Edge detection based on background change

• US 20040022437A1
• Filed: 07/30/2002
• Published: 02/05/2004
• Est. Priority Date: 07/30/2002
• Status: Active Grant

• Current Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Inventors: Beardsley, Paul
• Granted Patent: US 7,031,525 B2
• Time in Patent Office: Days
• US Class Current: 382/199
• CPC Class Codes: G06T 2207/10021 Stereoscopic video; Stereoscopic image sequence
• CPC Class Codes: G06T 2207/20224 Image subtraction
• CPC Class Codes: G06T 2207/30236 Traffic on road, railway or crossing
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/174 involving the use of two or more images
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments
• CPC Class Codes: G06T 7/593 from stereo images
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Abstract

A method detects edges of an object in a scene by first acquiring a static image of the scene when the scene is static, and a live image of the scene including the object. A reference image is constructed from the static image. Then, image gradients of the reference image are compared with image gradients of the live image to identify edges of the object in the live image.

Claims

1. A method for detecting edges for an object in a scene, comprising:

acquiring a static image of the scene when the scene is static;

acquiring a live image of the scene including the object;

constructing a reference image from the static image; and

comparing image gradients of the reference image with image gradients of the live image to identify edges of the object in the live image.

2. The method of claim 1 further comprising:

linking the edges of the object to form edge chains;

3. The method of claim 1 further comprising:

measuring a distance between the image gradient of each pixel in the reference image and the image gradient of the corresponding pixel in the live image to identify the edges of the object.

4. The method of claim 3 wherein the distance measurement is statistical.

5. The method of claim 4 wherein the reference image is constructed from multiple static images, and further comprising:

determining mean and variance values for the image gradients, and wherein the distance measurement is a Mahalanobis distance.

6. The method of claim 3 further comprising, for each pixel in the live image:

(a) testing if a gradient magnitude is greater than a first threshold and continuing to (b) if true, else labeling the pixel as not a object edge;

(b) testing if the gradient magnitude is a local maximum and continuing to (c) if true, else labeling the pixel as not a object edge;

(c) testing if a distance between the gradient magnitude at the pixel and a corresponding gradient magnitude of a corresponding pixel in the reference image is less than a second threshold and continuing to (d) if true, else labeling the pixel as a object edge;

(d) testing if a distance between the gradient direction at the pixel and a corresponding gradient direction of a corresponding pixel in the reference image is less than a third threshold and continuing to (e) if true, else labeling the pixel as a object edge;

(e) testing if a distance between an intensity at the pixel and a corresponding intensity of a neighboring pixel in a gradient direction of lower intensity is less than a fourth threshold and continuing to (f) if true, else labeling the pixel as a object edge; and

(f) testing if a distance between the intensity at the pixel and a corresponding intensity of a neighboring pixel in a gradient direction of higher intensity is less than a fifth threshold and labeling the pixel as not a object edge if true, else labeling the pixel as a object edge.

7. The method of claim 1 wherein the edges of the object are detected in stereo images to provide a 3D reconstruction for the object.

8. The method of claim 7 further comprising:

analyzing the 3D reconstruction to determine whether the object is a wheelchair user or an ambulatory person.

9. The method of claim 8 further comprising:

measuring a height of the 3D reconstruction to determine whether the object is a wheelchair user.

10. The method of claim 8 further comprising:

measuring a shape of the 3D reconstruction to determine whether the object is a wheelchair user.

11. The method of claim 8 further comprising:

measuring a rigidity at the base of the 3D reconstruction to determine whether the object is a wheelchair user.

12. The method of claim 8 wherein a signal is generated if the object is the wheelchair user.

13. The method of claim 7 wherein there are multiple objects in the scene, further comprising:

segmenting the 3D reconstruction into an individual 3D reconstruction for each object in a scene. 14. A system for detecting an object in an image, comprising;

a camera configured to acquire a static image of a scene when the scene is static and a live image of the scene when the scene includes an object;

means for constructing a reference image from image gradients in the static image; and

a comparator configured to compare the image gradients of the reference images with image gradients of the live image to locate edges of the object in the live image.