CN112509285B - Global typhoon message collection method and system based on convolutional neural network CNN - Google Patents
Global typhoon message collection method and system based on convolutional neural network CNN Download PDFInfo
- Publication number
- CN112509285B CN112509285B CN202011273051.3A CN202011273051A CN112509285B CN 112509285 B CN112509285 B CN 112509285B CN 202011273051 A CN202011273051 A CN 202011273051A CN 112509285 B CN112509285 B CN 112509285B
- Authority
- CN
- China
- Prior art keywords
- typhoon
- message
- cnn
- classification
- module
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000013527 convolutional neural network Methods 0.000 title claims description 85
- 238000003860 storage Methods 0.000 claims abstract description 17
- 230000002085 persistent effect Effects 0.000 claims abstract description 8
- 230000008569 process Effects 0.000 claims description 28
- 238000012549 training Methods 0.000 claims description 27
- 238000000605 extraction Methods 0.000 claims description 20
- 238000004458 analytical method Methods 0.000 claims description 19
- 230000003993 interaction Effects 0.000 claims description 15
- 238000011176 pooling Methods 0.000 claims description 14
- 230000002688 persistence Effects 0.000 claims description 13
- 238000012795 verification Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 6
- 230000009193 crawling Effects 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000015572 biosynthetic process Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 230000008030 elimination Effects 0.000 claims description 3
- 238000003379 elimination reaction Methods 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 230000008520 organization Effects 0.000 claims description 2
- 238000002360 preparation method Methods 0.000 claims description 2
- 230000005540 biological transmission Effects 0.000 abstract description 9
- 238000012544 monitoring process Methods 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 14
- 238000009826 distribution Methods 0.000 description 8
- 238000001228 spectrum Methods 0.000 description 7
- 230000004913 activation Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000001186 cumulative effect Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000029087 digestion Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- JTJMJGYZQZDUJJ-UHFFFAOYSA-N phencyclidine Chemical class C1CCCCN1C1(C=2C=CC=CC=2)CCCCC1 JTJMJGYZQZDUJJ-UHFFFAOYSA-N 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/02—Alarms for ensuring the safety of persons
- G08B21/10—Alarms for ensuring the safety of persons responsive to calamitous events, e.g. tornados or earthquakes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A50/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE in human health protection, e.g. against extreme weather
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Environmental & Geological Engineering (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geology (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a CNN-based global typhoon message collection method and a CNN-based global typhoon message collection system, and belongs to the field of typhoon monitoring. The global typhoon message collection method based on the CNN is used for classifying and monitoring typhoons on the basis of constructing a CNN-based typhoon classification and identification model, when tropical cyclone is identified to be generated, the tropical cyclone generation sea area is positioned, message data are actively extracted from a reporting center to which the positioning sea area belongs, and then according to the actively extracted message data, typhoon messages are compiled and sent to a typhoon early warning release platform, and when effective typhoon messages are judged, complete typhoon messages are collected and stored persistently. The invention can actively identify the typhoon early warning information, and acquire the typhoon message from the WMO after identifying the typhoon early warning, thereby saving the transmission distance, improving the transmission timeliness, simultaneously improving the integrity of the message and reducing the resource consumption of the message persistent storage.
Description
Technical Field
The invention belongs to the field of meteorological observation, and particularly relates to a global typhoon message collection method and system based on a Convolutional Neural Network (CNN).
Background
Like the observation data of the satellite and the meteorological station, the typhoon message is a kind of data for recording the observation data of the occurrence, development and ending states of typhoon, the intensity and path prediction, and the typhoon process is analyzed according to the data in the typhoon message, so that the typhoon initial value quality in future climate condition simulation is improved, and the typhoon message is an important data for improving the typhoon prediction accuracy.
The world weather organization (WMO) records 31 reporting centers for reporting the typhoon process occurring in 8 sea areas around the world. With the refinement of the forecast work, the WMO Tropical Cyclone Plan (TCP) divides the global sea area into 13 areas in total. Table 1 shows the ranges and names of the 13 sea areas.
TABLE 1
As shown in table 1, the typhoon warning information of 13 sea areas is respectively responsible for 6 regional centers (RSMCs) and 6 tropical warning centers (TCWCs), which ensures that each part contains 1-2 RSMCs or TCWCs, and for the sea area involved in each typhoon process, each member country and regional center (NMHSs) can also compile typhoon warnings and warnings to form a regional coordination system, so as to ensure that the damage caused by life loss and tropical cyclones is minimized.
