Method and apparatus for differentiating pedestrians, vehicles, and other objects

• US 7,403,659 B2
• Filed: 08/30/2004
• Issued: 07/22/2008
• Est. Priority Date: 08/28/2003
• Status: Active Grant

• Current Assignee: ZAMA Innovations LLC
• Original Assignee: Sarnoff Corporation (SRI International, Inc.)
• Inventors: Das, Aveek Kumar, Camus, Theodore Armand, Chang, Peng
• Primary Examiner(s): Bella, Matthew C.
• Assistant Examiner(s): RAHMJOO, MANUCHER
• Application Number:
  US10/930,735
• Publication Number:
  US 20050084156A1
• Time in Patent Office: 1,422 Days
• Field of Search: 382/224, 382/168, 382/158
• US Class Current: 382/224
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/46 Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features colour feature extraction G06V10/56
• 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

Abstract

A method and apparatus for classifying an object in an image is disclosed. Edge energy information and/or intensity variation information is generated for an object in a region of interest of the image. The object occupying the region of interest is classified based on the edge energy information and/or intensity variation information.

Claims

1. A method of classifying an object in an image, comprising:

building a histogram of pixel intensity values in a region of interest of the image;

locating maxima and minima in the histogram;

generating a binary image in accordance with said maxima and minima;

assigning a contour score to each pixel intensity value associated with at least one said local minimum;

segmenting the region of interest using two of the contour scores to produce a binary image; and

classifying said object using said binary image.

2. The method of claim 1, wherein the histogram is smoothed using an averaging filter.

3. The method of claim 2, wherein a mean-shift filter is applied to locate the maxima and minima.

4. The method of claim 3, wherein the contour score is obtained in accordance with:

contour_score=(y1−

y2)+(y3−

y2)+abs((x2−

x1)−

(x3−

x2))/(x3−

x−

1) where local maxima comprise (x1, y1) and (x3, y3) and a local minimum comprises (x2, y2), and where x1, x2, and x3 refer to intensity bin values and y1, y2, and y3 refer to image pixel counts in x1, x2, and x3, respectively.

5. The method of claim 4, wherein a rank ordered list of each of the minima and contour scores associated with each of the minima is generated.

6. The method of claim 5, wherein two highest contour scores are selected from the rank ordered list.

7. The method of claim 6, wherein segmenting the image further comprises thresholding the pixel intensity values of the region of interest using the two highest contour scores to create said binary image.

8. The method of claim 7, wherein row sum vectors and column sum vectors are generated from the binary image.

9. A computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to perform the steps of a method for classifying an object in an image, comprising:

building a histogram of pixel intensity values in a region of interest of the image;

locating maxima and minima in the histogram;

generating a binary image in accordance with said maxima and minima;

assigning a contour score to each pixel intensity value associated with at least one said local minimum;

segmenting the region of interest using two of the contour scores to produce a binary image; and

classifying said object using said binary image.

10. An apparatus for classifying an object in an image, comprising:

means for building a histogram of pixel intensity values in a region of interest of the image;

means for locating maxima and minima in the histogram;

means for generating a binary image in accordance with said maxima and minima;

generating a binary image in accordance with said maxima and minima;

means for assigning a contour score to each pixel intensity value associated with at least one said local minimum;

means for segmenting the region of interest using two of the contour scores to produce a binary image; and

means for classifying said object using said binary image.

11. A method of classifying an object in an image, comprising:

building a histogram of pixel intensity values in a region of interest of the image;

locating maxima and minima in the histogram;

generating a binary image in accordance with said maxima and minima;

assigning a contour score to each pixel intensity value associated with at least one said local minima;

segmenting the region of interest using two of the contour scores to produce a binary image;

determining an object score using the binary image; and

classifying the object in accordance with a threshold for the object score.

12. The method of claim 11, wherein the object score is determined by computing a mean deviation in values of row sum vectors and colum sum vectors and scaling the mean deviation with an aspect ratio of the binary image.