Multi-wavelength video image fire detecting system

• US 20090315722A1
• Filed: 06/20/2008
• Published: 12/24/2009
• Est. Priority Date: 06/20/2008
• Status: Active Grant

• Current Assignee: InnoSys Industries Limited
• Original Assignee: Billy Hou
• Inventors: Hou, Billy
• Granted Patent: US 7,786,877 B2
• Time in Patent Office: Days
• US Class Current: 340/578
• CPC Class Codes: G01J 2005/0077 Imaging
• CPC Class Codes: G01J 5/0018 Flames, plasma or welding
• CPC Class Codes: G01J 5/0066 for hot spots detection
• CPC Class Codes: G01J 5/025 Interfacing a pyrometer to an external device or network; User interface
• CPC Class Codes: G01J 5/602 using selective, monochromatic or bandpass filtering
• CPC Class Codes: G08B 17/125 by using a video camera to detect fire or smoke

Abstract

To eliminate the drawbacks of false alarm problems due to mimic flame features of sunlight or moonlight reflecting on wave or mimic flame features of fabrics, animals or stage flames and alarm failures of methanol or ethanol burning flames or flames under a high illumination background condition, a multi-wavelength video image fire detecting system is disclosed consisting of a multi-wavelength video image fire detector, which is comprised of a color and near IR dual-mode camera, a color and B/W dual-mode camera, an image capture module, a processor and controller, an I/O module, a housing and a visible light and near IR view window, an alarm and indicate equipment, a setting and debugging system, a communication equipment, an area alarm and monitor system, and a pan & title and control system.

Claims

1. A multi-wavelength video image fire detecting system comprised of a multi-wavelength video image fire detector, an alarm and indicate equipment, a setting and debugging system, a communication equipment, an area alarm and monitor system, and a pan &

title and control system, said multi-wavelength video image fire detector being comprised of a color and near IR dual-mode camera, a color and B/W dual-mode camera, an image capture module, a processor and controller, an I/O module, a housing and a visible light and near IR view window, wherein;

said image capture module is adapted for converting analog video images into digital images and processing the digital images through a filtering process;

said processor and controller is responsible for fire recognition, switching control of said color and near IR dual-mode camera and said color and B/W dual-mode camera, pan and tilt control and back light control, and provides hovering and flying attitude adjustment controls when used in a UAV (unmanned aerial vehicle) for forest fire detection;

said I/O module comprises a plurality of hardware and software interfaces and a power supply system for system internal and external communication;

said alarm and indicate equipment is adapted for indicating pre-alarm, fire alarm, and failure;

said setting and debugging system is adapted for setting sensitivity, selection of different types of detection zones and self learning;

said communication equipment is for connection to local and remote alarm monitoring systems to upload image and fire probability and alarm information and to receive control information;

said area alarm and monitor system is adapted for monitoring remote detectors, giving control commands subject to operator'"'"'s requirements, providing a secondary image fire recognition algorithm.

2. The multi-wavelength video image fire detecting system as claimed in claim 1, which utilizes system algorithms to analyze color, B/W and near IR images fire features of a fire obtained through said color and B/W dual-mode camera and said color and near IR dual-mode camera under different aperture conditions, said color, B/W and near IR images fire features of a fire including images of smoke and flame obtained under an illumination of light, images of smoke and flame obtained under a bright light condition, images of smoke and flame obtained under a dark condition, and images of a transparent flame.

3. The multi-wavelength video image fire detecting system as claimed in claim 2, wherein said color and near IR dual-mode camera utilizes a CCD or CMOS type image sensor of spectrum within 400 nm˜

1200 nm and a high-pass near IR filter of cut-off frequency 850 nm (or 950 nm and up) and is switchblade between a color mode and a near IR mode;

said color and B/W dual-mode camera utilizes a CCD or CMOS type image sensor of spectrum within 400 nm˜

1200 nm and is switchblade between a color mode when at daylight or under the radiation of light and a B/W mode when under a no light or weak light environment;

said color and near IR dual-mode camera is adapted for picking up flame images under different illumination conditions and flame images of burning of alcohol-like fuel and substances and flicker images of a flame;

said color and B/W dual-mode camera is adapted for catching clear fire smoke images and flicker images of a flame under different illumination conditions.

4. The multi-wavelength video image fire detecting system as claimed in claim 2, wherein said color and near IR dual-mode camera and said color and B/W dual-mode camera use same or different focal distance lenses;

said color and near IR dual-mode camera adopts a small aperture and a low compensation to fit a high back light condition;

said color and B/W dual-mode camera adopts a great aperture and a high compensation to catch smoke motion and diffusion features under different illumination conditions.

