CN201993875U - Combustion-detecting system based on color and dynamic characteristic - Google Patents
Combustion-detecting system based on color and dynamic characteristic Download PDFInfo
- Publication number
- CN201993875U CN201993875U CN2010206555472U CN201020655547U CN201993875U CN 201993875 U CN201993875 U CN 201993875U CN 2010206555472 U CN2010206555472 U CN 2010206555472U CN 201020655547 U CN201020655547 U CN 201020655547U CN 201993875 U CN201993875 U CN 201993875U
- Authority
- CN
- China
- Prior art keywords
- flame
- color
- smog
- detecting
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Fire-Detection Mechanisms (AREA)
Abstract
The utility model discloses a combustion-detecting system based on colors and dynamic characteristics, wherein the target colors and the dynamic characteristics are simultaneously utilized. The detecting system comprises a visible-light camera, a flame-detecting and smog-detecting module and an alarming output module, wherein the visible-light camera is installed on a supporting device, a video signal is output to the flame-detecting and smog-detecting module by the visible-light camera, and the alarming output module is driven to give an alarm by the flame-detecting and smog-detecting module. In the utility model, the target colors and the dynamic characteristics are utilized for small-scale flame and smog detection, and a fire is monitored and prevented through the simultaneous working of two kinds of detection, so that the fire is identified and control at the initial stage of origination, and the damage to the lives and properties of people is reduced.
Description
Technical field
The utility model discloses a kind of fire alarm detection system, particularly a kind of burning detection system based on color and dynamic perfromance of utilizing color of object and dynamic perfromance simultaneously.
Background technology
Fire causes serious threat and harm to people's life, property safety, be strict prevention and control in people's productive life always, flame that produces in the object combustion process and smog all are that human body directly endangers the source, therefore, detect reducing fire hazard significant to the initial stage of flame in the combustion process and smog.At present, not to the effective detection means of the detection of flame and smog, all be to adopt the thermal sensing element flame detection usually in the prior art, simultaneously, detect smog by smoke transducer, the two divides body running, still, generally, during by thermal sensing element and the early warning of smoke transducer detection of fires, have only the intensity of a fire to acquire a certain degree, could trigger warning, at this moment, certain harm and loss have been caused.
Summary of the invention
At there not being effective shortcoming of fire being carried out monitoring alarm in the above-mentioned prior art of mentioning, the utility model provides a kind of new burning detection system based on color and dynamic perfromance, the utility model has designed on the basis of having carried out big quantity research at small scale flame and Smoke Detection and has utilized color of object and dynamic perfromance to carry out the method and system of small scale flame and Smoke Detection simultaneously, it is worked simultaneously by two kinds of detections, and fire is monitored, prevented.
The technical scheme that its technical matters that solves the utility model adopts is: a kind of burning detection system based on color and dynamic perfromance, detection system comprises that the visible image capturing head, the flame that are installed on the bracing or strutting arrangement detect and Smoke Detection module and alarm output module, a visible image capturing outputting video signal detects and the Smoke Detection module to flame, and flame detects with Smoke Detection module drive alarm output module and reports to the police.
The technical scheme that its technical matters that solves the utility model adopts further comprises:
The angle of described visible image capturing head and vertical direction is the 40-60 degree, and towards ground.
The setting height(from bottom) of described visible image capturing head is apart from ground 2.6-3.5 rice.
The beneficial effects of the utility model are: the utility model utilizes color of object and dynamic perfromance to carry out small scale flame and Smoke Detection, it is worked simultaneously by two kinds of detections, fire is monitored, prevented, make fire promptly be identified, control, reduce infringement people's life, property at early period of origination.
Below in conjunction with the drawings and specific embodiments the utility model is described further.
Description of drawings
Fig. 1 is a system block diagram of the present utility model.
Fig. 2 is the utility model structural representation.
Fig. 3 is the utility model flame and Smoke Detection procedures system process flow diagram.
Fig. 4 tentatively declares the process flow diagram in knowledge stage for the utility model flame.
Fig. 5 is the utility model Smoke Detection process flow diagram.
Embodiment
Present embodiment is the utility model preferred implementation, and other all its principles are identical with present embodiment or approximate with basic structure, all within the utility model protection domain.
