CN108717582A - Forest fire prediction technique, device, computer equipment and storage medium - Google Patents
Forest fire prediction technique, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN108717582A CN108717582A CN201810345317.7A CN201810345317A CN108717582A CN 108717582 A CN108717582 A CN 108717582A CN 201810345317 A CN201810345317 A CN 201810345317A CN 108717582 A CN108717582 A CN 108717582A
- Authority
- CN
- China
- Prior art keywords
- fire
- data
- forest fire
- forest
- forecasting model
- 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 49
- 239000002023 wood Substances 0.000 claims abstract description 90
- 238000012300 Sequence Analysis Methods 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 13
- 238000012731 temporal analysis Methods 0.000 claims description 11
- 238000000700 time series analysis Methods 0.000 claims description 11
- 238000012512 characterization method Methods 0.000 abstract description 4
- 239000002689 soil Substances 0.000 description 20
- 230000008569 process Effects 0.000 description 13
- 238000004458 analytical method Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- 230000012010 growth Effects 0.000 description 8
- 238000002310 reflectometry Methods 0.000 description 8
- 230000003595 spectral effect Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 6
- 230000008859 change Effects 0.000 description 4
- 238000013480 data collection Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 230000007613 environmental effect Effects 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 230000002123 temporal effect Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 231100000241 scar Toxicity 0.000 description 3
- 239000002028 Biomass Substances 0.000 description 2
- 230000003698 anagen phase Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000032258 transport Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 239000003643 water by type Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
This application discloses a kind of forest fire prediction technique, device, computer equipment and storage medium, wherein method includes:Obtain the specified data of the wood land of forest fire to be predicted;The specified data are input to the preset Forecasting Model of Forest Fire based on ConvLSTM models and carry out operation;The fire coefficient of the wood land is determined according to operation result.The application makes full use of the time series modeling ability and space-time characterisation of ConvLSTM models, establishes a fire prediction model for taking into account time series and spatial sequence, to more accurately carry out forest fire prediction.
Description
Technical field
This application involves to field of computer technology, especially relate to a kind of forest fire prediction technique, device, calculating
Machine equipment and storage medium.
Background technology
Existing fire prediction is substantially according to the air themperature in region residing for forest, air humidity, precipitation, even dry days
The natural parameters such as number, wind are predicted, but prediction result accuracy is relatively low.For example, Australian fire danger prediction system be through
It crosses what more than 800 field igniting experiments were set up, although there is solid field experimental basis, is only applicable to single
Fuel type etc..So a kind of prediction practicability higher of offer and the better forest fire prediction technique of accuracy are that there is an urgent need for solutions
Certainly the problem of.
Invention content
The main purpose of the application is to provide a kind of practicability and accuracy higher forest fire compared with the existing technology
Prediction technique, device, computer equipment and storage medium.
In order to realize that above-mentioned application purpose, the application propose a kind of forest fire prediction technique, which is characterized in that including:
Obtain the specified data of the wood land of forest fire to be predicted;
The specified data are input to the preset Forecasting Model of Forest Fire based on ConvLSTM models and carry out operation;
The fire coefficient of the wood land is determined according to operation result.
Further, the specified data include at least the fire point information data of the corresponding wood land, fiery pixel number
According to, the burn wound data of the date data that catches fire, fire season, live fire alarm record data, terrain data and landform vector
Data.
Further, described that the specified data are input to the preset forest fire prediction based on ConvLSTM models
Model carries out the step of operation, including:
The fire is put into information data, fiery pel data, the date data that catches fire, the burn wound data of fire season and existing
Fire alarm record data in field carry out time series analysis in the Forecasting Model of Forest Fire, by terrain data and landform vector number
According to the progress spatial sequence analysis in the Forecasting Model of Forest Fire.
Further, described that the fire is put into information data, the burning of fiery pel data, the date data that catches fire, fire season
Scar trace data and live fire alarm record data and carry out time series analysis in the Forecasting Model of Forest Fire, by ground figurate number
According to before carrying out the step of spatial sequence analysis in the Forecasting Model of Forest Fire with landform vector data, including:
Judge in the live fire alarm record data with the presence or absence of the record data artificially set fire;
If there is the record data artificially set fire, then artificially set fire described from the live fire alarm record data
Record data dump.
Further, described that the specified data are input to the preset forest fire prediction based on ConvLSTM models
Before model carries out the step of operation, including:
The specified data of acquisition are normalized, the standardized specified data are obtained.
Further, described that the specified data are input to the preset forest fire prediction based on ConvLSTM models
Before model carries out the step of operation, including:
Obtain the geographical location of the wood land;
The Forest Fire in the geographical location of the corresponding wood land is called according to the geographical location of the wood land
Calamity prediction model, wherein the Forecasting Model of Forest Fire includes that multiple, different Forecasting Model of Forest Fire is for difference
Geographical location, and based on the Forecasting Model of Forest Fire of ConvLSTM model foundations.
Further, described that the specified data are input to the preset forest fire prediction based on ConvLSTM models
Before model carries out the step of operation, including:
Obtain the landforms attribute of the wood land;
The Forest Fire of the landforms attribute of the corresponding wood land is called according to the landforms attribute of the wood land
Calamity prediction model, wherein the Forecasting Model of Forest Fire includes that multiple, different Forecasting Model of Forest Fire is for difference
Landforms attribute, and based on the Forecasting Model of Forest Fire of ConvLSTM model foundations.
The application also provides a kind of forest fire prediction meanss, including:
Acquiring unit, the specified data of the wood land for obtaining forest fire to be predicted;
Arithmetic element is predicted for the specified data to be input to the preset forest fire based on ConvLSTM models
Model carries out operation;
Determination unit, the fire coefficient for determining the wood land according to operation result.
The application also provides a kind of computer equipment, including memory and processor, and the memory is stored with computer
The step of program, the processor realizes any of the above-described the method when executing the computer program.
The application also provides a kind of computer readable storage medium, is stored thereon with computer program, the computer journey
The step of method described in any one of the above embodiments is realized when sequence is executed by processor.
Forest fire prediction technique, device, computer equipment and the storage medium of the application, makes full use of ConvLSTM moulds
The time series modeling ability and space-time characterisation of type establish a fire prediction model for taking into account time series and spatial sequence, from
And more accurately carry out forest fire prediction.
Description of the drawings
Fig. 1 is the flow diagram of the forest fire prediction technique of one embodiment of the application;
Fig. 2 is the internal structure schematic diagram of the ConvLSTM of one embodiment of the application;
Fig. 3 is the structural schematic diagram of the coding prediction ConvLSTM networks of one embodiment of the application;
Fig. 4 is the structural schematic block diagram of the forest fire prediction meanss of one embodiment of the application;
Fig. 5 is the structural schematic block diagram of the arithmetic element of one embodiment of the application;
Fig. 6 is the structural schematic block diagram of the arithmetic element of one embodiment of the application;
Fig. 7 is the structural schematic block diagram of the forest fire prediction meanss of one embodiment of the application;
Fig. 8 is the structural schematic block diagram of the computer equipment of one embodiment of the application.
Specific implementation mode
It is with reference to the accompanying drawings and embodiments, right in order to make the object, technical solution and advantage of the application be more clearly understood
The application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, not
For limiting the application.
Referring to Fig.1, the embodiment of the present application proposes a kind of forest fire prediction technique, including step:
S1, obtain forest fire to be predicted wood land specified data;
S2, it the specified data is input to the preset Forecasting Model of Forest Fire based on ConvLSTM models transports
It calculates;
S3, the fire coefficient that the wood land is determined according to operation result.
As described in above-mentioned steps S1, above-mentioned wood land is to grow the region for having the plants such as trees, is generally people's class
Rareness, and the region that the plants such as trees are more, such as the virgin forest in Daxing'an Mountainrange, it is long white on some region of forest.On
It refers to that may reflect the data of wood land fire condition to state specified data, is specifically included:The fiery point of corresponding above-mentioned wood land
Information data, fiery pel data, the date data that catches fire, the burn wound data of fire season, live fire alarm record data, landform
Data and landform vector data etc..It in other embodiments, can also be including meteorological data etc..
