CN100559887C - Utilize flame color template to carry out the method for detection - Google Patents

Abstract

The present invention relates to a kind of method of utilizing flame color template to carry out detection, it adopts the colored CCD camera system, according to the regularity of distribution of flame three primary colors at color space, gather several flame samples in advance, color value to flame carries out analytic statistics, obtain the sample average and the variance of three primary colors, statistics is imported computer as parameter.During detector work, colored CCD is a digital signal with the image transitions at scene, send in the host memory by image card, after processor obtains view data, compare with pre-determined parameter, judge whether it meets the Gaussian Profile rule, the zone that satisfies condition is divided as suspicious region, and it is done difference processing, further analyze this target and whether satisfy the fire jump over time, flicker, unsettled characteristics, improve the reliability of detection, can satisfy detection, interlink warning, picture control, different demands such as fire-proof and theft-proof.

Description

Utilize flame color template to carry out the method for detection

Technical field

The present invention relates to a kind of fire detecting method, specifically a kind of method of utilizing flame color template to carry out detection.

Background technology

We know, the sense cigarette of traditional type, temperature-sensitive, sensitive detector have been widely used in the structural fire protection, smoke detector surveys whether there is the smoke particle that is produced by fire in the monitor area usually, flue gas is the omen that fire produces, thereby the solid phase of the suspension that is comprised in the flue gas and liquid-phase particle just become a foundation of detection, but the size of particle, the dust in the environment, electromagnetism, factors such as air-flow tend to cause wrong judgement.Another feature of fire is followed the rising of temperature, also has uniform temperature in the flue gas of turbulent flow, when fire temperature parameter in official hour surpasses certain value, point type or line-type heat detector have just been triggered, occasion or the fire low in some temperature are in the stage of glowing, because still reaching the boundary that triggers warning, the rising of temperature often do not cause incuring loss through delay time of fire alarming, and the acute variation of temperature, as use in the process of air-conditioning or heating system, air-flow is directly blown over the detector that contains temperature-sensing element just might cause false alarm; Sensitive detector has ultraviolet flame type and infra red flame type, utilize ultra-violet radiation that flame produces and infrared radiation to survey flame and responded respectively, there is similar defective in sensitive detector, be not to have only fire just to launch the ultraviolet infrared light, there is X ray in the environment, electro-welding arc, perhaps detector is subjected to sunlight or the direct or indirect irradiation of other light sources all might produce wrong report.

Summary of the invention

The present invention is for overcoming above-mentioned the deficiencies in the prior art, provide a kind of colored CCD gamma camera that utilizes to take monitoring environment, according to the color model parameter of setting up in advance, isolate suspicious flame region, wait fire behavior to be further analyzed to its jump flicker then, can realize the fire detecting method of the accurate judgement of fire.

The objective of the invention is to adopt following technical proposals to realize:

A kind of method of utilizing flame color template to carry out detection may further comprise the steps:

A. the collection of flame sample;

B. the statistical analysis of flame model color parameter;

C. statistics is stored in the computer;

D. by the image signal of CCD continuous acquisition monitoring scene;

E. image signal is carried out processed and make fire and judge unusually.

The flame sample collection is to use the colored CCD camera system to gather the some width of cloth of colored flames sample in the described steps A.

The statistical analysis of the flame model color parameter among the described step B is, some width of cloth colored flames sample background of gathering in the steps A are treated to constant color, adds up the average and the variance of all visual flame pixel three primary colors.

May further comprise the steps in the described step e:

(1). the just colored video of the first step is decomposed into red (R), green (G), blue (B) three primary colors component;

Second step was calculated the distribution of color probability of each picture element

$p\left(A\right)=\frac{1}{{\left(2\mathrm{\π}\right)}^{\frac{3}{2}}\·{\left|\mathrm{\ΣA}\right|}^{\frac{1}{2}}}\·e\{-\frac{1}{2}{(A-\stackrel{\‾}{A})}^T{\left(\mathrm{\ΣA}\right)}^{-1}(A-\stackrel{\‾}{A})\}$

A=in the formula (R, G, B); A=(R, G, B); ∑ A=E ((A-A) ^T(A-A)), ∑ A is the matrix of a 3*3;

The 3rd step compared judgement with result of calculation and preset threshold T, obtained the flame color template binary picture

$F_C=\left\{\begin{array}{cc}1& p\left(A\right)\&GreaterEqual;T_1\\ 0& p\left(A\right)<T_1\end{array}\right.$

The 4th step was calculated the flame template bianry image in a period of time

${\mathrm{\&Delta;F}}_c=\underset{i=1}{\overset{n}{\mathrm{\&cup;}}}{F}_c\left(i\right)$

(2) the brightness rate of change of calculating color template image:

The first step is transformed to gray scale image g with colored video.

