CN112198483A - Data processing method, device and equipment for satellite inversion radar and storage medium - Google Patents
Data processing method, device and equipment for satellite inversion radar and storage medium Download PDFInfo
- Publication number
- CN112198483A CN112198483A CN202011042984.1A CN202011042984A CN112198483A CN 112198483 A CN112198483 A CN 112198483A CN 202011042984 A CN202011042984 A CN 202011042984A CN 112198483 A CN112198483 A CN 112198483A
- Authority
- CN
- China
- Prior art keywords
- data
- satellite
- inversion
- radar echo
- network
- 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
- 238000003672 processing method Methods 0.000 title claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 113
- 238000000034 method Methods 0.000 claims abstract description 44
- 238000012545 processing Methods 0.000 claims description 41
- 230000006870 function Effects 0.000 claims description 14
- 238000010606 normalization Methods 0.000 claims description 14
- 230000017105 transposition Effects 0.000 claims description 7
- 208000037516 chromosome inversion disease Diseases 0.000 description 156
- 108091006146 Channels Proteins 0.000 description 82
- 238000010586 diagram Methods 0.000 description 21
- 238000013528 artificial neural network Methods 0.000 description 13
- 238000007781 pre-processing Methods 0.000 description 12
- 238000001556 precipitation Methods 0.000 description 11
- 230000008569 process Effects 0.000 description 9
- 238000006243 chemical reaction Methods 0.000 description 8
- 230000008859 change Effects 0.000 description 5
- 241000208818 Helianthus Species 0.000 description 4
- 235000003222 Helianthus annuus Nutrition 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 238000002310 reflectometry Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000006855 networking Effects 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000003042 antagnostic effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/35—Details of non-pulse systems
- G01S7/352—Receivers
- G01S7/354—Extracting wanted echo-signals
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/95—Radar or analogous systems specially adapted for specific applications for meteorological use
- G01S13/955—Radar or analogous systems specially adapted for specific applications for meteorological use mounted on satellite
-
- 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
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The embodiment of the invention provides a data processing method, a device, equipment and a storage medium for a satellite inversion radar, wherein the method comprises the following steps: acquiring first satellite data of a preset channel of a target area; determining a satellite data sequence with a preset frame number according to the first satellite data of each preset channel; and generating and outputting target radar echo data of the target area by adopting an inversion model obtained by pre-training according to the satellite data sequence with the preset frame number. The method can effectively improve the accuracy of the inversion result, reduce the influence of data irrelevant to radar in the satellite data on the inversion result, and solve the problem that the inversion result is often not accurate enough when the satellite data at a certain time is only adopted to generate radar echo data at a corresponding time in the prior art.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a data processing method, a data processing device, data processing equipment and a storage medium for a satellite inversion radar.
Background
At present, meteorological observation related to precipitation is usually realized based on ground radar networking, however, the detection range of ground radar is limited, a gap exists between radar and radar, and radar data cannot be obtained for an area where radar cannot be installed, such as the sea surface. The detection range of the meteorological satellite is very wide, and the satellite data with high space-time resolution and multi-channel information provides convenience for inverting the radar echo diagram from the satellite data.
In the prior art, neural networks are increasingly used in the inversion of radar echo maps. However, in the existing inversion method based on the neural network, radar echo data of corresponding time is usually generated only according to satellite data of a certain time, and an inversion result is often not accurate enough.
Disclosure of Invention
The embodiment of the invention provides a data processing method, a data processing device, data processing equipment and a data processing storage medium for a satellite inversion radar, and aims to overcome the defects that the inversion result is not accurate in the prior art and the like.
In a first aspect, an embodiment of the present invention provides a data processing method for a satellite inversion radar, including:
acquiring first satellite data of a preset channel of a target area;
determining a satellite data sequence with a preset frame number according to the first satellite data of each preset channel;
and generating and outputting target radar echo data of the target area by adopting an inversion model obtained by pre-training according to the satellite data sequence with the preset frame number.
Optionally, the network architecture of the inversion model at least comprises: the system comprises a 3D convolutional network module, an R (2+1) D residual error network module, a 3D non-local network module and a 3D transposition convolutional network module.
Optionally, the determining a satellite data sequence with a preset frame number according to the first satellite data of each preset channel includes:
generating a preset number of channels according to a preset rule aiming at each frame of first satellite data of each preset channel, and superposing the channels according to the channel direction to form a frame of satellite input data;
and forming the satellite input data of each frame into a continuous satellite data sequence with a preset frame number.