As described in above-mentioned steps S2, above-mentioned ConvLSTM models its not only with LSTM time series modeling ability, but also
Local feature can be portrayed as CNN, it may be said that be to be provided simultaneously with space-time characterisation.The internal structure of ConvLSTM as shown in Fig. 2,
Its operation principle can be indicated by following formula:
In above-mentioned each formula, * indicates that convolution, hollow small circle representing matrix corresponding element are multiplied, also known as Hadamard
Product.It is worth noting that, χ here, C, H, i, f, o are the tensor (tensor) of three-dimensional, their latter two dimension generation
The spatial information of table row and column can be imagined as ConvLSTM models to be the model for handling the feature vector in two-dimensional grid,
It can predict the feature of central gridding according to the feature that surrounding is put in grid.Coding prediction ConvLSTM networks are then such as Fig. 3
It is shown.
As described in above-mentioned steps S3, above-mentioned fire coefficient refers to the coefficient that fire occurs for above-mentioned wood land, that is, is occurred gloomy
The probability of forest fires calamity, generally forest fire danger class.
In the present embodiment, above-mentioned that the specified data are input to the preset forest fire based on ConvLSTM models
Prediction model carries out the step S2 of operation, including:
S21, the fire is put to information data, the burn wound data of fiery pel data, the date data that catches fire, fire season
Time series analysis is carried out in the Forecasting Model of Forest Fire with live fire alarm record data, and terrain data and landform are sweared
Amount data carry out spatial sequence analysis in the Forecasting Model of Forest Fire.
As described in above-mentioned steps S21, above-mentioned fire point information data, fiery pel data, the date data that catches fire, fire season
Burn wound data and live fire alarm record data etc. are related to date of occurrence, that is, have the property in sequential, therefore carried out
Analysis in terms of sequential.Above-mentioned terrain data and landform vector data are generally spatial data, such as height, area, so will
Terrain data and landform vector data carry out spatial sequence analysis.In one embodiment, first to above-mentioned fire point Information Number
According to, the burn wound data of fiery pel data, the date data that catches fire, fire season, live fire alarm record data, terrain data with
And each data such as landform vector data are classified, and the data with temporal information are divided into a kind of (fire point information data, fiery picture
Metadata, the date data that catches fire, the burn wound data of fire season and live fire alarm record data), with spatial information
Data are divided into a kind of (terrain data and landform vector data), and it is pre- that corresponding data are then input to above-mentioned forest fire respectively
It surveys in model and is calculated, it, can be pointedly right because Forecasting Model of Forest Fire is the model based on ConvLSTM models
Data with temporal information carry out the analysis in terms of sequential, and the data with spatial information are carried out with point in terms of spatial sequence
Analysis is input to the common knowledge for carrying out that calculating is this field in ConvLSTM models about data, is not repeating herein.
In a specific implementation, above-mentioned fire point information data is obtained by parsing the fiery mask data of above-mentioned wood land
's.Mask is a kind of template of image filters, is remote sensing images through commonly use its processing.In the application, above-mentioned fire mask data
As it is used to extract the image filters template of ignition point, detailed process is:When extract satellite remote sensing images terrestrial object information when,
Pixel filtering is carried out to image by the matrix of a n*n, then comes out the fire point presentation of information that we need.Fiery mask
MOD14A1 points of data set is 10 ranks, as shown in the table:
The pixel for extracting attribute value 8 and 9, it is respectively overlapped with the vegetative coverage figure of survey region, is removed pseudo- gloomy
Then forest fires point information is compared with the vegetation information of corresponding period, by the pixel of vegetation significant change before and after the date of catching fire
It is determined as fiery point, 8+9 data sets is finally used to determine fire point information.
Above-mentioned fire pel data and the date data that catches fire are as obtained from analyzing brulee data set.Above-mentioned baked wheaten cake
Slash refers to the soil for not yet having grown new woods in forest after fire is burnt.Above-mentioned brulee data set refers to above-mentioned forest district
The set of the brulee in domain, each brulee is corresponded in the set of brulee, and record has fiery pel data and catches fire
Date data.The area that pixel represents ground is how many.The flying height of remote sensing satellite generally the km of 4000 kms~600 it
Between, image resolution ratio is generally between 1 km~1 meter.(TV is by several to the point that pixel is equivalent on video screen
The image frame of point composition), a pixel being equivalent on computer display screens is equivalent to a group and lifts different colour tables and is combined into
One in the people of picture.When resolution ratio is 1 km, a pixel represents the area of 1 km X1 kms of ground, i.e., 1 square
Km;When resolution ratio is 30 meters, a pixel represents the area on 30 meters × 30 meters of ground;When resolution ratio is 1 meter, that is,
It says, on image a pixel is equivalent to the area of 1 meter of ground, 1 meter of x, i.e., 1 square metre.Above-mentioned fire pixel is to indicate to catch fire
One unit of area in slash can describe brulee if a fiery pixel is 1 square metre by how many fiery pixels
Area.The above-mentioned date data that catches fire refers to date data when fire occurs for brulee.
The burn wound data of above-mentioned fire season are obtained by analyzing the vegetation index data of above-mentioned wood land.
Vegetation index is the indicator of vegetation distribution density and growth conditions, with vegetative coverage correlation.And above-mentioned fire
The burn wound data in calamity season refer to the trace that forest fire leaves, i.e., the vegetation index on the mark region will be far below it
Around vegetation index, but excessive with the time, above-mentioned mark region can constantly restore the growth conditions of vegetation, i.e. trace
The vegetation index in mark region increases, it is possible to obtain the burn of fire season by the vegetation index data analysis of wood land
Scar data.Above-mentioned vegetation index data are the spectral characteristics according to vegetation, and satellite visible and near infrared band are carried out group
It closes, forms various vegetation indexs.Specifically:Vegetation index, according to the spectral characteristic of vegetation, by satellite visible and near-infrared
Wave band is combined, and forms various vegetation indexs.Vegetation index is simple, effective and experience degree to earth's surface vegetation state
Amount.In remote sensing application field, vegetation index has been widely used for qualitative and quantitative assessment vegetative coverage and its growth vigor.Due to planting
Vegetation, soil lightness, environment influence, shade, soil color and the reaction of humidity COMPLEX MIXED are shown as by spectrum, and by air
The influence of space-Temporal variation, therefore the value that vegetation index neither one is universal, research often show different results.It should
Index is increased rapidly with the increase of biomass.During vegetation is in, low cover degree when, the index is fast with the increase of coverage
Speed increases, and increasess slowly after reaching certain coverage, so early, mid-term growth phase the dynamic monitoring suitable for vegetation.It plants
It is specifically included by index:
Normalized differential vegetation index (NDVI):
Wherein, ρNIRAnd ρREDRespectively represent the reflectivity of near infrared band and red spectral band.NDV1 values between -1 and 1 it
Between.
Enhancement mode meta file (EVI):
ρNIR、ρRED、ρBLUERespectively represent the reflectivity of near infrared band, infrared band, blue wave band.
Ratio vegetation index (RVI):
ρNIRAnd ρREDRespectively represent the reflectivity of near infrared band, infrared band.
Difference vegetation index (DVI):
DVI=ρNIR-ρRED
ρNIRThe reflectivity of near infrared band, infrared band is respectively represented with ρ RED.Difference vegetation index is to Soil Background
Variation is extremely sensitive, is conducive to the monitoring to vegetation ecological environment, therefore the environmental vegetation index that is otherwise known as (EVI).
Soil adjusts vegetation index (SAVI):
Wherein, L is a soil adjustment factor, and the coefficient is related with vegetation concentration, is determined, is used for by actual area condition
Reduce sensibility of the vegetation index to different soils change of reflection.As L=0, SAVI=NDVI;For medium vegetation cover
Area, the value of L is normally close in 0.5.Multiplication factor (1+L) is primarily used to ensure last SAVI values between -1 and 1.It should
Index can reduce the influence of Soil Background, but possible lost part vegetation signal, keep vegetation index relatively low.
Above-mentioned scene fire alarm record data are the responding record data of the fire brigade of above-mentioned wood land, when generally comprising
Between, place, ignition cause etc..
