CN110473375A - Monitoring method, device, equipment and the system of forest fire - Google Patents
Monitoring method, device, equipment and the system of forest fire Download PDFInfo
- Publication number
- CN110473375A CN110473375A CN201910750219.6A CN201910750219A CN110473375A CN 110473375 A CN110473375 A CN 110473375A CN 201910750219 A CN201910750219 A CN 201910750219A CN 110473375 A CN110473375 A CN 110473375A
- Authority
- CN
- China
- Prior art keywords
- fire
- target
- target identification
- identification object
- forest
- 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
- 238000012544 monitoring process Methods 0.000 title claims abstract description 64
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000001514 detection method Methods 0.000 claims abstract description 69
- 230000008569 process Effects 0.000 claims abstract description 9
- 238000012806 monitoring device Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 19
- 238000003062 neural network model Methods 0.000 claims description 17
- 230000001629 suppression Effects 0.000 claims description 12
- 230000005540 biological transmission Effects 0.000 claims description 4
- 238000004891 communication Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 2
- 230000005055 memory storage Effects 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 abstract description 9
- 238000013473 artificial intelligence Methods 0.000 abstract description 8
- 230000000694 effects Effects 0.000 abstract description 5
- 238000012360 testing method Methods 0.000 abstract description 3
- 230000007306 turnover Effects 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 10
- 239000000779 smoke Substances 0.000 description 10
- 230000006870 function Effects 0.000 description 7
- 230000006399 behavior Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 235000013399 edible fruits Nutrition 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000009434 installation Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 241000406668 Loxodonta cyclotis Species 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 235000019504 cigarettes Nutrition 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000009432 framing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 239000003595 mist Substances 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- 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
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B17/00—Fire alarms; Alarms responsive to explosion
- G08B17/12—Actuation by presence of radiation or particles, e.g. of infrared radiation or of ions
- G08B17/125—Actuation by presence of radiation or particles, e.g. of infrared radiation or of ions by using a video camera to detect fire or smoke
Landscapes
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Alarm Systems (AREA)
- Fire-Detection Mechanisms (AREA)
Abstract
This application involves a kind of monitoring method of forest fire, device, equipment and systems.Wherein this method comprises: obtaining the video information that video camera is sent；Each frame picture is extracted from the video information and is sequentially input to fire detection model trained in advance, and the quantitative value of target identification object in each frame picture is respectively obtained；Wherein, the target identification object includes target flame object and target smog object；If the quantitative value of the target identification object is greater than 0, it is concluded that the monitoring result of fire has occurred.It is arranged such, monitor video picture is identified by artificial intelligence technology, as long as the sample size in detection model is enough, the result so identified is just accurate enough, detection process carries out automatically simultaneously, it operates, therefore can be saved labour turnover under the premise of guaranteeing that monitoring effect is good without personnel.In addition, testing result can be made more reliable in conjunction with two kinds of identification objects of flame and smog.
Description
Technical field
This application involves technical field of computer vision more particularly to a kind of monitoring methods of forest fire, device, equipment
And system.
Background technique
The forest reserves are very important natural resources in today's society, and forest fire once occurs will be to forest sheet
Other animal and plant resources and atmospheric environment in body, forest etc. bring serious destruction, while large-scale forest fire can also
Threat can be brought to the security of the lives and property of neighbouring personnel, therefore forest fire protection is all paid much attention in various regions, but, no matter forest
How perfect fire prevention done, it is also difficult to avoid the generation of forest fire completely, therefore, the initial moment of fire can occur in forest
Discovery is just significant in time, and such forester can prevent fire behavior from spreading in time and extinguish the blaze rapidly, to protect money
Source, environment and personal safety etc..
Currently, including the modes such as fire defector and Smoke Detection to the monitoring of fire in various situations.Smoke Detection is usual
It is realized using smoke alarm, but the smokescope that smoke alarm alarm needs is larger, therefore applies in open forest
Detection effect in environment is unsatisfactory.And fire defector includes being carried out by manually checking the video pictures of video camera shooting
Monitoring, but forest biggish for area, since the number of cameras for needing to be arranged is more, it is therefore desirable to largely manually look into
See monitoring screen, it is higher so as to cause cost of labor.
Summary of the invention
The application provides monitoring method, device, equipment and the system of a kind of forest fire, to solve in the related technology to gloomy
The problems such as effect existing for the monitoring method of forest fires calamity is undesirable higher with the cost of labor of needs.
The above-mentioned purpose of the application is achieved through the following technical solutions:
In a first aspect, the embodiment of the present application provides a kind of monitoring method of forest fire, comprising:
Obtain the video information that video camera is sent；Wherein, the video camera setting is aerial in the height of forest, gloomy for shooting
The video information of woods；
Each frame picture is extracted from the video information and is sequentially input to fire detection model trained in advance, difference
Obtain the quantitative value of target identification object in each frame picture；Wherein, the target identification object include target flame object and
Target smog object；
If the quantitative value of the target identification object is greater than 0, the monitoring result that fire has occurred is obtained.
