Tactical Object Finder

• US 20130322742A1
• Filed: 05/14/2013
• Published: 12/05/2013
• Est. Priority Date: 05/29/2012
• Status: Active Grant

• Current Assignee: Johns Hopkins University
• Original Assignee: Johns Hopkins University
• Inventors: Walton, William C., Kolodner, Marc A., Gifford, Christopher M., Ruffle, Donald A.
• Granted Patent: US 8,908,919 B2
• Time in Patent Office: Days
• US Class Current: 382/159
• CPC Class Codes: G06F 18/217 Validation; Performance evaluation; Active pattern learning techniques
• CPC Class Codes: G06F 18/25 Fusion techniques
• CPC Class Codes: G06V 10/776 Validation; Performance evaluation
• 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
• CPC Class Codes: G06V 20/13 Satellite images

Abstract

A detection system includes processing circuitry configured to receive overhead image data divided into a plurality of image chips and receive metadata associated with the image data. The metadata includes ground sample distance information associated with the image data and provides an indication of ground area represented by each pixel within the image chips. The processing circuitry is further configured to screen the image chips for candidate detections based on a multi-stage screening process and determine whether to classify candidate detections as target detections. The process includes an intensity based screening stage, an object extraction stage that employs binary shape features to extract objects from detect positions identified based on an output of the intensity based screening stage, and a candidate detection identification stage employing template based and structural feature criteria to identify candidate detections from an output of the object extraction stage.

Claims

1. A detection system comprising processing circuitry configured to:

receive overhead image data divided into a plurality of image chips;

receive metadata associated with the image data, the metadata including ground sample distance information associated with the image data and providing an indication of ground area represented by each pixel within the image chips;

screen the image chips for candidate detections based on a multi-stage screening process, the multi-stage process including;

an intensity based screening stage,

an object extraction stage that employs binary shape features to extract objects from detect positions identified based on an output of the intensity based screening stage, and

a candidate detection identification stage employing template based and structural feature criteria to identify candidate detections from an output of the object extraction stage; and

determine whether to classify candidate detections as target detections.

2. The detection system of claim 1, wherein the processing circuitry is configured to screen the image chips by employing a plurality of trained models, each of the trained models outputting a label and corresponding confidence level based on an output of the multi-stage process.

3. The detection system of claim 2, wherein the processing circuitry is configured to determine whether to classify candidate detections as target detections based on fusion of the labels and corresponding confidence levels generated by each respective one of the trained models.

4. The detection system of claim 3, wherein the fusion is conducted on the basis of:

selecting classifier output based on testing accuracy of the trained models;

selecting classifier output based on which of the trained models generates a maximum confidence level;

generating classifier output based on average confidence level of the trained models;

or

employing a voting strategy to determine classifier output.

5. The detection system of claim 3, wherein the fusion is conducted on the basis of employing a weighted voting strategy to determine classifier output.

6. The detection system of claim 2, wherein the processing circuitry is configured to determine whether to classify candidate detections as target detections based on employing a meta classifier to evaluate decision patterns to provide a consensus label and confidence value for each candidate detection.

7. The detection system of claim 2, wherein the processing circuitry is further configured to enable an operator to provide feedback on label assignment for use in model training.

8. The detection system of claim 1, wherein the processing circuitry is further configured to output a label of a classification of the candidate detection and provide corresponding confidence level based on a result of the determining.

9. The detection system of claim 8, wherein the label defines the classification as a target or clutter, and wherein the confidence level is defined as a value between a maximum and minimum confidence level.

10. The detection system of claim 1, wherein the processing circuitry is configured to employ template based and structural feature criteria to identify candidate detections by employing deformable templates, Hough Transforms, Fast Fourier Transform (FFT) based angle measures, parallel line based measures, or edge directivity measures.

11. The detection system of claim 1, wherein the processing circuitry is configured to employ the ground sample distance to compute normalized features for scale invariant classification.

12. The detection system of claim 1, wherein the metadata further includes sensor model information, and wherein the processing circuitry is configured to employ the sensor model information in association with a mensuration library to assign accuracy information to a target detection.

13. A method comprising:

receiving overhead image data divided into a plurality of image chips;

receiving metadata associated with the image data, the metadata including ground sample distance information associated with the image data and providing an indication of ground area represented by each pixel within the image chips;

screening the image chips for candidate detections based on a multi-stage screening process, the multi-stage process including;

an intensity based screening stage,

an object extraction stage that employs binary shape features to extract objects from detect positions identified based on an output of the intensity based screening stage, and

a candidate detection identification stage employing template based and structural feature criteria to identify candidate detections from an output of the object extraction stage; and

determining whether to classify candidate detections as target detections.

14. The method of claim 13, wherein the screening the image chips comprises employing a plurality of trained models to screen the image chips, each of the trained models outputting a label and corresponding confidence level based on an output of the multi-stage process.

15. The method of claim 14, wherein determining whether to classify candidate detections as target detections comprises employing fusion of the labels and corresponding confidence levels generated by each respective one of the trained models.

16. The method of claim 15, wherein the fusion is conducted on the basis of:

selecting classifier output based on testing accuracy of the trained models;

selecting classifier output based on which of the trained models generates a maximum confidence level;

generating classifier output based on average confidence level of the trained models;

or

employing a voting strategy to determine classifier output.

17. The method of claim 15, wherein the fusion is conducted on the basis of employing a weighted voting strategy to determine classifier output.

18. The method of claim 14, wherein determining whether to classify candidate detections as target detections comprises employing a meta classifier to evaluate decision patterns to provide a consensus label and confidence value for each candidate detection.

19. The method of claim 14, further comprising enabling an operator to provide feedback on label assignment for use in model training.

20. The method of claim 13, further comprising outputting a label of a classification of the candidate detection and providing corresponding confidence level based on a result of the determining.