Fig. 1 shows a flow chart of global typhoon message collection in the prior art. As shown in fig. 1, when there is a first time of generating tropical cyclone, the RSMCs, TCWCs, and NMHSs in the sea area compile and send early warning information, and upload a typhoon message to a superior WMO sub-node to which the compiling and reporting center belongs through the respective compiling and reporting center; the sub-nodes are uploaded to directly-affiliated WMO main nodes, the main nodes are uploaded to a WMO typhoon early warning release platform, the main nodes complete information summarization on the WMO platform, exchange and acquire message information of other main nodes in the sea area and then send the message information to the sub-nodes, and the sub-nodes receive the information and then send the message information to each compiling center; and each editing center extracts, analyzes and stores the received data messages in a warehouse and completes persistent storage.
As can be seen from fig. 1, in the prior art, a typhoon message undergoes a process of repeated forwarding and distribution, such as WMO subnode, master node, subnode, and the like, for a weather phenomenon that the process changes rapidly, such as typhoon, path prediction needs to be performed by highly time-efficient application of observation data, after the typhoon is formed and enters a 24-hour warning area, current time positioning and next time positioning alarm need to be performed every 1 hour, time delay caused by reciprocating transmission is caused, and timeliness of information transmission is low; the issuing of the main node and the sub-node of the GTS is limited by the communication bandwidth, the message is selectively transmitted with a selection, is selective, and is not issued with all information, so that the types of the collected information of the typhoon message are incomplete and the message is discontinuous; the collected typhoon messages are only distributed by a main node of the international communication system (GTS), the collection channel is single and passive, the most needed message data cannot be collected, and the data collection is not comprehensive; meanwhile, the message is acquired by timing triggering of the GTS node, and although the implementation is simple, the computing resources are excessively wasted in the event weather process such as typhoon.
Disclosure of Invention
In view of the above problems, embodiments of the present invention provide a global typhoon message collection method and a global typhoon message collection system based on a convolutional neural network CNN, which actively identify typhoon early warning information, and acquire a typhoon message from a WMO after identifying the typhoon early warning, so as to improve transmission timeliness and improve message integrity.
In order to achieve the above purpose, the embodiment of the present invention adopts the following technical solutions:
in a first aspect, an embodiment of the present invention provides a global typhoon message collection method and a global typhoon message collection system based on a convolutional neural network CNN, where the method includes the following steps:
step S1, collecting infrared IR images of multi-source satellites (MSG, Meteosat 5, MTSAT and GOES-W, GOES-E satellites) and storing the images into big typhoon image data;
step S2, constructing a typhoon classification and identification model based on CNN, and training and verifying the model based on typhoon image big data to obtain a typhoon classification and identification model based on CNN;
step S3, capturing early warning images from information issued by all global early warning information platforms of the typhoon websites at regular time;
step S4, according to the constructed typhoon classification recognition model based on the CNN, the captured early warning image is used as model input to monitor whether tropical cyclone is generated; when it is recognized that the tropical cyclone is generated, proceeding to step S5; when no thermal zone is generated, the process proceeds to step S3;
step S5, positioning the tropical cyclone generation sea area, and actively extracting message data from the positioning sea area affiliated reporting centers RSMCs, TCWCs and NMHSs;
and step S6, according to the actively extracted message data, compiling and sending the typhoon message, storing the typhoon message into a persistent database, and uploading the typhoon message to a typhoon early warning and issuing platform.
As a preferred embodiment of the present invention, the model building process of step S2 specifically includes:
step S21, constructing a typhoon classification recognition model based on the CNN, wherein the model comprises: a convolution layer, a pooling layer, and a full-link layer;
step S22, performing data set input preparation based on typhoon image big data, and defining train and validation; preprocessing the image by using ImageDataenerator class in tf.keras, converting the image into a floating point tensor, and using the floating point tensor as a training set and a verification set of an input model;
and step S23, training and verifying the CNN-based typhoon classification and identification model by using a fit _ generator function with the training set and the verification set as input, and obtaining a mature typhoon classification and identification model.
As a preferred embodiment of the present invention, in the model construction process of step S21, a replanning of CNN pattern determination recognition classification is performed based on the tropical cyclone stage of the satellite image.
As a preferred embodiment of the present invention, the replanning is as follows:
according to the wind power grade and name, the generation and the elimination of typhoon in the west tai region are defined as 8 formation stages, which are respectively: tropical disturbance-wind power 6 level, tropical low-voltage-wind power 7 level, tropical storm-wind power 8-9 level, strong tropical storm-wind power 10-11 level, typhoon-wind power 12-13 level, strong typhoon-wind power 14-16 level, super typhoon-wind power 17 level or above, and typhoon ending.
As a preferred embodiment of the present invention, the classification, which achieves two classification goals, the first is classified into 2 categories according to the presence or absence of typhoon; the second classification is classified into 3 types according to typhoon generation, strongest typhoon and no typhoon; and selecting a CNN-based typhoon classification recognition model which is more suitable for the service situation by comparing the two classification targets.
