Motor vehicle occupant detection system employing ellipse shape models and bayesian classification

• US 6,493,620 B2
• Filed: 04/18/2001
• Issued: 12/10/2002
• Est. Priority Date: 04/18/2001
• Status: Expired due to Fees

• Current Assignee: Board of Trustees of The University of Illinois
• Original Assignee: Eaton Corp. (Wisconsin) (Eaton Corporation PLC.)
• Inventors: Zhang, Jun
• Primary Examiner(s): LOUIS JACQUES, JACQUES H
• Application Number:
  US09/837,012
• Publication Number:
  US 20020169532A1
• Time in Patent Office: 601 Days
• Field of Search: 701/45, 701/46, 701/28, 280/734, 280/735, 382/100, 382/104, 382/156, 382/190, 382/173, 382/175, 382/180, 382/209, 382/224, 382/103, 382/107, 382/203, 382/123, 382/199, 356/4.07, 356/5.05, 348/77, 348/155, 180/271, 180/273, 340/438
• US Class Current: 701/45
• CPC Class Codes: B60R 21/01538 for image processing, e.g. cameras or sensor arrays
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Abstract

A object on a seat of a motor vehicle is classified by creating a video image of the area and forming a silhouette of the object. The silhouette is divided into two segments and a separate ellipse is positioned to approximate the shape of each segment. The parameters that define the location and size of the two ellipses form a feature vector for the object. A Bayesian classification function utilizes the feature vector to determine the probability that the object fits which each of a plurality of classes. A class for the object is determined based on the probabilities. This method can be used to control operation of an air bag in the motor vehicle in response to the class of the object on the seat.

Claims

1. A method for classifying a object that is present within a motor vehicle, that method comprising:

acquiring an image of the object;

segmenting the object into a first segment and a second segment;

approximating the first segment with a first geometric shape which is selected from a group consisting of a circle, an ellipse, and a regular polygon and which is defined by a first set of parameters;

approximating the second segment with a second geometric shape which is selected from a group consisting of a circle, an ellipse, and a regular polygon and which is defined by a second set of parameters;

forming a feature vector from the first and second sets of parameters;

comparing the feature vector to plurality of template vectors, each representing different class of objects, to determine probabilities that the object falls within each class; and

classifying the object in response to the probabilities.

2. The method as recited in claim 1 wherein comparing the feature vector to plurality of template vectors utilizes a Bayesian classification function.

3. The method recited in claim 1 further comprising classifying the object in a given object class when none of the probabilities exceeds a predetermined value.

4. The method as recited in claim 1 wherein the first geometric shape and the second geometric shape are both ellipses.

5. The method as recited in claim 4 wherein the first set of parameters and the second set of parameters each comprises a location of a center, a major axis length, a minor axis length, and an rotation amount for the respective ellipse.

6. The object detection system as recited in claim 1 wherein segmenting the object comprises defining a portion of the object on one side of a demarcation line in the image as the first segment, and defining another portion of the object on another side of the demarcation line as the second segment.

7. The method as recited in claim 6 further comprising predefining the demarcation line for use with all images produced by the camera.

8. The method as recited in claim 1 further comprising defining each of the plurality of template vectors by producing a feature vector for at least one exemplary object in a respective class of objects.

9. The method as recited in claim 1 wherein classifying the object comprises determining whether the object is an adult person, child or an infant seat.

10. The method recited in claim 9 wherein classifying the object further comprises determining whether the infant seat is facing forward or rearward in the motor vehicle.

11. The method as recited in claim 1 further comprising controlling a mechanism in the motor vehicle in response to classifying the object.

12. The method as recited in claim 1 further comprising controlling an air bag in the motor vehicle in response to classifying the object.

13. A method for classifying a object that is present on a seat of a motor vehicle and controlling an air bag in response to the classifying, that method comprising:

acquiring an image of an interior of the motor vehicle;

extracting a portion of the image which corresponds to an object on the seat;

dividing the object into a first segment and a second segment;

substantially enclosing the first segment with a first ellipse that has a shape specified by a first set of parameters;

substantially enclosing the second segment with a second ellipse that has a shape specified by a second set of parameters;

forming a feature vector from the first and second sets of parameters;

comparing the feature vector to a plurality of template vectors representing different classes of objects to determine probabilities that the object falls within each class;

classifying the object into one of the different classes of objects in response to the probabilities; and

controlling operation of the air bag in response to classifying the object.

14. The method as recited in claim 13 wherein the plurality of template vectors represents an adult person, a child, and an infant seat.

15. The method recited in claim 13 wherein the plurality of template vectors represents an adult person, child, a forward facing infant seat and a rearward facing infant seat.

16. The method recited in claim 13 wherein extracting a portion of the image which corresponds to an object comprises creating a silhouette of the object.

17. The method as recited in claim 13 wherein the first set of parameters and the second set of parameters each comprises a location of a center, a major axis length, a minor axis length, and an rotation angle for the respective ellipse.

18. The method as recited in claim 13 further comprising predefining the demarcation line within all images produced by the camera;

and wherein dividing the object defines the first segment as a portion of the object on one side of a demarcation line, and defines the second segment as a portion of the object on another side of a demarcation line.

19. The method as recited in claim 13 further comprises defining each of the plurality of template vectors by producing a feature vector from at least one exemplary object in a respective class of objects.

20. The method recited in claim 13 further comprising classifying the object is a given object class when none of the probabilities exceeds a predetermined value.