IMAGE PROCESSING EQUIPMENT

• US 20070047767A1
• Filed: 08/24/2006
• Published: 03/01/2007
• Est. Priority Date: 08/30/2005
• Status: Active Grant

• Current Assignee: Subaru Corp.
• Original Assignee: Fuji Heavy Industries Limited (Subaru Corp.)
• Inventors: WATABE, Naoki
• Granted Patent: US 7,689,039 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06V 20/58 Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
• CPC Class Codes: G06V 40/103 Static body considered as a whole, e.g. static pedestrian or occupant recognition

Abstract

A design for the high speed identification of a pedestrian at an image is provided, having an input feature vector generation unit 23b for generating, based on an image of an object, an input feature vector u that includes as elements the object'"'"'s size, the object'"'"'s upper portion shape, and the object'"'"'s side portion shape; a kernel discriminant analysis operations unit 23c for generating a vector y mapped to the discriminant space, through performing operations for kernel discriminant analysis based on the input feature vector, and an object determination unit 23d for determining whether the object is a pedestrian or not, depending on whether or not this vector y is within a fixed area at the discriminant space.

Claims

1. An image processing equipment for performing kernel discriminant analysis based on a photographed image of an object comprising:

an input feature vector generation unit for generating, based on data related to said image, an input feature vector including as elements a size of said object, a side portion shape of said object, and an upper portion shape of said object;

a discriminant analysis operations unit for generating a vector mapped to a discriminant space, through performing the operations for kernel discriminant analysis based on said input feature vector; and

an object determination unit for determining whether said object is a pedestrian or not, according to whether or not the vector mapped to said discriminant space exists within a fixed area at the discriminant space.

2. The image processing equipment of claim 1, wherein the size of said object includes an absolute size width and an absolute size length, calculated based on an area in which the concerned object was photographed and on a distance to the concerned object.

3. The image processing equipment of claim 1, wherein at respective sections from dividing at a fixed width in a horizontal direction, an area in which the concerned object was photographed, the upper portion shape of said object is characterized through representative values of the object'"'"'s lengths in a vertical direction from a bottom edge of said area to a top edge of said object.

4. The image processing equipment of claim 3, wherein said input feature vector generation unit normalizes said representative values for height in the vertical direction, and normalizes a number of the sections to a fixed number, as elements of said input feature vector.

5. The image processing equipment of claim 1, wherein at respective sections from dividing at a fixed width in a vertical direction, an area in which the concerned object was photographed, the side portion shape of said object is characterized through representative values of the object'"'"'s lengths in a horizontal direction, from a right edge of said area to a left edge of said object, and through representative values of the object'"'"'s lengths from a left edge of said area to a right edge of said object.

6. The image processing equipment of claim 5, wherein said input feature vector generation unit normalizes said representative values of the object'"'"'s length from the right edge of said area to the left edge of said object, for width in the horizontal direction,

normalizes said representative values of the object'"'"'s length from the left edge of said area to the right edge of said object, for width in the horizontal direction, and

normalizes a number of the sections to a fixed number.

7. The image processing equipment of claim 1, wherein said discriminant space is configured for said input feature vector using, as a base, an eigenvector of a fixed matrix related to a first training data of said objects being a pedestrian, and to a second training data of said objects being other than a pedestrian, and

wherein said discriminant analysis operations unit generates a fixed kernel feature vector, based on said input feature vector related to said object which is generated through said input feature vector generation unit, and based on said first and said second training data stored in advance, and

generates a vector mapped to said discriminant space, based on the concerned generated kernel feature vector and said matrix.

8. The image processing equipment of claim 2, further comprising a stereo camera for photographing a basis image and a comparison image corresponding to said basis image, and a stereo image processing unit for calculating parallax error for each fixed small area configuring said basis image, based on the photographed said basis image and said comparison image, and

further comprising an object recognition unit for recognizing the existence of said object based on said calculated parallax error for the small areas, where said input feature vector generation unit generates said input feature vector based on said calculated parallax error regarding said recognized object.

9. The image processing equipment of claim 4, further comprising a stereo camera for photographing a basis image and a comparison image corresponding to said basis image, and a stereo image processing unit for calculating parallax error for each fixed small area configuring said basis image, based on the photographed said basis image and said comparison image, and

further comprising an object recognition unit for recognizing the existence of said object based on said calculated parallax error for the small areas, where said input feature vector generation unit generates said input feature vector based on said calculated parallax error regarding said recognized object.