As a preferred embodiment of the invention, a typhoon classification and identification model is built by tf.keras.models.S-equal, the model is subjected to convolution for 3 times and pooling for 2 times, and dropout is introduced to prevent overfitting; after completing the convolution 3 times and pooling 2 times, the multidimensional matrix is compressed into one dimension using the Flatten () function as an input to the Dense () function to generate the fully connected layer.
The typhoon images are pictures captured from the early warning information publishing platforms of all the typhoon websites through picture timing crawling or timing specified extraction.
In a second aspect, an embodiment of the present invention further provides a global typhoon message collection system, where the global typhoon message collection system includes: the system comprises an infrared IR image acquisition module, a crawler module, a typhoon classification and identification module based on a CNN model, a message active extraction module, a message analysis module, a GTS typhoon message interaction module and a message persistence storage module;
the infrared IR image acquisition module is connected with the typhoon classification and identification module based on the CNN model and is used for collecting infrared IR images of multi-source satellites (MSG, Metasat 5, MTSAT and GOES-W, GOES-E satellites), storing the infrared IR images into typhoon image big data and using the typhoon image big data as a typhoon classification and identification model training and verification data set based on the CNN;
the crawler module is connected with the typhoon classification and identification module based on the CNN model and is used for capturing early warning images from information issued by early warning information platforms of all global typhoon websites at regular time;
the typhoon classification and identification module based on the CNN model is connected with the message active extraction module and used for constructing and training a CNN-based typhoon classification and identification model and judging whether tropical cyclone is generated or not according to the model by taking an early warning image captured at fixed time as input; when tropical cyclone is generated, sending a typhoon generation instruction to a message active extraction module; when no tropical cyclone is generated, sending a typhoon-free instruction to the GTS typhoon message interaction module, and continuing to capture the early warning image;
the message active extraction module is connected with the message analysis module and used for positioning the tropical cyclone generation sea area when receiving a typhoon generation instruction, actively extracting message data from the positioning sea area affiliated reporting centers RSMCs, TCWCs and NMHSs and sending the message data to the message analysis module;
the message analysis module is connected with the GTS typhoon message interaction module and the message persistence storage module, and when receiving message data actively extracted by the message active extraction module, analyzes the message data to generate a typhoon message and sends the typhoon message to the GTS typhoon message interaction module and the message persistence storage module;
the GTS typhoon message interaction module is used for receiving the typhoon message which is analyzed by the message analysis module and contains the actively extracted message data when the tropical cyclone is generated, and uploading the typhoon message to the typhoon early warning and issuing platform;
the message persistence storage module is used for persistently storing the typhoon message.
The invention has the following beneficial effects:
the global typhoon message collection method and system based on the convolutional neural network CNN provided by the embodiment of the invention adopt typhoon image big data to construct a CNN-based typhoon classification recognition model, classify and monitor typhoons, when tropical cyclone generation is recognized, position the tropical cyclone generation sea area, actively extract message data from the positioning sea area affiliated reporting centers RSMCs, TCWCs and NMHSs, then compile and send typhoon messages according to the actively extracted message data and upload the typhoon messages to a typhoon early warning release platform, and when the valid typhoon messages are judged, collect complete typhoon messages and carry out persistent storage. According to the invention, typhoon early warning information is actively identified through the typhoon classification identification model, and the typhoon message is acquired from the WMO after typhoon early warning is identified, so that the transmission distance is saved, the transmission timeliness is improved, and the message integrity is improved; the model identification process and the message acquisition process are matched, so that the resource consumption of message persistent storage is reduced.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a flow chart of a global typhoon message collection in the prior art;
fig. 2 is a flowchart of a global typhoon message collection method and a global typhoon message collection system based on a convolutional neural network CNN according to an embodiment of the present invention;
FIG. 3 is a sample cumulative distribution graph of 1351 typhoon events in one embodiment of the invention;
FIG. 4 is an exemplary graph of a grade 6-11 wind chart image which does not reach the grade of typhoon in the 2018 Tanmamei typhoon monitoring;
fig. 5 is an exemplary graph of a 12-grade and above wind power image reaching a typhoon grade in 2018 tan america typhoon monitoring;
fig. 6 is a structural diagram of a classification recognition model of typhoon based on CNN constructed in the embodiment of the invention;
fig. 7 is a schematic view of a global typhoon message collection system structure and the overall situation according to an embodiment of the present invention;
FIG. 8 is a crawler module crawling schematic in an embodiment of the present invention.
Detailed Description
The technical problems, aspects and advantages of the invention will be apparent from the following detailed description, which proceeds with reference to the accompanying drawings, when taken in conjunction with the accompanying exemplary embodiments. The following exemplary embodiments are merely illustrative of the present invention and are not to be construed as limiting the invention. It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The embodiment of the invention provides a global typhoon message collecting method and system based on a Convolutional Neural Network (CNN), which can actively acquire a received typhoon message and fulfill the aims of collecting and supplementing the collected message in advance by deep learning and training of a typhoon recognition module.
