Image processing apparatus

• US 5,835,614 A
• Filed: 11/29/1994
• Issued: 11/10/1998
• Est. Priority Date: 06/26/1992
• Status: Expired due to Term

• Current Assignee: Honda Giken Kogyo Kabushiki Kaisha (Honda Motor Company)
• Original Assignee: Honda Giken Kogyo Kabushiki Kaisha (Honda Motor Company)
• Inventors: Sekine, Takaharu, Aoyama, Chiaki
• Primary Examiner(s): Mancuso, Joseph
• Application Number:
  US08/346,354
• Time in Patent Office: 1,442 Days
• Field of Search: 382/173, 382/199, 382/313, 382/274, 382/271, 382/272, 382/270, 382/104, 358/457, 358/566, 358/518, 358/528, 364/478
• US Class Current: 382/104
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06V 20/588 Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road

Abstract

The image processing apparatus comprises: a pixel train setting unit for setting a pixel train; a moving average calculating unit for estimating the intensity of light illuminating the object as a moving average of luminance of the pixel train; a reflectance calculating unit for converting the luminance of the pixel into reflectance, with the luminance of the pixel divided by the estimated intensity of light; a digitizing unit for digitizing the reflectance of the pixel; an edge detecting means for detecting an edge at which the luminance of the image changes sharply; and a comparing means for comparing the digitized reflectance signal with the edge signal to detect a boundary of regions having different luminance from each other, to eliminate it from the digitized reflectance signal for being a false edge.

Claims

1. An image processing apparatus for processing an image of a roadway, said image including a plurality of pixels, each pixel having a luminance representative of an intensity of reflected light from said roadway, wherein said roadway is exposed to and illuminated by a source of non-stroboscopic ambient light having a diurnal variation, said apparatus comprising:

estimating means for estimating an intensity of said ambient light illuminating the roadway at a location of a pixel of attention;

means for outputting an image signal representative of a luminance of said pixel of attention;

reflectance calculating means for converting the luminance of said pixel of attention into a reflectance, wherein said reflectance is determined by dividing the luminance of said pixel of attention by said estimated intensity of said ambient light illuminating the roadway at the location of said pixel of attention;

digitizing means for binary-digitizing the reflectance of said pixel of attention by comparing the reflectance of said pixel of attention with a threshold value, and for outputting a digitized reflectance signal; and

boundary recognizing means for recognizing a boundary of said roadway, said boundary of said roadway being defined by regions having digitized reflectance signal values different from each other in the image.

2. An apparatus according to claim 1, wherein said estimating means comprises:

pixel train setting means for setting a pixel train having therein the pixel of attention and a specified number of pixels adjacent to the pixel; and

moving average calculating means for estimating the intensity of light illuminating the roadway at the location of the pixel as a moving average of luminance calculated for said pixel train.

3. An apparatus according to claim 2, wherein said boundary recognizing means comprises:

edge detecting means for detecting, at the location of the pixel of attention, an edge as a succession of pixels in front and in rear of which the luminance of the image changes sharply, and for outputting an edge signal; and

comparing means for comparing said digitized reflectance signal with said edge signal to detect a boundary of regions in the image having different luminance from each other, thereby to eliminate a part of the digitized reflectance signal corresponding to the detected boundary due to the difference of luminance.

4. An apparatus according to claim 3, wherein said edge detecting means is sensitive to vertical edges.

5. An apparatus according to claim 3, wherein said boundary recognizing means comprises normalizing means for normalizing said edge signal at the location of the pixel of attention, with the edge signal divided by the estimated intensity of light illuminating the roadway at the location of the pixel, and for outputting a normalized edge signal.

6. An apparatus according to claim 5, wherein said boundary recognizing means comprises digitizing means for binary-digitizing said normalized edge signal by comparing the normalizing edge signal with a threshold value, and for outputting a normalized and digitized edge signal.

7. An apparatus according to claim 3, wherein said edge detecting means determines if the moving average of luminance calculated with respect to the pixel train at the location of the pixel of attention is smaller than a judging value, and then replaces the edge signal at the location of the pixel of attention with an average of the edge signal at the location of the pixel of attention and edge signals respectively at the location of a few pixels adjacent to the pixel of attention when the determination is affirmative.

