry temperature drifts. Therefore, the passage of the projection of the slit ray across a cell may remain undetected.
U.S. Pat. No. 5,408,324 to K. Sato et al. shows another VLSI implementation is of the same method whereas each cell in the sensor array includes two photosensors disposed side by side in the direction of the slit ray sweep. By comparing the photocurrents, the passage of the image of the slit ray is detected when the appreciable difference between the photocurrents is observed. Yokoyama et al. U.S. Pat. No. 5,436,727 discloses that such a detection of the slit ray passage is more robust to varying object reflectance and temperature variations, and remedies one deficiency of the implementation by Gruss et al. This approach can produce a new problem. While the pair of photosensors is waiting to “see” the image of the slit ray, their sensory signals are of similar intensities, thus making it difficult for the comparator in each cell to determine which signal is greater. In fact, due to the noise and the limited resolution of the comparator, the comparator's output is very likely to transition erratically before the image of the slit ray actually passes cross the photosensors. A more recent patent by the same group of inventors, U.S. Pat. No. 5,847,833 to Yokoyama et al., introduces a hysteresis to the comparison process. The area of one of the two photosensors is sized a few percent larger than the area of the other. The smaller photosensor is the one that is to receive the image of the slit ray first, while the larger photosensor receives it second. The object is to prevent faulty and premature transitions of the comparator's output. Due to the ambient illumination and the reflectivity patterns of the scene, however, one might have such a light distribution over the two photosensors that could nullify the hysteresis produced by different area size, thus still causing unreliable performance. This disclosure also records time in a memory array whose cells have one-to-one correspondence to the detector array cells.
Several deficiencies of the above-described prior art have already been mentioned. The main deficiency of these three methods stems from the fact that they are cell-parallel. That is, the range sensor is an array of mutually independent cells that are able to detect the slit ray as it sweeps across the scene and record the time when it is detected. These approaches, therefore, require one-to-one correspondence between the memory array cells and the detector array cells. This deficiency of these methods is manifested in at least two ways. Large cell size is required if the memory is located in the cell together with the slit ray detector (see U.S. Pat. No. 5,107,103 to Gruss et al.). The large cell size limits the spatial resolution of the range sensor. If the memory cell is not located in the cl
Eckert Seamans Cherin & Mellott , LLC
Lang IV William F.
Pham Hoa Q.
Silverman Arnold B.
LandOfFree
Method and apparatus for rapid range imaging does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for rapid range imaging, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for rapid range imaging will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2884256