Fig. 2 shows a flowchart of a global typhoon message collection method and a global typhoon message collection system based on the convolutional neural network CNN according to this embodiment. As shown in fig. 2, the method comprises the steps of:
step S1, collecting infrared IR images of multi-source satellites (MSG, Meteosat 5, MTSAT and GOES-W, GOES-E satellites) and storing the images into big typhoon image data;
step S2, constructing a typhoon classification and identification model based on CNN, and training and verifying the model based on typhoon image big data to obtain a typhoon classification and identification model based on CNN;
step S3, capturing early warning images from information issued by all global early warning information platforms of the typhoon websites at regular time;
step S4, according to the constructed typhoon classification recognition model based on the CNN, the captured early warning image is used as model input to monitor whether tropical cyclone is generated; when it is recognized that the tropical cyclone is generated, proceeding to step S5; when no thermal zone is generated, the process proceeds to step S3;
step S5, positioning the tropical cyclone generation sea area, and actively extracting message data from the positioning sea area affiliated reporting centers RSMCs, TCWCs and NMHSs;
and step S6, according to the actively extracted message data, compiling and sending the typhoon message, storing the typhoon message into a persistent database, and uploading the typhoon message to a typhoon early warning and issuing platform.
As described above, the model building process of step S2 specifically includes:
and step S21, constructing a typhoon classification recognition model based on the CNN.
In the actual typhoon monitoring, the naming and division of tropical cyclone digestion stages in different sea areas are different. Table 2 shows a summary of prior art wind levels for different tropical cyclone strengths, central maximum wind speed and the name of the sea area to which it belongs. As shown in table 2, since the units used in the respective reporting centers are different, 3 units are used for the maximum wind speed near the center, and the division and the naming of each stage are also different in each sea area. For example, when the wind rating is class 6, low pressure is called in north indian ocean area v, tropical disturbance in south-west indian ocean area vi, and the rest tropical low pressure; when the wind power level is 7; in the north indian ocean area v is called deep hypotony, in the south west indian ocean area vi is called tropical hypotony, and the rest is still called tropical hypotony.
TABLE 2
Therefore, when collecting typhoon images based on satellite views to form big data for model training, in addition to considering the application of CNN patterns to perform feature spectrum extraction on the shapes and features of the images, the CNN pattern recognition classification number is determined according to the classification rules of table 2 in combination with the research targets, that is: replanning of the CNN pattern determination recognition classification is performed based on the tropical cyclone phase to which the satellite image belongs.
In this step, the replanning is as follows:
firstly, according to the wind power grade and name, the definition of typhoon generation and elimination in the west tai region is 8 formation stages, which are respectively: the method comprises the steps of carrying out 1351 sample cumulative distribution analysis on typhoon processes according to the wind power level when typhoon is generated and the wind power level when typhoon is developed to the strength peak value, wherein the sample cumulative distribution analysis comprises tropical disturbance (wind power level 6), tropical low pressure (wind power level 7), tropical storm (wind power level 8-9), strong tropical storm (wind power level 10-11), strong tropical storm (wind power level 12-13), strong typhoon (wind power level 14-16), super strong typhoon (wind power level 17 and above), typhoon end and the like, and a statistical distribution graph is shown in fig. 3. As shown in fig. 3, the sample distribution analysis of the 1351 typhoon process is performed, and the sample distribution of the initial wind power level generated by each typhoon is concentrated in 6-9 levels of wind power, accounting for 99.4% of the total samples of the typhoon; typhoons develop to peak intensity, mostly distributed in levels of 12-17 and above, accounting for about 52.8% of the total samples.
Meanwhile, by taking the typhoon of Tantanei (Trami) in 2018 as an example, as shown in fig. 4 and 5, it can be seen from the image characteristics and the stage division of the typhoon that the image with the wind power of 6-11 levels is compared with the image with the wind power of more than 12 levels, and no obvious structure of the eye wall of the typhoon is formed because the stage of the typhoon is not reached.
Therefore, the CNN model is adopted in the step, two classification targets are realized, and the first classification target is classified into 2 types according to the existence of typhoon; the second classification is classified into 3 types according to typhoon generation, strongest typhoon and no typhoon; by comparing the two classification targets, a CNN-based typhoon classification recognition model more suitable for service situations is selected, so that the deep learning model is better adapted, and the image characteristic spectrum is more accurately analyzed.
