CN107301653A - Video image fire disaster flame detection method based on BP neural network - Google Patents
Video image fire disaster flame detection method based on BP neural network Download PDFInfo
- Publication number
- CN107301653A CN107301653A CN201710351239.7A CN201710351239A CN107301653A CN 107301653 A CN107301653 A CN 107301653A CN 201710351239 A CN201710351239 A CN 201710351239A CN 107301653 A CN107301653 A CN 107301653A
- Authority
- CN
- China
- Prior art keywords
- flame
- video image
- fire
- detection method
- fire disaster
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Multimedia (AREA)
- Fire-Detection Mechanisms (AREA)
Abstract
The present invention relates to a kind of detection method, especially a kind of video image fire disaster flame detection method based on BP neural network belongs to the technical field of fire disaster flame detection.Flame image is split by the background modeling of mixed Gauss model, the geometric properties such as the color and area growth rate, circularity, fire angle of flame are analyzed, flame can be effectively recognized in the interference environments such as background light change.While detection accuracy is ensured compared with traditional flame detection method, false drop rate is reduced.
Description
Technical field
The present invention relates to a kind of detection method, especially a kind of video image fire disaster flame detection based on BP neural network
Method, belongs to the technical field of fire disaster flame detection.
Background technology
In early days, people use various electron detectors, and gathered data judges whether fire occurs.But because of its information list
One, and be easy to be influenceed by environment temperature, humidity, corrosivity etc., easily situations such as generation flase drop, missing inspection, its reliability,
Sensitivity, sustainability etc. all have much room for improvement.Spreading velocity of fire is exceedingly fast, and traditional fire detecting system can not meet fire completely
Calamity prevention is required.With the development of computer vision technique, fire image identification technology receives the highest attention of people with grinding
Study carefully.Fire image technology pointedly overcomes the main weakness of conventional fire detecting system, is adopted with reference to the image of high development
Truck and computer process ability so that visualization fire detection technology can combine the largely dynamic static nature of flame, greatly
Reliability, the real-time of fire identification are strengthened, there is important breakthrough to fire detection prevention.
Ma Zongfang etc. proposes the Image Fire Detection Technology based on SVMs.Vector machine is internal strict because of it
Mathematical modeling, makes it have amount of calculation small in any case, the characteristics of precision is high.But current corresponding mathematical modeling is not
Motion feature during flame combustion can be described accurately.
But existing video fire hazard flame identification technology still has following deficiency：1st, mathematical modeling is complicated, computationally intensive；
2nd, the degree of accuracy of flame identification is not high.
The content of the invention
The purpose of the present invention be overcome the deficiencies in the prior art there is provided it is a kind of based on BP neural network video fire
Calamity flame detecting method, it can lift the speed of flame identification and can effectively realize the detection of meeting flame, and accuracy of detection is high.
The technical scheme provided according to the present invention, a kind of video image fire disaster flame detection side based on BP neural network
Method, the fire disaster flame detection method comprises the following steps：
Step 1, offer video image to be detected, and extract the flame image region in video image；
Step 2, the flame image region to extraction, carry out flame characteristic extraction, and the flame characteristic of the extraction includes face
Product growth rate, circularity and fire angle；
Step 3, it regard area growth rate, circularity and the fire angle of said extracted as the defeated of ant colony neutral net
Enter, and utilization ant colony neutral net judges the probability of flame.
In step 1, the flame region in video image is extracted using background modeling method and mixed Gauss model.
Advantages of the present invention：Flame image is split by the background modeling of mixed Gauss model, analyze flame color and
The geometric properties such as area growth rate, circularity, fire angle, fire can be effectively recognized in the interference environments such as background light change
Flame.While detection accuracy is ensured compared with traditional flame detection method, false drop rate is reduced.
Brief description of the drawings
Fig. 1 is flow chart of the invention.
Embodiment
With reference to specific drawings and examples, the invention will be further described.
As shown in Figure 1：In order to be able to lift the speed of flame identification and can effectively realize the detection of meeting flame, fire of the invention
Calamity flame detecting method comprises the following steps：
Step 1, offer video image to be detected, and extract the flame image region in video image；
When it is implemented, in the video sequence of shot by camera, compared to object static in background, flame combustion
With obvious kinetic characteristic.There are many methods for the detection of moving target, the present invention uses background modeling method, and using mixed
Gauss model is closed, there is good discrimination under the disturbance of various light conditions.
In the RGB color for the video image that camera is shot, X represents single pixel (pixel) point value, then probability
Density function can be described with K Gaussian function：
Wherein：K is distribution number, makes K=3；K-th distribution Mean Matrix, covariance matrix and weight coefficient be successively
uk、ΣkAnd ωk。xnIt is natural number for sample n values
One group of image sequence in the case where different illumination are not disturbed in the same time is randomly selected, an experimental image vector F is constituted
={ f1,f2,…,fn, using the distributed constant of greatest hope method initialization sample sequence, then iteration is until convergence
Obtain distributed constant θ=[ω of each pixel1,ω2,ω3,u1,u2,u3,∑1,∑2,∑3]。
Binaryzation present image and template image, set up picture element matrix θ and θ ', subtract each other each pixel value successively, take its inclined
The absolute value of difference.| E |=θ-θ ', suspicious region is screened after diagonalization according to threshold formula E '≤5E12.
