Method of classifying vehicle occupants

• US 20070055428A1
• Filed: 09/02/2005
• Published: 03/08/2007
• Est. Priority Date: 09/02/2005
• Status: Active Grant

• Current Assignee: Aptiv Technologies AG
• Original Assignee: Delphi Technologies, Inc. (BorgWarner Incorporated)
• Inventors: Sun, Qin, Kong, Hongzhi, Nieto, Victor, Eiche, David
• Granted Patent: US 7,472,007 B2
• Time in Patent Office: Days
• US Class Current: 701/45
• CPC Class Codes: B60R 21/01538 for image processing, e.g. cameras or sensor arrays
• CPC Class Codes: G06V 40/103 Static body considered as a whole, e.g. static pedestrian or occupant recognition

Abstract

A method of classifying vehicle occupants utilizes a neural network engine having a state machine for determining if an occupant has changed preferably between an adult and adult, an adult and child, and a child and child from a pair of images. If no change has occurred, the method utilizes the prior occupant type and then decides if the occupant has changed in position. If no, the occupant is deemed static and the prior type is valid as a classification or output to preferably a vehicle restraint system. If the occupant has changed in position, a dynamic classification process is initiated by either an adult or a child dynamic classifier as dictated by the state machine. Valid dynamic classifier outputs or classifications can be sent to the restraint system and invalid dynamic classifier outputs are sent to a static classifier for update of the occupant type.

Claims

1. A method of classifying vehicle occupants comprising the steps of:

generating a stereo image pair of the occupant from two cameras;

inputting the image pair into an occupant classification controller;

processing the stereo image pair into a differential edge density feature by the controller;

inputting the differential edge density feature into a state machine of the controller;

determining, by the state machine and using the differential edge density feature, if the occupant has changed;

if the occupant has changed, applying a static classification process having a prior knowledge of the occupant category in memory of the controller;

determining if prior occupant type has changed by the state machine and between an adult and a child;

establishing whether the occupant is an adult or a child;

switching on a corresponding adult or child dynamic classifier of the controller;

switching off a static classifier of the static classification process; and

utilizing the differential edge density features by the corresponding adult or child dynamic classifier to determine an occupant class output which signifies occupant position.

2. The method of classifying vehicle occupants set forth in claim 1 wherein the occupant class output is outputted from a neural network engine having the adult and child dynamic classifiers, and is inputted into a restraint controller of a restraint system.

3. The method of classifying vehicle occupants set forth in claim 2 comprising the further steps of:

disarming the restraint system by the classification controller if the occupant class output signifies occupant is out of position; and

not disarming the restraint system by the classification controller if the occupant class output signifies occupant is in a normal position.

4. The method of classifying vehicle occupants set forth in claim 1 comprising the further step of combining wavelet features with the differential edge density features as an input into the adult and child dynamic classifiers to determine the occupant class output.

5. The method of classifying vehicle occupants set forth in claim 4 wherein the occupant class output is either out of position or normal position.

6. The method of classifying vehicle occupants set forth in claim 4 comprising the further step of applying a classification threshold to the occupant class output to determine validity of the output.

7. The method of classifying vehicle occupants set forth in claim 6 comprising the further step of training the adult and child classifiers with three prescribed input occupant categories signifying occupant out of position, occupant in normal position and empty seat.

8. The method of classifying vehicle occupants set forth in claim 7 comprising the further steps of:

forcing a low confidence value for invalidity of the output by the prescribed input occupant category signifying empty seat if an empty seat is present;

switching on the static classifier for receipt of the low confidence output; and

updating the occupant type.

9. The method of classifying vehicle occupants set forth in claim 1 wherein the occupant change is between adult and adult, adult and child, child and child, adult and empty seat, and child and empty seat.

10. The method of classifying vehicle occupants set forth in claim 1 comprising the further step of utilizing the prior occupant type if the occupant has not changed, as detected by the state machine.

11. The method of classifying vehicle occupants set forth in claim 10 comprising the further step of outputting the prior occupant type as the classification if the occupant remains in position as indicated by the differential edge density feature and thus the dynamic classification process is not required.

12. The method of classifying vehicle occupants set forth in claim 11 comprising the further step of initializing the corresponding adult or child dynamic classifier if the occupant has not remained in position.

13. The method of classifying vehicle occupants set forth in claim 12 comprising the further steps of:

comparing the output of the adult or child dynamic classifier to a classification threshold;

recording the output as an occupant class output if the output passes the classification threshold; and

recording the output as an invalid output if the output does not pass the classification threshold.

14. The method of classifying vehicle occupants set forth in claim 13 comprising the further steps of:

switching on the static classifier for receipt of the invalid output; and

updating of the occupant type by the static classifier.

15. A method of classifying vehicle occupants comprising the steps of:

generating a stereo image pair of the occupant from two cameras;

determining if the occupant has changed generally between an adult and adult, an adult and child, and a child and child by a state machine of a neural network engine;

determining if the adult or child has altered position within the vehicle by the state machine;

if position is altered, initiating a corresponding adult or child dynamic classifier to compute a classification output; and

if the position is not altered, recording a previous classification as the current classification.

16. The method of classifying vehicle occupants set forth in claim 15 comprising the further steps of:

processing the classification output through a pre-established confidence threshold to determine validity of the classification output;

if valid, releasing the classification output to a restraint system; and

if invalid, sending the classification output to a static classifier for updating of an occupant type.