Based on this, the CNN-based typhoon classification and identification model constructed in this step, as shown in fig. 6, includes: conv represents the convoluting layer, MaxPholing represents the pooling layer, Dropout is for model overfitting, Flatten is the transition layer from the convoluting layer to the fully-connected layer, and Dense represents the fully-connected layer. As shown in fig. 6, according to the collected multi-source satellite infrared images, classifying the satellite images according to 3 types of typhoons, which are generated, have the strongest typhoons and have no typhoons, as preprocessing, then inputting the images into a convolution layer, a pooling layer, a transition layer and a full connection layer with a target of 'minimized prediction error', extracting image geometric features, spectral features and the like of 3 types of typhoons in each layer, and finally obtaining a trained typhoon classification recognition model CNN-typhoon.
Preferably, a typhoon classification and identification model is built by tf, keras, models, S-equivalent, the model is subjected to rolling for 3 times and pooling for 2 times, and dropout is introduced to prevent overfitting; after completing the convolution 3 times and pooling 2 times, the multidimensional matrix is compressed into one dimension using the Flatten () function as an input to the Dense () function to generate a fully connected layer, and the model is compiled using the model's build function.
Step S22, preprocessing typhoon image big data;
in this step, the typhoon image big data is from the satellite IR image as shown in fig. 6. For the prepared typhoon image big data, preparing data set input, and defining train and validation; and preprocessing the image by using an imagedataactor class in tf.keras, and converting the image into a floating point tensor to be used as a training set and a verification set of the input model.
And step S23, training and verifying the CNN-based typhoon classification and identification model to obtain a mature typhoon classification and identification model.
And training the CNN-based typhoon classification and identification model by using a fit _ generator function by taking the training set and the verification set as input.
In the training process, after a typhoon infrared satellite image is given, the typhoon infrared satellite image enters a CNN network through image preprocessing, the characteristics of geometric characteristics, spectrum and the like of the typhoon image with certain invariance are extracted by taking the minimum prediction error as a target, and finally the final result of typhoon classification and identification is obtained on an output layer.
The convolutional neural network updates the weight of each neuron by applying a back propagation rule, so that the overall error of the model is continuously reduced. The convolution layer performs dot product operation with the window coverage range of the input image in a sliding window mode by using the convolution kernel of the layer, then adds offset, activates the output result of the convolution through an activation function, outputs the characteristic spectrum of the layer and realizes image characteristic extraction. Expression (1) of the convolutional layer back propagation rule is:
wherein l represents the number of layers of the convolutional layer,a feature spectrum representing the l-th layer, the j-th input image, the i-th neuron convolution output, f (x) represents an activation function,which represents a weight parameter that is representative of the weight,represents the bias parameters, namely: performing 2-dimensional convolution operation on the extracted characteristic spectrum by using the convolution layer, and further extracting typhoon through an infrared satellite imageGeometric features, spectral features; the sizes of convolution kernels are all 3 x 3, point operation is carried out from the top right corner of the matrix to the bottom left corner in a mode of sliding a window and step length being 1, and element values corresponding to positions smaller than 0 in the output tensor are all changed into 0 through nonlinear calculation by utilizing a ReLU activation function. Here, the activation function is shown in expression (2):
the pooling layer is used for reducing the size of the convolutional layer output characteristic diagram and reducing the number of parameters needing to be learned during convolutional neural network training; and the translation, scaling and rotation are invariant, that is, the pooling layer only reduces the size of the feature map, and does not change the feature spectrum. In the method for selecting the maximum value of the sampling window for the pooling layer in the embodiment, the size of the sampling window is 2 × 2, so that the image characteristics of the typhoon process can be well extracted. Wherein, the expression of the maximum pooling is shown in (3):
according to the living and disappearing process of the tropical cyclone, the classification of the images is realized, and a mechanism for judging whether to trigger message collection is completed; namely: after the image passes through a convolutional layer and a pooling layer, the dimension of the image is reduced through full-connection layer processing, the one-dimensional characteristic is input into a softmax classifier after the two-dimensional characteristic is changed into the one-dimensional characteristic, an objective function is constructed corresponding to cross entropy loss in the training process, and the distance between the probability distribution predicted by the classifier and a true value is calculated through cross entropy (formula 4), so that the degree of proximity between the actual output and the expected output is judged; meanwhile, the Adam optimization algorithm is a first-order optimization algorithm which can replace the traditional random gradient descent process and can iteratively update the weight of the neural network based on training data. And continuously adjusting the learning rate along with the training process through an Adam algorithm, accelerating the optimization process, finally obtaining the classification probability and completing the prediction work of the image category. Wherein, the cross entropy loss function formula is as follows:
here, N is the number of classifications, p (x)a) Is an indicator variable (0 or 1), is 1 if the class is the same as that of sample a, and is 0 otherwise; q (x)a) Representing the predicted probability of belonging to a certain class for the observed sample a.
