CN110334660A - A kind of forest fire monitoring method based on machine vision under the conditions of greasy weather - Google Patents
A kind of forest fire monitoring method based on machine vision under the conditions of greasy weather Download PDFInfo
- Publication number
- CN110334660A CN110334660A CN201910612245.2A CN201910612245A CN110334660A CN 110334660 A CN110334660 A CN 110334660A CN 201910612245 A CN201910612245 A CN 201910612245A CN 110334660 A CN110334660 A CN 110334660A
- Authority
- CN
- China
- Prior art keywords
- image
- forest fire
- smog
- formula
- defogging
- 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
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000012544 monitoring process Methods 0.000 title claims abstract description 37
- 239000000779 smoke Substances 0.000 claims abstract description 31
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 29
- 238000001514 detection method Methods 0.000 claims abstract description 28
- 239000003595 mist Substances 0.000 claims abstract description 24
- 230000003287 optical effect Effects 0.000 claims abstract description 20
- 238000012545 processing Methods 0.000 claims abstract description 18
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 230000005540 biological transmission Effects 0.000 claims description 7
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 2
- 230000008030 elimination Effects 0.000 claims description 2
- 238000003379 elimination reaction Methods 0.000 claims description 2
- 239000008264 cloud Substances 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 abstract description 2
- 238000010835 comparative analysis Methods 0.000 abstract 1
- 238000002474 experimental method Methods 0.000 abstract 1
- 238000012795 verification Methods 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 6
- 238000009792 diffusion process Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 239000000284 extract Substances 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 235000019504 cigarettes Nutrition 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000029052 metamorphosis Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/269—Analysis of motion using gradient-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B17/00—Fire alarms; Alarms responsive to explosion
- G08B17/005—Fire alarms; Alarms responsive to explosion for forest fires, e.g. detecting fires spread over a large or outdoors area
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of forest fire monitoring method based on video that cloud and mist and Smoke Detection are gone in combination.Firstly, extracting several frame images in video carries out defogging processing to sample image using the defogging algorithm based on Haze-Line as sample image.Then, Smoke Detection is carried out using the Smoke Detection algorithm based on Horn-Schunck optical flow method, and chooses suitable threshold using maximum variance between clusters to remove influence of the pixel qualities difference to Smoke Detection between adjacent two field pictures.Finally, utilizing the correctness of diffusivity analysis verification algorithm.Emulation experiment and comparative analysis the result shows that, context of methods is capable of detecting when the trend that smoke region gradually increases at any time, to obtain the testing result of accurate forest fire.The forest fire monitoring under the conditions of the greasy weather can be effectively performed in this method, have higher accuracy, validity and robustness.
Description
Technical field
The invention belongs to fire monitoring technical fields, and in particular to the Forest Fire based on machine vision under the conditions of a kind of greasy weather
Calamity monitoring method.
Background technique
(such as greasy weather, rainy day etc.) can generate influence to a certain extent to forest fire monitoring under severe weather conditions, very
It is judged by accident to being likely to result in.Research for the forest fire monitoring technology under severe weather conditions at this stage is simultaneously few, therefore
Influence for the reduction greasy weather to forest fire monitoring herein proposes the defogging algorithm based on Haze-Line and is based on Horn-
The forest fire monitoring method that the Smoke Detection algorithm of Schunck optical flow method blends.