Second step is with t ₁, t ₂Different gray scale image g constantly carry out calculus of differences, obtain brightness and change binary picture

$D_{t_2{t}_1}=\left\{\begin{array}{cc}1& g_{t2}-g_{t1}\&GreaterEqual;T_2\\ 0& g_{t2}-g_{t1}<T_2\end{array}\right.$

The 3rd step statistics [t ₁, t _n] interior brightness rate of change of time

${\mathrm{\&Delta;D}}_{t_n{t}_1}=\frac{\underset{i=2}{\overset{n}{\mathrm{\&Sigma;}}}{D}_{t_i{t}_{i-1}}}{t_n-t_1}$

The 4th step calculated by the flame region S after the brightness variation judgement

$S=\left\{\begin{array}{cc}1& \mathrm{\&Delta;D}\&GreaterEqual;T_3\mathrm{and\&Delta;}{F}_c=1\\ 0& \mathrm{\&Delta;D}<T_3\mathrm{or\&Delta;}{F}_c=0\end{array}\right.$

(3) it is unusual to judge whether there is fire: if flame region thinks that to have fire unusual during greater than given threshold value M.

T in described (1) in the 3rd step ₁Be arranged between 0.3～0.6.

T in described (2) in the 3rd step ₂Be 50～80.

Brightness rate of change determining step based on grey scale difference is as follows:

(1) the colored video with a width of cloth scene reads in calculator memory;

(2) reading image data is decomposed into the three primary colors component A with it (R, G B), calculates the distribution of color probability of each pixel, ask for t ₁Flame color template bianry image F constantly _c(t ₁), deposit in calculator memory or the hard disk;

(3) ask for t ₁Gray scale image g (t constantly ₁), deposit in calculator memory or the hard disk;

(4) repetitive process (1) (2) (3) obtains t ₂Flame color template bianry image F constantly _c(t ₂) and gray scale image g (t ₂), calculate t ₁, t ₂Flame template bianry image Δ F constantly _cChange bianry image with brightness

(5) repeat (1)-(4) process, up to moment t _n, zoning S supposes S greater than given threshold value M, then output signal thinks that to have fire unusual, otherwise repeats the step of (1)-(5).

Determining of described Fujian value M value is relevant with the focal length of visual video input window and gamma camera, and focal length increase or input window reduce the M value and should suitably reduce, on the contrary M can increase, specifically determine according to actual conditions.

The present invention proposes to utilize the colored CCD gamma camera to take monitoring environment, utilizes the color model parameter of setting up in advance, isolates suspicious flame region, and fire behaviors such as the flicker of then it being jumped are further analyzed, and have realized the accurate judgement of fire.

The present invention has the following advantages:

1. contactless, adopt the common color gamma camera that environment is monitored, as a kind of contactless detecting devices, the differentiation of fire is to be based upon the live video video is carried out on the basis of data analysis, overcome some detector and responded insensitive defective, improved the actual effect and the range of application of detector at open space.

2. dual criterion, the present invention adopts flame color template and brightness differential pair fire video imagery to discern, the color characteristic that utilizes fire is to carry out the mode of thinking that fire is judged according to the mankind by vision, there is the scholar to point out that the people is when observation one images judges whether to have fire, mainly being to differentiate according to the color of image, secondly is texture and seldom according to shape.This paper utilizes the brightness rate of change to do further to judge, is this thought of out of control burning according to fire, and flame has the characteristic of the flicker of beating, and the combination of above-mentioned two kinds of criterions has greatly improved the reliability of detection.

3. need not to add filtering apparatus, need not to add extra filtering apparatus, can carry out fire monitoring automatically by computer by existing surveillance.

4. interlink warning, the present invention can connect conventional linkage alarm device, afterwards in time takes measures finding that fire is unusual, can be connected with automatic fire protection equipment, realizes the target of in time finding, in time reporting to the police, in time put out a fire to save life and property.

5. monitoring owing to adopted the common color gamma camera to carry out fire monitoring, can be included this device in existing closed monitor system, realizes that conventional monitoring and fire alarm unite two into one.

6. keeping records, the present invention can write down on-the-spot fire image automatically after finding fire, provide the firsthand information for analyzing cause of fire.