After the training and verification are completed, calling the save function to store the obtained mature CNN-based typhoon classification and identification model.
As described above, the picture capturing process in step S3 is realized by the time crawling or the time-specified extraction. Wherein preferably a timed crawl is employed. And the timed crawling regularly triggers the crawler through a timed task to acquire WMO and the early warning information release pictures of all global wind turbine sites.
An embodiment of the present invention further provides a typhoon service processing system, as shown in fig. 7, the typhoon service processing system includes: the system comprises an infrared IR image acquisition module, a crawler module, a typhoon classification and identification module based on a CNN model, a message active extraction module, a message analysis module, a GTS typhoon message interaction module and a message persistence storage module.
The infrared IR image acquisition module is connected with the typhoon classification and identification module based on the CNN model and is used for collecting infrared IR images of multi-source satellites (MSG, Metasat 5, MTSAT and GOES-W, GOES-E satellites) and storing the infrared IR images into typhoon image big data to serve as a typhoon classification and identification model training and verification data set based on the CNN.
And the crawler module is connected with the typhoon classification and identification module based on the CNN model and is used for capturing early warning images from the WMO and information issued by each typhoon website early warning information platform at regular time.
The typhoon classification and identification module based on the CNN model is connected with the message active extraction module and used for constructing and training a CNN-based typhoon classification and identification model and judging whether tropical cyclone is generated or not according to the model by taking an early warning image captured at fixed time as input; when tropical cyclone is generated, sending a typhoon generation instruction to a message active extraction module; and when no thermal zone cyclone is generated, sending a typhoon-free instruction to the GTS typhoon message interaction module, and continuing to capture the early warning image.
The message active extraction module is connected with the message analysis module and used for positioning the tropical cyclone generation sea area when receiving a typhoon generation instruction, actively extracting message data from the positioning sea area affiliated reporting centers RSMCs, TCWCs and NMHSs, and sending the message data to the message analysis module.
The message analysis module is connected with the GTS typhoon message interaction module and the message persistence storage module, and when receiving message data actively extracted by the message active extraction module, the message analysis module analyzes the message data to generate a typhoon message and sends the typhoon message to the GTS typhoon message interaction module and the message persistence storage module.
And the GTS typhoon message interaction module is used for receiving the typhoon message which is analyzed by the message analysis module and contains the active extracted message data when the tropical cyclone is generated, and uploading the typhoon message to the typhoon early warning and issuing platform.
The message persistence storage module is used for persistently storing typhoon messages, as shown in fig. 8, the crawler module is triggered according to a time interval or a self-defined rule set by a timing task, the crawler cluster acquires pictures and message files from a WMO early warning information publishing website and all typhoon center early warning information publishing websites around the world, the original files are placed in a message queue cluster, a consumer cluster monitors the message queue cluster, the acquired original files are processed in a certain format and are persisted to a big data cloud platform, and after persistence, the persisted messages and picture files can be searched by using local search service, so that corresponding typhoon information search service is provided for the outside.
The information crawled by the crawler mainly comprises two aspects: (1) typhoon early warning picture includes: WMO typhoon early warning information picture; WMO early warning information pictures of each typhoon center sea area; and (4) warning website pictures of all wind centers in the world. (2) Typhoon message: and all the global typhoon centers synchronize to related data messages such as typhoon early warning information and the like issued by WMO, and the messages can be mutually verified with GTS system messages.
According to the technical scheme, the global typhoon message collection method and system based on the convolutional neural network CNN provided by the embodiment of the invention actively identify typhoon early warning information through the typhoon classification identification model, and actively acquire typhoon messages from WMO and various reporting centers after typhoon early warning is identified, so that the transmission distance is saved, the transmission timeliness is improved, and the message integrity is improved; the model identification process and the message acquisition process are matched, so that the resource consumption of message persistent storage is reduced.
While the foregoing is directed to the preferred embodiment of the present invention, it is understood that the invention is not limited to the exemplary embodiments disclosed, but is made merely for the purpose of providing those skilled in the relevant art with a comprehensive understanding of the specific details of the invention. It will be apparent to those skilled in the art that various modifications and adaptations of the present invention can be made without departing from the principles of the invention and the scope of the invention is to be determined by the claims.
Claims (8)
1. A global typhoon message collection method based on a Convolutional Neural Network (CNN) is characterized by comprising the following steps:
step S1, collecting infrared IR images of the multisource satellite, and storing the infrared IR images into big typhoon image data;
step S2, constructing a typhoon classification and identification model based on CNN, and training and verifying the model based on typhoon image big data to obtain a typhoon classification and identification model based on CNN;
step S3, capturing early warning images from information issued by all global early warning information platforms of the typhoon websites at regular time;
step S4, according to the constructed typhoon classification recognition model based on the CNN, the captured early warning image is used as model input to monitor whether tropical cyclone is generated; when it is recognized that the tropical cyclone is generated, proceeding to step S5; when no thermal zone is generated, the process proceeds to step S3;
step S5, positioning the tropical cyclone generation sea area, and actively extracting message data from the compilation center of the positioning sea area;
and step S6, according to the actively extracted message data, compiling and sending the typhoon message, storing the typhoon message into a persistent database, and uploading the typhoon message to a typhoon early warning and issuing platform.
