Imaging flame detection system

• US 5,937,077 A
• Filed: 04/25/1996
• Issued: 08/10/1999
• Est. Priority Date: 04/25/1996
• Status: Expired due to Term

• Current Assignee: General Monitors, Inc. (MSA Safety, Inc.)
• Original Assignee: General Monitors, Inc. (MSA Safety, Inc.)
• Inventors: Chan, William S., Burge, John W.
• Primary Examiner(s): CHANG, JON CARLTON
• Application Number:
  US08/637,557
• Time in Patent Office: 1,202 Days
• Field of Search: 382/100, 382/180, 382/203, 348/161, 348/143, 348/152, 348/164, 340/577, 340/578, 340/584, 340/587, 340/588, 250/339.15, 169/61
• US Class Current: 382/100
• CPC Class Codes: G06V 20/40 in video content extracting overlay text G06V20/62; video retrieval G06F16/70; processing of video elementary streams in video servers H04N21/234; processing of video elementary streams in video clients H04N21/44
• CPC Class Codes: G08B 17/125 by using a video camera to detect fire or smoke

Abstract

An imaging flame detection system for detecting flames of fires against false alarms using continuous imaging and pattern recognition of infrared (IR) images of a protected area under surveillance by the system. The system uses a Silicon (Si) Charge-Coupled-Device (CCD) array to detect the IR images, which first have passed through a wide field of view lens to establish a wide surveillance of the protected area, and have been filtered by a narrow band IR filter centered at 1,140 nm wavelength to remove false alarms originated from background sources of the protected area. Then, the system digitizes and thresholds these images into binarized images for pattern recognition to remove any remaining false alarms. To accomplish this function, this process forms blobs of suspect flames from the binarized images, extracts from these blobs characteristic parameters for statistical evaluation of the suspect blobs, matches these parameters with those stored in a system database to establish a vote for a flame or a false alarm, and then declares the suspect blob as a flame when the voting value exceeds an empirically-determined threshold. On a conclusive declaration, the system automatically activates a warning for suppressant release to put out the flame.

Claims

1. A flame detection system for detecting the presence of flame within an observed area while discriminating against false alarms, comprising:

infrared (IR) camera apparatus sensitive to energy in an infrared spectrum for producing infrared images of the observed area at a predetermined frame rate;

an optical apparatus for directing ambient spectral energy onto the camera apparatus, the optical apparatus including a wide field of view lens and a narrow band optical filter having a narrow passband centered in the near infrared spectrum, so that the spectral energy directed onto the IR camera apparatus is limited in its spectral content to the energy passed by the narrow band filter, thereby improving discrimination against false alarms due to background sources of infrared energy, said filter passband centered at 1140 nm and having a bandwidth which does not exceed about 50 nm for enhanced detection of flames and for suppression of background false alarm sources of infrared and ultraviolet energy outside its bandwidth;

pattern recognition apparatus responsive to the infrared images for discriminating flames from false alarms, the pattern recognition apparatus including;

feature extraction apparatus for determining blobs representing hot spots comprising each image, the feature extraction apparatus providing blob data regarding one or more features of the blobs;

decision apparatus for analyzing the blob data from the feature extraction apparatus to decide whether a blob is due to a flame or a false alarm, the decision apparatus updating the blob data continuously for an improved evaluation of whether a blob is due to a flame or a false alarm, the decision apparatus generating a flame detection signal indicating an alarm if the evaluation indicates a flame.

2. The flame detection system of claim 1 wherein the wide field of view lens has a field of view of at least 110 degrees.

3. The flame detection system of claim 1 wherein the wide field of view lens is fabricated of an optical material which passes only radiation in the visible and near infrared spectra.

4. The flame detection system of claim 1 wherein the IR camera apparatus includes a two-dimensional charge-coupled-device (CCD) array sensitive to near IR energy to reject background noise energy outside the near IR spectrum, and said system further includes analog-to-digital (A/D) converter apparatus for converting the analog signal of said array to digital form so that the signal level of each pixel is represented by a digital value, and wherein the pattern recognition apparatus includes binarizing apparatus for mapping each pixel value to a black value if the digital pixel value is below a threshold or otherwise to a white value, so that each image from the IR camera is represented by only black and white pixel values.