Optionally, described to extract each frame picture from the video information and sequentially input to fire inspection trained in advance
Model is surveyed, the quantitative value of target identification object in each frame picture is respectively obtained, comprising:
Each frame picture is extracted from the video information；
Each frame picture is sequentially input to the fire detection model based on deep neural network model training；
Obtain the quantitative value of target identification object in each frame picture respectively by object detection algorithms.
Optionally, described sequentially input each frame picture to the fire based on deep neural network model training is examined
It surveys before model, further includes:
Each frame picture is scaled presetted pixel；
It is described to sequentially input each frame picture to the fire detection model based on deep neural network model training,
Include:
The each frame picture for being scaled presetted pixel is sequentially input to the fire based on deep neural network model training
Detection model.
Optionally, the method also includes:
If the quantitative value of the target identification object is greater than 0, every place's target identification object is assessed, is obtained respectively
Confidence level scoring, and non-maxima suppression is carried out to target complete identification object；
Compare the size of the confidence level scoring and preset score threshold, if confidence level scoring is less than preset point
Number threshold value, then give up corresponding target identification object；
It obtains in each frame picture by the assessment and remaining target identification object after the non-maxima suppression
Quantitative value obtains all remaining target identification objects and exists if the quantitative value of the remaining target identification object is greater than 0
Position in the frame picture；
Quantitative value, position and the confidence level scoring of the remaining target identification object are exported, while obtaining and having occurred
The monitoring result of fire.
Optionally, the method also includes:
Obtain the target flame object and the target smog object for including in the remaining target identification object
Corresponding quantitative value, position and confidence level scoring；
It exports the target flame object and the corresponding quantitative value of the target smog object, position and confidence level is commented
Point.
Optionally, the training process of the fire detection model includes:
Obtain the fire picture sample of preset quantity；
The fire picture sample is input to the deep neural network model constructed in advance, and to the fire picture sample
Target identification object in this is identified one by one, to obtain the fire detection model.
Optionally, it is described export the monitoring result of fire has occurred after, further includes:
Warning message and the target flame object and the target smog pair are sent to pre-set smart machine
As corresponding quantitative value, position and confidence level score.
Second aspect, the embodiment of the present application also provide a kind of monitoring device of forest fire, which includes:
Module is obtained, for obtaining the video information of video camera transmission；Wherein, the video camera setting is in the high-altitude of forest
In, for shooting the video information of forest；
Detection module, for extracting each frame picture from the video information and sequentially inputting to fire trained in advance
Detection model respectively obtains the quantitative value of target identification object in each frame picture；Wherein the target identification object includes mesh
Mark flame object and target smog object；
Output module obtains the monitoring knot that fire has occurred if the quantitative value for the target identification object is greater than 0
Fruit.
The third aspect, the embodiment of the present application also provide a kind of monitoring device of forest fire, which includes:
Memory and the processor being connected with the memory；
For the memory for storing program, described program is at least used to execute forest fire described in any of the above item
Monitoring method；
The processor is used to call and execute the described program of the memory storage.
Fourth aspect, the embodiment of the present application also provide a kind of monitoring system of forest fire, which includes:
Multiple video cameras of the high aerial different location of forest and the forest with each video camera communication connection are set
The monitoring device of fire.
The technical solution that embodiments herein provides can include the following benefits:
When using technical solution provided by the embodiments of the present application, the view for the video camera shooting being arranged in forest is obtained first
Frequently, the picture in video and as unit of frame is extracted, the picture of extraction is inputted into fire detection model later, to pass through calculating
Machine vision technique (artificial intelligence technology) identifies in the picture of input whether include target identification object (i.e. after generation fire
Flame and smoke characteristics), finally export the result of identification.So set, not using traditional artificial checking monitoring picture to sentence
It is disconnected that fire whether occurs, but monitored picture is identified by artificial intelligence, as long as the sample size in detection model is enough,
The result so identified is just accurate enough, and the detection model learning ability based on artificial intelligence is very strong, therefore with using
The time accuracy of increasingly longer identification also can be higher and higher.That is, using the technical solution of the application to forest fire
When being monitored, it can save labour turnover under the premise of guaranteeing that monitoring effect is good.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
The application can be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the application
Example, and together with specification it is used to explain the principle of the application.
Fig. 1 is a kind of flow diagram of the monitoring method of forest fire provided by the embodiments of the present application；
Fig. 2 is the flow diagram of the monitoring method of another forest fire provided by the embodiments of the present application；
Fig. 3 is a kind of specific implementation procedure schematic diagram of the monitoring method of forest fire provided by the embodiments of the present application；
Fig. 4 is a kind of structural schematic diagram of the monitoring device of forest fire provided by the embodiments of the present application；
Fig. 5 is a kind of structural schematic diagram of the monitoring system of forest fire provided by the embodiments of the present application；
Fig. 6 is the structural schematic diagram of the monitoring system of another forest fire provided by the embodiments of the present application.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistent with the application.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the application.
For forest fire monitoring, the monitor mode of country's mainstream is video monitoring system at present, this is traditional cities prison
The simple extension of control summarizes acquisition video image by microwave, and by being accomplished manually centralized watch, but, direct surveillance is easily made
At naked eyes fatigue, the fire behavior in video is caused to be not easy to be found, in addition, the video line of monitoring center is more, direct surveillance
It not can guarantee and supervise one by one, therefore easily cause and fail to report.That is, the disadvantage of traditional video surveillance is that rate of failing to report is very high.