10. The image processing equipment of claim 6, further comprising a stereo camera for photographing a basis image and a comparison image corresponding to said basis image, and a stereo image processing unit for calculating parallax error for each fixed small area configuring said basis image, based on the photographed said basis image and said comparison image, and

further comprising an object recognition unit for recognizing the existence of said object based on said calculated parallax error for the small areas, where said input feature vector generation unit generates said input feature vector based on said calculated parallax error regarding said recognized object.

11. The image processing equipment of claim 2, further comprising a score accumulation unit for accumulating fixed scores for the same object appearing at a series of plural frames, wherein a score expresses a likeness to a pedestrian of said object at an image of 1 frame and depends on a determination result related to said object through said object determination unit, and

further comprising a score judgment unit for judging whether that same object at said plural frames is a pedestrian or not based on said accumulated score.

12. The image processing equipment of claim 4, further comprising a score accumulation unit for accumulating fixed scores for the same object appearing at a series of plural frames, wherein a score expresses a likeness to a pedestrian of said object at an image of 1 frame and depends on a determination result related to said object through said object determination unit, and

further comprising a score judgment unit for judging whether that same object at said plural frames is a pedestrian or not, based on said accumulated score.

13. The image processing equipment of claim 6, further comprising a score accumulation unit for accumulating fixed scores for the same object appearing at a series of plural frames, wherein a score expresses a likeness to a pedestrian of said object at an image of 1 frame and depends on a determination result related to said object through said object determination unit, and

further comprising a score judgment unit for judging whether that same object at said plural frames is a pedestrian or not, based on said accumulated score.

14. The image processing equipment of claim 9, wherein said score accumulation unit adds or subtracts a fixed value depending on the determination result of said object determination unit, and accumulates said scores, and

if said accumulated score is larger than a fixed first threshold value, said score judgment unit judges that same object at said plural frames to be a pedestrian, and

if said accumulated score is smaller than a fixed second threshold value, which is smaller than said first threshold value, said score judgment unit judges that same object to not be a pedestrian.

15. The image processing equipment of claim 9, further comprising a precondition judgment unit, for judging preconditions of whether or not to carry out pedestrian identification through kernel discriminant analysis regarding said object, and

if a judgment result of said precondition judgment unit is that said object satisfies the preconditions and also a judgment result of said object determination unit is that said object is a pedestrian, then said score accumulation unit adds a first fixed value regarding said object;

and

if a judgment result of said precondition judgment unit is that said object satisfies the preconditions and also a judgment result of said object determination unit is that said object is other than a pedestrian, then said score accumulation unit subtracts a second fixed value, which is smaller than said first fixed value, regarding said object; and

if a judgment result of said precondition judgment unit is that said object does not satisfy the preconditions, then said score accumulation unit subtracts a third fixed value which is larger than said second fixed value and smaller than said first fixed value, regarding said object.

16. The image processing equipment of claim 9, wherein the score accumulation regarding said object through said score accumulation unit and the judgment based on said accumulated scores through said score judgment unit, are carried out for each successive frame, and

if the accumulated score for said same object up to a previous frame is larger than a fixed third threshold value, said accumulation of scores through said score accumulation unit is not performed, and said score judgment unit judges that the same object at said plural frames is a pedestrian.

17. The image processing equipment of claim 9, wherein the score accumulation regarding said object through said score accumulation unit and the judgment based on said accumulated scores through said score judgment unit, are carried out for each successive frame, and

said score judgment unit prepares separately the respective threshold value for when the accumulated score is increasing and the threshold value for when the accumulated score is decreasing, and establishes the threshold value for when the accumulated score is decreasing to be smaller than the threshold value for when the accumulated score is increasing.

18. An image processing equipment for distinguishing pedestrians based on a photographed image of an object comprising:

an input feature vector generation unit for generating, based on data related to said image, an input feature vector including as elements said objects size, said object'"'"'s upper portion shape, and said object'"'"'s side portion shape;

an operations unit for generating a vector mapped to a specified space, based on said input feature vector; and

an object determination unit for determining whether said object is a pedestrian or not, depending on whether or not the vector mapped to said specified space is within a specified area at the space, which has established an identification boundary.