Please referring to accompanying drawing 1, the utility model comprises that mainly the visible image capturing head, the flame that are installed on the bracing or strutting arrangement detect and Smoke Detection module and alarm output module, the visible image capturing head obtains monitor video in real time, and outputting video signal detects and the Smoke Detection module to flame, in the present embodiment, flame detects and the Smoke Detection module is a MCU module, flame detects with Smoke Detection module drive alarm output module and reports to the police, as trigger warning, and alerting signal is sent to Surveillance center or 119.Please referring to accompanying drawing 2, in the present embodiment, camera 1 is fixedly mounted on the support column 2, and the setting height(from bottom) of camera 1 is apart between the 2.6-3.5 rice of ground, and and the angle theta of vertical direction be between the 40-60 degree, and towards ground.
Please referring to accompanying drawing 3, in the utility model, the main flow process and the step of flame and smoke detection system are as follows:
The first step, flame detects with the Smoke Detection module and reads real-time video simultaneously, carries out flame and Smoke Detection respectively;
In second step, when the testing result of Smoke Detection module is a true time, system triggers is reported to the police, and alerting signal is sent to Surveillance center or 119, reads in video then again and continue to detect.
Wherein, the flame detection comprises tentatively to be declared knowledge and finally declares two stages of knowledge, and the testing result of tentatively declaring the knowledge stage when flame is a fictitious time, transfers to reading in video and continuation detection again; The testing result of tentatively declaring the knowledge stage when flame transfers execution flame to and finally declares the knowledge stage for true.The result's continuous three times (at every turn utilizing 50 frame video images) who finally declares the knowledge stage when flame is true time, and system triggers is reported to the police, and alerting signal is sent to Surveillance center or 119, reads in video then again and continue to detect.
Please referring to accompanying drawing 4, the process flow diagram that the flame in the utility model is tentatively declared the knowledge stage as shown in Figure 3, the key step that flame is tentatively declared in the knowledge stage process flow diagram is described as follows:
The first step is the calculating of foreground image, and its computation process is as follows:
Wherein
With
Represent the m+1 frame in the original video and the matrix of m frame gray level image respectively.The RGB color conversion is that the computing formula of gray scale adopts Gray=R*0.299+G*0.587+B*0.114.I
mBe called m frame frame difference image.I
m(p q) represents I
mIn the pixel value of the capable q of p row.Utilize big Tianjin method to I
mCarry out the image I that binaryzation obtains '
mBe called m frame binaryzation foreground image.
Second step was the calculating of the MHI image of binaryzation foreground image (motion history image).The MHI image helps following the tracks of general motion.The computing formula of the MHI image of binaryzation foreground image is
I
MHI(p，q)＝I′
1(p，q)AND?I′
2(p，q)...AND?I′
10(p，q). (4)
P=1 wherein ..., P, q=1 ..., Q.
IMHI(p q) represents the pixel value that the capable q of p is listed as in the MHI image.AND represents " logical and ", and in other words, formula (4) expression is carried out the logical and operation to continuous 10 frame binaryzation foreground images, and the result of operation is the MHI image of binaryzation foreground image.
The 3rd step was the calculating of two-value flame color image.Make F
1With F
2Be respectively corresponding
With
Two-value flame color image.The concrete computation process of two-value flame color image is as follows:
To the video that the visible image capturing head is gathered, we at first judge the pixel that satisfies following color condition in its RGB color space and the hsv color space:
R＞G＞B (1)
R＞R
t (2)
S＞＝(255-R)*S
t/R
t (3)
Wherein, R, G, B represent the red, green, blue component value in the RGB color space respectively.R
tWith S
tBe two following threshold value: R is set
t=65, S
t=135.S is the S component in hsv color space.The computing formula of S is as follows:
max＝max(R，G，B)，min＝min(R，G，B).
Utilize above-mentioned formula, can draw its two-value flame color image according to each two field picture.Order
(herein
Representative is from the m color image frame of input video) and F
mBe respectively original video frame and corresponding two-value flame color image.
With F
mBetween the corresponding relation of pixel as follows: if
In certain pixel satisfy formula (1), (2), (3), then F simultaneously
mIn respective pixel values be set to 1, otherwise be set to zero.
The 4th step was carried out following two logical ands operation:
F
3＝F
1?AND?F
2.?(5)
F＝F
3?AND?I
MHI.(6)
F
3Can be described as " accumulation " flame region, tentatively declare the flame region that the knowledge stage determines and F is a flame.The rationality of formula (5) is as follows: if occur the colour of flame in the initial frame in the video, and such colour no longer occurs in the subsequent frame, then in the video corresponding target generally be one for flame but have the moving target of flame color.