In view of this, the application provides one kind based on video technique, the Forest Fire of computer vision technique is combined
The monitoring method of calamity, the virtual bench for applying this method and entity device and the monitoring system for realizing this method, it is specific interior
Appearance will be illustrated by following multiple embodiments.
Embodiment one
Referring to Fig. 1, Fig. 1 is a kind of flow diagram of the monitoring method of forest fire provided by the embodiments of the present application.
As shown in Figure 1, method includes the following steps:
S101: the video information that video camera is sent is obtained；Wherein, the video camera setting is aerial in the height of forest, is used for
Shoot the video information of forest；
Specifically, the mode for being similar to traditional video monitoring, manually checking, the present embodiment also need first in forest
It is middle to establish perfect video monitoring system, i.e., a large amount of camera shootings are set in forest high-altitude by modes such as plateau, high tower or electric poles
Head, so that monitoring range be made to cover entire forest location as far as possible.The video pictures of camera shooting can be by wired
Or it is wirelessly transmitted to specific monitoring device, it is handled for analysis.
In some embodiments, monitoring device can be integrated to video camera, i.e., comprising storage in each video camera
There is the memory of relative program and execute the processor of the program, to be the shooting that video can be achieved at the same time in video camera front end
It is handled with analysis.So set, being transmitted to same fixed location relative to traditional all videos for shooting each video camera
For the mode uniformly checked, it is not necessary that the data line or wireless network that are used for data transmission are specially arranged for each video camera,
Therefore installation cost can be reduced during installation, while being also convenient for safeguarding.
S102: each frame picture is extracted from the video information and is sequentially input to fire detection mould trained in advance
Type respectively obtains the quantitative value of target identification object in each frame picture；Wherein the target identification object includes target flame
Object and target smog object；
In the specific implementation, step S102 may is that extracts each frame picture from the video information；It will be described every
One frame picture is sequentially input to the fire detection model based on deep neural network model training；Distinguished by object detection algorithms
Obtain the quantitative value of target identification object in each frame picture.
Specifically, what fire detection model was realized particularly directed to static picture when being detected, therefore, it is necessary to first
Each frame tableaux is first extracted from dynamic video.Above mentioned each frame picture is exactly minimum unit in image animation
Single width image frame, a frame are exactly a static picture.After one frame picture is input to fire detection model, fire detection mould
Whether type in picture to detect input includes target identification object (i.e. flame or cigarette automatically according to preset sample and mark
Mist), if the quantity for counting target identification object comprising if.
Wherein, fire detection model is formed based on deep neural network model training, more specifically, used depth
Neural network can be convolutional neural networks (Convolutional Neural Networks, CNNs), and CNNs is that one kind includes
Convolutional calculation and the feedforward neural network with depth structure, are one of the representative algorithms of deep learning (deep learning),
Certainly, it in addition to this can also be realized using other algorithms of deep learning, to this without limiting.
Further, the training process of fire detection model may is that the fire picture sample for obtaining preset quantity；By institute
It states fire picture sample and is input to the deep neural network model constructed in advance, and the target in the fire picture sample is known
Other object is identified one by one, to obtain the fire detection model.
That is, it is necessary first to prepare a large amount of pictures for occurring to shoot after forest fire or video pictures etc. as sample
This, wherein sample can be oneself shooting and be also possible to obtain by internet or other approach, be then based on common
Deep neural network model, such as Mobilenet0.25 are identified the target identification object in sample one by one, i.e., will be every
Everywhere flame in one width picture (or picture) or the position where smoke characteristics are identified, wherein the work being identified
Work can realize (such as SSD algorithm) by object detection algorithms, object detection be classical problem in computer vision it
One, task is the position for removing to mark objects in images with frame, and provides the classification of object.That is, object detection algorithms
Exactly specific identification object is marked by frame, to provide position and the classification of identification object.
In the specific implementation, when training fire detection model, need to mark the flame and smog in sample in the form of frame,
And provide different signature identifications respectively to flame and smog, it is popular for, i.e., " inform " model being trained to: including
The information of certain signature identification belongs to " flame " and includes that the information of another signature identification belongs to " smog ", thus fire detection
Model can identify pair between identification object by deep neural network model and object detection algorithms " study " this feature
It should be related to；And in the good fire detection model of application training, be based on object detection algorithms detection picture in whether include
Flame or smog (whether being substantially in detection picture comprising corresponding signature identification), if comprising will be in picture by frame
Flame and smog mark (position), and can designate that the identification object in frame is flame or smog (classifying).
In addition, being sequentially input by each frame picture to based on deep neural network model training in some embodiments
It can also include: that each frame picture is zoomed into pixel identical with training sample before fire detection model.
Specifically, in order to shorten the detection time of the fire detection model trained and improve accuracy in detection, in training
Before the fire detection model, the pixel of all fire picture samples can be zoomed to all unanimously, such as 300*300, because
This also needs first to zoom to the pixel of picture to be measured and training sample when being detected using the fire detection model
This is consistent, consequently facilitating fire detection model is detected.