Step 2, the flame image region to extraction, carry out flame characteristic extraction, and the flame characteristic of the extraction includes face
Product growth rate, circularity and fire angle；
Specifically, the feature of image flame is divided into color characteristic, physical geometric.Physical geometric can be segmented again
For, edge feature, textural characteristics etc..Flame is recognized according to the characteristic parameter of flame, developed in computer vision field
Rapidly.In the embodiment of the present invention, using the feature of area growth rate, circularity and fire angle as flame identification.
Firstly for area growth rate, flame starts in the very short time of burning, with the burning degree of flame, its area
Become larger, compared to static object, its rate of change has very high discrimination.Due to being the image on same video flowing
(the flame space physical location of adjacent two frame can't differ too big), the matching of flame region need to be only obtained by foregoing
Pretreatment obtains binary image, and white portion is flame region, can be carried out by calculating the ratio of change of its area
The judgement of doubtful flame.With fire area growth rate GiIt is used as criterion.Fire area growth rate G can be calculated by following formulai：
Formula (3) is the area difference in two moment flame image areas, S (Ri)tFor the area of t, S (Ri)t0For t0Moment
Area, Ri is that the mikey distance between adjacent two bright spot of flame radius estimated mean value at the moment is unit length 1.Its
The distance of remaining bright spot is tried to achieve by Pythagorean theorem.Girth can be tried to achieve by Boundary algorithm, specially known to those skilled in the art, this
Place is repeated no more.
Calculated for circularity：Irregular, and the shape of part interference source (street lamp, car light etc.) of authority fire disaster flame shape
Shape alignment degree is high, so by circularity, it is used as flame distinguishing rule.The definition of circularity such as formula (4), be specially：
Wherein：Ck、Ak、PkIt is followed successively by, the circularity of k-th of unit, area and girth, n is unit number, if outside object
Shape is closer to circle, then CkIt is bigger, if conversely, its profile is more complicated, CkIt is smaller, CkValue between 0 and 1.Draft a threshold
Value C0, work as Ck＞ C0When, then it is assumed that figure circle appearance profile is more regular, assert nonflame；Work as Ck＜ C0When, then the image appearance
Profile is very irregular, meets flame contours feature.
The kinetic characteristic of flame clearly have wedge angle：The wedge angle number of flame, it is irregular in time domain to follow, therefore, adopt
Take the characteristic of edge shake, it is possible to identify suspicious burning things which may cause a fire disaster.Analyze wedge angle number when under early stage flame, noise jamming, and doubtful thing
The edge changing rule of body, can distinguish other interference.In order to improve the accuracy of flame judgement, still asked using multi-group data
The method of average.A frame is taken out from each group image sequence at random, 5 groups of fire angle number experiment collection are obtained within the unit time limit,
Take its desired valueWith the threshold value J of priori0Compare, whenWhen, assert k sequence number atlas, meet fire
The edge trembling feature of flame, conversely, not possessing the edge features of fire disaster flame then.
Step 3, it regard area growth rate, circularity and the fire angle of said extracted as the defeated of ant colony neutral net
Enter, and utilization ant colony neutral net judges the probability of flame.
In the embodiment of the present invention, it is assumed that all weights of BP neural network and threshold value have m, setting weights interval [Wmin,
Wmax], s equal portions are evenly dividing into, set is set upInclude the neural network parameter p after decilei(1≤i≤m).Every ant (1
≤ k≤m) from setJth (1≤j≤m) individual elementSet out, according to the pheromones ρ and Path selection of each element
New probability formula (formula (5)) is in each setIt is middle to select an element as next target,
After all ants complete to select, it is believed that one time algorithm is completed, according to Pheromone update formula (formula (6)), adjustment
The pheromones of all elements, iterate this process.
Think the generation of optimal solution when all ants converge to same path, or reach maximum cycle NcmaxWhen,
Algorithm terminates.
Claims (2)
1. a kind of video image fire disaster flame detection method based on BP neural network, it is characterized in that, the fire disaster flame detection
Method comprises the following steps：
Step 1, offer video image to be detected, and extract the flame image region in video image；
Step 2, the flame image region to extraction, carry out flame characteristic extraction, and the flame characteristic of the extraction increases including area
Long rate, circularity and fire angle；
Step 3, using area growth rate, circularity and the fire angle of said extracted as ant colony neutral net input, and
The probability of flame is judged using ant colony neutral net.