2. The global typhoon message collection method based on convolutional neural network CNN as claimed in claim 1, wherein the model construction process of step S2 specifically includes:
step S21, constructing a typhoon classification recognition model based on the CNN, wherein the model comprises: a convolution layer, a pooling layer, and a full-link layer;
step S22, performing data set input preparation based on typhoon image big data, and defining train and validation; preprocessing the image by using ImageDataenerator class in tf.keras, converting the image into a floating point tensor, and using the floating point tensor as a training set and a verification set of an input model;
and step S23, training and verifying the CNN-based typhoon classification and identification model by using a fit _ generator function with the training set and the verification set as input, and obtaining a mature typhoon classification and identification model.
3. The global typhoon message collection method based on Convolutional Neural Network (CNN) as claimed in claim 2, wherein in the model construction process of step S21, the re-planning of CNN mode determination recognition classification is performed based on the tropical cyclone stage of the satellite image.
4. The global typhoon message collection method based on Convolutional Neural Network (CNN) as claimed in claim 3, wherein the replanning is as follows:
according to the wind power grade and name, the generation and the elimination of typhoon in the west tai region are defined as 8 formation stages, which are respectively: tropical disturbance-wind power 6 level, tropical low-voltage-wind power 7 level, tropical storm-wind power 8-9 level, strong tropical storm-wind power 10-11 level, typhoon-wind power 12-13 level, strong typhoon-wind power 14-16 level, super typhoon-wind power 17 level or above, and typhoon ending.
5. The global typhoon message collection method based on Convolutional Neural Network (CNN) as claimed in claim 4, wherein said classification realizes two classification targets, the first one is classified into 2 types according to existence of typhoon; the second classification is classified into 3 types according to typhoon generation, strongest typhoon and no typhoon; and selecting a CNN-based typhoon classification recognition model which is more suitable for the service situation by comparing the two classification targets.
6. The global typhoon message collection method based on the convolutional neural network CNN as claimed in any one of claims 2 to 5, characterized in that a typhoon classification recognition model is built by using tf.keras.models.S-equal, the model is rolled up 3 times and pooled 2 times, and dropout is introduced to prevent overfitting; after completing the convolution 3 times and pooling 2 times, the multidimensional matrix is compressed into one dimension using the Flatten () function as an input to the Dense () function to generate the fully connected layer.
7. The global typhoon message collection method based on Convolutional Neural Network (CNN) as claimed in any one of claims 2-5, wherein the typhoon image is a picture captured from world weather organization WMO and early warning information publishing platform of each typhoon website by picture timing crawling or timing specified extraction.
8. A global typhoon message collection system, the global typhoon message collection system comprising: the system comprises an infrared IR image acquisition module, a crawler module, a typhoon classification and identification module based on a Convolutional Neural Network (CNN) model, a message active extraction module, a message analysis module, a GTS typhoon message interaction module and a message persistence storage module; wherein,
the infrared IR image acquisition module is connected with the typhoon classification and identification module based on the CNN model and used for collecting infrared IR images of the multisource satellite, storing the infrared IR images into typhoon image big data and using the typhoon image big data as a training and verification data set of the typhoon classification and identification module based on the CNN;
the crawler module is connected with the typhoon classification and identification module based on the CNN model and is used for capturing early warning images from information issued by early warning information platforms of all global typhoon websites at regular time;
the typhoon classification and identification module based on the CNN model is connected with the message active extraction module and used for constructing and training a CNN-based typhoon classification and identification model and judging whether tropical cyclone is generated or not according to the model by taking an early warning image captured at fixed time as input; when tropical cyclone is generated, sending a typhoon generation instruction to a message active extraction module; when no tropical cyclone is generated, sending a typhoon-free instruction to the GTS typhoon message interaction module, and continuing to capture the early warning image;
the message active extraction module is connected with the message analysis module and used for positioning the tropical cyclone generation sea area when receiving a typhoon generation instruction, actively extracting message data from an editing and reporting center to which the positioning sea area belongs and sending the message data to the message analysis module;
the message analysis module is connected with the GTS typhoon message interaction module and the message persistence storage module, and when receiving message data actively extracted by the message active extraction module, analyzes the message data to generate a typhoon message and sends the typhoon message to the GTS typhoon message interaction module and the message persistence storage module;
