Hybrid architecture for acquisition, recognition, and fusion

• US 7,536,365 B2
• Filed: 12/08/2005
• Issued: 05/19/2009
• Est. Priority Date: 12/08/2005
• Status: Active Grant

• Current Assignee: Northrop Grumman Systems Corporation (Northrop Grumman Corporation)
• Original Assignee: Northrop Grumman Corporation
• Inventors: Aboutalib, Omar
• Primary Examiner(s): STARKS, WILBERT L
• Application Number:
  US11/297,679
• Publication Number:
  US 20080162389A1
• Time in Patent Office: 1,258 Days
• Field of Search: 706/15, 706/45
• US Class Current: 706/15
• CPC Class Codes: G06F 18/25 Fusion techniques
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/80 Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level multimodal speaker identification or verification G10L17/10

Abstract

A method comprises the steps of: producing one or more signals representative of a feature of a target or area of interest, statistically processing the signals to produce one or more first hypotheses of the target or area of interest, comparing the first hypotheses to one or more templates in a template library to produce one or more second hypotheses, and comparing the second hypotheses to one or more models in a model library to produce a target decision. An apparatus that performs the method is also provided.

Claims

1. A method comprising the steps of:

producing one or more signals representative of a feature of a target or area of interest;

statistically processing the signals to produce one or more first hypotheses of the target or area of interest;

comparing the first hypotheses to one or more templates in a template library to produce one or more second hypotheses, wherein if at least one first hypothesis did not match at least one of the templates, then retrieving at least one additional first hypothesis for comparison with the one or more templates;

comparing the second hypotheses to one or more models in a model library to produce a target decision, wherein if at least one second hypothesis does not match at least one of the models, then retrieving at least one additional second hypothesis for comparison with the one or more models; and

outputting the target decision identifying the target.

2. The method of claim 1, where the templates selected for comparison are selected based on the first hypotheses.

3. The method of claim 1, where the models selected for comparison are selected based on the second hypotheses.

4. The method of claim 1, where the step of statistically processing the signals uses one or more of:

a neural network, a decision tree, and fuzzy logic.

5. The method of claim 1, where the first hypotheses include a target orientation angle.

6. The method of claim 1, where the step of statistically processing the signals uses decision-level fusion.

7. The method of claim 1, wherein the signals are produced using different types of sensors.

8. An apparatus comprising:

a plurality of sensors for producing one or more signals representative of a feature of a target or area of interest; and

a processor for statistically processing the signals to produce one or more first hypotheses of the target or area of interest, for comparing the first hypotheses to one or more templates in a template library to produce one or more second hypotheses, wherein if at least one hypothesis did not match at least one of the templates, then retrieving at least one additional first hypothesis for comparison with the one or more templates, and for comparing the second hypotheses to one or more models in a model library to produce a target decision output, wherein if at least one second hypothesis does not match at least one of the models, then retrieving at least one additional second hypothesis for comparison with the one or more models.

9. The apparatus of claim 8, where the templates selected for companson are selected based on the first hypotheses.

10. The apparatus of claim 8, where the models selected for comparison are selected based on the second hypotheses.

11. The apparatus of claim 8, where the processor statistically processes the signals using one or more of:

a neural network, a decision tree, and fuzzy logic.

12. The apparatus of claim 8, where the first hypotheses include a target orientation angle.

13. The apparatus of claim 8, where the processor statistically processes the signals using decision-level fusion.

14. An apparatus comprising:

means for producing one or more signals representative of a feature of a target or area of interest;

means for statistically processing the signals to produce one or more Image Page 3 first hypotheses of the target or area of interest;

means for comparing the first hypotheses to one or more templates in a template library to produce one or more second hypotheses, wherein if at least one first hypothesis did not match at least templates, then retrieving at least one additional first hypothesis for comparison with the one or more templates; and

means for comparing the second hypotheses to one or more models in a model library to produce a target decision output identifying the target, wherein if at least one second hypothesis does not match at least one of the models, then retrieving at least one additional second hypothesis for comparison with the one or more models.

15. The apparatus of claim 14, where the templates selected for companson are selected based on the first hypotheses.

16. The apparatus of claim 14, where the models selected for comparison are selected based on the second hypotheses.

17. The apparatus of claim 14, where the means for statistically processing the signals uses one or more of:

a neural network, a decision tree, and fuzzy logic.

18. The apparatus of claim 14, where the first hypotheses include a target orientation angle.

19. The apparatus of claim 14, where the means for statistically processing the signals uses decision-level fusion.