Image analysis – Color image processing – Pattern recognition or classification using color
Reexamination Certificate
1998-02-02
2001-05-01
Grant, II, Jerome (Department: 2624)
Image analysis
Color image processing
Pattern recognition or classification using color
C382S153000, C382S154000, C382S163000, C358S520000
Reexamination Certificate
active
06226401
ABSTRACT:
BACKGROUND
1. Field of the Invention
This invention relates to a system and apparatus of image recognition, and particularly to that having a simplified structure using active eyes for quick, easy, and accurate image processing.
2. Background of the Art
Picture information is a function of time and space and is characterized by open-ended characteristics (i.e., changing with time), multidimensional data, large scale data, discontinuity such as edges, and noise and uncertainty created during sampling and quantifying processes. Accordingly, an image processing system needs adaptive control processing which follows time change, real time processing which processes a large quantity of data in a real time, non-linear processing which deals with discontinuity, and robust characteristics for noise and fluctuation.
A neural network comprised of arithmetic elements called “neurons” has: mathematical characteristics and information processing capacity, which realize simultaneous parallel processing by using a number of neurons; learning capacity in which weight coupling ratio between neurons change plastically, and optimization capacity for minimizing evaluation equations under complex restraint conditions. By using the above capacities, it is possible to solve problems of large-scale optimization in image recognition, at a high speed. Also, by using the learning capacity, a system, which enables recognition of various images by changing the weight coupling ratio in an adaptive manner, has been suggested.
FIG. 9
shows one example of a conventional image recognition system. In the figure, A is image data to be processed, B is an HSI transformation unit, C is a filter, D is a binarization unit, and E is an outline extracting unit.
As described above, heretofore, image recognition systems use a particular hardware which satisfies requirements for image information processing, resulting in large-scale and costly machinery. On the other hand, without such particular hardware, it takes an extremely long time to recognize images. That is, in the conventional image recognition systems, elimination of normal noises is necessary, and processing of, for example, color transformation must be performed pixel by pixel (picture element by picture element), leading to long processing time for image recognition. Thus, heretofore, in order to shorten time for image recognition, there was no way other than reduction of the number of pixels. However, when reducing the number of pixels, accuracy suffers or is sacrificed.
In addition, the conventional image recognition systems as shown in
FIG. 9
cannot provide a result until the last processing step is complete, since each processing is performed in sequence in one direction. Thus, if a processing step ceases for some reason, the interrupted step influences the whole system until the end of the process. Accurate image recognition cannot, therefore, be guarantied.
SUMMARY OF THE INVENTION
The present invention has exploited an image recognition system. An objective of the present invention is to provide an image recognition system which enables rapid, easy, and accurate image processing without using special hardware, i.e., based on a technological concept of image recognition which is distinct from the conventionally founded concept.
Namely, one important aspect of the present invention is an image recognition system for recognizing a target having a color and a pattern, comprising: a display screen for displaying a scene including a target and its background, using a CCD camera; a plurality of active eyes for recognizing the target, each active eye being an information processing unit formed in a closed frame defined on the display screen, said active eye being movable on the display screen; a memory for storing information, wherein the color and pattern of the target is stored in the memory; a controlling unit for controlling movement of the active eyes, wherein the controlling unit moves each active eye until a color within the closed frame of the active eye is the color of the target in the memory, wherein plural active eyes are located at a boundary between the target and the background on the display screen, thereby forming a colony having a shape, wherein the controlling unit further moves each active eye until the shape of the colony matches the pattern of the target in the memory; and a processing unit for receiving data from the active eyes to identify the target, wherein the processing unit calculates the location of the target based on the data from the active eyes.
In the above, preferably, the image recognition system further comprises a color processing unit, which transforms a RGB system of a color from the CCD camera into a combination of hue, saturation, and brightness, and which compares the combination of hue, saturation, and brightness with the color data of the target in the memory, thereby outputting a value indicating the likelihood of color image in each active eye matching the target.
The present invention can equally be adapted to a method.
According to the present invention, by storing information on color and pattern (shape) of an object, the location of the object can be identified regardless of the size of the object on a display screen. Thus, it is not necessary to be concerned with the location or angle of a camera, as required in a pattern matching method. The time of each processing cycle is as short as 0.03 seconds, and thus, the movement of the object can be traced without a time lag (real time tracing). Accordingly, the present invention can be adapted to a mobile robot as its eye, and further, the present invention can be applied to an intelligent mobile robot for service as a means enabling autonomous movement. It can also be adapted for detecting the location of a door knob when a robot opens or closes the door, detecting the location of a button of an elevator, detecting the location of an object such as documents when a robot hands them over, or the like.
REFERENCES:
patent: 4926251 (1990-05-01), Sekizawa et al.
patent: 5023724 (1991-06-01), Ferren
patent: 5155775 (1992-10-01), Brown
patent: 5621824 (1997-04-01), Ijir et al.
patent: 5638461 (1997-06-01), Fridge
patent: 5703961 (1997-12-01), Rogina et al.
patent: 5793501 (1998-08-01), Murakami
patent: 5917936 (1999-06-01), Katto
patent: 5920477 (1999-07-01), Hoffberg et al.
patent: 6005959 (1999-12-01), Mohan et al.
patent: 6028955 (2000-02-01), Cohen et al.
patent: 6057909 (2000-05-01), Yahav et al.
patent: 6058209 (2000-05-01), Vaidyanathan et al.
Junji Ohwi, et al., “Development of Intelligent Mobil Robot for Service Use”, The 14th Annual Conference of RSJ, Nov. 1-2-3, 1996, pp. 295-296, Nov. 1, 1996.
Ooi Junji
Yamafuji Kazuo
Grant II Jerome
Kianni Kevin Cyrus
Knobbe Marten Olson & Bear LLP
Yamaha Hatsudoki Kabushiki Kaisha
LandOfFree
Image recognition system and apparatus using active eyes does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image recognition system and apparatus using active eyes, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image recognition system and apparatus using active eyes will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2499969