5. The multi-wavelength video image fire detecting system as claimed in claim 2, wherein said processor and controller comprises a flame image recognition algorithm, a fire general probability algorithm a light source control algorithm, a pan control algorithm, a dual-mode camera switching control algorithm, and a fire positioning algorithm.

6. The multi-wavelength video image fire detecting system as claimed in claim 5, wherein said flame image recognition algorithm is to learn one long cycle background T_L and one short cycle background T_S, and to compare a current image to a last image or last few images for analysis, and also to compare a current image to a short cycle background or long cycle background for analysis, obtaining an index of flame and smoke related characteristic parameter.

7. The multi-wavelength video image fire detecting system as claimed in claim 5, wherein said flame image recognition algorithm comprises a flame recognition flow and a smoke recognition flow performed simultaneously to analyze series of images obtained through said color and near IR dual-mode camera and said color and B/W dual-mode camera;

to a complicated environment condition or suspected fire image, said flame image recognition algorithm drives a fire synthetical detecting module to compare and analyze second spectrum images so as to further obtain the fire character value and data fusion probability.

8. The multi-wavelength video image fire detecting system as claimed in claim 5, wherein said fire general probability algorithm integrates flame probability and smoke probability into a fire general probability subject to the probability fusion method of:

P_f1(t)=P_Smoke(t) [1+(P_Flame(t)−

K)]

P_f2(t)=P_Flame(t) [1+(P_Smoke(t)−

K)]

P_Fire(t)=max{P_f1(t)′

P_f2(t)}

in which, P_Flame(t) and P_Smoke(t) are flame occurrence probability and smoke occurrence probability at time t respectively;

P_Fire(t) is the fire general probability at time t;

K is the dead zone value that can be 25% or other value.

9. The multi-wavelength video image fire detecting system as claimed in claim 5, wherein said fire synthetical detecting module determines warning, pre-alarm and alarm subject to the following rules:

Warning;

if P_Smoke(t)≧

Atten1 or P_Flame(t)≧

Atten2 or P_Fire(t)≧

Atten3

Pre-alarm;

if P_Smoke(t)≧

Warn1 or P_Flame(t)≧

Warn2 or P_Fire(t)≧

Warn3

Alarm;

if P_Smoke(t)≧

Alarm1 or P_Flame(t)≧

Alarm 2 or P_Fire(t)≧

Alarm 3

in which, Atten1, Atten2, . . . , Alarm 3 are determination threshold values.

10. The multi-wavelength video image fire detecting system as claimed in claim 1, wherein said processor and controller controls switching of said color and B/W dual-mode camera between the color mode and the B/W mode and said color and near IR dual-mode camera between the color mode and the near IR mode subject to the light intensity of the field being monitored to obtain optimal fire video images.

11. The multi-wavelength video image fire detecting system as claimed in claim 1, wherein said processor and controller sets the monitoring area into 64 different detection zones, including high sensitivity smoke and flame detection zones, low sensitivity smoke and flame detection zones, failure detection zones and non-detection zones so that images of more than three failure detection zones assures rapid determination of field deviation and failure conditions.

12. The multi-wavelength video image fire detecting system as claimed in claim 1, wherein said processor and controller switch on/off an infrared backlight source subject to the intensity and distribution of light in the field being monitored.

13. The multi-wavelength video image fire detecting system as claimed in claim 1, wherein said processor and controller controls rotation of pan in each requested detection zone through a routing scanning or variable scanning control, and controls the pan in position in the detection zone where a fire is suspected or recognized until elimination of the fire suspect or system reset.

14. The multi-wavelength video image fire detecting system as claimed in claim 13, wherein the pan control of said processor and controller comprises a simple control to divide the camera field angle into predetermined n number of positions and to control rotation of the pan subject to the predetermined positions, and a continuously complicated control to use an angle feedback type pan.

15. The multi-wavelength video image fire detecting system as claimed in claim 1, which analyzes color, B/W and near IF video images through a recognition algorithm procedures:

Algorithm procedure I when the intensity of back light is below a predetermined threshold value, i.e., when L_LOW, the system outputs a control command to start the infrared back light source or other type of light source;

when the intensity of back light surpasses the predetermined threshold value, i.e., when L_NORMAL, the system outputs a control command to turn off the infrared back light source or other type of light source;

when the intensity of back light is high or the intensity of a certain ratio of the background area is excessively high, i.e., when L_HIGH, the system algorithm controls the two dual-mode cameras to work, enabling the field image obtained through the color and near IR dual-mode camera to be used as the main image sequence and the field image obtained through the color and B/W dual-mode camera to be used as a supplementary image sequence.