Above-mentioned terrain data parses to obtain generally by the satellite remote sensing images to above-mentioned wood land, specific to wrap
Include data as shown in the table:
Above-mentioned landform vector data (Digital Line Graphic, abbreviation DLG) is again by above-mentioned forest district
The satellite remote sensing images in domain parse to obtain.Landform vector data is the vector data collection of basic geographic elements on existing topographic map,
Including 9 water system, settlement place and facility, traffic, pipeline, boundary and administrative division, landforms and soil property, vegetation, place name and annotation data
Collection, and preserve spatial relationship and relevant attribute information between element.In the present embodiment, the element that landform vector data includes is
Spatial relationship etc. between water system, pipeline, boundary and administrative division, landforms and soil property, vegetation, place name and annotation and element.
It is above-mentioned that the fire is put into information data, fiery pel data, the date data that catches fire, fire season in the present embodiment
Burn wound data and live fire alarm record data and carry out time series analysis in the Forecasting Model of Forest Fire, by landform
Before data and landform vector data carry out the step S21 of spatial sequence analysis in the Forecasting Model of Forest Fire, including:
S22, judge in the live fire alarm record data with the presence or absence of the record data artificially set fire;
S23, if there is the record data artificially set fire, then artificially put described from the live fire alarm record data
The record data dump of fire.
As described in above-mentioned steps S22, the ignition cause of forest fire can be recorded as in fire alarm record, as caused by dry
Fire caused by fire, thunder and lightning, and artificial caused fire etc..For example, fire alarm with being recorded as the * * on the 1st of August in 2016 responding,
Cause of fire is that peasant household burns stalk and causes, and the meaning that can parse fire alarm record by semantic analysis at this time is artificial
It burns stalk and causes fire, and then be denoted as the record data artificially set fire.
As described in above-mentioned steps S23, the record data dump that above-mentioned correspondence is artificially set fire retains going out for natural fire
Alert data, human factor is uncontrollable enchancement factor, is rejected the accuracy that can improve forest fire prediction.
In the present embodiment, above-mentioned that the specified data are input to the preset forest fire based on ConvLSTM models
Before prediction model carries out the step S2 of operation, including:
S201, the specified data of acquisition are normalized, obtain the standardized specified data.
As described in above-mentioned steps S201, above-mentioned normalized refers to the expression formula that will have dimension, by transformation, is turned to
Nondimensional expression formula, becomes scalar.In the present embodiment, variant data are as converted to the scalar of identical expression, with
Facilitate calculating.
In one embodiment, above-mentioned that the specified data are input to the preset forest based on ConvLSTM models
Before fire prediction model carries out the step S2 of operation, including:
S202, the geographical location for obtaining the wood land;
S203, the described gloomy of the geographical location for corresponding to the wood land is called according to the geographical location of the wood land
Forest fires calamity prediction model, wherein the Forecasting Model of Forest Fire includes that multiple, different Forecasting Model of Forest Fire is to be directed to
Diverse geographic location, and based on the Forecasting Model of Forest Fire of ConvLSTM model foundations.
As described in above-mentioned steps S202 and S203, above-mentioned geographical location is specific longitude and latitude, can be by gloomy
Longitude and latitude residing for the domain of forest zone determines the environmental quality of wood land, for example, the gas in the south and the north of the People's Republic of China (PRC)
Difference is waited, is predicted if carrying out forest fire using the same Fire Model, there may be differences for the result of prediction
Different, in the present embodiment, setting is two or more with the Forecasting Model of Forest Fire based on ConvLSTM model foundations, and different is gloomy
Forest fires calamity prediction model, be the wood land based on different geographical locations specified data be trained study obtain, i.e.,
The relevant modeling data for acquiring different geographical locations respectively by same modeling method, is then modeled.In forest
When fire alarm is predicted, for different areas, corresponding fire prediction model is called, is predicted with obtaining more accurate forest fire
As a result.
In another specific embodiment, it is above-mentioned the specified data are input to it is preset based on the gloomy of ConvLSTM models
Before forest fires calamity prediction model carries out the step S2 of operation, including:
S204, the landforms attribute for obtaining the wood land;
S205, the described gloomy of the landforms attribute for corresponding to the wood land is called according to the landforms attribute of the wood land
Forest fires calamity prediction model, wherein the Forecasting Model of Forest Fire includes that multiple, different Forecasting Model of Forest Fire is to be directed to
Different landforms attribute, and based on the Forecasting Model of Forest Fire of ConvLSTM model foundations.
As described in above-mentioned steps S204 and S205, the general name of above-mentioned landforms, that is, earth surface various forms also can referred to as
Shape.Surface configuration is diversified, and the origin cause of formation is also not quite similar, and is result of the inside and outside power geologic process to earth's crust comprehensive function.
Internal force transmission causes the fluctuating of earth's surface, control Land-sea Distributions outline and mountainous region, plateau, basin and Plain region
Configuration, determines the tectonic framework of landforms.And exogenic force (the growth and work of flowing water, wind-force, solar radiant energy, air and biology
It is dynamic) geologic process constantly carries out the earth's crust entry material weathering, degrades, carries and accumulate, to be formed in several ways
The various forms on modern ground.Landforms attribute is the specific form of landforms, such as mountainous region, plateau, basin.The present embodiment,
Corresponding Forecasting Model of Forest Fire can be called to use by wood land landforms attribute feature, for example, Daxing'an Mountainrange
Wood land landforms attribute it is different from the landforms attribute of the wood land in Xishuangbanna, if using the same forest fire
Model carries out forest fire prediction, and the result of prediction is there may be difference, and in the present embodiment, setting is two or more for not
With landforms attribute Forecasting Model of Forest Fire, different Forecasting Model of Forest Fire is the wood land based on different landforms
Specified data be trained study and obtain, i.e., acquire the correlation of different landforms attributes respectively by same modeling method
Modeling data, then modeled.When forest fire alarm is predicted, for different landforms, corresponding fire prediction mould is called
Type, to obtain more accurate forest fire prediction result.
In the present embodiment, the process for establishing above-mentioned Forecasting Model of Forest Fire is as follows:
Extraction is specified in wood land, specifies the historical satellite remote sensing images in historical time section, and respectively to each
It opens historical satellite remote sensing images to be parsed, to obtain corresponding terrain data, landform vector data, vegetation index data and fire
Mask data.Then corresponding fire point information is obtained by fiery mask data;Fire season is obtained by vegetation index data
Burn wound data.Also need to obtain the live fire alarm record data in the above-mentioned specified historical time section of the specified wood land
With brulee data set;Wherein, the fiery pel data of each brulee is parsed by brulee data set and caught fire day
Issue evidence;The history fire coefficient data of each time in above-mentioned specified historical time section can also be obtained.
The modeling data of acquisition is divided into training set and test set, is then input to by the corresponding sample data of training set
It is trained in ConvLSTM models, in the training process, by the corresponding fire point information data of each history fire coefficient data, fire
The burn wound data progress time series analysis of pel data, catch fire date data and fire season, terrain data and landform
Vector data carries out spatial sequence analysis, finally obtains a Forecasting Model of Forest Fire based on ConvLSTM models, i.e., defeated
Enter fire point information data, burn wound data, terrain data and the landform of fiery pel data, the date data that catches fire, fire season
In vector data to Forecasting Model of Forest Fire, which can export a fire coefficient.Then pass through survey
The validity of examination collection verification Forecasting Model of Forest Fire can carry out the Forecasting Model of Forest Fire real if being verified
Border is applied, that is, predicts forest fire.
In the present embodiment, when establishing model, climatic data when shooting each historical satellite remote sensing images, such as season are also obtained
The data such as section, air themperature, air humidity, fine day, rainy day, in modeling, these data are substantially carried out the analysis of time series.
It is corresponding to obtain climatic data when carrying out forest fire prediction, it is then input in Forecasting Model of Forest Fire.
The forest fire prediction technique of the present embodiment makes full use of the time series modeling ability and space-time of ConvLSTM models
Characteristic establishes a fire prediction model for taking into account time series and spatial sequence, to more accurately carry out forest fire
Prediction.