The purpose of formula (6) is to make the dynamic perfromance of considering flame in definite process of flame, its concrete effect is as follows: if there is the target with flame color of a static state in the video, then the MHI image of the binaryzation foreground image of target area should be null value basically, therefore, flame tentatively declare the knowledge stage can be with its erroneous judgement for flame.Obviously, the use of formula (5) and (6) can get rid of respectively have flame color moving target in the interference of static object.
The 5th step was determined for " really " flame region.This step is searching of 1 connected region to the F value and judges.Be no less than 81 connected region if wherein existence value is 1 number of pixels, think that then this zone is " really " flame region (adopting 350 * 288 image in different resolution in the present embodiment).
Though flame is initially declared the knowledge stage real flame there is very high verification and measurement ratio, also has many wrong reports.Flame finally declare the knowledge stage can eliminate these the wrong report in major part.Flame is finally declared the detected real flame zone that the knowledge stage at first utilizes a part of flame initially to declare the knowledge stage and is done training sample with false flame region, designs an improved k nearest neighbor sorter then on this basis.To a video to be detected, whole flame testing process is as follows: at first move flame and initially declare the knowledge stage, if the knowledge result that declares that this stage provides is true, then move flame and finally declare the knowledge stage, be true time only, just carry out flame and report to the police, otherwise do not report to the police finally declaring knowledge stage output result.The key step that flame is finally declared the knowledge stage is as follows:
The first step, training sample collection
Choose a part of flame and initially declare knowledge stage output identification value detected " flame " zone for " 1 ", some is the real flame zone for these zones, and some is false flame region.This step is calculated these regional Hu seven rank squares, and the Hu seven rank squares in a real flame zone are called a positive sample, and the Hu seven rank squares of false flame region are called a negative sample.Obviously, each positive sample and negative sample are a 7 degree of freedom vector.The set that positive sample and negative sample are formed is called training set.
The design of second step, improved k nearest neighbor sorter
The starting point of the design of improved k nearest neighbor sorter is: the classifying rules of practical k nearest neighbor sorter all need be trained and be drawn in the actual classification decision-making.After the training sample collection of the first step was finished, second step at first gathered other positive samples and negative sample is formed the checking collection, and the sample that checking is concentrated is called the checking sample.The classifying rules that need determine in this step is as follows: at first be that a checking sample is determined apart from its nearest K sample as its k nearest neighbor from training set, when the t of being no less than neighbour arranged among K the neighbour of this checking sample for positive sample, will verify that sample classification is a flame, otherwise it will be categorized as the nonflame target.This step is classified to verifying all concentrated samples at some different K and t value.Because checking concentrates the classification of sample known, we can calculate the classification accuracy of the concentrated sample of checking that K and t value arbitrarily draw.Classification accuracy is defined as the concentrated number of samples of checking of correct classification and the ratio of the sum of the concentrated sample of checking herein.This step will have the K of maximum classification accuracy and the t value optimal design as improved k nearest neighbor sorter, and use this design (corresponding K and t are called Ko and optimum to) in practice.Therefore,, at first move flame and initially declare the knowledge stage, if the knowledge result that declares that this stage provides is true to a video to be detected; Then move flame and finally declare the knowledge stage, at first calculate Hu seven rank squares that flame is initially declared " flame " zone that the knowledge stage obtains in this stage, calculation training is concentrated the distance of all samples and current Hu seven rank squares then, and determines its Ko neighbour; When have among the Ko neighbour be no less than to neighbour when the positive sample with current video in detected " flame " zone as real flame region.Ko that present embodiment draws and to are respectively 11 and 4.
Please referring to accompanying drawing 5, the key step of the Smoke Detection in the utility model is as follows:
The first step, determine possible smog zone in the following manner and draw respective image I ':
The judgement of at first, possible smog pixel is based upon on the basis of following rule:
R±α＝G±α＝B±α. (7)
L
1≤I≤L
2 (8)
D
1≤I≤D
2 (9)
Herein, R, G, B still represent the red, green, blue component value in the RGB color space respectively.And I represents the I component of HSI colour, and by formula
Calculate.α, D
1, D
2, L
1, L
2Be parameter, and be set to α=16, D
1=90, D
2=130, L
1=180, L
2=220.Use A respectively, on behalf of certain pixel, B, C satisfy criterion (7), (8) or (9) this incident.When incident took place, the respective symbol value was true.For example, when incident A took place, the value of A was very (TRUE).If the following logic operation result of certain pixel correspondence is true, then this pixel is judged as possible smog pixel: A AND (B OR C).Each pixel to picture frame all as above judges, judged result will be formed a bianry image I ' (I ' in be 1 the possible smog pixel of pixel representative).