Algorithm procedure II;

To a pan system, it is necessary to calculate the installation location or the angle of rotation at every location subject to the field angle of the camera;

the conversion cycle of the pan at each location can be a fixed cycle or variable cycle;

the length of a fixed cycle must make sure that sufficient length series of images can be obtained for determination of fire;

the control of variable cycle is subject to condition of real time analysis on every angle or predetermined location;

then there is no any suspect or trend existed in the field, the algorithm outputs a newly calculated rotation cycle value to control rotation of the pan, and the rotation cycle should not less than a predetermined value P;

Algorithm procedure III;

To a system using a pan, the system runs matching and correction on the corresponding long cycle background when rotated to a predetermined angle or bit, facilitating further calculation;

Algorithm procedure IV;

In the image recognition algorithm, either on color, B/W or near IR image, it is necessary to learn one long cycle background T_L and one short cycle background T_S;

the length of time of the long cycle background T_L is divided into several minutes or hours;

the length of time of the short cycle background T_S is divided into several seconds or minutes;

Algorithm procedure V;

The image recognition algorithm should compare the current image to the last image or last few images for analysis, and also compare the current image to the short cycle background or long cycle background for analysis, reflecting the movement speed characteristic index I_SPEED=(change area between current image and last one or last few images)/(change area between current image and short cycle or long cycle background). This index is for smoke detection;

Algorithm procedure VI;

The algorithm divides field image into n number of equal or unequal monitoring areas, and calculates the related coefficient R₁, gradient variation G₁, saturation variation S₁, texture variation T₁ and optical flow field F₁ of every monitoring area, and also calculates a smoke diffusion characteristic index I_Disp=f₁{R_i, G_i, S_i, T_i, F_i, . . . };

Algorithm procedure VII;

The algorithm analyzes color video at first, calculating flame characteristics, i.e., the parameters of flicker frequency F_qi at 2˜

12 Hz, flame area variation rate A_i, and long cycle and short cycle expansion rate PR_i, obtaining a flame characteristic index I_Chara1=f₃{Fq_i, A_i, PR_i, . . . };

Algorithm procedure VIII;

The algorithm analyzes flicker video of blocked fire appeared in the color, B/W or near IR image. Normally, B/W and near IR image catching is employed to pick up flicker image of flicker frequency at 2˜

6 Hz;

obtaining an index I_Freq^—

indicative of the flicker characteristics;

Algorithm procedure IX;

The algorithm performs a simultaneous or time-shared calculation on the two parallel image series (color, B/W, or near IR);

the two calculation results are fused into a flame probability, in which, P_Flame(t)=F₂{I_Chara1, I_Chara2, I_Freq}; and

Algorithm procedure X;

The algorithm performs a cluster analysis on the flame/smoke appeared in the field, marks the fire appeared at different locations or areas, and uses two cameras to calculate the coordinates of the location of the fire,

16. The multi-wavelength video image fire detecting system as claimed in claim 1, wherein the fire detection sensitivity is divided into levels 1˜

9, being selectable, and adjustable by setting.

17. The multi-wavelength video image fire detecting system as claimed in claim 1, wherein the alarm output of the system includes flame, smoke and fire probability, flame, smoke and fire location coordinates, fire area and location, address code of the detector in the alarmed area and other necessary information.

18. The multi-wavelength video image fire detecting system as claimed in claim 1, wherein said alarm and indicate equipment comprises a fire alarm button and a failure button operable to give an analog fire alarm signal and a failure signal to the multi-wavelength video image fire detector respectively for checking functioning of every detector signal transmission path to facilitate system maintenance.

19. The multi-wavelength video image fire detecting system as claimed in claim 1, which runs self-learning by means of said setting and debugging system subject the steps of:

Step 1, the video image of the false alarm or alarm failure is used as a sample and inputted into said setting and debugging system for calculating data fusion weight parameter;

Step 2, input the obtained standard fire video images corresponding to the calculated weight parameter into said setting and debugging system and check whether system response ability reaches the expected level or not, and then use the standard fire video image databank sample to run self learning and to regulate data fusion weight if the system response ability deviated from the expected level; and

Step 3, download the newly learned data fusion weight onto said multi-wavelength video image fire detector or computer for running through one run, and for further use as a reference parameter in any similar place when checked to be correct.

20. The multi-wavelength video image fire detecting system as claimed in claim 1, wherein the system simply uses a color and near IR dual-mode camera to eliminate transparent flame, fire blocking and other false alarm and alarm failure problems.