S103: if the quantity of the target identification object is greater than 0, the monitoring result that fire has occurred is obtained.
Specifically, the quantity of target identification object refers to the sum of flame quantity and smog quantity, as long as detecting fire
Flame or smog it is any, that is, be considered as generation fire.It should be noted that since flame and smog are uncountable noun, because
This, the flame quantity and smog quantity being previously mentioned in the application are actually referred to when detecting using object detection algorithms,
The quantity for the frame that flame or smog are marked.
In addition, it should be noted that, forest fire usually has very high concealment, therefore only rely on flame identification fire
There is significant limitation, i.e., is usually in ground location when just occurring due to fire, and video camera is in order to covering biggish model
Enclose, be generally arranged at high aerial, therefore blocking due to trees, fire just occurred and the intensity of a fire it is smaller when video camera be difficult in time
Shooting has occurred that sprawling in actually discovery so as to cause fire.In consideration of it, would generally be with big when since fire occurs
The generation of smog is measured, and smog is shot (relatively fiery earlier because the lighter reason of its density can flow upwards convenient for video camera
Flame), therefore, identify that fire behavior is of great significance by smog when fire occurs.But, traditional video monitoring, artificial
The mode checked is usually predominantly flame characteristic when fire occurs for identification, and can not utilize smoke characteristics well, and reason exists
There is very strong similitude in the visual signature of smog and sky medium cloud, while can also be influenced by the greasy weather, therefore shoots into video
It is visually difficult to accurately distinguish after picture, additionally due to the video line of monitoring center is more, i.e., everyone needs while checking more
A monitored picture, the accuracy and timeliness of manual identified smog further reduce.
Based on this, using in deep neural network algorithm and computer vision technique in the technical solution of the embodiment of the present application
Object detection algorithms, so that identification to flame characteristic and smoke characteristics is realized by artificial intelligence technology, so that identification knot
Fruit has very high robustness relative to conventional method.In addition, by combining two kinds of recognition results of smoke characteristics and flame characteristic,
Keep whole detection result more accurate and reliable.
The technical solution that embodiments herein provides can include the following benefits:
When using technical solution provided by the embodiments of the present application, the view for the video camera shooting being arranged in forest is obtained first
Frequently, the picture in video and as unit of frame is extracted, the picture of extraction is inputted into fire detection model later, to pass through calculating
Machine vision technique (artificial intelligence technology) identifies in the picture of input whether include target identification object (i.e. after generation fire
Flame and smoke characteristics), finally export the result of identification.So set, not using traditional artificial checking monitoring picture to sentence
It is disconnected that fire whether occurs, but monitored picture is identified by artificial intelligence, as long as the sample size in detection model is enough,
The result so identified is just accurate enough, and the detection model learning ability based on artificial intelligence is very strong, therefore with using
The time accuracy of increasingly longer identification also can be higher and higher.That is, using the technical solution of the application to forest fire
When being monitored, it can save labour turnover under the premise of guaranteeing that monitoring effect is good.
In order to improve the forest fire in above-described embodiment monitoring method practicability, the application also provides following improvement
Scheme.
Embodiment two
Referring to Fig. 2, Fig. 2 is the process signal of the monitoring method of another forest fire provided by the embodiments of the present application
Figure.As shown in Fig. 2, method includes the following steps:
S201: the video information that video camera is sent is obtained；Wherein, the video camera setting is aerial in the height of forest, is used for
Shoot the video information of forest；
S202: each frame picture is extracted from the video information and is sequentially input to based on deep neural network model and is instructed
Experienced fire detection model respectively obtains the quantitative value of target identification object in each frame picture by object detection algorithms；Its
Described in target identification object include target flame object and target smog object；
S203: if the quantitative value of the target identification object is greater than 0, assessing every place's target identification object, point
It does not show that confidence level scores, and non-maxima suppression is carried out to target complete identification object；
S204: the size of the confidence level scoring and preset score threshold, if confidence level scoring is less than in advance
If score threshold, then give up corresponding target identification object；
S205: it obtains in each frame picture by the assessment and remaining target identification pair after the non-maxima suppression
The quantitative value of elephant obtains all remaining target identifications pair if the quantitative value of the remaining target identification object is greater than 0
As the position in the frame picture；
S206: quantitative value, position and the confidence level scoring of the output remaining target identification object, while obtaining
The monitoring result of fire occurs.
Specifically, the process of the specific implementation of step S201 and S202 in the present embodiment is referred in embodiment one
Identical content realizes that and will not be described here in detail.And the difference between this embodiment and the first embodiment lies in, it is being detected in the present embodiment
There are when target identification object in video pictures, a series of calibrations have been carried out to testing result, specific as follows:
For step S203 and S204, the target identification in each frame can be provided by common method in the prior art
The confidence level of object scores, i.e., it is required target identification object " possibility " that the object identified to every place, which provides it,.This can
Confidence score can be indicated with a decimal between 0~1, such as 0.35 indicates that the object in the frame identified is mesh
" confidence level " (or being " possibility ") for identifying other object (flame or smog) is 35%, in addition, in order to guarantee not obtain standard
The lower recognition result of true property, can be set a score threshold, such as 0.60, in this case, all scores are no more than
The frame (i.e. frame of the confidence level lower than 0.60) of the score threshold, will be rejected, other frames are then retained.It should be noted that
It must be 0.60 that the score threshold, which does not limit, but can go to adjust according to actual needs, but it is not recommended that be arranged too
Greatly, because score threshold setting it is bigger, it is concluded that the time that the recognition result of fire has occurred is more late, be so unfavorable for early
It was found that fire behavior.