the GTS typhoon message interaction module is used for receiving the typhoon message which is analyzed by the message analysis module and contains the actively extracted message data when the tropical cyclone is generated, and uploading the typhoon message to the typhoon early warning and issuing platform;
the message persistence storage module is used for persistently storing the typhoon message.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011273051.3A CN112509285B (en) | 2020-11-14 | 2020-11-14 | Global typhoon message collection method and system based on convolutional neural network CNN |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011273051.3A CN112509285B (en) | 2020-11-14 | 2020-11-14 | Global typhoon message collection method and system based on convolutional neural network CNN |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112509285A CN112509285A (en) | 2021-03-16 |
CN112509285B true CN112509285B (en) | 2021-07-09 |
Family
ID=74957774
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011273051.3A Active CN112509285B (en) | 2020-11-14 | 2020-11-14 | Global typhoon message collection method and system based on convolutional neural network CNN |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112509285B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112801227B (en) * | 2021-04-06 | 2021-09-28 | 航天宏图信息技术股份有限公司 | Typhoon identification model generation method, device, equipment and storage medium |
CN113671599A (en) * | 2021-08-25 | 2021-11-19 | 华北电力大学 | Global climate mode-based login cyclone identification method |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6752077B2 (en) * | 2016-08-08 | 2020-09-09 | 矢崎エナジーシステム株式会社 | Management device and management system |
CN106599158A (en) * | 2016-12-07 | 2017-04-26 | 国家海洋局第二海洋研究所 | Quick query method of typhoon sea area Argo information based on space-time dual approximate index |
KR102044161B1 (en) * | 2017-05-31 | 2019-11-13 | 한국생명공학연구원 | Method for controlling pathogenicity of pathogenic bacteria by treating 2R,3R-Butanediol or 2S,3S-Butanediol |
CN109325960B (en) * | 2018-11-20 | 2021-07-09 | 南京信息工程大学 | Infrared cloud chart cyclone analysis method and analysis system |
CN110837137A (en) * | 2019-11-07 | 2020-02-25 | 刘健华 | Typhoon prediction alarm method |
CN111551964B (en) * | 2020-05-27 | 2023-03-31 | 中国气象局气象探测中心 | Water vapor observation method based on navigation satellite system |
CN111695473B (en) * | 2020-06-03 | 2023-12-19 | 中国人民解放军国防科技大学 | Tropical cyclone strength objective monitoring method based on long-short-term memory network model |
-
2020
- 2020-11-14 CN CN202011273051.3A patent/CN112509285B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN112509285A (en) | 2021-03-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113449680B (en) | Knowledge distillation-based multimode small target detection method | |
CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
CN107609525A (en) | Remote Sensing Target detection method based on Pruning strategy structure convolutional neural networks | |
CN107392901A (en) | A kind of method for transmission line part intelligence automatic identification | |
CN111929748A (en) | Meteorological element forecasting method and system | |
CN109508360A (en) | A kind of polynary flow data space-time autocorrelation analysis method of geography based on cellular automata | |
CN112509285B (en) | Global typhoon message collection method and system based on convolutional neural network CNN | |
CN114240000A (en) | Air quality prediction method based on space-time graph convolution network | |
CN114169445A (en) | Day-ahead photovoltaic power prediction method, device and system based on CAE and GAN hybrid network | |
CN112232328A (en) | Remote sensing image building area extraction method and device based on convolutional neural network | |
Han et al. | Kill two birds with one stone: A multi-view multi-adversarial learning approach for joint air quality and weather prediction | |
CN112966853A (en) | Urban road network short-term traffic flow prediction method based on space-time residual error mixed model | |
CN114139760A (en) | Method, system, storage medium and equipment for predicting typhoon path | |
CN117556197B (en) | Typhoon vortex initialization method based on artificial intelligence | |
CN113496310A (en) | Atmospheric pollutant prediction method and system based on deep learning model | |
CN111488974B (en) | Ocean wind energy downscaling method based on deep learning neural network | |
Alburshaid et al. | Palm trees detection using the integration between gis and deep learning | |
CN118131365A (en) | Meteorological monitoring device and method for new energy power generation prediction | |
CN118070010A (en) | Renewable energy scene generation method based on generation countermeasure network | |
CN117392547A (en) | Remote sensing image scene classification method integrating Swin transducer and visual map neural network | |
CN115456238A (en) | Urban trip demand prediction method based on dynamic multi-view coupling graph convolution | |
CN117175535A (en) | Wind power group power prediction method, system, equipment and medium | |
Lestari et al. | Rainfall Prediction using Spatial Convolutional Neural Networks and Recurrent Neural Networks | |
Fan et al. | Construction of a neural network and its application on target classification | |
Xu et al. | Remote Sensing Image Change Detection Based on General Deep Forest Module |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |