Infrared camera detection systems and methods

• US 9,247,159 B2
• Filed: 02/07/2013
• Issued: 01/26/2016
• Est. Priority Date: 08/11/2010
• Status: Active Grant

• Current Assignee: Teledyne FLIR LLC (Teledyne Technologies Incorporated)
• Original Assignee: FLIR Systems, Inc. (Teledyne Technologies Incorporated)
• Inventors: Brown, Andrew Phillip, Hgasten, Nicholas, Richards, Austin A.
• Primary Examiner(s): Elahi, Shan
• Application Number:
  US13/762,255
• Publication Number:
  US 20130147951A1
• Time in Patent Office: 1,083 Days
• Field of Search: 348/143
• US Class Current: 1/1
• CPC Class Codes: G01N 25/18 by investigating thermal conductivity by calorimetry G01N25/20; by measuring change of resistance of an electrically-heated body G01N27/18
• CPC Class Codes: G01N 33/1833 Oil in water water in oil G01N33/2847
• CPC Class Codes: G06F 2218/08 Feature extraction
• CPC Class Codes: G06T 2207/10048 Infrared image
• CPC Class Codes: G06T 2207/20076 Probabilistic image processing
• CPC Class Codes: G06T 2207/30108 Industrial image inspection
• CPC Class Codes: G06T 2207/30181 Earth observation
• CPC Class Codes: G06T 7/0004 Industrial image inspection
• CPC Class Codes: H04N 5/33 Transforming infrared radiation cameras or camera modules for generating image signals from infrared radiation H04N23/20; circuitry of SSIS for transforming infrared radiation into image signals H04N25/20

Abstract

Infrared imaging systems and methods disclosed herein, in accordance with one or more embodiments, provide for detecting petroleum in an infrared image. The detection involves the evaluation of local background model likelihoods in light of extracted pixel features as well as the evaluation of a global outlier model in light of extracted pixel features. Those pixels having a low likelihood for both their local background model as well as the outlier model are identified as corresponding to oil-contaminated water within the infrared image.

Claims

1. An infrared camera system, comprising:

a sensor adapted to capture a current image of a body of water and a plurality of preceding images of the body of water, wherein the current image and each preceding image all comprise a plurality of digital pixels; and

a processor adapted to;

determine a plurality of local background models for the plurality of digital pixels in the plurality of preceding images such that each digital pixel corresponds on a one-to-one basis with the corresponding local background model;

evaluate each digital pixel in the current image with the corresponding local background model; and

detect a petroleum-containing contaminant on the body of water responsive to at least the local background model evaluation for each digital pixel.

2. The infrared camera system of claim 1, wherein the processor is further adapted to:

evaluate a second set of features from the each digital pixel with regard to an outlier model; and

detect the petroleum-containing contaminant further responsive to at least the outlier model evaluation.

3. The infrared camera system of claim 2, wherein the outlier model comprises a single outlier model.

4. The infrared camera system of claim 2, wherein the first set of features comprises an intensity for each pixel, and wherein the evaluation with regard to the local background model comprises determining a likelihood of the pixel intensity in light of the local background model.

5. The infrared camera system of claim 4, wherein the processor is further adapted to determine a likelihood of the extracted features in light of the outlier model.

6. The infrared camera system of claim 5, wherein the processor is further adapted to identify pixels whose extracted features are relatively unlikely under both the outlier model and the local background model, wherein the identified pixels correspond to the detected petroleum-containing contaminant.

7. The infrared camera system of claim 1, wherein the plurality of digital pixels is downsampled pixels.

8. A method of detecting a petroleum-containing contaminant within an infrared image comprised of a plurality of digital pixels, the method comprising:

developing a plurality of local background models for a plurality of digital pixels in a plurality of proceeding images, wherein each digital pixel corresponds on a one-to-one basis with the corresponding local background model and the local background model models statistical behavior for at least one feature of the pixel;

extracting the at least one feature from each of the pixels;

determining a likelihood for the extracted feature in light of each corresponding local background model; and

detecting contaminant-containing pixels within the infrared image responsive to at least the determined likelihoods for each digital pixel.

9. The method of claim 8, wherein the at least one extracted feature comprises a pixel intensity.

10. The method of claim 9, wherein the at least one extracted feature further comprises a pixel contrast.

11. The method of claim 10, wherein the at least one extracted feature further comprises a pixel texture.

12. The method of claim 8, further comprising:

developing an outlier model, wherein the outlier model models the statistical behavior for the at least one extracted feature for pixels identified as corresponding to relatively unlikely to occur in light of the local background models; and

determining a likelihood for the feature in light of the outlier model for each of the pixels, wherein the detection of the contaminant-containing pixels is further responsive to outlier model likelihoods.

13. The model of claim 12, wherein the statistical behavior modeled by each of the local background models and by the outlier model is in the absence of contaminant.

14. The model of claim 12, further comprising updating each of the local background models responsive to the extracted features.

15. The model of claim 14, further comprising weighting each pixel according to the likelihood of the extracted feature in light of the corresponding local background model.

16. The model of claim 15, further comprising updating the outlier model responsive to pixel weights.

17. The method of claim 8, wherein the plurality of digital pixels is downsampled pixels.

18. A method, comprising

sensing a first infrared image frame of an ocean surface;

wherein the infrared image frame comprises a plurality of digital pixels;

for each pixel, modeling a statistical behavior of at least one extracted image feature using a mixture of Gaussians model (MGM), each pixel thereby corresponding to a local background model that corresponds to the modeled statistical behavior;

sensing a second infrared image frame of the ocean surface;

for each pixel in the second infrared image frame, extracting the at least one image feature;

for each pixel in the second infrared image frame, determining a likelihood of the at least one image feature in light of corresponding local background model, wherein each pixel in the second infrared image frame corresponds on a one-to-one basis with the corresponding local background model; and

detecting any oil-containing pixels in the second infrared image responsive to the determined likelihoods.

19. The method of claim 18, wherein the MGM comprise three Gaussians.

20. The method of claim 18, further comprising updating each local background model if no oil-containing pixels are detected.