In second step, utilize following formula to calculate background subtraction partial image D:
Wherein, I
N-1(p, q) and I
n(p q) represents n-1 frame and n two field picture (the n two field picture also is " current " frame) respectively.To D utilize big Tianjin method draw difference bianry image D ' (p, q), p=1 ..., P, q=1...Q.
In the 3rd step, utilize following formula to calculate the MHI image S of D '
MHI:
S
MHI(p，q)＝D′
1(p，q)?AND?D′
2(p，q)...AND?D′
10(p，q)，p＝1，...，P，q＝1...Q(11)
In other words, the utility model carries out the logical and operation to continuous 10 width of cloth difference bianry images, and its operating result is S
MHI
In the 4th step, utilize following formula to determine smog areal map S:
S＝S
MHI?AND?I‘?AND?D. (12)
Wherein, AND still represents the logical and operation.The S intermediate value is that 1 pixel is promptly represented the flame pixels of determining.
In the 5th step,, then trigger smog alarm if to have pixel value in the smog zone of determining be 1 and be no less than the connected region of 20 pixels.
The utility model use 40 sections flame smog videos (some videos include only flame and smog the two one of, and the minority video comprises flame and smog simultaneously) and 100 sections indoor videos that do not contain flame or smog test.Test shows that failing to report with rate of false alarm that flame of the present utility model detects is respectively: 10% and 15%; Fail to report and the rate of false alarm of smog are respectively: 20% and 16%
The big Tianjin method that adopts in the utility model is described below:
Big Tianjin method is proposed in 1979 by big Tianjin.Big Tianjin method claims maximum variance between clusters again, also is written as the Otsu method.The ultimate principle of big Tianjin method is when piece image is carried out binaryzation, should should maximize this principle according to the variance between two classes (being that prospect and background pixel respectively are a class) and select segmentation threshold, and respectively will be greater than being quantified as 1 and 0 (or being quantified as 0 and 1) with pixel value less than segmentation threshold.
Big Tianjin method is implemented as follows: to piece image, note g is the segmentation threshold of prospect and background, and prospect is counted and accounted for image scaled is w0, and average gray is u0; Background is counted and accounted for image scaled is w1, and average gray is u1.The overall average gray scale of image is: u=w0*u0+w1*u1.Travel through t from the minimum gradation value to the maximum gradation value, g is the optimal threshold of cutting apart when t makes value g=w0* (u0-u) 2+w1* (u1-u) 2 maximums.The pixel value of the bianry image after cutting apart is as follows: the some value of foreground pixel is 1 in the corresponding original image, and the some value of foreground pixel is 0 in the corresponding original image.
Can do following understanding to big Tianjin method: because of variance is the inhomogeneity a kind of tolerance of intensity profile, variance yields is big more, two parts difference that composing images is described is big more, be divided into target and all can cause two parts difference to diminish when part target mistake is divided into background or part background mistake, therefore make to mean the misclassification probability minimum cutting apart of inter-class variance maximum.
Claims (3)
1. burning detection system based on color and dynamic perfromance, it is characterized in that: described detection system comprises that the visible image capturing head, the flame that are installed on the bracing or strutting arrangement detect and Smoke Detection module and alarm output module, a visible image capturing outputting video signal detects and the Smoke Detection module to flame, and flame detects with Smoke Detection module drive alarm output module and reports to the police.
2. the burning detection system based on color and dynamic perfromance according to claim 1 is characterized in that: the angle of described visible image capturing head and vertical direction is the 40-60 degree, and towards ground.
