Image analysis – Pattern recognition – Feature extraction
Reexamination Certificate
2000-10-18
2004-09-07
Mariam, Daniel (Department: 2621)
Image analysis
Pattern recognition
Feature extraction
C382S190000, C382S291000
Reexamination Certificate
active
06788817
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to an object recognition system which is mounted on a vehicle and recognizes an object ahead of the vehicle, and more particularly to an object recognition system which is mounted on a vehicle and recognizes the outline of an object ahead of the vehicle.
BACKGROUND OF THE INVENTION
In recent years, devices which determine the distance and size of objects in front of a vehicle, and which appropriately control the vehicle in accordance with this determination, have been proposed for improving the safety of vehicle operation. A typical method for obtaining information about an object ahead includes the steps of: extracting horizontal edges and vertical edges from a captured image; determining whether each of the extracted edges belongs to the object such as a vehicle ahead; and recognizing the object based on the edges determined to belong to the object.
Japanese Patent Application Kokai No. Hei 10-97699 describes a device for determining whether or not an object exists. The device recognizes left and right guidelines such as white lines in a captured image, and extracts horizontal edges from an area between the left and right guidelines. A probable area in which an obstacle may exist is estimated based on the extracted horizontal edges. Vertical edges are then extracted in the probable area. The device judges that no object exists if the number of vertical edges whose lengths exceed a predetermined value is less than two.
Japanese Patent Application Kokai No. Hei 9-16782 describes a device for determining whether or not an object exists. According to the device, a captured image is divided into a plurality of small areas. An area containing detected vertical edge points is defined as a vertical edge area, and an area containing detected horizontal edge points is defined as a horizontal edge area. Intersections of a group of continuous vertical edge areas and a group of continuous horizontal edge areas are determined. If four or more intersections are detected and they form a rectangle, then the rectangle is recognized as the object.
However, according to Japanese Patent Application Kokai No. Hei 10-97699, if the color of a part of an object is similar to, and therefore has not a sufficient contrast to the color of the background, and two or more vertical edges cannot be obtained in the probable area, then the recognition error may take place that ‘no object’ exists even if there is actually an object. According to Japanese Patent Application Kokai No. Hei 9-16782, if four or more intersections of vertical edges and horizontal edges cannot be obtained because of an insufficient contract, then the recognition error may take place that ‘no object’ exists even if there is actually an object.
Accordingly, one object of the invention is to provide a system capable of recognizing the outline of an object from horizontal edges and/or vertical edges even if desired horizontal edges or vertical edges cannot be obtained.
SUMMARY OF THE INVENTION
According to one aspect of the invention, an object recognition system including a position sensor, an image sensor, and a controller is provided. The position sensor determines the position of an object, and the image sensor captures an image of the object. The position sensor can be implemented with radar or a pair of image sensors. The controller sets a processing area within the image captured by the image sensor based on the position of the object determined by the position sensor and a predetermined size for the object to be recognized. It is preferable that the predetermined size is set to surround the object to be recognized.
The controller extracts horizontal edges from the processing area. Extraction of edges can be carried out by detecting a variation of intensity values of pixels contained in the processing area. The controller then identifies horizontal edges indicating or belonging to the outline of the object from the extracted horizontal edges. Thus, the object can be recognized based on the horizontal edges only. Similarly, the controller can extract vertical edges from the processing area, and identify vertical edges belonging to the outline of the object.
According to one aspect of the invention, the outline of the object is recognized by determining upper, lower, left, and right ends of the object. Thus, the vehicle ahead traveling in front of the vehicle mounting the system of the invention is recognized by the positions of the upper, lower, left, and right ends of the outline of the vehicle ahead.
According to one aspect of the invention, the controller selects upper, lower, left, and right candidate ends from the identified horizontal edges, and selects upper, lower, left, and right candidate ends from the identified vertical edges. The controller then determines the upper, lower, left, and right ends of the object based on the upper, lower, left, and right candidate ends selected from the horizontal edges and the upper, lower, left, and right candidate ends selected from the vertical edges.
Preferably, the left and right candidate ends selected from the vertical edges are chosen as the left and right ends of the object respectively. For the upper end of the object, the higher one of the upper candidate end selected from the horizontal edges and the upper candidate end selected from the vertical edges is chosen as the upper end of the object. For the lower end of the object, the lower one of the lower candidate end selected from the horizontal edges and the lower candidate end selected from the vertical edges is chosen as the lower end of the object.
According to one aspect of the invention, if either one or both of the left and right candidate ends cannot be selected from the vertical edges, then the left or right end of the object is determined by the left or right candidate end selected from the horizontal edges in lieu of the left or right candidate end selected from the vertical edges.
According to another aspect of the invention, if either one or both of the left and right candidate ends cannot be selected from the horizontal edges, then the left and right ends of the object are determined by estimated left and right candidate ends in lieu of the left and right candidate ends selected from the horizontal edges. The estimated left and right candidate ends can be determined based on the position of the object recognized in a previous recognition cycle.
The controller can comprise a micro-controller which typically includes a central unit (CPU), or a micro-processor, a read-only memory (ROM) containing control programs that when executed by the processor performs respective functions which are to be described hereafter. The controller also includes a random-access memory (RAM) that provides a working area for the CPU and temporary storage for various data and programs.
REFERENCES:
patent: 5487116 (1996-01-01), Nakano et al.
patent: 5617085 (1997-04-01), Tsutsumi et al.
patent: 5892855 (1999-04-01), Kakinami et al.
patent: 5987174 (1999-11-01), Nakamura et al.
patent: 6005492 (1999-12-01), Tamura et al.
patent: 6191704 (2001-02-01), Takenaga et al.
patent: 9-16782 (1997-01-01), None
patent: 10-97699 (1998-04-01), None
Bennamoun, et al. “A structural-description-based vision system for automatic object recognition”, IEEE, pp. 893-906, 1997.
Aoki Tomoyoshi
Nishigaki Morimichi
Saka Masakazu
Yuhara Hiromitsu
Honda Giken Kogyo Kabushikikaisha
Lahive & Cockfield LLP
Laurentano, Esq. Anthony A.
Mariam Daniel
LandOfFree
Object recognition system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Object recognition system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Object recognition system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3220291