5. The system of claim 4 wherein said feature extraction apparatus includes apparatus for resolving the frame of binarized pixel data into blobs of adjacent white pixels.

6. The system of claim 4 wherien the CCD arry has sensitivety to inferared energy at 1140 nm sufficient to image flames, but which is insufficient to image background infrared sources emitting radiation at wavelengths longer than 1140 nm.

7. The system of claim 1 wherein the feature extraction apparatus further provides a predetermined parameter value describing each blob, wherein said parameter is an A/B ratio of the number of pixels in the blob to the number of pixels in the smallest upright rectangle which includes every pixel in the blob.

8. The flame detection system of claim 1 wherein said IR camera has sufficient sensitivity to capture images of flames at 1140 nm, and is without sufficient sensitivity to detect background sources emitting energy above 1140 nm.

9. The flame detection system of claim 1 wherein said narrow passband has a width of 17 nm.

10. A flame detection system for detecting the presence of flame within an observed area while discriminating against false alarms, comprising:

infrared (IR) camera apparatus sensitive to energy in an infrared spectrum for producing infrared images of the observed area at a predetermined frame rate;

pattern recognition apparatus responsive to the successive infrared images for producing flame detection signals, the pattern recognition apparatus including;

feature extraction apparatus for finding and determining predetermined features of connected sets of pixels representing a hot area in an image, said features including an A/B ratio of the number of pixels in a hot area to the number of pixels in the smallest upright rectangle which includes every pixel in the hot area, an A/B ratio difference value representing the difference in A/B ratios between successive inflection points, and a frame difference value representing the time between the occurrences of said successive inflection points; and

decision apparatus for analyzing the feature data from the feature extraction apparatus to decide whether a hot area is due to a flame or a false alarm, the decision apparatus employing updated feature data from images being constantly produced by the IR camera apparatus for an improved evaluation of whether a hot area is due to a flame or a false alarm, the decision apparatus generating a flame detection signal indicating an alarm if the evaluation indicates a fire.

11. The system of claim 10 wherein the decision apparatus applies feature data from the feature extraction apparatus in a decision algorithm to provide a likelihood of flame parameter value, the decision apparatus updating the flame parameter value as successive digital image representations are provided by the IR camera apparatus and comparing the flame parameter value for a given hot area to a threshold value, the decision apparatus generating a flame detection signal indicating an alarm if the flame parameter value exceeds the threshold value.

12. The flame detection system of claim 10 further comprising an optical apparatus for directing ambient spectral energy onto the IR camera apparatus, the optical apparatus including a wide field of view lens and a narrow band optical filter having a narrow passband centered in the near infrared spectrum, so that the spectral energy directed onto the IR camera apparatus is limited in its spectral content to the energy passed by the narrow band filter, thereby improving discrimination against false alarms due to background sources of infrared energy.

13. The flame detection system of claim 12 wherein the filter has a passband centered at 1140 nm, the passband having a bandwidth which does not exceed about 50 nm for enhanced detection of flames and for suppression of background false alarm sources of infrared and ultraviolet energy outside its bandwidth.

14. The flame detection system of claim 10 wherein the IR camera apparatus includes a two-dimensional charge-coupled-device (CCD) array sensitive to near IR energy to reject background noise energy outside the near IR spectrum, and said system further includes analog-to-digital (A/D) converter apparatus for converting the analog signal of said IR camera apparatus to digital form so that the signal level of each pixel is represented by a digital value, and wherein the pattern recognition apparatus includes binarizing apparatus for mapping each pixel value to a black value if the digital pixel value is below a threshold or otherwise to a white value, such that each image from the IR camera is represented by only black and white pixel values.

15. The system of claim 10 wherein said feature extraction apparatus further includes apparatus for resolving each frame of binarized pixel data into areas of adjacent white pixels.