In the related technology, the process of object detection algorithms identification special object can usually pass through classifier
(Classifier) Lai Shixian directly can be to mark while positioning and classification when carrying out Classification and Identification by classifier
Each frame provide confidence level scoring.
In addition, non-maxima suppression (Non-Maximum Suppression, NMS), also referred to as non-maximum suppression care for
Name Si Yi be exactly inhibit be not maximum element, it can be understood as local maxima search, is a kind of edge thinning technology, one
As be applied to " thinned " edge.
For the application, by taking flame as an example, since flame belongs to uncountable noun and it does not have well-regulated shape, because
This, should be considered as the continuous flame that quantity is 1 for certain in video pictures, when being detected using object detection algorithms,
It may be marked by the frame of multiple and different sizes, and generally will appear between multiple frames and completely include or largely intersect
The case where, and the flame that substantially multiple collimation marks go out is same place flame, therefore in this case it is necessary to passes through NMS technology
The maximum frame of range is found out as the indicia framing at this.
Specifically, being to the algorithm of the progress non-maxima suppression of each pixel in gradual change image: by the side of current pixel
Edge intensity is compared (for example, for being directed toward the direction y with the edge strength of the pixel on positive gradient direction and negative gradient direction
It is then compared by pixel with the pixel above and below it)；If the edge strength of current pixel with have the same direction
Mask in other pixels compared to being the largest, which will be retained, and otherwise, which will be suppressed that (repressed value is usual
It is arranged to 0).
After carrying out confidence level scoring and non-maxima suppression, if the quantity of remaining frame is greater than 0, it can be concluded that every
Locate the position of target identification object (each frame), it later can be by the quantity of remaining target identification object (frame), view where it
Position and confidence level scoring output in frequency picture.It in the specific implementation, can be with after fire detection model output recognition result
It is stored into monitoring device, monitoring device finally obtains the monitoring result that fire has occurred, and then can be to preset intelligence
Energy equipment, such as the smart phone or computer of related personnel, send warning message.Further, monitoring device is also if necessary
Above-mentioned recognition result can be sent to smart machine in the form of picture or video, in order to manually check, thus relevant people
Member can intuitively observe the details of fire behavior.It further, can also be by the flame in recognition result in some embodiments
The quantity of frame and smog frame is counted and is exported respectively.
When using technical solution provided in this embodiment, relative to embodiment one, pass through confidence level scoring and non-maximum
Inhibit that the lower recognition result of confidence level can be excluded and repeat the recognition result of statistics, to further increase testing result
Accuracy.
In order to which the technical solution to the application is described in detail, will be said below by a specific example
It is bright.
Referring to Fig. 3, Fig. 3 is a kind of specific implementation procedure of the monitoring method of forest fire provided by the embodiments of the present application
Schematic diagram.As shown in figure 3, video pictures are input to fire detection model first, to obtain the quantity of target identification object
Num_detetc, judges whether num_detetc is greater than 0 later, continues input video picture if num_detetc is equal to 0, if
Num_detetc is greater than 0, then confidence level scoring and non-maxima suppression is carried out to recognition result, in conjunction with preset score threshold
The quantity num_all_detetc for obtaining remaining target identification object (not distinguishing flame or smog), finally according to num_
All_detetc whether be greater than 0 come determine obtain occur fire result or need to continue input video picture.
In order to which the technical solution to the application is more fully introduced, provided corresponding to the above embodiments of the present application gloomy
The monitoring method of forest fires calamity, the embodiment of the present application also provide a kind of monitoring device of forest fire.
Referring to Fig. 4, Fig. 4 is a kind of structural schematic diagram of the monitoring device of forest fire provided by the embodiments of the present application.
As shown in figure 4, the device includes:
Module 41 is obtained, for obtaining the video information of video camera transmission；Wherein, height of the video camera setting in forest
In the air, for shooting the video information of forest；
Detection module 42, for extracting each frame picture from the video information and sequentially inputting to fire trained in advance
Calamity detection model respectively obtains the quantitative value of target identification object in each frame picture；Wherein the target identification object includes
Target flame object and target smog object；
Output module 43 obtains the monitoring knot that fire has occurred if the quantitative value for the target identification object is greater than 0
Fruit.
Specifically, the concrete methods of realizing of each functional module of the device please refers to the forest fire in above-described embodiment
Related content in monitoring method realizes that this will not be detailed here.
In order to which the technical solution to the application is more fully introduced, provided corresponding to the above embodiments of the present application gloomy
The monitoring method of forest fires calamity, the embodiment of the present application also provide a kind of monitoring system of forest fire.
Fig. 5 and Fig. 6 are please referred to, Fig. 5 and Fig. 6 are the monitoring system of two different forest fires provided by the embodiments of the present application
The structural schematic diagram of system.System as shown in Figure 5 and Figure 6 includes:
The forest fire that multiple video cameras 5 of the high aerial different location of forest are set and are communicated to connect with video camera 5
Monitoring device 6；Wherein, monitoring device 6 includes: memory 61 and the processor being connected with memory 61 62；
For memory 61 for storing program, described program is at least used to execute the forest fire in any above-described embodiment
Monitoring method；
Processor 62 is used to call and execute the described program of the storage of memory 61.
Wherein, monitoring system shown in fig. 5 is similar with traditional video monitoring method, it includes multiple video cameras 5 it is equal
It is in communication with each other and connect with the monitoring device 6 for being set to locality (such as the monitoring station being specially arranged), so that each video camera 5 is clapped
The monitoring device 6 that the video taken the photograph is transmitted to same position carries out united analysis processing, consequently facilitating integrated management.And Fig. 6 institute
In the monitoring system shown, each video camera 5, which is directly integrated monitoring device 6, (can be considered that every video camera individually connects one
Monitoring device), to be the shooting and analysis processing that video can be achieved at the same time in video camera front end, to reduce data line
The installation cost on road and convenient for being separately maintained respectively.In practical application, user can according to itself actual needs from
It is selected in monitoring system shown in fig. 5 or monitoring system shown in fig. 6, on the whole, the area of forest is bigger, Fig. 6 institute
The advantage for the monitoring system shown is bigger.
Specifically, the concrete methods of realizing of the function program in monitoring device in the system please refers in above-described embodiment
Forest fire monitoring method in related content realize that this will not be detailed here.
It is understood that same or similar part can mutually refer in the various embodiments described above, in some embodiments
Unspecified content may refer to the same or similar content in other embodiments.
It should be noted that term " first ", " second " etc. are used for description purposes only in the description of the present application, without
It can be interpreted as indication or suggestion relative importance.In addition, in the description of the present application, unless otherwise indicated, the meaning of " multiple "
Refer at least two.
Any process described otherwise above or method description are construed as in flow chart or herein, and expression includes
It is one or more for realizing specific logical function or process the step of executable instruction code module, segment or portion
Point, and the range of the preferred embodiment of the application includes other realization, wherein can not press shown or discussed suitable
Sequence, including according to related function by it is basic simultaneously in the way of or in the opposite order, Lai Zhihang function, this should be by the application
Embodiment person of ordinary skill in the field understood.
It should be appreciated that each section of the application can be realized with hardware, software, firmware or their combination.Above-mentioned
In embodiment, software that multiple steps or method can be executed in memory and by suitable instruction execution system with storage
Or firmware is realized.It, and in another embodiment, can be under well known in the art for example, if realized with hardware
Any one of column technology or their combination are realized: having a logic gates for realizing logic function to data-signal
Discrete logic, with suitable combinational logic gate circuit specific integrated circuit, programmable gate array (PGA), scene
Programmable gate array (FPGA) etc..
Those skilled in the art are understood that realize all or part of step that above-described embodiment method carries
It suddenly is that relevant hardware can be instructed to complete by program, the program can store in a kind of computer-readable storage medium
In matter, which when being executed, includes the steps that one or a combination set of embodiment of the method.
It, can also be in addition, can integrate in a processing module in each functional unit in each embodiment of the application
It is that each unit physically exists alone, can also be integrated in two or more units in a module.Above-mentioned integrated mould
Block both can take the form of hardware realization, can also be realized in the form of software function module.The integrated module is such as
Fruit is realized and when sold or used as an independent product in the form of software function module, also can store in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only memory, disk or CD etc..
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is contained at least one embodiment or example of the application.In the present specification, schematic expression of the above terms are not
Centainly refer to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be any
One or more embodiment or examples in can be combined in any suitable manner.
Although embodiments herein has been shown and described above, it is to be understood that above-described embodiment is example
Property, it should not be understood as the limitation to the application, those skilled in the art within the scope of application can be to above-mentioned
Embodiment is changed, modifies, replacement and variant.
Claims (10)
1. a kind of monitoring method of forest fire characterized by comprising
Obtain the video information that video camera is sent；Wherein, the video camera setting is aerial in the height of forest, for shooting forest
The video information；
Each frame picture is extracted from the video information and is sequentially input to fire detection model trained in advance, is respectively obtained
The quantitative value of target identification object in each frame picture；Wherein, the target identification object includes target flame object and target
Smog object；
If the quantitative value of the target identification object is greater than 0, the monitoring result that fire has occurred is obtained.
2. the method according to claim 1, wherein described extract each frame picture simultaneously from the video information
It sequentially inputs to fire detection model trained in advance, respectively obtains the quantitative value of target identification object in each frame picture, wrap
It includes:
Each frame picture is extracted from the video information；
Each frame picture is sequentially input to the fire detection model based on deep neural network model training；
Obtain the quantitative value of target identification object in each frame picture respectively by object detection algorithms.
3. according to the method described in claim 2, it is characterized in that, described sequentially input each frame picture to based on deep
It spends before the fire detection model of neural network model training, further includes:
Each frame picture is scaled presetted pixel；
It is described to sequentially input each frame picture to the fire detection model based on deep neural network model training, packet
It includes:
The each frame picture for being scaled presetted pixel is sequentially input to the fire detection based on deep neural network model training
Model.
4. according to the method described in claim 2, it is characterized by further comprising:
If the quantitative value of the target identification object is greater than 0, every place's target identification object is assessed, is obtained respectively credible
Degree scoring, and non-maxima suppression is carried out to target complete identification object；
Compare the size of the confidence level scoring and preset score threshold, if confidence level scoring is less than preset score threshold
Value, then give up corresponding target identification object；
Obtain the quantity of the remaining target identification object after the assessment and the non-maxima suppression in each frame picture
Value obtains whole remaining target identification objects in the frame if the quantitative value of the remaining target identification object is greater than 0
Position in picture；
Quantitative value, position and the confidence level scoring of the remaining target identification object are exported, while obtaining and fire has occurred
Monitoring result.
5. according to the method described in claim 4, it is characterized by further comprising:
Obtain the target flame object for including in the remaining target identification object and the target smog object respectively
Corresponding quantitative value, position and confidence level scoring；
Export the target flame object and the corresponding quantitative value of the target smog object, position and confidence level scoring.
6. the method according to claim 1, wherein the training process of the fire detection model includes:
Obtain the fire picture sample of preset quantity；
The fire picture sample is input to the deep neural network model constructed in advance, and in the fire picture sample
Target identification object identified one by one, to obtain the fire detection model.
7. according to the method described in claim 5, it is characterized in that, it is described export the monitoring result of fire has occurred after, also
Include:
Warning message and the target flame object are sent to pre-set smart machine and the target smog object is each
Self-corresponding quantitative value, position and confidence level scoring.
8. a kind of monitoring device of forest fire characterized by comprising
Module is obtained, for obtaining the video information of video camera transmission；Wherein, the video camera setting is aerial in the height of forest,
For shooting the video information of forest；
Detection module, for extracting each frame picture from the video information and sequentially inputting to fire detection trained in advance
Model respectively obtains the quantitative value of target identification object in each frame picture；Wherein the target identification object includes target fire
Flame object and target smog object；
Output module obtains the monitoring result that fire has occurred if the quantitative value for the target identification object is greater than 0.
9. a kind of monitoring device of forest fire characterized by comprising
Memory and the processor being connected with the memory；
For storing program, described program is at least used to execute such as claim 1-7 described in any item forests the memory
The monitoring method of fire；
The processor is used to call and execute the described program of the memory storage.
10. a kind of monitoring system of forest fire characterized by comprising
Multiple video cameras of the high aerial different location of forest are set and with each video camera communication connection as right is wanted
The monitoring device of forest fire described in asking 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910750219.6A CN110473375A (en) | 2019-08-14 | 2019-08-14 | Monitoring method, device, equipment and the system of forest fire |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910750219.6A CN110473375A (en) | 2019-08-14 | 2019-08-14 | Monitoring method, device, equipment and the system of forest fire |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110473375A true CN110473375A (en) | 2019-11-19 |
Family
ID=68510719
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910750219.6A Pending CN110473375A (en) | 2019-08-14 | 2019-08-14 | Monitoring method, device, equipment and the system of forest fire |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110473375A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111080955A (en) * | 2019-12-30 | 2020-04-28 | 重庆市海普软件产业有限公司 | Forest fire prevention intelligent control system and method |
| CN111145275A (en) * | 2019-12-30 | 2020-05-12 | 重庆市海普软件产业有限公司 | Intelligent automatic control forest fire prevention monitoring system and method |
| CN111739249A (en) * | 2020-06-20 | 2020-10-02 | 深泽县联宇电子科技有限公司 | Fire monitoring method, device and system |
| CN112132090A (en) * | 2020-09-28 | 2020-12-25 | 天地伟业技术有限公司 | Smoke and fire automatic detection and early warning method based on YOLOV3 |
| CN113836967A (en) * | 2020-06-08 | 2021-12-24 | 阿里巴巴集团控股有限公司 | Data processing method, device, storage medium and computer equipment |
| CN113877124A (en) * | 2021-11-15 | 2022-01-04 | 应急管理部天津消防研究所 | Intelligent control system for jet flow falling point of fire monitor |
| CN114446002A (en) * | 2022-01-17 | 2022-05-06 | 厦门理工学院 | Fire on-line monitoring method, device, medium and system |
| CN114998843A (en) * | 2022-08-04 | 2022-09-02 | 深圳市海清视讯科技有限公司 | Fire detection method and related device |
| CN115035447A (en) * | 2022-06-09 | 2022-09-09 | 中国工商银行股份有限公司 | Fire detection processing method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778582A (en) * | 2016-12-07 | 2017-05-31 | 哈尔滨工业大学 | Flame/smog recognition methods after forest map picture cutting based on RGB reconstruct |
| CN108564065A (en) * | 2018-04-28 | 2018-09-21 | 广东电网有限责任公司 | A kind of cable tunnel open fire recognition methods based on SSD |
| CN109147254A (en) * | 2018-07-18 | 2019-01-04 | 武汉大学 | A kind of video outdoor fire disaster smog real-time detection method based on convolutional neural networks |
| JP2019079445A (en) * | 2017-10-27 | 2019-05-23 | ホーチキ株式会社 | Fire monitoring system |
| CN109842779A (en) * | 2017-11-28 | 2019-06-04 | 沈阳益泰科信息咨询有限公司 | Multifunctional video supervisory system |
-
2019
- 2019-08-14 CN CN201910750219.6A patent/CN110473375A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778582A (en) * | 2016-12-07 | 2017-05-31 | 哈尔滨工业大学 | Flame/smog recognition methods after forest map picture cutting based on RGB reconstruct |
| JP2019079445A (en) * | 2017-10-27 | 2019-05-23 | ホーチキ株式会社 | Fire monitoring system |
| CN109842779A (en) * | 2017-11-28 | 2019-06-04 | 沈阳益泰科信息咨询有限公司 | Multifunctional video supervisory system |
| CN108564065A (en) * | 2018-04-28 | 2018-09-21 | 广东电网有限责任公司 | A kind of cable tunnel open fire recognition methods based on SSD |
| CN109147254A (en) * | 2018-07-18 | 2019-01-04 | 武汉大学 | A kind of video outdoor fire disaster smog real-time detection method based on convolutional neural networks |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111080955A (en) * | 2019-12-30 | 2020-04-28 | 重庆市海普软件产业有限公司 | Forest fire prevention intelligent control system and method |
| CN111145275A (en) * | 2019-12-30 | 2020-05-12 | 重庆市海普软件产业有限公司 | Intelligent automatic control forest fire prevention monitoring system and method |
| CN113836967A (en) * | 2020-06-08 | 2021-12-24 | 阿里巴巴集团控股有限公司 | Data processing method, device, storage medium and computer equipment |
| CN111739249A (en) * | 2020-06-20 | 2020-10-02 | 深泽县联宇电子科技有限公司 | Fire monitoring method, device and system |
| CN111739249B (en) * | 2020-06-20 | 2023-08-11 | 深泽县联宇电子科技有限公司 | Fire monitoring method, device and system |
| CN112132090A (en) * | 2020-09-28 | 2020-12-25 | 天地伟业技术有限公司 | Smoke and fire automatic detection and early warning method based on YOLOV3 |
| CN113877124A (en) * | 2021-11-15 | 2022-01-04 | 应急管理部天津消防研究所 | Intelligent control system for jet flow falling point of fire monitor |
| CN114446002A (en) * | 2022-01-17 | 2022-05-06 | 厦门理工学院 | Fire on-line monitoring method, device, medium and system |
| CN114446002B (en) * | 2022-01-17 | 2023-10-31 | 厦门理工学院 | Fire online monitoring methods, devices, media and systems |
| CN115035447A (en) * | 2022-06-09 | 2022-09-09 | 中国工商银行股份有限公司 | Fire detection processing method and device |
| CN114998843A (en) * | 2022-08-04 | 2022-09-02 | 深圳市海清视讯科技有限公司 | Fire detection method and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110473375A (en) | Monitoring method, device, equipment and the system of forest fire | |
| EP3869459B1 (en) | Target object identification method and apparatus, storage medium and electronic apparatus | |
| CN111062429A (en) | Chef cap and mask wearing detection method based on deep learning | |
| US10956753B2 (en) | Image processing system and image processing method | |
| CN110969205A (en) | Forest smoke and fire detection method based on target detection, storage medium and equipment | |
| CN111814638B (en) | Security scene flame detection method based on deep learning | |
| CN111091098A (en) | Training method and detection method of detection model and related device | |
| CN106097346A (en) | A kind of video fire hazard detection method of self study | |
| Lim et al. | Gun detection in surveillance videos using deep neural networks | |
| CN105741477B (en) | Aircraft with intelligent fire voice assistant | |
| Wang et al. | Early forest fire region segmentation based on deep learning | |
| CN111339997A (en) | Method and apparatus for determining ignition region, storage medium, and electronic apparatus | |
| CN105139429A (en) | Fire detecting method based on flame salient picture and spatial pyramid histogram | |
| CN116129490A (en) | A monitoring device and monitoring method for behavior recognition in complex environments | |
| CN111539325A (en) | Forest fire detection method based on deep learning | |
| CN111325133A (en) | Image processing system based on artificial intelligence recognition | |
| KR101366198B1 (en) | Image processing method for automatic early smoke signature of forest fire detection based on the gaussian background mixture models and hsl color space analysis | |
| CN111985331B (en) | Detection method and device for preventing trade secret from being stolen | |
| CN114067244A (en) | A method and system for video analysis of safety work violations | |
| CN115880765A (en) | Method and device for detecting abnormal behavior of regional intrusion and computer equipment | |
| CA3119574A1 (en) | System and method for using artificial intelligence to enable elevated temperature detection of persons using commodity-based thermal cameras | |
| CN111860187A (en) | High-precision worn mask identification method and system | |
| Wang et al. | Forest fire detection method based on deep learning | |
| CN111126411A (en) | Abnormal behavior identification method and device | |
| CN118038148A (en) | Intelligent recognition method for personnel invasion of coal mine belt conveyor |
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: 20191119 |