2. the video image fire disaster flame detection method according to claim 1 based on BP neural network, it is characterized in that：Step
In rapid 1, the flame region in video image is extracted using background modeling method and mixed Gauss model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710351239.7A CN107301653A (en) | 2017-05-18 | 2017-05-18 | Video image fire disaster flame detection method based on BP neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710351239.7A CN107301653A (en) | 2017-05-18 | 2017-05-18 | Video image fire disaster flame detection method based on BP neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107301653A true CN107301653A (en) | 2017-10-27 |
Family
ID=60138096
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710351239.7A Pending CN107301653A (en) | 2017-05-18 | 2017-05-18 | Video image fire disaster flame detection method based on BP neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107301653A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111639825A (en) * | 2020-07-01 | 2020-09-08 | 广东工业大学 | Method and system for indicating escape path of forest fire based on A-Star algorithm |
| CN111986436A (en) * | 2020-09-02 | 2020-11-24 | 成都指码科技有限公司 | Comprehensive flame detection method based on ultraviolet and deep neural networks |
| CN116843628A (en) * | 2023-06-15 | 2023-10-03 | 华中农业大学 | Lotus root zone nondestructive testing and grading method based on machine learning composite optimization |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101231756A (en) * | 2008-01-30 | 2008-07-30 | 安防科技（中国）有限公司 | Method and apparatus for detecting moving goal shade |
| CN101751679A (en) * | 2009-12-24 | 2010-06-23 | 北京中星微电子有限公司 | Sorting method, detecting method and device of moving object |
| CN105788142A (en) * | 2016-05-11 | 2016-07-20 | 中国计量大学 | Video image processing-based fire detection system and detection method |
-
2017
- 2017-05-18 CN CN201710351239.7A patent/CN107301653A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101231756A (en) * | 2008-01-30 | 2008-07-30 | 安防科技（中国）有限公司 | Method and apparatus for detecting moving goal shade |
| CN101751679A (en) * | 2009-12-24 | 2010-06-23 | 北京中星微电子有限公司 | Sorting method, detecting method and device of moving object |
| CN105788142A (en) * | 2016-05-11 | 2016-07-20 | 中国计量大学 | Video image processing-based fire detection system and detection method |
Non-Patent Citations (1)
| Title |
|---|
| 熊和金 等: "《智能信息处理 第2版》", 31 August 2012, 《国防工业出版社》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111639825A (en) * | 2020-07-01 | 2020-09-08 | 广东工业大学 | Method and system for indicating escape path of forest fire based on A-Star algorithm |
| CN111639825B (en) * | 2020-07-01 | 2024-02-23 | 广东工业大学 | Forest fire indication escape path method and system based on A-Star algorithm |
| CN111986436A (en) * | 2020-09-02 | 2020-11-24 | 成都指码科技有限公司 | Comprehensive flame detection method based on ultraviolet and deep neural networks |
| CN111986436B (en) * | 2020-09-02 | 2022-12-13 | 成都视道信息技术有限公司 | Comprehensive flame detection method based on ultraviolet and deep neural networks |
| CN116843628A (en) * | 2023-06-15 | 2023-10-03 | 华中农业大学 | Lotus root zone nondestructive testing and grading method based on machine learning composite optimization |
| CN116843628B (en) * | 2023-06-15 | 2024-01-02 | 华中农业大学 | Lotus root zone nondestructive testing and grading method based on machine learning composite optimization |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110516609B (en) | Fire disaster video detection and early warning method based on image multi-feature fusion | |
| CN107609470B (en) | Method for detecting early smoke of field fire by video | |
| Khatami et al. | A new PSO-based approach to fire flame detection using K-Medoids clustering | |
| CN110135269B (en) | Fire image detection method based on mixed color model and neural network | |
| CN105404847B (en) | A kind of residue real-time detection method | |
| CN103942557B (en) | A kind of underground coal mine image pre-processing method | |
| CN113537099B (en) | Dynamic detection method for fire smoke in highway tunnel | |
| CN105260731A (en) | Human face living body detection system and method based on optical pulses | |
| CN112069975A (en) | Comprehensive flame detection method based on ultraviolet, infrared and vision | |
| CN110222644A (en) | Forest fire smoke detection method based on image segmentation | |
| CN108074234A (en) | A kind of large space flame detecting method based on target following and multiple features fusion | |
| CN104616006B (en) | A kind of beard method for detecting human face towards monitor video | |
| CN108257347B (en) | Flame image sequence classification method and device by using convolutional neural network | |
| CN108921215A (en) | A kind of Smoke Detection based on local extremum Symbiotic Model and energy spectrometer | |
| CN107301653A (en) | Video image fire disaster flame detection method based on BP neural network | |
| CN108171241B (en) | IFCS/Otsu-based IOFR flame identification method | |
| Yoshinaga et al. | Object detection based on spatiotemporal background models | |
| CN106570490A (en) | Pedestrian real-time tracking method based on fast clustering | |
| TW201308254A (en) | Motion detection method for comples scenes | |
| CN110276776A (en) | A kind of adaptive targets detection method based on SPCNN | |
| Chen et al. | Fire detection using spatial-temporal analysis | |
| Filippidis et al. | Fusion of intelligent agents for the detection of aircraft in SAR images | |
| CN107274374A (en) | A kind of smoke monitoring method based on computer vision technique | |
| CN107729811B (en) | Night flame detection method based on scene modeling | |
| Asatryan et al. | Method for fire and smoke detection in monitored forest areas |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171027 |
|
| RJ01 | Rejection of invention patent application after publication |