16. The system of claim 10 wherein the IR camera apparatus includes a sensor array which has sensitivity to infrared energy at 1140 nm sufficient to image flames, but which is insufficient to image background infrared sources.

17. A method for detecting the presence of flame in an observed area while providing discrimination against false alarms, comprising the following steps:

forming successive digitized images of the observed area at a frame rate, each image representing only a narrow band of spectral components of ambient energy within a near infrared spectrum;

analyzing the digitized images to recognize blobs due to hot areas within the digitized images and extract features of the blobs;

in successive images of the observed area, matching blobs and determining points of inflection of a predetermined measure of the blobs representing change of the blob feature as a function of time;

using differences between adjacent points of inflection to determine a likelihood of flame parameter value;

comparing the flame parameter value to a threshold value and setting a flame detection signal if the threshold is exceeded.

18. The method of claim 17 further comprising the steps of updating the parameter value for successive digital image representations, comparing the updated parameter value to the threshold value, and setting the flame detection signal if the threshold is exceeded.

19. The method of claim 17 wherein the predetermined measure is an A/B ratio of the number of pixels in a blob to the number of pixels in the smallest upright rectangle which includes every pixel in the blob.

20. The method of claim 19 wherein the step of using temporal changes in said parameter includes using differences in the A/B ratio values between points of inflection and a frame difference value representing the number of image frames between those points of inflection as data values for determining whether a blob indicates a flame or a false alarm.

21. The method of claim 20 wherein corresponding ones of said A/B ratio difference values and said frame difference value are respectively employed as address values to look up a corresponding database entry having two values, a first database value representing a probability of flame and a second database value representing a probability of false alarm.

22. The method of claim 21 wherein said first database values are predetermined values obtained by processing images of known flame sources, and said second database values are obtained from known false-alarm sources.

23. The method of claim 22 wherein each of said first database values represent the fraction of all flame cases in which the particular A/B ratio difference value and frame difference value was observed, and each of said second database values represent the fraction of all false alarm cases in which the particular A/B difference value and frame difference values are observed.

24. The method of claim 17 wherein the step of analyzing the digitized images includes binarizing each pixel of said digitized images to either a black or a white pixel value.

25. The method of claim 17 wherein the step of analyzing the digitized images includes the steps of processing successive rows of pixels comprising each digitized image to find runs of adjacent white pixels, and comparing runs in adjacent rows of pixels to find overlapping runs to be resolved into blobs.

26. A flame detection system for detecting the presence of flame within an observed area while discriminating against false alarms, comprising:

infrared (IR) camera apparatus sensitive to energy in an infrared spectrum for producing infrared images of the observed area at a predetermined frame rate;

an optical apparatus for directing ambient spectral energy onto the camera apparatus, the optical apparatus including a wide field of view lens and a narrow band optical filter having a narrow passband centered in the near infrared spectrum, so that the spectral energy directed onto the IR camera apparatus is limited in its spectral content to the energy passed by the narrow band filter, thereby improving discrimination against false alarms due to background sources of infrared energy;

pattern recognition apparatus responsive to the infrared images for discriminating flames from false alarms, the pattern recognition apparatus including;

feature extraction apparatus for determining blobs representing hot spots comprising each image, the feature extraction apparatus providing blob data regarding one or more features of the blobs, the feature extraction apparatus further adapted to provide a predetermined parameter value describing each blob, wherein said parameter is an A/B ratio of the number of pixels in the blob to the number of pixels in the smallest upright rectangle which includes every pixel in the blob, wherein the blob data from said feature extraction apparatus includes an A/B ratio difference value representing the difference in A/B ratios between successive inflection points, and a frame difference value representing the time between the occurrences of said successive inflection points;

decision apparatus for analyzing the blob data from the feature extraction apparatus to decide whether a blob is due to a flame or a false alarm, the decision apparatus updating the blob data continuously for an improved evaluation of whether a blob is due to a flame or a false alarm, the decision apparatus generating a flame detection signal indicating an alarm if the evaluation indicates a flame.