Description of drawings

Fig. 1 is a workflow diagram of the present invention;

Fig. 2 is that hardware of the present invention is formed schematic diagram.

Embodiment

The present invention is further described below in conjunction with drawings and Examples.

As shown in Figure 1 and Figure 2, the present invention adopts mini-computer, and the Pentium series processors adopts the multimedia image capture card of company of Daheng, and as DH-CG series, color video camera adopts the LTC series gamma camera of Bosch.

Gather the flame combustion picture array under the different light situation, comburant is based on wood materials, remove background parts except that flame by program or existing software, add up the average and the variance of all visual flame part three primary colors, average deposits computer in as an important parameter of follow-up computing, variance is represented the fluctuating range of each color component of sample image around average, and it has reference role for definite following threshold value.

Many color video camera framves are loaded on around the monitored area, field range according to every gamma camera, whole system can be covered effectively to monitor area, multichannel color video camera continuous acquisition monitored area influence signal, sent into the image collection card of computer in turn by video switcher, the image that image collection card will collect is saved in the hard disc of computer frame by frame continuously, store with binary file, pre-determine the time interval that every paths is observed, desirable 2～3 seconds, after collection finishes, from hard disk, read a frame color image data, the three primary colors component of supposing pixel A for (R, G, B), A=(R, G B) is flame color average through calculating, according to the covariance ∑ A of following formula compute matrix A:

∑ A=E ((A-A) ^T(A-A)), ∑ A is the matrix of a 3*3, and then calculates the Probability p that this pixel is the flame color pixel (A):

$p\left(A\right)=\frac{1}{{\left(2\mathrm{\&pi;}\right)}^{\frac{3}{2}}\&CenterDot;{\left|\mathrm{\&Sigma;A}\right|}^{\frac{1}{2}}}\&CenterDot;e\{-\frac{1}{2}{(A-\stackrel{\&OverBar;}{A})}^T{\left(\mathrm{\&Sigma;A}\right)}^{-1}(A-\stackrel{\&OverBar;}{A})\}$

A=in the formula (R, G, B); A=(R, G, B);

And according to given threshold value T ₁Compare judgement, obtain the binary picture of the flame color template of present frame image

$F_C=\left\{\begin{array}{cc}1& p\left(A\right)\&GreaterEqual;T_1\\ 0& p\left(A\right)<T_1\end{array}\right.$

Threshold value T ₁Selection can influence the size of binary picture target area, T ₁The too small condition that makes is relaxed, and is differentiated for the number of picture elements of flame color increases, and crosses senior general and makes the condition harshness, may foreclose indivedual real flame pixels, and suggestion is with T ₁Be arranged between 0.3～0.6, differentiating F as a result _cDeposit in the computer, read in another frame image and repeat said process, preserve result of calculation, travel through all images after, statistics all possible flame region in this time period forms a total flame color two-value template Δ F _c, suppose to have gathered altogether n width of cloth image data, so ${\mathrm{\&Delta;F}}_c=\underset{i=1}{\overset{n}{\mathrm{\&cup;}}}{F}_c\left(i\right).$

Next change speed according to brightness again and determine suspicious region:

Read different t constantly ₁, t ₂Two frame chromatic imagies, make calculus of differences behind the gray processing, the brightness that obtains two frame images changes template

$D_{t_2{t}_1}=\left\{\begin{array}{cc}1& g_{t2}-g_{t1}\&GreaterEqual;T_2\\ 0& g_{t2}-g_{t1}<T_2\end{array}\right.$

T ₂Can get 50～80, The above results is saved in computer, read t ₃Chromatic image constantly calculates t ₂, t ₃Brightness constantly changes template

And the like after the whole picture group of traversal resembles, statistics [t ₁, t _n] interior brightness rate of change of time

${\mathrm{\&Delta;D}}_{t_n{t}_1}=\frac{\underset{i=2}{\overset{n}{\mathrm{\&Sigma;}}}{D}_{t_i{t}_{i-1}}}{t_n-t_1},$

The brightness rate of change is greater than set-point T under the calculating color template overlay area ₃Image area S,

$S=\left\{\begin{array}{cc}1& \mathrm{\&Delta;D}\&GreaterEqual;T_3\mathrm{and\&Delta;}{F}_c=1\\ 0& \mathrm{\&Delta;D}<T_3\mathrm{or\&Delta;}{F}_c=0\end{array}\right.$

S as a result judged wherein whether there is the target area (target area thresholding be represent to exist) earlier at 1 o'clock,, add up every region area S if exist _iSize, work as S _iThink during greater than given threshold value M that to have fire unusual.