China is the country that a forest-covered area lacks more Forest Fire Disasters, therefore in order to realize that forest fire is early sent out
Target that is existing, early saving, the early warning and monitoring of forest fire have caused people and have adequately paid attention to[1]。
In terms of China's video Forest Fire Monitoring status, using based in a manner of operator on duty's visual monitoring.But artificial observation regards
The efficiency of frequency monitored picture is not high, and it is gloomy only to monitor that a large amount of image and monitored picture are not able to satisfy round the clock by operator on duty
Woods fire monitoring is efficient, real-time demand[2].Especially severe weather conditions can bring bigger difficulty to monitoring.And it is based on machine
The forest fire monitoring of device vision is due to can effectively find early stage forest fire, it has also become the research heat of field of forest fire prevention
One of point.Forest fire monitoring technology based on video be collection video image processing, Video Remote monitoring, machine recognition technology and
Computer technology is in the comprehensive monitoring mode of one[3].It is that digitlization will be carried out by video imaging apparatus captured image first
Reason, and by Pattern recognition and image processing technology extract in doubtful fire generation image area-of-interest (flame region or
Doubtful generation conflagration area), then, by discriminant classification, the methods of feature extraction determines whether the region that is taken has generated
Fire realizes intellectual monitoring and the early warning of forest fire with this[2].It is advantageous that monitoring range is more widened than traditional sensors
Extensively, the not too many particular/special requirement of camera installation site, it is smaller etc. that detection system builds difficulty[4]。
Fire defector and Smoke Detection can be classified as generally, based on the forest fire detection method of video.Due to flame
It is very small and blocked in forest fire by trees, therefore smog is the good index of forest fire[5].Domestic and international video
Forest fire monitoring technology was risen at the beginning of 21 century, and regard the smog of outdoor fire and fire defector identification as research object.
Document[6]Successfully forest fire feature is extracted from smog image with the methods of time domain wavelet transformation, elimination background.
But since the huge calculation amount of wavelet transformation causes to handle overlong time, it is unable to satisfy the real-time of forest fire monitoring.Text
It offers[7]Median background calculus of finite differences is carried out to doubtful smoke region and extracts monitoring moving target, and carries out reverse neural network, mostly spy
Sign fusion is extracted so that carrying out smog identification etc. all obtains good performance.Document[8]It is primarily based on Weber('s)law analysis pair
Than degree, then the transformed energy feature of analysis wavelet, finally calculates the moving characteristic of doubtful cigarette district.If gray value changes
Become, the background of next frame will be estimated referring to the original background of present frame and next frame image.Existing gray difference detection zone
Domain histogram judges detection method and according to smog color characteristic, detection method of wavelet transformation etc., will usually wave
It is smog that cloud and mist etc., which interferes erroneous detection,[10].Document[9]Propose the Gabor network for smog identification.Firstly, building one
The concentration response diagram that Gabor convolution unit exports is carried out across channel coding and counts histogram spy by Gabor convolution unit
Sign can form a feedforward network structure by stacking above-mentioned Gabor basal layer, and each layer of feature, which is joined end to end, can be obtained
The multi-layer feature of smog.But since the training sample of forest fire is less, cause this method that can not be widely applied to respectively
In the different forest environment of class.Therefore herein for the monitoring of forest fire under the conditions of the greasy weather, cloud and mist and Smoke Detection two will be removed
Kind of method effective integration proposes the forest fire detection method based on video under the conditions of a kind of greasy weather.
Cloud and mist is gone mainly to use the method based on image restoration at present.Document[11]It observes outdoor in bright day gas
At least one obvious dark Color Channel of object proposes a kind of dark model then to estimate transmission plot.This method
Defect is that generalization ability difference is not applied for various scenes.Document[12]It is artificially exposed by a series of operation of gamma corrections
Obtained multiple-exposure image is combined into fogless result by multiple dimensioned Laplce's hybrid plan by original blurred picture.Text
It offers[13]Optimize transmissivity using regularization, dark channel image merged to acquisition fusion dark with Morphological Gradient linearity,
Adaptive Gauss weight parameter is constructed by the air light value of atmospherical scattering model and reparation, then to the dark of fusion
Image carries out processing pixel-by-pixel and finally restores fog free images to obtain thick transmissivity.It is more multiple that this method compares other methods
It is miscellaneous, it is longer the time required to processing.Document[14]It is proposed that a fogless color image can be replaced with several hundred a rgb values
Image after mist.Document[15]The concept of Haze-line is clustered and proposed first on this basis to the Similar color point in image.
This method can preferably restore the color of fog free images, therefore can get preferably in forest compared to other defogging methods
Defog effect.So the defogging algorithm based on Haze-Line is used to carry out defogging processing herein.
Summary of the invention
Since burning can generate incessantly partial size in 25nm-100 μm, temperature in 100-1000 DEG C of range when fire occurs
Tiny particles, be academicly known as being corpuscular cloud, i.e. vision smog.These corpuscular clouds burning generate heat effect under,
It can be moved as a certain specific direction, the forest fire smoke of early stage and cloud and mist have a visibly different feature, i.e. smog is from bottom
It is mobile to top.Secondly, the behavioral characteristics of forest fires smog have the characteristics that unstructured, the movement of particle shows a kind of nothing
Rule variation, and the metamorphosis of atmosphere cloud and mist is then relatively slow.So selecting diffusion velocity and the side of smog and cloud and mist herein
To difference as feature, determined using the Vector Message that light stream vector is reflected.But at single use Smoke Detection algorithm
It is simultaneously unreliable to manage forest fire video, to avoid the other interference of cloud and mist and still life etc., by doubtful generation smog picture into
Row go cloud and mist to handle, can be more accurate judge whether there is fire behavior.This method is divided into three steps, is primarily based on Haze-
Doubtful generation smog image is carried out cloud and mist and handled by Line defogging algorithm.Secondly according to based on Horn-Schunck optical flow method
Smoke Detection algorithm identifies the smog in forest fire image.Fire monitoring is finally carried out, method is using binaryzation to smog
The analysis of being diffused property.
The technical scheme is that a kind of forest fire monitoring method under the conditions of greasy weather based on machine vision, including with
Lower step:
(1) it doubtful generation smog image is subjected to cloud and mist handles according to based on Haze-Line defogging algorithm;
(2) according to the smog in the Smoke Detection algorithm identification forest fire image based on Horn-Schunck optical flow method;
(3) fire monitoring is finally carried out, method is to analyze using binaryzation being diffused property of smog.
Doubtful generation smog image is carried out at cloud and mist according to based on Haze-Line defogging algorithm in the step (1)
Reason includes the following contents:
It can be indicated by the model of the foggy image of Influence of cloud are as follows:
I (x)=t (x) J (x)+[1-t (x)] A (1)
A. Haze-Line is clustered, I is definedAAre as follows:
IA(x)=I (x) (2)-A
Wherein: I (x) is foggy image, and IA (x) indicates the color pixel values of foggy image and the difference of atmosphere luminous intensity;
That is, 3D RGB coordinate system is translated, so that atmosphere light A is in origin;
Bringing formula (1) into formula (2) can obtain:
IA(x)=t (x) [J (x)-A] (3)
B. estimate transmissivity:
In the formula (4) and formula (5) defined by J and A, γ (x) depends on object distance, transmissivity t (x) and object to phase
The distance of machine is related;
So γ (x) is indicated are as follows:
γ (x)=t (x) | | J (x)-A | |, 0≤t (x)≤1 (4)
Therefore, as t=1, radius is up toThen it is based on Haze-Line maximum radius γmaxTransmission
Expression formula are as follows: t (x)=γ (x)/γmax;
Maximum radius γmaxMathematic(al) representation are as follows:
Wherein: H Haze-Line, and the initially mathematic(al) representation of perspective rate are as follows:
Since radiance J is non-negative (that is, J >=0), formula (11) provides the lower limit of transmissivity:
Compare transmission lower limit and initial transmission, and takes in the two biggish one, expression formula are as follows:
C. regularization:
The initial transmission calculated above is based on entire flat image, in order to increase the accuracy of calculating using neighbours
The pixel in domain is balanced;
Minimize below in relation toFunction, expression formula are as follows:
D. defogging:
Final step carries out defogging, the transmissivity being calculated by formula (9)It brings formula (10) into and calculates defogging figure
Picture:
The step (2) is according in the Smoke Detection algorithm identification forest fire image based on Horn-Schunck optical flow method
Smog include the following contents:
Horn-Schunck optical flow method introduces the global restriction of smoothness on the flow field of calculating, and this method will be desired
Vector field h is defined as the minimum value of a certain energy functional K (h);
Energy functional is constituted by two, as shown in formula (11);
The data accessory item that energy functional provides and the regular terms based on optical flow gradient:
K (h)=∫ ∫Ω((Ixu+Iyv+It)+α2(|Δu|2+|Δv|2))dxdy (11)
Wherein: the horizontal direction light stream value u at u representative image pixel (i, j)ij, at v representative image pixel (i, j)
Vertical direction light stream value vij, IxAnd IyImage is respectively indicated about the gradient and image of x (horizontal direction) about y (Vertical Square
To) gradient, ItIndicate that the derivative between 2 frame images about the time, α are to join for controlling with the weight of the smooth item of optical flow constraint
Number, in which:
Step (3) smoke monitoring includes the following contents to the analysis of being diffused property of smog using binaryzation:
The calculating of growth rate and relative growth rate is carried out to the obtained doubtful smoke region by optical flow method processing:
Wherein: Δ PiFor growth rate, Pi+kFor the sum of the target white pixel of the i-th+k frame, PiFor the object pixel of the i-th frame
The sum of point, k is interval frame number, IPFor relative growth rate.
Beneficial effects of the present invention: many chaff interferents of cloud and mist and similar smog significantly reduce forest fire identification
Precision, this paper presents a kind of forest fire monitoring methods under the conditions of greasy weather based on video.On the one hand, it is based on using introducing
The defogging algorithm of Haze-Line carries out defogging processing to the picture of the Sample video interception of forest fire, to exclude cloud and mist pair
The influence of Smoke Detection.On the other hand, the Horn-Schunck light stream of light stream vector length is defined by maximum variance between clusters
Method carries out smog identification to going result after cloud and mist.Finally whether fire is analyzed to identify using being diffused property of binary conversion treatment
It generates.Smog the experimental results showed that, this method can significantly improve the Accuracy and high efficiency of smoke detection.
Detailed description of the invention
The comparison of Fig. 1: four kinds of defogging methods:
(a) original image;
(b)multi-scale Laplacian;
(c) dark channel prior;
(d) this paper algorithm;
Fig. 2: defogging result is shown: (a) first frame;(b) the second frame;(c) third frame;(d) the 4th frame;
Fig. 3: optical flow method processing result comparison diagram under non smoke state:
(a) video interception when for one group without defogging under two frame non smoke states;
It (b) is processing result;
(a) it is identical one group of original image with (c), but is cut through the video under defogging treated non smoke state
Figure (d) is processing result;
Fig. 4: optical flow method processing result comparison diagram under smog disperse state:
(a) for one group not defogging when two frame smog disperse states under video interception;
It (b) is processing result;
(a) it is identical one group of original image with (c), but is cut through the video under defogging treated smog disperse state
Figure;
It (d) is processing result;
Fig. 5: binary conversion treatment result.
Specific embodiment
Below by attached drawing, the present invention is further illustrated.The embodiment of the present invention is to preferably make this field
Technical staff more fully understand the present invention, not to the present invention make any limitation.
Embodiment 1: the defogging process based on Haze-Line defogging algorithm:
A frame image is chosen from the Sample video of forest fire carries out defogging processing, and and multi-scale
The single image defogging algorithm of Laplacian algorithm, dark channel prior is compared, and the result after defogging is as shown in Figure 1.
Image after multi-scale Laplacian algorithm removal cloud and mist seems dim, and smog is made to become light, influences to judge
As a result.Dark channel prior defogging algorithm has certain defog effect, also can preferably retain smog, but image totally whitens, can
To see that remote areas has apparent fog to remain.And in this paper defogging algorithm, global linear contrast is executed in output
It stretches, the pixel value of shearing percent 0.5 in shadow and highlight, so that result is more bright, the contrast of smog is higher.Phase
Than under, algorithm used herein has restored more details, and due to this method be based on pixel rather than it is block-based,
This makes the algorithm process speed of this paper faster, more stable, and in most cases can effectively handle noise, thus
Ideal defog effect is obtained.
The smog as caused by fire is in white colour in the picture, so each pixel in restoring foggy image
When rgb value, smog will not be caused excessively to influence, can preferably retain smog, so using the defogging algorithm to doubtful production
Raw smog picture carries out cloud and mist and handles.Fig. 2 is that video interception is gone under one group of interval 40 seconds four frame smog disperse states
Mist result.
Embodiment 2: the Smoke Detection process based on Horn-Schunck algorithm:
During using optical flow method identification smog, the diffusion velocity of smog is differentiated as main feature, is
Influence of the pixel qualities difference to Smoke Detection between removal two field pictures, will treated that sample image utilizes through optical flow method
Maximum variance between clusters retain suitable light stream vector value.When generating smog in sample image, light can be passed through in effect picture
The length of flow vector and the difference in direction discriminate whether the position that smog generates and smog generates.Two frames are free of in selecting video
The forest image of smog and two frame forest images containing smog use light to the image before defogging and the image after defogging respectively
Stream method carries out Smoke Detection processing.Since the presence of cloud and mist causes processing result much noise occur, through defogging treated knot
Fruit, which has, to be significantly improved.Compare Fig. 3 and Fig. 4 can be seen that and carry out to before two groups of defoggings with the optical flow method testing result after defogging
Compare, it can be found that the result effect for carrying out optical flow method Smoke Detection after defogging is more obvious, successfully avoids cloud and mist and deposit
In generated noise.
Embodiment 3: fire monitoring:
The diffusion property of smog can be used as effective foundation of identification Forest Fire smog, and in order to which more objectively verifying is calculated
The accuracy of method determines whether smog diffusion to being diffused property of the result analysis obtained after Smoke Detection.Specific method is
After Smoke Detection, the region spread to four frames there are smog carries out binary conversion treatment, then calculate white pixel point number and
Growth rate.
It should be understood that embodiment and example discussed herein simply to illustrate that, to those skilled in the art
For, it can be improved or converted, and all these modifications and variations all should belong to the protection of appended claims of the present invention
Range.
Bibliography:
[1] Zhang Heng, Ma Yunjia, Peng Xujian wait the North China .2003-2016 area forest fire space-time characteristic to study
[J] Xibei Forest College journal, 2019 (1): 24.
[2] forest fire of the Zhan Qi based on video image monitors the Chengdu identification technology research [D] automatically: electronics technology is big
It learns, 2017.
[3]Zhao Y,Li Q,Gu Z.Early smoke detection of forest fire video using
CS Adaboost algorithm[J].Optik-International Journal for Light and Electron
Optics,2015,126(19):2121-2124.
[4] Zeng Zhiqiang, He little Dong, Wang Ying wait to mix big data analysis system based on the forest fire of Hadoop and Spark
System research [J] World Forestry research, 2018,31 (2): 55-59.
[5] present Research [J] the China image graphics journal of Luo Sheng, Jiang Y Z. video detection smog, 2013,18
(10):1225-1236.
[6]B.U.Toreyin,Y.Dedeoglu,A.E.Cetin,Dynamic texture detection
segmentationand analysis[J].Conf.Image and Video Retrieval 2007:131–134.
[7]D,M,L.Integration of forest fire video
monitoring system and geographic information system[C].2009 International
Symposium ELMAR.IEEE,2009:49-52.
[8]Ye S P,Bai Z C,Chen H F,et al.An effective algorithm to detect
both smoke and flame using color and wavelet analysis[J].Pattern Recognition
and Image Analysis,2017,27(1):131-138.
[9] Yuan Feiniu, Xia Xue, Li Gang wait towards Gabor network [J] the China image of smog identification and Texture classification
Figure journal, 2019,24 (2): 269-281.
[10] Wu Mengyu, Luo Qinjuan;Forest fire smoke image recognition algorithm [J] north of the Han Ning based on feature of doing more physical exercises
Capital Forestry University journal, 2013,35 (3): 154-158.
[11]K.He,J.Sun,and X.Tang.Single image haze removal using dark
channel prior[J].IEEE Trans on PAMI,2011,33(12):2341–2353.
[12]Galdran,Adrian.Image dehazing by artificial multiple-exposure
image fusion.Signal Processing,2018,149:135-147.
[13] Zhang Chen, Yang Yan weight single image defogging algorithm [J] photonics journal of dark based on fusion with Gauss,
2018,48(1):110002-0110002.
[14]M.T.Orchard and C.A.Bouman.Color quantization of images.Signal
Processing,IEEE Transactions.1991,39(12):2677–2690.
[15]Berman D,Treibitz T,Avidan S.Non-local image dehazing[C]
.ComputerVision and Pattern Recognition.IEEE,2016:1674-1682.
Claims (4)
1. a kind of forest fire monitoring method under the conditions of greasy weather based on machine vision, it is characterised in that: the following steps are included:
(1) doubtful generation smog image cloud and mist is carried out based on Haze-Line defogging algorithm to handle;
(2) smog in the Smoke Detection algorithm identification forest fire image based on Horn-Schunck optical flow method;
(3) fire monitoring is finally carried out, method is to analyze using binaryzation being diffused property of smog.
2. the forest fire monitoring method under the conditions of the greasy weather according to claim 1 based on machine vision, it is characterised in that:
In the step (1) according to doubtful generations smog image carried out by cloud and mist based on Haze-Line defogging algorithm handle include with
Lower content:
It can be indicated by the model of the foggy image of Influence of cloud are as follows:
I (x)=t (x) J (x)+[1-t (x)] A (1)
A. Haze-Line is clustered, I is definedAAre as follows:
IA(x)=I (x) (2)-A
Wherein: I (x) is foggy image, and IA (x) indicates the color pixel values of foggy image and the difference of atmosphere luminous intensity;
That is: 3D RGB coordinate system is translated, so that atmosphere light A is in origin;
Bringing formula (1) into formula (2) can obtain:
IA(x)=t (x) [J (x)-A] (3)
B. estimate transmissivity:
In the formula (4) and formula (5) defined by J and A, γ (x) depends on object distance, and transmissivity t (x) and object arrive camera
Distance is related;
So γ (x) is indicated are as follows:
γ (x)=t (x) | | J (x)-A | |, 0≤t (x)≤1 (4)
Therefore, as t=1, radius is up toThen it is based on Haze-Line maximum radius γmaxTransmission expression
Formula are as follows: t (x)=γ (x)/γmax;
Maximum radius γmaxMathematic(al) representation are as follows:
Wherein: H Haze-Line, and the initially mathematic(al) representation of perspective rate are as follows:
Since radiance J is non-negative (that is, J >=0), formula (11) provides the lower limit of transmissivity:
Compare transmission lower limit and initial transmission, and takes in the two biggish one, expression formula are as follows:
C. regularization:
The initial transmission calculated above is based on entire flat image, in order to increase the accuracy of calculating using four neighborhoods
Pixel is balanced;
Minimize below in relation toFunction, expression formula are as follows:
D. defogging:
Final step carries out defogging, the transmissivity being calculated by formula (9)It brings formula (10) into and calculates mist elimination image:
3. the forest fire monitoring method under the conditions of the greasy weather according to claim 1 based on machine vision, it is characterised in that:
The step (2) is according to the smog packet in the Smoke Detection algorithm identification forest fire image based on Horn-Schunck optical flow method
Include the following contents:
Desired vector field h is defined as to the minimum value of a certain energy functional K (h);
Energy functional is constituted by two, as shown in formula (11);
The data accessory item that energy functional provides and the regular terms based on optical flow gradient:
K (h)=∫ ∫Ω((Ixu+Iyv+It)+α2(|Δu|2+|Δv|2))dxdy (11)
Wherein: the horizontal direction light stream value u at u representative image pixel (i, j)ij, hanging down at v representative image pixel (i, j)
Histogram is to light stream value vij, IxAnd IyImage is respectively indicated about the gradient and image of x (horizontal direction) about y's (vertical direction)
Gradient, ItIndicate the derivative between 2 frame images about the time, α be for control with the weight parameter of the smooth item of optical flow constraint,
In:
4. the forest fire monitoring method under the conditions of the greasy weather according to claim 1 based on machine vision, it is characterised in that:
Step (3) smoke monitoring includes the following contents to the analysis of being diffused property of smog using binaryzation:
The calculating of growth rate and relative growth rate is carried out to the obtained doubtful smoke region by optical flow method processing:
Wherein: Δ PiFor growth rate, Pi+kFor the sum of the target white pixel of the i-th+k frame, PiFor the target pixel points of i-th frame
With k is interval frame number, IPFor relative growth rate.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910612245.2A CN110334660A (en) | 2019-07-08 | 2019-07-08 | A kind of forest fire monitoring method based on machine vision under the conditions of greasy weather |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910612245.2A CN110334660A (en) | 2019-07-08 | 2019-07-08 | A kind of forest fire monitoring method based on machine vision under the conditions of greasy weather |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110334660A true CN110334660A (en) | 2019-10-15 |
Family
ID=68143315
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910612245.2A Pending CN110334660A (en) | 2019-07-08 | 2019-07-08 | A kind of forest fire monitoring method based on machine vision under the conditions of greasy weather |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110334660A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111047565A (en) * | 2019-11-29 | 2020-04-21 | 南京恩博科技有限公司 | Method, storage medium and equipment for forest cloud image segmentation |
CN111274896A (en) * | 2020-01-15 | 2020-06-12 | 深圳市守行智能科技有限公司 | Smoke and fire identification algorithm |
CN111414969A (en) * | 2020-03-26 | 2020-07-14 | 西安交通大学 | Smoke detection method in foggy environment |
CN111523528A (en) * | 2020-07-03 | 2020-08-11 | 平安国际智慧城市科技股份有限公司 | Strategy sending method and device based on scale recognition model and computer equipment |
CN111784994A (en) * | 2020-07-14 | 2020-10-16 | 中国民航大学 | Fire detection method and device |
CN112132124A (en) * | 2020-11-30 | 2020-12-25 | 江苏久智环境科技服务有限公司 | Real-time VR video detection method for garden monitoring |
CN112614302A (en) * | 2020-12-03 | 2021-04-06 | 杭州海康微影传感科技有限公司 | Fire detection method, device and system and electronic equipment |
CN115100555A (en) * | 2022-07-27 | 2022-09-23 | 南京理工大学 | Forest fire smoke detection method based on multi-motion characteristics in cloud and fog environment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050069207A1 (en) * | 2002-05-20 | 2005-03-31 | Zakrzewski Radoslaw Romuald | Method for detection and recognition of fog presence within an aircraft compartment using video images |
KR20140111426A (en) * | 2013-03-11 | 2014-09-19 | 주식회사 브이아이티시스템 | System for enhancing the foggy luminance image using the estimation model of haze removal |
CN104715556A (en) * | 2015-04-02 | 2015-06-17 | 无锡桑尼安科技有限公司 | Fire alarm method based on aerial photography |
-
2019
- 2019-07-08 CN CN201910612245.2A patent/CN110334660A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050069207A1 (en) * | 2002-05-20 | 2005-03-31 | Zakrzewski Radoslaw Romuald | Method for detection and recognition of fog presence within an aircraft compartment using video images |
KR20140111426A (en) * | 2013-03-11 | 2014-09-19 | 주식회사 브이아이티시스템 | System for enhancing the foggy luminance image using the estimation model of haze removal |
CN104715556A (en) * | 2015-04-02 | 2015-06-17 | 无锡桑尼安科技有限公司 | Fire alarm method based on aerial photography |
Non-Patent Citations (4)
Title |
---|
BERTHOLD K.P. HORN等: "Determining Optical Flow", 《ARTIFICIAL INTELLIGENCE》 * |
DANA BERMAN等: "Non-Local Image Dehazing", 《2016IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION(CVPR)》 * |
E. MOMMA等: "Detection of fog and smoke particles with discrete near infrared light", 《ELECTRONICS AND COMMUNICATIONS IN JAPAN》 * |
吴梦宇等: "基于多运动特征的森林火灾烟雾图像识别算法", 《北京林业大学学报》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111047565A (en) * | 2019-11-29 | 2020-04-21 | 南京恩博科技有限公司 | Method, storage medium and equipment for forest cloud image segmentation |
CN111274896A (en) * | 2020-01-15 | 2020-06-12 | 深圳市守行智能科技有限公司 | Smoke and fire identification algorithm |
CN111274896B (en) * | 2020-01-15 | 2023-09-26 | 深圳市守行智能科技有限公司 | Smoke and fire recognition algorithm |
CN111414969A (en) * | 2020-03-26 | 2020-07-14 | 西安交通大学 | Smoke detection method in foggy environment |
CN111523528A (en) * | 2020-07-03 | 2020-08-11 | 平安国际智慧城市科技股份有限公司 | Strategy sending method and device based on scale recognition model and computer equipment |
CN111523528B (en) * | 2020-07-03 | 2020-10-20 | 平安国际智慧城市科技股份有限公司 | Strategy sending method and device based on scale recognition model and computer equipment |
CN111784994A (en) * | 2020-07-14 | 2020-10-16 | 中国民航大学 | Fire detection method and device |
CN112132124A (en) * | 2020-11-30 | 2020-12-25 | 江苏久智环境科技服务有限公司 | Real-time VR video detection method for garden monitoring |
CN112614302A (en) * | 2020-12-03 | 2021-04-06 | 杭州海康微影传感科技有限公司 | Fire detection method, device and system and electronic equipment |
CN112614302B (en) * | 2020-12-03 | 2022-10-04 | 杭州海康微影传感科技有限公司 | Fire detection method, device and system and electronic equipment |
CN115100555A (en) * | 2022-07-27 | 2022-09-23 | 南京理工大学 | Forest fire smoke detection method based on multi-motion characteristics in cloud and fog environment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110334660A (en) | A kind of forest fire monitoring method based on machine vision under the conditions of greasy weather | |
CN109543606B (en) | Human face recognition method with attention mechanism | |
Tao et al. | Smoke detection based on deep convolutional neural networks | |
Chakma et al. | Image-based air quality analysis using deep convolutional neural network | |
CN107145846B (en) | A kind of insulator recognition methods based on deep learning | |
CN107609470B (en) | Method for detecting early smoke of field fire by video | |
CN110909690B (en) | Method for detecting occluded face image based on region generation | |
CN106296612B (en) | Hierarchical monitoring video sharpening system and method for image quality evaluation and weather condition guidance | |
CN104050471B (en) | Natural scene character detection method and system | |
CN108038846A (en) | Transmission line equipment image defect detection method and system based on multilayer convolutional neural networks | |
CN108985192A (en) | A kind of video smoke recognition methods based on multitask depth convolutional neural networks | |
CN107609512A (en) | A kind of video human face method for catching based on neutral net | |
CN110490043A (en) | A kind of forest rocket detection method based on region division and feature extraction | |
CN106339657B (en) | Crop straw burning monitoring method based on monitor video, device | |
CN111046827A (en) | Video smoke detection method based on convolutional neural network | |
CN111666944A (en) | Infrared weak and small target detection method and device | |
CN103729620B (en) | A kind of multi-view pedestrian detection method based on multi-view Bayesian network | |
CN114202646A (en) | Infrared image smoking detection method and system based on deep learning | |
CN106778488B (en) | Low-light (level) smog video detecting method based on image correlation | |
CN111582074A (en) | Monitoring video leaf occlusion detection method based on scene depth information perception | |
CN111046728A (en) | Straw combustion detection method based on characteristic pyramid network | |
Su et al. | A new local-main-gradient-orientation HOG and contour differences based algorithm for object classification | |
CN114627269A (en) | Virtual reality security protection monitoring platform based on degree of depth learning target detection | |
CN111461076A (en) | Smoke detection method and smoke detection system combining frame difference method and neural network | |
CN110751667A (en) | Method for detecting infrared dim small target under complex background based on human visual system |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191015 |