8. An apparatus according to claim 2, wherein said pixel train setting means sets the pixel train having the length which varies depending on the location of the pixel of attention in an image plane.

9. An apparatus according to claim 8, wherein said pixel train setting means sets the pixel train having the length which varies in relation to the widths of regions whose boundary said recognizing means is recognizing.

10. An apparatus according to claim 9, wherein said pixel train setting means divides the image plane into upper and lower portions at the image of a hypothetical horizon, and sets the pixel train having the length which varies depending on in which portion of the upper and lower portions of the image plane the pixel of attention falls.

11. An apparatus according to claim 8, wherein said pixel train setting means sets the pixel train having a constant minimum length on the image of the hypothetical horizon and in the upper portion of the image plane, and the pixel train having an increasingly large length as the pixel of attention is at an increasingly low location in the lower portion of the image plane.

12. An apparatus according to claim 2, wherein said reflectance calculating means determines if the moving average of luminance calculated for the pixel train at the location of the pixel of attention is smaller than a judging value, and then replaces the luminance of the pixel of attention with an average of luminance with respect to the pixel of attention and a few pixels adjacent thereto when the determination is affirmative.

13. An apparatus according to claim 2 wherein said reflectance calculating means determines if the moving average of luminance calculated for the pixel train at the location of the pixel of attention is smaller than a standard value, and then replaces the moving average of luminance with the standard value when the determination is affirmative.

14. An apparatus according to claim 2, wherein said edge detecting means determines if the moving average of luminance calculated with respect to the pixel train at the location of the pixel of attention is smaller than a standard value, and then replaces the moving average of luminance with the standard value when the determination is affirmative.

15. An image processing apparatus for processing a signal representing a video image of a roadway, said video image including a plurality of pixels, each pixel having a luminance representative of an intensity of light reflected from said roadway, wherein a plurality of said pixels are exposed to and illuminated by a source of non-stroboscopic ambient light having a diurnal variation in intensity, said apparatus comprising:

a video camera for generating an image signal representing the luminance of each of a plurality of pixels located at a plurality of pixel locations within said video image;

a reflectance determination circuit for receiving said image signal and converting the luminance of each of said plurality of pixels to a reflectance value, wherein said reflectance value of a selected pixel is determined by dividing the luminance of the selected pixel by a value representing the intensity of said ambient light at the location of the selected pixel; and

a boundary detection circuit for generating a signal representative of a boundary between pixel locations having reflectance signal values different from each other, said boundary defining a boundary of said roadway.

16. An image processing apparatus for processing a signal representing a video image of a roadway, said apparatus comprising:

a video camera for generating an image signal representing a luminance of each of a plurality of pixels located respectively at a plurality of pixel locations within an image of said roadway;

a processing unit for receiving said image signal and converting the luminance of each of said plurality of pixels to a reflectance value, wherein the reflectance value of a selected pixel is determined by dividing the luminance of the selected pixel by a value representing an intensity of non-stroboscopic ambient light illuminating said roadway at the location of the selected pixel, said non-stroboscopic ambient light having a diurnal variation; and

a boundary detection circuit for generating a signal representative of a boundary between pixel locations having reflectance signal values different from each other, said boundary defining a boundary of said roadway.

17. The image processing apparatus of claim 16 wherein said processing unit generates said value representing said intensity of ambient light at a selected pixel location by calculating a moving average of the luminance of each selected pixel and a predetermined number of pixels adjacent to each selected pixel.

18. A machine implemented process for identifying a boundary of a roadway, wherein said roadway includes regions illuminated by varying amounts of non-stroboscopic ambient light, said process comprising the steps of:

generating an image signal representing an image of said roadway, said image of said roadway comprising a plurality of pixels located respectively at a plurality of pixel locations;

generating a luminance value for each of said pixels comprising said image of said roadway;

converting each said luminance value to a reflectance value by dividing the luminance value of a selected pixel by an average luminance of said selected pixel and a predetermined number of adjacent pixels; and

generating a signal representing a boundary of said roadway by identifying adjacent pixel locations having different reflectance values.