Determining of M value is relevant with the focal length of visual video input window and gamma camera, focal length increase or input window reduce the M value and should suitably reduce, otherwise M can increase, specifically determine according to actual conditions, for example focal length is 12mm, the video output window is 300*300, and M gets 1 ‰ of input window area.

If judge to have fire, the influence of this passage immediately switches to the current picture of closed monitor system, starts linkage alarm device simultaneously, starts the intelligent fire device, thus the purpose that reaches timely discovery, in time handles.

Claims (2)

1, a kind of method of utilizing flame color template to carry out detection is characterized in that taking following steps:

(1) uses the colored CCD gamma camera to gather the some width of cloth of flame sample image, add up the average and the variance of all visual flame pixel three primary colors;

(2) gather the on-site supervision signal of video signal, as follows, calculate flame color template according to distribution of color probability and known average, variance parameter;

The first step with colored video be decomposed into the red R of being, the green G of being, indigo plant is B three primary colors components;

Second step was calculated the distribution of color probability of each picture element

$p\left(A\right)=\frac{1}{{\left(2\mathrm{\&pi;}\right)}^{\frac{3}{2}}\&CenterDot;{\left|\mathrm{\&Sigma;A}\right|}^{\frac{1}{2}}}\&CenterDot;e\{-\frac{1}{2}{(A-\stackrel{\&OverBar;}{A})}^T{\left(\mathrm{\&Sigma;A}\right)}^{-1}(A-\stackrel{\&OverBar;}{A})\}$

A=in the formula (R, G, B); A=(R, G, B); ∑ A=E ((A-A) ^T(A-A)), ∑ A is the matrix of a 3*3;

The 3rd step compared judgement with result of calculation and preset threshold T, obtained the flame color template binary picture

$F_C=\left\{\begin{array}{cc}1& p\left(A\right)\&GreaterEqual;T_1\\ 0& p\left(A\right)<T_1\end{array}\right.$

The 4th step was calculated the flame color template bianry image in a period of time

${\mathrm{\&Delta;F}}_c=\underset{i=1}{\overset{n}{\&cup;}}{F}_c\left(i\right)$

(3) the brightness rate of change of the image under the calculating flame color template correspondence:

The first step is transformed to gray scale image g with colored video

Second step is with t ₁, t ₂Different gray scale image g constantly carry out calculus of differences, obtain brightness and change binary picture

$D_{t_2{t}_1}=\left\{\begin{array}{cc}1& g_{t2}-g_{t1}\&GreaterEqual;T_2\\ 0& g_{t2}-g_{t1}<T_2\end{array}\right.$

The 3rd step statistics [t ₁, t _n] interior brightness rate of change of time

${\mathrm{\&Delta;D}}_{t_n{t}_1}=\frac{\underset{i=2}{\overset{n}{\mathrm{\&Sigma;}}}{D}_{t_i{t}_{i-1}}}{t_n-t_1}$

The 4th step calculated by the flame region S after the judgement of brightness rate of change

$S=\left\{\begin{array}{cccc}1& \mathrm{\&Delta;D}\&GreaterEqual;T_3& \mathrm{and}& \mathrm{\&Delta;}{F}_c=1\\ 0& \mathrm{\&Delta;D}<T_3& \mathrm{or}& {\mathrm{\&Delta;F}}_c=0\end{array}\right..$

2. the method for utilizing flame color template to carry out detection according to claim 1 is characterized in that, following steps are taked in the brightness rate of change judgement in the 4th step in the described step (3):

(1) the colored video with a width of cloth scene reads in calculator memory;

(2) read image data it is decomposed into the three primary colors component A, calculate the distribution of color probability of each pixel, ask for t ₁Flame color template bianry image F constantly _c(t ₁), deposit in calculator memory or the hard disk;

(3) ask for t ₁Gray scale image g (t constantly ₁), deposit in calculator memory or the hard disk;

(4) repetition abovementioned steps (1), (2) and (3) obtain t ₂Flame color template bianry image F constantly _c(t ₂) and gray scale image g (t ₂), calculate t respectively ₁And t ₂Flame color template bianry image Δ F constantly _cChange bianry image with brightness

(5) repeat (1)-(4) process, up to moment t _n, zoning S supposes S greater than given threshold value M, then output signal thinks that to have fire unusual, otherwise repeats the step of (1)-(5).

Images