3. the burning detection system based on color and dynamic perfromance according to claim 1 and 2 is characterized in that: the setting height(from bottom) of described visible image capturing head is apart from ground 2.6-3.5 rice.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010206555472U CN201993875U (en) | 2010-12-13 | 2010-12-13 | Combustion-detecting system based on color and dynamic characteristic |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010206555472U CN201993875U (en) | 2010-12-13 | 2010-12-13 | Combustion-detecting system based on color and dynamic characteristic |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN201993875U true CN201993875U (en) | 2011-09-28 |
Family
ID=44670356
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2010206555472U Expired - Fee Related CN201993875U (en) | 2010-12-13 | 2010-12-13 | Combustion-detecting system based on color and dynamic characteristic |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN201993875U (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105741480A (en) * | 2016-03-17 | 2016-07-06 | 福州大学 | Fire and smoke detection method based on image identification |
| CN105788142A (en) * | 2016-05-11 | 2016-07-20 | 中国计量大学 | Video image processing-based fire detection system and detection method |
| CN106373320A (en) * | 2016-08-22 | 2017-02-01 | 中国人民解放军海军工程大学 | Fire identification method based on flame color dispersion and continuous frame image similarity |
| CN106781210A (en) * | 2016-12-23 | 2017-05-31 | 安徽工程大学机电学院 | The recognition methods of cigarette in fire based on spatial closure |
| CN109344683A (en) * | 2018-08-03 | 2019-02-15 | 昆明理工大学 | A kind of the fire-smoke detection system and its control method of artificial intelligence |
| CN113033505A (en) * | 2021-05-20 | 2021-06-25 | 南京甄视智能科技有限公司 | Flame detection method, device and system based on dynamic classification detection and server |
-
2010
- 2010-12-13 CN CN2010206555472U patent/CN201993875U/en not_active Expired - Fee Related
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105741480A (en) * | 2016-03-17 | 2016-07-06 | 福州大学 | Fire and smoke detection method based on image identification |
| CN105788142A (en) * | 2016-05-11 | 2016-07-20 | 中国计量大学 | Video image processing-based fire detection system and detection method |
| CN105788142B (en) * | 2016-05-11 | 2018-08-31 | 中国计量大学 | A kind of fire detection system and detection method based on Computer Vision |
| CN106373320A (en) * | 2016-08-22 | 2017-02-01 | 中国人民解放军海军工程大学 | Fire identification method based on flame color dispersion and continuous frame image similarity |
| CN106781210A (en) * | 2016-12-23 | 2017-05-31 | 安徽工程大学机电学院 | The recognition methods of cigarette in fire based on spatial closure |
| CN106781210B (en) * | 2016-12-23 | 2019-05-10 | 安徽信息工程学院 | The recognition methods of cigarette in fire based on spatial closure |
| CN109344683A (en) * | 2018-08-03 | 2019-02-15 | 昆明理工大学 | A kind of the fire-smoke detection system and its control method of artificial intelligence |
| CN113033505A (en) * | 2021-05-20 | 2021-06-25 | 南京甄视智能科技有限公司 | Flame detection method, device and system based on dynamic classification detection and server |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN201993875U (en) | Combustion-detecting system based on color and dynamic characteristic | |
| CN110135269A (en) | A kind of fire image detection method based on blend color model and neural network | |
| CN101908142A (en) | Feature analysis-based video flame detecting method | |
| CN105744232B (en) | A kind of method of the transmission line of electricity video external force damage prevention of Behavior-based control analytical technology | |
| CN101777114B (en) | Intelligent analysis system and intelligent analysis method for video monitoring, and system and method for detecting and tracking head and shoulder | |
| CN106373320B (en) | Method for recognizing fire disaster based on flame color dispersion and sequential frame image similarity | |
| CN101395643B (en) | Detection of smoke with a video camera | |
| JP4705090B2 (en) | Smoke sensing device and method | |
| CN103400111A (en) | Method for detecting fire accident on expressway or in tunnel based on video detection technology | |
| CN104902265B (en) | A kind of video camera method for detecting abnormality and system based on background edge model | |
| CN101639983B (en) | Multilane traffic volume detection method based on image information entropy | |
| KR100918436B1 (en) | Fire detection system and method basedon visual data | |
| CN105575034B (en) | A kind of two waveband forest fire protection intelligent monitoring software image processing analysis method | |
| CN104780361B (en) | A kind of quality evaluating method of city video monitoring system | |
| CN101493980A (en) | Rapid video flame detection method based on multi-characteristic fusion | |
| CN101872418A (en) | Detection method based on group environment abnormal behavior | |
| JP2010238032A (en) | Smoke detection device | |
| CN102034110B (en) | Detection method of flame | |
| CN113449675B (en) | Method for detecting crossing of coal mine personnel | |
| CN105160326A (en) | Automatic highway parking detection method and device | |
| CN108052865A (en) | A kind of flame detecting method based on convolutional neural networks and support vector machines | |
| CN103456024A (en) | Moving object line crossing judgment method | |
| Hossain et al. | Wildfire flame and smoke detection using static image features and artificial neural network | |
| CN102496275A (en) | Method for detecting overload of coach or not | |
| CN104966059A (en) | Method for detecting phoning behavior of driver during driving based on intelligent monitoring system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110928 Termination date: 20121213 |