With reference to Fig. 4, the embodiment of the present application also provides a kind of forest fire prediction meanss, including step:
Acquiring unit 10, the specified data of the wood land for obtaining forest fire to be predicted;
Arithmetic element 20, it is pre- for the specified data to be input to the preset forest fire based on ConvLSTM models
It surveys model and carries out operation;
Determination unit 30, the fire coefficient for determining the wood land according to operation result.
Such as above-mentioned acquiring unit 10, above-mentioned wood land is to grow the region for having the plants such as trees, is generally people's class
Rareness, and the region that the plants such as trees are more, such as the virgin forest in Daxing'an Mountainrange, it is long white on some region of forest.On
It refers to that may reflect the data of wood land fire condition to state specified data, is specifically included:The fiery point of corresponding above-mentioned wood land
Information data, fiery pel data, the date data that catches fire, the burn wound data of fire season, live fire alarm record data, landform
Data and landform vector data etc..It in other embodiments, can also be including meteorological data etc..
Such as above-mentioned arithmetic element 20, its not only time series modeling ability with LSTM of above-mentioned ConvLSTM models, but also
Local feature can be portrayed as CNN, it may be said that be to be provided simultaneously with space-time characterisation.The internal structure of ConvLSTM models such as Fig. 2
Shown, operation principle can be indicated by following formula:
In above-mentioned each formula, * indicates that convolution, hollow small circle representing matrix corresponding element are multiplied, also known as Hadamard
Product.It is worth noting that, χ here, C, H, i, f, o are three-dimensional tensor, their latter two dimension represent row and
The spatial information of row can be imagined as ConvLSTM models to be the model for handling the feature vector in two-dimensional grid, can be with
The feature of central gridding is predicted according to the feature that surrounding is put in grid.Coding prediction ConvLSTM networks are then as shown in Figure 3.
Such as above-mentioned determination unit 30, above-mentioned fire coefficient refers to the coefficient that fire occurs for above-mentioned wood land, that is, is occurred gloomy
The probability of forest fires calamity, generally forest fire danger class.
With reference to Fig. 5, in the present embodiment, above-mentioned arithmetic element 20, including:
Computing module 21, for the fire to be put information data, fiery pel data, the date data that catches fire, fire season
Burn wound data and live fire alarm record data and carry out time series analysis in the Forecasting Model of Forest Fire, by landform
Data and landform vector data carry out spatial sequence analysis in the Forecasting Model of Forest Fire.
Such as above-mentioned computing module 21, above-mentioned fire point information data, fiery pel data, the date data that catches fire, fire season
Burn wound data and live fire alarm record data etc. are related to date of occurrence, that is, have the property in sequential, therefore carried out
Analysis in terms of sequential.Above-mentioned terrain data and landform vector data are generally spatial data, such as height, area, so will
Terrain data and landform vector data carry out spatial sequence analysis.In one embodiment, first to above-mentioned fire point Information Number
According to, the burn wound data of fiery pel data, the date data that catches fire, fire season, live fire alarm record data, terrain data with
And each data such as landform vector data are classified, and the data with temporal information are divided into a kind of (fire point information data, fiery picture
Metadata, the date data that catches fire, the burn wound data of fire season and live fire alarm record data), with spatial information
Data are divided into a kind of (terrain data and landform vector data), and it is pre- that corresponding data are then input to above-mentioned forest fire respectively
It surveys in model and is calculated, it, can be pointedly right because Forecasting Model of Forest Fire is the model based on ConvLSTM models
Data with temporal information carry out the analysis in terms of sequential, and the data with spatial information are carried out with point in terms of spatial sequence
Analysis is input to the common knowledge for carrying out that calculating is this field in ConvLSTM models about data, is not repeating herein.
In a specific implementation, above-mentioned fire point information data is obtained by parsing the fiery mask data of above-mentioned wood land
's.Mask is a kind of template of image filters, is remote sensing images through commonly use its processing.In the application, above-mentioned fire mask data
As it is used to extract the image filters template of ignition point, detailed process is:When extract satellite remote sensing images terrestrial object information when,
Pixel filtering is carried out to image by the matrix of a n*n, then comes out the fire point presentation of information that we need.Fiery mask
MOD14A1 points of data set is 10 ranks, as shown in the table:
Rank | Corresponding particular content |
0 | Untreated pixel |
1 | Untreated pixel |
2 | Untreated pixel |
3 | Waters |
4 | Cloud |
5 | Not fiery bare area |
6 | Unknown pixel |
7 | The fiery point of low confidence detection |
8 | The middle fiery point of confidence level detection |
9 | The fiery point of high confidence level detection |
The pixel for extracting attribute value 8 and 9, it is respectively overlapped with the vegetative coverage figure of survey region, is removed pseudo- gloomy
Then forest fires point information is compared with the vegetation information of corresponding period, by the pixel of vegetation significant change before and after the date of catching fire
It is determined as fiery point, 8+9 data sets is finally used to determine fire point information.
Above-mentioned fire pel data and the date data that catches fire are as obtained from analyzing brulee data set.Above-mentioned baked wheaten cake
Slash refers to the soil for not yet having grown new woods in forest after fire is burnt.Above-mentioned brulee data set refers to above-mentioned forest district
The set of the brulee in domain, each brulee is corresponded in the set of brulee, and record has fiery pel data and catches fire
Date data.The area that pixel represents ground is how many.The flying height of remote sensing satellite generally the km of 4000 kms~600 it
Between, image resolution ratio is generally between 1 km~1 meter.(TV is by several to the point that pixel is equivalent on video screen
The image frame of point composition), a pixel being equivalent on computer display screens is equivalent to a group and lifts different colour tables and is combined into
One in the people of picture.When resolution ratio is 1 km, a pixel represents the area of 1 km X1 kms of ground, i.e., 1 square
Km;When resolution ratio is 30 meters, a pixel represents the area on 30 meters × 30 meters of ground;When resolution ratio is 1 meter, that is,
It says, on image a pixel is equivalent to the area of 1 meter of ground, 1 meter of x, i.e., 1 square metre.Above-mentioned fire pixel is to indicate to catch fire
One unit of area in slash can describe brulee if a fiery pixel is 1 square metre by how many fiery pixels
Area.The above-mentioned date data that catches fire refers to date data when fire occurs for brulee.
The burn wound data of above-mentioned fire season are obtained by analyzing the vegetation index data of above-mentioned wood land.
Vegetation index is the indicator of vegetation distribution density and growth conditions, with vegetative coverage correlation.And above-mentioned fire
The burn wound data in calamity season refer to the trace that forest fire leaves, i.e., the vegetation index on the mark region will be far below it
Around vegetation index, but excessive with the time, above-mentioned mark region can constantly restore the growth conditions of vegetation, i.e. trace
The vegetation index in mark region increases, it is possible to obtain the burn of fire season by the vegetation index data analysis of wood land
Scar data.Above-mentioned vegetation index data are the spectral characteristics according to vegetation, and satellite visible and near infrared band are carried out group
It closes, forms various vegetation indexs.Specifically:Vegetation index, according to the spectral characteristic of vegetation, by satellite visible and near-infrared
Wave band is combined, and forms various vegetation indexs.Vegetation index is simple, effective and experience degree to earth's surface vegetation state
Amount.In remote sensing application field, vegetation index has been widely used for qualitative and quantitative assessment vegetative coverage and its growth vigor.Due to planting
Vegetation, soil lightness, environment influence, shade, soil color and the reaction of humidity COMPLEX MIXED are shown as by spectrum, and by air
The influence of space-Temporal variation, therefore the value that vegetation index neither one is universal, research often show different results.It should
Index is increased rapidly with the increase of biomass.During vegetation is in, low cover degree when, the index is fast with the increase of coverage
Speed increases, and increasess slowly after reaching certain coverage, so early, mid-term growth phase the dynamic monitoring suitable for vegetation.It plants
It is specifically included by index:
Normalized differential vegetation index (NDVI):
Wherein, ρNIRAnd ρREDRespectively represent the reflectivity of near infrared band and red spectral band.NDV1 values between -1 and 1 it
Between.
Enhancement mode meta file (EVI):
ρNIR、ρRED、ρBLUERespectively represent the reflectivity of near infrared band, infrared band, blue wave band.
Ratio vegetation index (RVI):
ρNIRAnd ρREDRespectively represent the reflectivity of near infrared band, infrared band.
Difference vegetation index (DVI):
DVI=ρNIR-ρRED
ρNIRAnd ρREDRespectively represent the reflectivity of near infrared band, infrared band.Difference vegetation index is to Soil Background
Variation is extremely sensitive, is conducive to the monitoring to vegetation ecological environment, therefore the environmental vegetation index that is otherwise known as (EVI).
Soil adjusts vegetation index (SAVI):
Wherein, L is a soil adjustment factor, and the coefficient is related with vegetation concentration, is determined, is used for by actual area condition
Reduce sensibility of the vegetation index to different soils change of reflection.As L=0, SAVI=NDVI;For medium vegetation cover
Area, the value of L is normally close in 0.5.Multiplication factor (1+L) is primarily used to ensure last SAVI values between -1 and 1.It should
Index can reduce the influence of Soil Background, but possible lost part vegetation signal, keep vegetation index relatively low.
Above-mentioned scene fire alarm record data are the responding record data of the fire brigade of above-mentioned wood land, when generally comprising
Between, place, ignition cause etc..
Above-mentioned terrain data parses to obtain generally by the satellite remote sensing images to above-mentioned wood land, specific to wrap
Include data as shown in the table:
Above-mentioned landform vector data (Digital Line Graphic, abbreviation DLG) is again by above-mentioned forest district
The satellite remote sensing images in domain parse to obtain.Landform vector data is the vector data collection of basic geographic elements on existing topographic map,
Including 9 water system, settlement place and facility, traffic, pipeline, boundary and administrative division, landforms and soil property, vegetation, place name and annotation data
Collection, and preserve spatial relationship and relevant attribute information between element.In the present embodiment, the element that landform vector data includes is
Spatial relationship etc. between water system, pipeline, boundary and administrative division, landforms and soil property, vegetation, place name and annotation and element.
With reference to Fig. 6, in the present embodiment, above-mentioned arithmetic element 20, including:
Judgment module 22, for judging in the live fire alarm record data with the presence or absence of the record data artificially set fire;
Module 23 is removed, for if there is the record data artificially set fire, then being recorded in data from the live fire alarm
By the record data dump artificially set fire.
Such as above-mentioned judgment module 22, the ignition cause of forest fire can be recorded as in fire alarm record, as caused by drying
Fire caused by fire, thunder and lightning, and artificial caused fire etc..For example, fire alarm with being recorded as the * * on the 1st of August in 2016 responding,
Cause of fire is that peasant household burns stalk and causes, and the meaning that can parse fire alarm record by semantic analysis at this time is artificial
It burns stalk and causes fire, and then be denoted as the record data artificially set fire.
Such as above-mentioned removing module 23, the record data dump that above-mentioned correspondence is artificially set fire retains going out for natural fire
Alert data, human factor is uncontrollable enchancement factor, is rejected the accuracy that can improve forest fire prediction.
With reference to Fig. 7, in the present embodiment, above-mentioned forest fire prediction meanss, including:
Normalization unit 201 obtains standardized described for the specified data obtained to be normalized
Specified data.
Such as above-mentioned normalization unit 201, above-mentioned normalized refers to the expression formula that will have dimension, by transformation, is changed
For nondimensional expression formula, become scalar.In the present embodiment, variant data are as converted to the scalar of identical expression,
It is calculated with facilitating.
In one embodiment, above-mentioned forest fire prediction meanss, including:
First acquisition unit, the geographical location for obtaining the wood land;
First call unit, the geography for calling the corresponding wood land according to the geographical location of the wood land
The Forecasting Model of Forest Fire of position, wherein the Forecasting Model of Forest Fire includes that multiple, different forest fires is pre-
It is to be directed to diverse geographic location to survey model, and based on the Forecasting Model of Forest Fire of ConvLSTM model foundations.
Such as above-mentioned first acquisition unit and the first call unit, above-mentioned geographical location is specific longitude and latitude, can
To determine the environmental quality of wood land by the longitude and latitude residing for wood land, for example, the People's Republic of China (PRC) south with
The weather in the north is different, is predicted if carrying out Fire Model using the same model, the result of prediction may deposit
In difference, the present embodiment, setting is two or more with the Forecasting Model of Forest Fire based on ConvLSTM model foundations, different
Forecasting Model of Forest Fire, be that the specified data of the wood land based on different geographical locations are trained study and obtain
, i.e., it acquires the relevant modeling data in different geographical locations respectively by same modeling method, is then modeled.?
When forest fire alarm is predicted, for different areas, corresponding fire prediction model is called, to obtain more accurate forest fire
Prediction result.
In another specific embodiment, above-mentioned forest fire prediction meanss, including:
Second acquisition unit, the landforms attribute for obtaining the wood land;
Second call unit, the landforms for calling the corresponding wood land according to the landforms attribute of the wood land
The Forecasting Model of Forest Fire of attribute, wherein the Forecasting Model of Forest Fire includes that multiple, different forest fires is pre-
It is to be directed to different landforms attribute to survey model, and based on the Forecasting Model of Forest Fire of ConvLSTM model foundations.
Such as above-mentioned second acquisition unit and the second call unit, the general name of above-mentioned landforms, that is, earth surface various forms,
It can be known as landform.Surface configuration is diversified, and the origin cause of formation is also not quite similar, and is that inside and outside power geologic process integrates work to the earth's crust
Result.Internal force transmission causes the fluctuating of earth's surface, control Land-sea Distributions outline and mountainous region, plateau, basin and
The ground configuration of territory in Plain, determines the tectonic framework of landforms.And exogenic force (flowing water, wind-force, solar radiant energy, air and biology
Growth and activity) geologic process constantly carries out the earth's crust entry material weathering, degrades, carries and heap in several ways
Product, so as to form the various forms on modern ground.Landforms attribute is the specific form of landforms, such as mountainous region, plateau, basin
Deng.The present embodiment can call corresponding Forecasting Model of Forest Fire to use by wood land landforms attribute feature, than
Such as, the landforms attribute of the wood land in Daxing'an Mountainrange is different from the landforms attribute of the wood land in Xishuangbanna, if using same
One Fire Model carries out forest fire prediction, and the result of prediction in the present embodiment, is arranged two there may be difference
Or it is multiple for different landforms attribute Forecasting Model of Forest Fire, different Forecasting Model of Forest Fire is based on different ground
The specified data of the wood land of looks are trained what study obtained, i.e., acquire different ground respectively by same modeling method
The relevant modeling data of looks attribute, is then modeled.When forest fire alarm is predicted, for different landforms, calls and correspond to
Fire prediction model, to obtain more accurate forest fire prediction result.
In the present embodiment, the process for establishing above-mentioned Forecasting Model of Forest Fire is as follows:
Extraction is specified in wood land, specifies the historical satellite remote sensing images in historical time section, and respectively to each
It opens historical satellite remote sensing images to be parsed, to obtain corresponding terrain data, landform vector data, vegetation index data and fire
Mask data.Then corresponding fire point information is obtained by fiery mask data;Fire season is obtained by vegetation index data
Burn wound data.Also need to obtain the live fire alarm record data in the above-mentioned specified historical time section of the specified wood land
With brulee data set;Wherein, the fiery pel data of each brulee is parsed by brulee data set and caught fire day
Issue evidence;The history fire coefficient data of each time in above-mentioned specified historical time section can also be obtained.
The modeling data of acquisition is divided into training set and test set, is then input to by the corresponding sample data of training set
It is trained in ConvLSTM models, in the training process, by the corresponding fire point information data of each history fire coefficient data, fire
The burn wound data progress time series analysis of pel data, catch fire date data and fire season, terrain data and landform
Vector data carries out spatial sequence analysis, finally obtains a Forecasting Model of Forest Fire based on ConvLSTM models, i.e., defeated
Enter fire point information data, burn wound data, terrain data and the landform of fiery pel data, the date data that catches fire, fire season
In vector data to Forecasting Model of Forest Fire, which can export a fire coefficient.Then pass through survey
The validity of examination collection verification Forecasting Model of Forest Fire can carry out the Forecasting Model of Forest Fire real if being verified
Border is applied, that is, predicts forest fire.
In the present embodiment, when establishing model, climatic data when shooting each historical satellite remote sensing images, such as season are also obtained
The data such as section, air themperature, air humidity, fine day, rainy day, in modeling, these data are substantially carried out the analysis of time series.
It is corresponding to obtain climatic data when carrying out forest fire prediction, it is then input in Forecasting Model of Forest Fire.
The forest fire prediction meanss of the present embodiment make full use of the time series modeling ability and space-time of ConvLSTM models
Characteristic establishes a fire prediction model for taking into account time series and spatial sequence, to more accurately carry out forest fire
Prediction.
With reference to Fig. 8, a kind of computer equipment is also provided in the embodiment of the present application, which can be server,
Its internal structure can be as shown in Figure 8.The computer equipment includes processor, memory, the network connected by system bus
Interface and database.Wherein, the processor of the Computer Design is for providing calculating and control ability.The computer equipment is deposited
Reservoir includes non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program
And database.The internal memory provides environment for the operation of operating system and computer program in non-volatile memory medium.It should
The database of computer equipment is used to store the data such as the Forecasting Model of Forest Fire based on ConvLSTM models.The computer is set
Standby network interface is used to communicate by network connection with external terminal.To realize when the computer program is executed by processor
A kind of forest fire prediction technique.
Above-mentioned processor executes the step of above-mentioned forest fire prediction technique:Obtain the wood land of forest fire to be predicted
Specified data;The specified data are input to the preset Forecasting Model of Forest Fire based on ConvLSTM models to transport
It calculates;The fire coefficient of the wood land is determined according to operation result.
In one embodiment, above-mentioned specified data include at least the fire point information data of the corresponding wood land, fire
Pel data, the date data that catches fire, the burn wound data of fire season, live fire alarm record data, terrain data and ground
Shape vector data.
In one embodiment, above-mentioned that the specified data are input to the preset Forest Fire based on ConvLSTM models
Calamity prediction model carries out the step of operation, including:The fire is put into information data, fiery pel data, the date data that catches fire, fire
The burn wound data in season and live fire alarm record data and carry out time series analysis in the Forecasting Model of Forest Fire,
Terrain data and landform vector data are subjected to spatial sequence analysis in the Forecasting Model of Forest Fire.
In one embodiment, above-mentioned that the fire is put into information data, fiery pel data, the date data that catches fire, fire season
The burn wound data of section and live fire alarm record data and carry out time series analysis in the Forecasting Model of Forest Fire, will
Before the step of terrain data and landform vector data carry out spatial sequence analysis in the Forecasting Model of Forest Fire, packet
It includes:Judge in the live fire alarm record data with the presence or absence of the record data artificially set fire;If there is the note artificially set fire
Data are recorded, then by the record data dump artificially set fire from the live fire alarm record data.
In one embodiment, above-mentioned that the specified data are input to the preset Forest Fire based on ConvLSTM models
Before calamity prediction model carries out the step of operation, including:The specified data of acquisition are normalized, standard is obtained
The specified data changed.
In one embodiment, above-mentioned that the specified data are input to the preset Forest Fire based on ConvLSTM models
Before calamity prediction model carries out the step of operation, including:Obtain the geographical location of the wood land;According to the wood land
Geographical location call the corresponding wood land geographical location the Forecasting Model of Forest Fire, wherein the forest
Fire prediction model includes that multiple, different Forecasting Model of Forest Fire is to be directed to diverse geographic location, and be based on ConvLSTM
The Forecasting Model of Forest Fire of model foundation.
In another embodiment, above-mentioned that the specified data are input to the preset forest based on ConvLSTM models
Before fire prediction model carries out the step of operation, including:Obtain the landforms attribute of the wood land;According to the forest district
The landforms attribute in domain calls the Forecasting Model of Forest Fire of the landforms attribute of the corresponding wood land, wherein described gloomy
Forest fires calamity prediction model includes that multiple, different Forecasting Model of Forest Fire is to be directed to different landforms attribute, and be based on
The Forecasting Model of Forest Fire of ConvLSTM model foundations.
It will be understood by those skilled in the art that structure shown in Fig. 8, is only tied with the relevant part of application scheme
The block diagram of structure does not constitute the restriction for the computer equipment being applied thereon to application scheme.
The computer equipment of the embodiment of the present application makes full use of the time series modeling ability of ConvLSTM models and space-time special
Property, a fire prediction model for taking into account time series and spatial sequence is established, it is pre- to more accurately carry out forest fire
It surveys.
One embodiment of the application also provides a kind of computer readable storage medium, is stored thereon with computer program, calculates
Machine program realizes a kind of forest fire prediction technique when being executed by processor, specially:Obtain the forest of forest fire to be predicted
The specified data in region;By the specified data be input to the preset Forecasting Model of Forest Fire based on ConvLSTM models into
Row operation;The fire coefficient of the wood land is determined according to operation result.
Above-mentioned domestic animal forest fire prediction technique makes full use of the time series modeling ability of ConvLSTM models and space-time special
Property, a fire prediction model for taking into account time series and spatial sequence is established, it is pre- to more accurately carry out forest fire
It surveys.
In one embodiment, above-mentioned specified data include at least the fire point information data of the corresponding wood land, fire
Pel data, the date data that catches fire, the burn wound data of fire season, live fire alarm record data, terrain data and ground
Shape vector data.
In one embodiment, the specified data are input to preset based on ConvLSTM models by above-mentioned processor
Forecasting Model of Forest Fire carries out the step of operation, including:By the fire point information data, fiery pel data, catch fire a day issue
Time sequence is carried out in the Forecasting Model of Forest Fire according to the burn wound data and live fire alarm record data of, fire season
Row analysis, spatial sequence analysis is carried out by terrain data and landform vector data in the Forecasting Model of Forest Fire.
In one embodiment, above-mentioned processor by the fire put information data, fiery pel data, the date data that catches fire,
The burn wound data of fire season and live fire alarm record data and carry out time series in the Forecasting Model of Forest Fire
Analysis, terrain data and landform vector data are carried out in the Forecasting Model of Forest Fire the step of spatial sequence analysis it
Before, including:Judge in the live fire alarm record data with the presence or absence of the record data artificially set fire;If there is artificially setting fire
Record data, then by the record data dump artificially set fire from the live fire alarm record data.
In one embodiment, above-mentioned that the specified data are input to the preset Forest Fire based on ConvLSTM models
Before calamity prediction model carries out the step of operation, including:The specified data of acquisition are normalized, standard is obtained
The specified data changed.
In one embodiment, the specified data are input to preset based on ConvLSTM models by above-mentioned processor
Before Forecasting Model of Forest Fire carries out the step of operation, including:Obtain the geographical location of the wood land;According to described gloomy
Call the Forecasting Model of Forest Fire in the geographical location of the corresponding wood land in the geographical location in forest zone domain, wherein institute
It is to be directed to diverse geographic location that state Forecasting Model of Forest Fire, which include multiple, different Forecasting Model of Forest Fire, and is based on
The Forecasting Model of Forest Fire of ConvLSTM model foundations.
In another embodiment, the specified data are input to preset based on ConvLSTM models by above-mentioned processor
Forecasting Model of Forest Fire carry out operation the step of before, including:Obtain the geographical location of the wood land;According to described
Call the Forecasting Model of Forest Fire in the geographical location of the corresponding wood land in the geographical location of wood land, wherein
The Forecasting Model of Forest Fire includes that multiple, different Forecasting Model of Forest Fire is to be directed to diverse geographic location, and be based on
The Forecasting Model of Forest Fire of ConvLSTM model foundations.
One of ordinary skill in the art will appreciate that realizing all or part of flow in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the flow of the embodiment of above-mentioned each method.Wherein,
Any reference to memory, storage, database or other media used in provided herein and embodiment,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double speed are according to rate SDRAM (SSRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
It should be noted that herein, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that process, device, article or method including a series of elements include not only those elements, and
And further include other elements that are not explicitly listed, or further include for this process, device, article or method institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including this
There is also other identical elements in the process of element, device, article or method.
The foregoing is merely the preferred embodiments of the application, are not intended to limit the scope of the claims of the application, every utilization
Equivalent structure or equivalent flow shift made by present specification and accompanying drawing content is applied directly or indirectly in other correlations
Technical field, include similarly in the scope of patent protection of the application.
Claims (10)
1. a kind of forest fire prediction technique, which is characterized in that including:
Obtain the specified data of the wood land of forest fire to be predicted;
The specified data are input to the preset Forecasting Model of Forest Fire based on ConvLSTM models and carry out operation;
The fire coefficient of the wood land is determined according to operation result.
2. forest fire prediction technique according to claim 1, which is characterized in that the specified data include described in correspondence
The fire point information data of wood land, fiery pel data, the date data that catches fire, the burn wound data of fire season, scene fire
Alert record data, terrain data and landform vector data.
3. forest fire prediction technique according to claim 2, which is characterized in that described to be input to the specified data
The step of preset Forecasting Model of Forest Fire based on ConvLSTM models carries out operation, including:
The fire is put into information data, fiery pel data, the date data that catches fire, the burn wound data of fire season and scene fire
Alert record data carry out time series analysis in the Forecasting Model of Forest Fire, and terrain data and landform vector data are existed
Spatial sequence analysis is carried out in the Forecasting Model of Forest Fire.
4. forest fire prediction technique according to claim 2, which is characterized in that described that the fire is put information data, fiery picture
Metadata, the date data that catches fire, the burn wound data of fire season and live fire alarm record data are pre- in the forest fire
It surveys in model and carries out time series analysis, terrain data and landform vector data are carried out in the Forecasting Model of Forest Fire
Before the step of spatial sequence is analyzed, including:
Judge in the live fire alarm record data with the presence or absence of the record data artificially set fire;
If there is the record data artificially set fire, then by the record artificially set fire from the live fire alarm record data
Data dump.
5. forest fire prediction technique according to claim 1, which is characterized in that described to be input to the specified data
Before the step of preset Forecasting Model of Forest Fire based on ConvLSTM models carries out operation, including:
The specified data of acquisition are normalized, the standardized specified data are obtained.
6. forest fire prediction technique according to claim 1, which is characterized in that described to be input to the specified data
Before the step of preset Forecasting Model of Forest Fire based on ConvLSTM models carries out operation, including:
Obtain the geographical location of the wood land;
Call the forest fire in the geographical location of the corresponding wood land pre- according to the geographical location of the wood land
Survey model, wherein the Forecasting Model of Forest Fire includes that multiple, different Forecasting Model of Forest Fire is to be directed to different geography
Position, and based on the Forecasting Model of Forest Fire of ConvLSTM model foundations.
7. forest fire prediction technique according to claim 1, which is characterized in that described to be input to the specified data
Before the step of preset Forecasting Model of Forest Fire based on ConvLSTM models carries out operation, including:
Obtain the landforms attribute of the wood land;
Call the forest fire of the landforms attribute of the corresponding wood land pre- according to the landforms attribute of the wood land
Survey model, wherein the Forecasting Model of Forest Fire includes that multiple, different Forecasting Model of Forest Fire is to be directed to different landforms
Attribute, and based on the Forecasting Model of Forest Fire of ConvLSTM model foundations.
8. a kind of forest fire prediction meanss, which is characterized in that including:
Acquiring unit, the specified data of the wood land for obtaining forest fire to be predicted;
Arithmetic element, for the specified data to be input to the preset Forecasting Model of Forest Fire based on ConvLSTM models
Carry out operation;
Determination unit, the fire coefficient for determining the wood land according to operation result.
9. a kind of computer equipment, including memory and processor, the memory are stored with computer program, feature exists
In when the processor executes the computer program the step of any one of realization claim 1 to 7 the method.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of method described in any one of claim 1 to 7 is realized when being executed by processor.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810345317.7A CN108717582A (en) | 2018-04-17 | 2018-04-17 | Forest fire prediction technique, device, computer equipment and storage medium |
PCT/CN2018/095485 WO2019200743A1 (en) | 2018-04-17 | 2018-07-12 | Forest fire prediction method, apparatus, computer device, and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810345317.7A CN108717582A (en) | 2018-04-17 | 2018-04-17 | Forest fire prediction technique, device, computer equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108717582A true CN108717582A (en) | 2018-10-30 |
Family
ID=63899097
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810345317.7A Pending CN108717582A (en) | 2018-04-17 | 2018-04-17 | Forest fire prediction technique, device, computer equipment and storage medium |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN108717582A (en) |
WO (1) | WO2019200743A1 (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110119682A (en) * | 2019-04-04 | 2019-08-13 | 北京理工雷科电子信息技术有限公司 | A kind of infrared remote sensing Image Fire point recognition methods |
CN110220554A (en) * | 2019-06-05 | 2019-09-10 | 北京百度网讯科技有限公司 | Method and apparatus for output information |
CN111523277A (en) * | 2020-05-19 | 2020-08-11 | 珠海格力电器股份有限公司 | Fire prediction method, storage device and processor |
CN111625999A (en) * | 2020-05-29 | 2020-09-04 | 中南林业科技大学 | Forest fire early warning model and system based on deep learning technology |
CN112950880A (en) * | 2021-01-26 | 2021-06-11 | 特斯联科技集团有限公司 | Fire early warning method and system based on big data |
CN113362553A (en) * | 2021-06-08 | 2021-09-07 | 内蒙古工业大学 | Grassland fire monitoring and early warning method and device based on multi-source heterogeneous data |
CN113408776A (en) * | 2020-12-21 | 2021-09-17 | 电子科技大学 | Sichuan wildfire risk early warning method based on time dimension characteristic enhancement |
CN113903136A (en) * | 2021-12-13 | 2022-01-07 | 环球数科集团有限公司 | Forest fire emergency situation modeling analysis system |
CN114005237A (en) * | 2021-10-30 | 2022-02-01 | 南京林业大学 | Forest fire identification method and equipment based on thermal imaging analysis technology and computer storage medium |
CN114298392A (en) * | 2021-12-23 | 2022-04-08 | 电子科技大学 | Fire wire longitude and latitude prediction algorithm for forest fire |
RU2797748C2 (en) * | 2021-09-30 | 2023-06-08 | Общество с ограниченной ответственностью "НЕДРЕМЛЮЩЕЕ ОКО" | Intelligent system for detecting and predicting events based on neural networks |
CN117575320A (en) * | 2023-11-29 | 2024-02-20 | 中国科学院空天信息创新研究院 | Forest fire risk prediction method based on graph convolution neural network |
CN117854212A (en) * | 2023-12-25 | 2024-04-09 | 湖南赛德雷特卫星科技有限公司 | Forest fire monitoring method based on deep convolutional neural network and remote sensing image |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR102473061B1 (en) * | 2020-10-23 | 2022-12-05 | 대한민국 | Apparatus and method for providing the forest fire risk index |
KR102496878B1 (en) * | 2021-01-13 | 2023-02-07 | 대한민국 | Apparatus and method for providing the forest fire risk index based on forest fire risk seasonal prediction model |
CN117113854B (en) * | 2023-10-18 | 2024-02-02 | 珠江水利委员会珠江水利科学研究院 | Salt tide forecasting method based on ConvLSTM and three-dimensional numerical simulation |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102663082A (en) * | 2012-04-06 | 2012-09-12 | 昆明理工大学 | Forest fire forecasting method based on data mining |
CN103605899A (en) * | 2013-11-29 | 2014-02-26 | 民政部国家减灾中心 | Regional fire risk assessment method |
JP2014093063A (en) * | 2012-11-07 | 2014-05-19 | Pasco Corp | Prediction device of forest fire and prediction program of forest fire |
CN104615848A (en) * | 2014-12-26 | 2015-05-13 | 中国南方电网有限责任公司 | Comprehensive application method for integrating forest fire danger forecasting and ground fire point monitoring |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
PT102617B (en) * | 2001-05-30 | 2004-01-30 | Inst Superior Tecnico | COMPUTER-CONTROLLED LIDAR SYSTEM FOR SMOKING LOCATION, APPLICABLE, IN PARTICULAR, TO EARLY DETECTION OF FIREFIGHTERS |
CN1529291A (en) * | 2003-10-16 | 2004-09-15 | 华南理工大学 | Forest fire intelligent moritoring and alarming device |
US20160034812A1 (en) * | 2014-07-31 | 2016-02-04 | Qualcomm Incorporated | Long short-term memory using a spiking neural network |
CN105551168B (en) * | 2015-12-03 | 2017-10-20 | 中国科学院遥感与数字地球研究所 | A kind of method and system of farmland remote sensing monitoring and early warning fire |
-
2018
- 2018-04-17 CN CN201810345317.7A patent/CN108717582A/en active Pending
- 2018-07-12 WO PCT/CN2018/095485 patent/WO2019200743A1/en active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102663082A (en) * | 2012-04-06 | 2012-09-12 | 昆明理工大学 | Forest fire forecasting method based on data mining |
JP2014093063A (en) * | 2012-11-07 | 2014-05-19 | Pasco Corp | Prediction device of forest fire and prediction program of forest fire |
CN103605899A (en) * | 2013-11-29 | 2014-02-26 | 民政部国家减灾中心 | Regional fire risk assessment method |
CN104615848A (en) * | 2014-12-26 | 2015-05-13 | 中国南方电网有限责任公司 | Comprehensive application method for integrating forest fire danger forecasting and ground fire point monitoring |
Non-Patent Citations (5)
Title |
---|
XINGJIAN SHI ET AL: "Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting", 《HTTPS://ARXIV.ORG/ABS/1506.04214V1》 * |
何诚 等: "基于MODIS数据的森林火险时空分异规律研究", 《光谱学与光谱分析》 * |
崔学明 等: "《农业气象学》", 31 August 2006 * |
闵连权: "《地理空间数据隐藏与数字水印》", 31 January 2015, 测绘出版社 * |
黄宝华 等: "基于综合火险指数的森林火险预报", 《消防科学与技术》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110119682A (en) * | 2019-04-04 | 2019-08-13 | 北京理工雷科电子信息技术有限公司 | A kind of infrared remote sensing Image Fire point recognition methods |
CN110220554A (en) * | 2019-06-05 | 2019-09-10 | 北京百度网讯科技有限公司 | Method and apparatus for output information |
CN111523277B (en) * | 2020-05-19 | 2022-08-16 | 珠海格力电器股份有限公司 | Fire prediction method, storage device and processor |
CN111523277A (en) * | 2020-05-19 | 2020-08-11 | 珠海格力电器股份有限公司 | Fire prediction method, storage device and processor |
CN111625999A (en) * | 2020-05-29 | 2020-09-04 | 中南林业科技大学 | Forest fire early warning model and system based on deep learning technology |
CN113408776B (en) * | 2020-12-21 | 2023-03-28 | 电子科技大学 | Sichuan wildfire risk early warning method based on time dimension feature enhancement |
CN113408776A (en) * | 2020-12-21 | 2021-09-17 | 电子科技大学 | Sichuan wildfire risk early warning method based on time dimension characteristic enhancement |
CN112950880A (en) * | 2021-01-26 | 2021-06-11 | 特斯联科技集团有限公司 | Fire early warning method and system based on big data |
CN113362553A (en) * | 2021-06-08 | 2021-09-07 | 内蒙古工业大学 | Grassland fire monitoring and early warning method and device based on multi-source heterogeneous data |
RU2797748C2 (en) * | 2021-09-30 | 2023-06-08 | Общество с ограниченной ответственностью "НЕДРЕМЛЮЩЕЕ ОКО" | Intelligent system for detecting and predicting events based on neural networks |
CN114005237A (en) * | 2021-10-30 | 2022-02-01 | 南京林业大学 | Forest fire identification method and equipment based on thermal imaging analysis technology and computer storage medium |
CN113903136B (en) * | 2021-12-13 | 2022-02-22 | 环球数科集团有限公司 | Forest fire emergency situation modeling analysis system |
CN113903136A (en) * | 2021-12-13 | 2022-01-07 | 环球数科集团有限公司 | Forest fire emergency situation modeling analysis system |
CN114298392A (en) * | 2021-12-23 | 2022-04-08 | 电子科技大学 | Fire wire longitude and latitude prediction algorithm for forest fire |
CN117575320A (en) * | 2023-11-29 | 2024-02-20 | 中国科学院空天信息创新研究院 | Forest fire risk prediction method based on graph convolution neural network |
CN117575320B (en) * | 2023-11-29 | 2024-04-30 | 中国科学院空天信息创新研究院 | Forest fire risk prediction method based on graph convolution neural network |
CN117854212A (en) * | 2023-12-25 | 2024-04-09 | 湖南赛德雷特卫星科技有限公司 | Forest fire monitoring method based on deep convolutional neural network and remote sensing image |
Also Published As
Publication number | Publication date |
---|---|
WO2019200743A1 (en) | 2019-10-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108717582A (en) | Forest fire prediction technique, device, computer equipment and storage medium | |
Arieira et al. | Integrating field sampling, geostatistics and remote sensing to map wetland vegetation in the Pantanal, Brazil | |
Fisher et al. | Remote sensing of trophic cascades: multi‐temporal landsat imagery reveals vegetation change driven by the removal of an apex predator | |
Wan et al. | Mapping aboveground biomass of four typical vegetation types in the Poyang Lake wetlands based on random forest modelling and landsat images | |
Pujiono et al. | A three decades assessment of forest cover changes in the mountainous tropical forest of Timor Island, Indonesia | |
Artan et al. | Testing NDVI, tree cover density and land cover type as fuel indicators in the wildfire spread capacity index (WSCI): case of Montenegro | |
Henbo et al. | Long‐term canopy dynamics in a large area of temperate old‐growth beech (Fagus crenata) forest: analysis by aerial photographs and digital elevation models | |
Nkundabose | Employing remote sensing tools for assessment of land use/land cover (LULC) changes in eastern Province, Rwanda | |
Wessels et al. | Spatial natural resource monitoring in Mpumalanga Province of South Africa | |
Prasad et al. | Assessing forest canopy closure in a geospatial medium to address management concerns for tropical islands—Southeast Asia | |
Kambombe | Impact of climate variability and land use change on stream flow in Lake Chilwa basin, Malawi | |
Sun et al. | Hazard or non-hazard flood: Post analysis for paddy rice, wetland, and other potential non-hazard flood extraction from the VIIRS flood products | |
Prior et al. | Multi‐decadal stability of woody cover in a mesic eucalypt savanna in the Australian monsoon tropics | |
Shahzad et al. | Does livelihood vulnerability index justify the socio-economic status of mountainous community? A case study of post-earthquake ecological adaptation of Balakot population | |
Valencia et al. | Water sufficiency for cacao production in the Sierra Nevada de Santa Marta (SNSM) region, Colombia | |
Yelwa | Broadscale vegetation change assessment across Nigeria from coarse spatial and high temporal resolution AVHRR data | |
Hamed et al. | Quantifying the Historical Development of Abugadaf Natural Forest Using GIS-Remote Sensing Analytical Techniques/Blue Nile State/Sudan | |
Rahman et al. | Land Use and Land Cover Change Using Remote Sensing and GIS: A Case Study on Bakerganj Upazila, Barishal, Bangladesh | |
Ostojić et al. | Geospatial analysis of land cover changes in Bačka from 1990. to 2018. | |
Jangra et al. | A study on deforestation in hilly areas of Haryana using Remote Sensing and GIS technique | |
Hasbrouck et al. | ABoVE: Environmental Conditions During Fall Moose Hunting Seasons, Alaska, 2000-2016. ORNL DAAC, Oak Ridge, Tennessee, USA | |
Koirala | land use/land cover change and its impact on soil erosion process in Begnas Tal Rupa Tal watershed using geospatial tools, Kaski district, Nepal | |
HOANG | Analysis of Natural Forest and Plantation Forest Changes in Vietnam Using Remote Sensing Data | |
Vetrita | Using New and Long-Term Multi-Scale Remotely Sensed Data to Detect Recurrent Fires and Quantify Their Relationship to Land Cover/Use in Indonesian Peatlands | |
Koukal et al. | Mapping and monitoring of tree resources outside the forest in central America |
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: 20181030 |