Optionally, the method further comprises:
normalizing each frame of satellite input data to obtain normalized satellite input data;
forming each frame of satellite input data into a continuous satellite data sequence with a preset frame number, wherein the satellite data sequence comprises:
and forming the normalized satellite input data of each frame into a continuous satellite data sequence with a preset frame number.
Optionally, the inverse model is obtained by training:
acquiring a training satellite data sequence and corresponding training radar echo data;
inputting the training satellite data sequence into a preset inversion network, inputting radar echo data output by the inversion network and corresponding training radar echo data into a first identification network, and performing countermeasure training on the inversion network;
and judging the end of training based on a preset loss function to obtain the inversion model.
Optionally, the inputting the training satellite data sequence into a preset inversion network, and inputting radar echo data output by the inversion network and the corresponding training radar echo data into a first identifying network, and performing countermeasure training on the inversion network includes:
and inputting the training satellite data sequence into a preset inversion network, inputting radar echo data output by the inversion network and corresponding training radar echo data into a first discrimination network, forming a radar echo data sequence by the radar echo data output by the inversion network and a first amount of adjacent radar echo data, and inputting the radar echo data sequence and the corresponding training radar echo data sequence into a second discrimination network for confrontation training.
Optionally, the acquiring first satellite data of a preset channel of the target area includes:
acquiring original satellite data of the target area;
and preprocessing the original satellite data to obtain first satellite data of each preset channel, wherein the preprocessing comprises coordinate system conversion and interpolation processing.
Optionally, the method further comprises:
performing inverse normalization processing on the target radar echo data of the target area to obtain first radar echo data;
and generating a target radar echo map of the target area based on the first radar echo data.
Optionally, the method further comprises:
and determining the weather condition of the target area according to the target radar echo map.
In a second aspect, an embodiment of the present invention provides a data processing apparatus for a satellite inversion radar, including:
the acquisition module is used for acquiring first satellite data of a preset channel of a target area;
the determining module is used for determining a satellite data sequence with a preset frame number according to the first satellite data of each preset channel;
and the processing module is used for generating and outputting target radar echo data of the target area by adopting an inversion model obtained by pre-training according to the satellite data sequence with the preset frame number.
Optionally, the network architecture of the inversion model at least comprises: the system comprises a 3D convolutional network module, an R (2+1) D residual error network module, a 3D non-local network module and a 3D transposition convolutional network module.
Optionally, the determining module is specifically configured to:
generating a preset number of channels according to a preset rule aiming at each frame of first satellite data of each preset channel, and superposing the channels according to the channel direction to form a frame of satellite input data;
and forming the satellite input data of each frame into a continuous satellite data sequence with a preset frame number.
Optionally, the determining module is further configured to normalize the satellite input data of each frame to obtain normalized satellite input data;
the determining module is specifically configured to combine each frame of normalized satellite input data into a continuous satellite data sequence with a preset frame number.
Optionally, the obtaining module is further configured to combine each frame of normalized satellite input data into a continuous satellite data sequence with a preset frame number;
the processing module is further configured to:
inputting the training satellite data sequence into a preset inversion network, inputting radar echo data output by the inversion network and corresponding training radar echo data into a first identification network, and performing countermeasure training on the inversion network;
and judging the end of training based on a preset loss function to obtain the inversion model.
Optionally, the processing module is specifically configured to:
and inputting the training satellite data sequence into a preset inversion network, inputting radar echo data output by the inversion network and corresponding training radar echo data into a first discrimination network, forming a radar echo data sequence by the radar echo data output by the inversion network and a first amount of adjacent radar echo data, and inputting the radar echo data sequence and the corresponding training radar echo data sequence into a second discrimination network for confrontation training.
Optionally, the obtaining module is specifically configured to:
acquiring original satellite data of the target area;
and preprocessing the original satellite data to obtain first satellite data of each preset channel, wherein the preprocessing comprises coordinate system conversion and interpolation processing.
Optionally, the processing module is further configured to:
performing inverse normalization processing on the target radar echo data of the target area to obtain first radar echo data;
and generating a target radar echo map of the target area based on the first radar echo data.
Optionally, the processing module is further configured to determine a weather condition of the target area according to the target radar echo map.
In a third aspect, an embodiment of the present invention provides an electronic device, including: at least one processor and memory;
the memory stores computer-executable instructions;
the at least one processor executes computer-executable instructions stored by the memory to cause the at least one processor to perform the method as set forth in the first aspect above and in various possible designs of the first aspect.
In a fourth aspect, the present invention provides a computer-readable storage medium, in which computer-executable instructions are stored, and when the computer-executable instructions are executed by a processor, the method according to the first aspect and various possible designs of the first aspect is implemented.
According to the data processing method, device, equipment and storage medium for the satellite inversion radar, radar echo data of a target area are inverted by adopting a satellite data sequence with preset frame numbers of the target area, in the inversion process, the characteristics of current frame satellite data and historical frame satellite data can be combined, the accuracy of an inversion result can be effectively improved, the influence of data irrelevant to precipitation in the satellite data on the inversion result is reduced, and the problem that in the prior art, radar echo data of corresponding time are generated by only adopting satellite data at a certain time, and the inversion result is often not accurate enough is solved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a block diagram of a processing system according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of a data processing method for a satellite inversion radar according to an embodiment of the present invention;
fig. 3 is a schematic flow chart of a data processing method for a satellite inversion radar according to another embodiment of the present invention;
FIG. 4 is a schematic diagram of a network architecture of an inversion model according to an embodiment of the present invention;
fig. 5 is a schematic diagram of an architecture of an R (2+1) D residual network module in fig. 4 according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of input and output of an authentication network according to an embodiment of the present invention;
fig. 7 is a schematic diagram of an architecture of an authentication network according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of a data processing apparatus for a satellite inversion radar according to an embodiment of the present invention;
fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
With the above figures, certain embodiments of the invention have been illustrated and described in more detail below. The drawings and written description are not intended to limit the scope of the disclosed concepts in any way, but rather to illustrate the concepts of the invention by those skilled in the art with reference to specific embodiments.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The terms to which the present invention relates will be explained first:
non-local network: a non-local neural network.
And (3) GAN: a countermeasure network is generated.
R (2+1) D ResnetBlock: the R (2+1) D residual error network module is a new space-time convolution block and decomposes 3-dimensional space-time convolution into 2-dimensional space convolution and 1-dimensional time convolution.
WGAN-GP: the Wasserstein GAN-Gradient Penalty is provided for the problems of WGAN, and introduces the Gradient Penalty, wherein the core of the Gradient Penalty is to set an additional loss term to realize the relation between the Gradient and K (discriminator Gradient threshold).
Batch GAN: referred to as the arbiter of GAN (also called discriminator), which is replaced by a full convolutional network. A common GAN discriminator is a Patch (matrix) that maps an input to a real number, i.e., the probability that the input sample is a true sample, and a Patch GAN maps the input to NxN.
VGG permanent loss: VGG perception loss.
At present, meteorological observation related to precipitation is usually realized based on ground radar networking, however, the detection range of ground radar is limited, a gap exists between radar and radar, and radar data cannot be obtained for an area where radar cannot be installed, such as the sea surface. The detection range of the meteorological satellite is very wide, and the satellite data with high space-time resolution and multi-channel information provides convenience for inverting the radar echo diagram from the satellite data. In the prior art, neural networks are increasingly used in the inversion of radar echo maps. However, in the existing inversion method based on the neural network, radar echo data of corresponding time is usually generated only according to satellite data of a certain time, and an inversion result is often not accurate enough.
Aiming at the problems in the prior art, the inventor conducts creative research and finds that in the research, the prior art generates radar echo data of corresponding time according to satellite data of certain time, because data irrelevant to precipitation in the satellite data cannot be completely ignored through single frame data, the radar echo data can be influenced to a certain degree, so that an inversion result is not accurate enough, and the existing neural network mainly adopts L1 or L2 as a loss function, so that images generated by training are easy to average, and a model is poor in scoring and visual effects, in order to solve the problems, the inventor creatively finds that a satellite data sequence is formed by combining the satellite data of a current frame and the satellite data of a historical frame to invert the radar echo data of the current time, so that the accuracy of the inversion result can be effectively improved, for example, the influence of data which is irrelevant to precipitation in satellite data on an inversion result is reduced. Therefore, the embodiment of the invention provides a data processing method of a satellite inversion radar, which combines satellite data of a current frame and satellite data of a historical frame to form a satellite data sequence to invert radar echo data of the current moment, can effectively improve the accuracy of an inversion result, for example, reduce the influence of data irrelevant to precipitation in the satellite data on the inversion result, simultaneously, an inversion model adopts a 3D convolution module, an R (2+1) D residual module, a 3D non-local module, a 3D transposition convolution module and other neural network sub-modules as a basis to construct a neural network suitable for generating a radar echo diagram, adopts GAN training, combines an image discriminator (can be called as a first discrimination network) and a video discriminator (can be called as a second discrimination network), discriminates a training result based on a Patch WGAN-GP loss function and a loss function combining VGG perception and L1, the quality of the radar echo map is further improved.
The data processing method for the satellite inversion radar provided by the embodiment of the invention is suitable for scenes needing to invert radar echo data. Such as the acquisition of radar echo data for areas where radar cannot be installed. Fig. 1 is a schematic diagram of an architecture of a processing system according to an embodiment of the present invention. The processing system may include: an electronic device, such as a server. The system can also comprise a satellite pot and weather service equipment. The satellite pan can receive satellite data and send to the electronic equipment, the electronic equipment can cut out satellite data of a target area after receiving the satellite data, and preprocesses the satellite data to obtain first satellite data of preset channels, and further determines a satellite data sequence of preset frame numbers according to the first satellite data of each preset channel, wherein the preset frame numbers can comprise satellite data of a current frame and satellite data of a historical frame, and radar echo data of the target area is obtained through inversion according to the satellite data sequence of the preset frame numbers, and the radar echo data are called as target radar echo data for distinguishing. After the target radar echo data is obtained, the electronic device may determine the weather condition, such as precipitation condition, of the target area according to the target radar echo data, or the weather service device may send the target radar echo data to the weather service device, and the weather service device determines the weather condition of the target area according to the target radar echo data.
Furthermore, the terms "first", "second", etc. are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. In the description of the following examples, "plurality" means two or more unless specifically limited otherwise.
The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present invention will be described below with reference to the accompanying drawings.
An embodiment of the invention provides a data processing method for a satellite inversion radar, which is used for obtaining radar echo data in a certain area. The execution subject of the embodiment is a data processing device of a satellite inversion radar, and the device may be disposed in an electronic device, and the electronic device may be a server, a notebook computer, a desktop computer, or the like.
As shown in fig. 2, a schematic flow chart of a data processing method for a satellite inversion radar provided in this embodiment is shown, where the method includes:
Specifically, satellite data (referred to as original satellite data) can be received through the satellite pan, the received satellite data can be acquired from the satellite pan, satellite data of a target area can be cut out from the satellite data, and preprocessing is performed to acquire first satellite data of a preset channel. The preprocessing may include coordinate system conversion and interpolation processing, and the satellite data of the target area is converted and interpolated onto the grid points of the preset space. The preset channel may include all or a portion of channels involved in satellite data.
For example, taking the new generation of stationary meteorological satellite sunflower No. 8 (Himapari-8, H8) as an example, the full-disk scanning can be completed within 10 minutes, the main channel of the sensor comprises 16 channels from 0.46 to 13.3 μm, wherein the resolution of the visible light and near infrared channels can reach 0.5 to 1km, and the resolution of the infrared channel can also reach 2 km. Satellite data with high spatial-temporal resolution and multi-channel information provides convenience for the inversion of radar echo maps from the satellite data. Taking the sunflower 8 original satellite data as an example, after cutting out the original satellite data of the target area, the original satellite data can be analyzed into data of 14 channels after operations such as coordinate system conversion, interpolation and the like, and satellite data of six channels including IR1, IR2, IR3, B10, B12 and VIS can be used as the first satellite data. Namely, the preset channels refer to six channels of IR1, IR2, IR3, B10, B12 and VIS. In practical application, which channels are selected can be set according to actual requirements, and this embodiment is not limited.
And 102, determining a satellite data sequence with a preset frame number according to the first satellite data of each preset channel.
Specifically, after the first satellite data of the preset channel of the target area is acquired, the satellite data sequence with the preset frame number may be determined according to the first satellite data of each preset channel, and the preset frame number may be set according to actual requirements, such as 5 frames, 6 frames, 7 frames, and so on. The satellite data sequence with the preset frame number comprises satellite data of a current frame and satellite data of a historical frame, and specifically, the satellite data sequence can comprise a brightness temperature value (BT value for short) of a part of channels and a brightness temperature difference value (BTD for short) of a part of channels.
Illustratively, the acquired first satellite data (under the current frame) of six channels including IR1, IR2, IR3, B10, B12 and VIS is used to form 5 channels (for daytime, VIS channels can be further superimposed in channel direction) from the IR1 channel BT value, BTD IR2-IR1, BTD IR3-IR1, BTD IR3-B10, and BTD B12-IR1, and form one frame of satellite input data from the channel direction, respectively form one frame of satellite input data from the first historical satellite data of the previous 4 frames of the current frame, and form a continuous 5-frame satellite data sequence from the current frame and each satellite input data of the historical frames. The satellite data sequence can be used as the input of an inversion model after being normalized.
And 103, generating and outputting target radar echo data of a target area by adopting an inversion model obtained by pre-training according to the satellite data sequence with the preset frame number.
Specifically, after the satellite data sequence with the preset frame number is determined, the radar can be inverted according to the satellite data sequence with the preset frame number to obtain radar echo data of a target area, and the radar echo data is called as target radar echo data for distinguishing.
Optionally, after the target radar echo data is obtained, the electronic device may determine a weather condition, such as a precipitation condition, of the target area according to the target radar echo data, or may send the target radar echo data to the weather service device, and the weather service device determines the weather condition of the target area according to the target radar echo data. Specifically, after the target radar echo data is obtained, inverse normalization needs to be performed on the target radar echo data, where the inverse normalization refers to amplifying the obtained target radar echo data back according to the proportion of normalization performed on the actual radar echo data in the training process, so as to form a corresponding radar echo map, so as to analyze the weather condition of the target area according to the radar echo map in the subsequent process.
Illustratively, different combinations of multiple infrared channels (including IR1, IR2, IR3, B10 and B12) and visible light channel VIS information of a sunflower 8 meteorological satellite are used, radar echo data with high spatial resolution (1km multiplied by 1km) of one frame in 10 minutes are generated in a multi-frame satellite data video input mode, accurate radar echo data of a corresponding ground radar station are used for supervision, the model can invert the radar echo data of a corresponding area end to end through the meteorological satellite data, the POD predicted by the inversion model on pixel points of a real radar echo map 35dbz is higher than 0.6, FAR is lower than 0.3, CSI is higher than 0.4, and the whole predicted radar echo map can visually reflect precipitation trends of all positions of the area, so that meteorological personnel can observe and make judgment conveniently.
It should be noted that the inverse model needs to be obtained through training, and when the inverse model is trained, the obtaining mode of the adopted training satellite data sequence is consistent with that described above, and is not repeated herein one by one, and meanwhile, the actual radar data paired with the training satellite data needs to be correspondingly processed, the actual radar data is converted into radar combined reflectivity, and then converted onto the grid points in the same space as the training satellite data, so as to obtain the training radar echo data for training. The radar data is the combined reflectivity obtained based on the radar base data interpolation, and the finally trained radar data and satellite data are radar images and satellite images in the same area. And inputting the training satellite data sequence into an inversion network, and judging the training end according to the output of the inversion network, the training radar echo data and a preset loss function to obtain an inversion model.
Optionally, because the satellite and the radar are not time-consistent, in the training, a pair of satellite frames and radar frames with time-consistent need to be selected, for example, the time difference between the paired satellite data and radar data does not exceed a preset time threshold, such as 1 minute. The method can be specifically set according to actual requirements.
Optionally, for training the inversion model, any implementable manner may be adopted for training, and this embodiment is not limited.
Illustratively, the GAN method can be used for training, that is, the training network architecture includes an inversion network and a discrimination network, which constitute an antagonistic network. The authentication network may include an image authentication network and a video authentication network. The method comprises the steps that the GAN loss (loss function) can adopt WGAN-GP loss, and is combined with the batch GAN loss to improve the effect of the GAN loss, and the batch GAN loss focuses more on matching of detail features, so that the quality of a generated image is higher. Besides, the distance between the generated radar echo map and the real value thereof can be calculated by using L1loss and VGG perceptual loss (VGG perceptual loss) for supervising the training of the inversion model, wherein L1loss calculates the average value of the absolute value of the pixel value difference of each corresponding pixel point between the generated value and the real value, VGG perceptual loss needs to input the generated value and the real value into the trained VGG network, and L1loss is calculated for the feature captured by the network, and optionally, the VGG19 network can be used as the VGG network.
For example, for a video identification network, the input of the video identification network is a radar data sequence, radar echo data output by the inversion network and real radar echo data of other frames can be combined into a radar data sequence, and the radar echo data output by the inversion network and real radar echo data of two frames before and two frames after the inversion network in the time dimension are combined into a continuous 5-frame radar data sequence, and the radar echo data are used as the input of the video identification network, compared with the corresponding 5-frame real radar data sequence, and training of the loss assistance network is calculated based on a preset loss function.
According to the data processing method for the satellite inversion radar, radar echo data of a target area are inverted by adopting a satellite data sequence with preset frame numbers of the target area, in the inversion process, the characteristics of current frame satellite data and historical frame satellite data can be combined, the accuracy of an inversion result can be effectively improved, the influence of data irrelevant to radar in the satellite data on the inversion result is reduced, and the problem that in the prior art, the radar echo data of corresponding time is generated by only adopting the satellite data at a certain time, and the inversion result is often inaccurate is solved.
The method provided by the above embodiment is further described in an additional embodiment of the present invention.
Fig. 3 is a schematic flow chart of the data processing method for satellite inversion radar according to this embodiment.
As a practical way, on the basis of the above embodiment, optionally, the network architecture of the inversion model at least includes: the system comprises a 3D convolutional network module, an R (2+1) D residual error network module, a 3D non-local network module and a 3D transposition convolutional network module.
It will be appreciated that the network architecture of the inverse model may also include an input layer, an output layer, and the like. The method can be specifically set according to actual requirements.
Illustratively, as shown in fig. 4, a schematic diagram of a network architecture of the inversion model provided for this embodiment is provided. I.e. an architectural diagram of the inversion network. As shown in fig. 5, a schematic diagram of an architecture of the R (2+1) D residual network module in fig. 4 is provided in this embodiment. The Element-wise sum, i.e., Element wise sum, indicates that addition is performed on each corresponding position of the tensor or matrix with the same size, RELU is an activation function, and Identity indicates that an input is copied. The network structure parameters of the inversion network are shown in table 1, wherein the number of convolution kernel channels of the 3D Conv1 layer is the same as the number of input satellite data sequence channels; the network structure parameters of the R (2+1) D residual network module are shown in table 2. Wherein, R (2+1) D ResBlock represents R (2+1) D residual network module. The method is only exemplary, and in practical application, the network structure parameters may be set according to actual requirements.
TABLE 1
TABLE 2
As another implementable manner, on the basis of the foregoing embodiment, optionally, determining the satellite data sequence with the preset frame number according to the first satellite data of each preset channel specifically includes:
in step 2011, for each frame of first satellite data of each preset channel, a preset number of channels are generated according to a preset rule, and are overlapped according to the channel direction to form a frame of satellite input data.
Step 2012, forming the satellite input data of each frame into a satellite data sequence with a continuous preset frame number.
Illustratively, the acquired first satellite data (under the current frame) of six channels including IR1, IR2, IR3, B10, B12 and VIS is used to form 5 channels (for daytime, VIS channels can be further superimposed in channel direction) from the IR1 channel BT value, BTD IR2-IR1, BTD IR3-IR1, BTD IR3-B10, and BTD B12-IR1, and form one frame of satellite input data from the channel direction, respectively form one frame of satellite input data from the first historical satellite data of the previous 4 frames of the current frame, and form a continuous 5-frame satellite data sequence from the current frame and each satellite input data of the historical frames. The satellite data sequence can be used as the input of an inversion model after being normalized.
Optionally, the method further comprises:
Correspondingly, the method for forming the satellite input data of each frame into a continuous satellite data sequence with a preset frame number comprises the following steps:
Specifically, normalizing the satellite input data of each frame means that each channel of the satellite input data is normalized to between [ -1,1] according to a respective threshold range. The specific normalization operation is prior art and is not described herein.
As another practical way, on the basis of the above embodiment, optionally, the inverse model is obtained by training in the following way:
Step 2032, inputting the training satellite data sequence into a preset inversion network, and inputting the radar echo data output by the inversion network and the corresponding training radar echo data into a first identification network, and performing countermeasure training on the inversion network.
Specifically, the inversion model needs to be obtained through training, a large amount of paired original satellite data and original radar data (namely, real radar data) can be obtained in advance, the paired original satellite data and original radar data are converted to grid points in the same space through certain processing, a training satellite data sequence is obtained according to the method for obtaining the satellite data sequence, and training radar echo data are determined. Inputting a training satellite data sequence into a preset inversion network, taking the output of the inversion network and training radar echo data as the input of a first identification network, carrying out countermeasure training on the identification network and the inversion network, calculating loss based on a preset loss function, judging that the training is finished when the loss meets a preset requirement, and obtaining an inversion model. Wherein the first authentication network is an image authentication network.
Optionally, inputting the training satellite data sequence into a preset inversion network, and inputting radar echo data output by the inversion network and corresponding training radar echo data into a first discrimination network, to perform countermeasure training on the inversion network, including:
Specifically, the identification network may further include a video identification network (i.e., a second identification network), identification is performed based on the radar data sequence, the training stability and the inversion model performance are further improved, and meanwhile, the accuracy of radar echo data generated by the obtained inversion model is higher by combining VGG probability loss and L1 loss. The input of the video identification network is a radar data sequence, and specifically comprises two inputs, wherein one input is a first radar data sequence formed by superimposing continuous frames consisting of radar echo data output by the inversion network and front and rear real frames thereof according to a channel direction, and the other input is a second radar data sequence formed by superimposing training radar echo data corresponding to the radar echo data output by the inversion network and continuous frames consisting of the front and rear real frames thereof according to the channel direction.
Exemplarily, as shown in fig. 6, a schematic diagram of input and output of an authentication network provided for the present embodiment; fig. 7 is a schematic diagram of an architecture of an authentication network provided in this embodiment. Wherein bn (batch normalization) represents batch normalization, or batch normalization, and LeakyRELU is an activation function. The network architecture of the image identification network (also called an image identifier) and the network architecture of the video identification network (also called a video identifier) are the same, only the input in the training process is different, the input of the image identifier is radar echo data (also called generated radar echo data) output by the inversion network and a corresponding real value thereof, and the input of the video identifier is a radar data sequence formed by the generated radar echo data and real values of frames before and after the generated radar echo data and a corresponding real value sequence of the sequence. The network parameters for authenticating the network are shown in table 3. Where the number of convolution kernel channels of Conv1 is the same as the number of input channels, and for the image discriminator, the value is 1 if the input has only one channel, and for the video discriminator, the number of channels is 3 if the input is a 3-frame radar echo map sequence.
TABLE 3
The training of the inversion model adopts a training mode of generating a confrontation network, and the GAN loss adopts WGAN-GP loss, which comprises the following steps:
wherein D represents the discrimination network, and for the image discriminator, x represents the true value of the radar echo data,
representing the radar echo data generated by the inversion network,
denotes x and
a random combination of (a); for a video discriminator, x represents the true value of a radar echo data sequence (i.e., a sequence of training radar echo data),
representing a radar data sequence containing radar echo data generated by the inversion network,
denotes x and
a random combination of (a). It should be noted that the loss is specifically related to the prior art, and is not described herein again. In order to improve the effect of the GAN loss, Patch GAN loss can be adopted, and the matching of detail features is focused more, so that the quality of the generated image is higher. Patch GAN loss causes the discriminator to output a feature map matrix (rather than a value of conventional GAN) for true and false determination of each element of the feature map. Meanwhile, the input resize of the discriminator is changed into different sizes, so that the characteristics of different receptive fields are obtained. Besides, the distance between the generated radar echo map and the real value thereof can be calculated by using L1loss and VGG perceptual loss (VGG perceptual loss) for supervising the training of the inversion model, wherein L1loss calculates the average value of the absolute value of the pixel value difference of each corresponding pixel point between the generated value and the real value, VGG perceptual loss needs to input the generated value and the real value into the trained VGG network, and L1loss is calculated for the feature captured by the network, and optionally, the VGG19 network can be used as the VGG network.
And the radar echo data output by the inversion network is a radar combined reflectivity echo diagram. Optionally, the parameters of the network are supervised by a real radar combined reflectivity echo diagram to perform back propagation gradient descent training, and when the loss of the network does not descend any more, the parameters of the network are fixed after the network converges, and the parameters are used as a final inversion model.
Optionally, after training is finished, the obtained inversion model can be tested, the inversion model is only needed to be used in the test inversion stage, a discriminator is not needed, specifically, the satellite data sequence with the preset frame number can be input into the inversion model, the output is corresponding radar echo data, the corresponding radar echo data can be compared with corresponding real radar echo data, and the inversion effect of the inversion model is judged.
Illustratively, data of 5 channels of continuous 5 frames of satellite data are input into an inversion model after being superposed according to time and channel directions, and corresponding radar echo data are output.
The test result proves that after a large amount of data are used for training, the inversion model can generate radar echo data with higher quality in an area without radar, and the vacancy of the radar data is filled.
As another implementable manner, on the basis of the foregoing embodiment, optionally, the acquiring of the first satellite data of the preset channel of the target area includes:
And step 2052, preprocessing the original satellite data to obtain first satellite data of each preset channel, wherein the preprocessing comprises coordinate system conversion and interpolation processing.
Specifically, satellite data (referred to as original satellite data) can be received by the satellite pan, the original satellite data can be obtained from the satellite pan, the original satellite data of a target area can be cut out from the original satellite data, and preprocessing is performed to obtain first satellite data of a preset channel. The preprocessing may include coordinate system conversion and interpolation processing, and the satellite data of the target area is converted and interpolated onto the grid points of the preset space. The preset channel may include all or a portion of channels involved in satellite data.
For example, taking the original sunflower 8 satellite data as an example, after cutting out the original satellite data of the target area, the original satellite data can be analyzed into data of 14 channels after performing operations such as coordinate system conversion and interpolation, and satellite data of six channels including IR1, IR2, IR3, B10, B12 and VIS can be used as the first satellite data. Namely, the preset channels refer to six channels of IR1, IR2, IR3, B10, B12 and VIS. In practical application, which channels are selected can be set according to actual requirements, and this embodiment is not limited.
As another practicable manner, on the basis of the foregoing embodiment, optionally, the method further includes:
Specifically, after the target radar echo data are obtained, inverse normalization needs to be performed on the target radar echo data, the inverse normalization refers to amplifying the obtained target radar echo data back according to the proportion when the actual radar echo data are normalized in the training process, and a corresponding radar echo map can be formed, so that the weather condition of a target area can be analyzed according to the radar echo map in the subsequent process.
Optionally, the method further comprises:
Specifically, after the target radar echo map of the target area is obtained, the weather condition of the target area can be determined according to the target radar echo map. Such as determining precipitation conditions in the target area.
According to the method provided by the embodiment of the invention, the satellite data sequence is used as input information to invert single-frame radar data, and data irrelevant to the radar data in the satellite is filtered by capturing the characteristics of different change directions, speeds and trends of the radar data in the satellite data, so that the inversion accuracy is improved. And based on neural network sub-modules such as a 3D convolution network module, an R2+1D residual error network module, a 3D non-local module, a 3D transposition convolution module and the like, an inversion network suitable for inputting multi-frame satellite data sequences and generating single-frame radar echo data is constructed, GAN loss training combining batch GAN loss with WGAN-GP loss is adopted, although single-frame radar echo data is generated, a video discriminator is adopted to further improve the training stability and the performance of an inversion model, and meanwhile, VGG temporal loss and L1loss are combined, so that the radar echo data generated by the inversion model in actual inversion application has high precision.
It should be noted that the respective implementable modes in the embodiment may be implemented individually, or may be implemented in combination in any combination without conflict, and the present invention is not limited thereto.
According to the data processing method for the satellite inversion radar, the satellite data sequence is used as input information, single-frame radar data are inverted, data irrelevant to the radar data in the satellite are filtered by capturing the characteristics of different change directions, different speeds and different trends of the radar data in the satellite data, and the inversion accuracy is improved. And an inversion network suitable for inputting multi-frame satellite data sequences to generate single-frame radar echo data is constructed by combining various neural network modules, and the Gate loss training combining batch GAN loss with WGAN-GP loss is adopted, so that although single-frame radar echo data is generated, the training stability and the performance of an inversion model are further improved by adopting a video discriminator, and meanwhile, the precision of the radar echo data generated by the inversion model in the actual inversion application is effectively improved by combining VGG performance loss and L1 loss.
Still another embodiment of the present invention provides a data processing apparatus for a satellite inversion radar, configured to perform the method of the foregoing embodiment.
Fig. 8 is a schematic structural diagram of a data processing apparatus for satellite inversion radar according to this embodiment. The data processing device 30 for satellite inversion radar comprises an acquisition module 31, a determination module 32 and a processing module 33.
The acquisition module is used for acquiring first satellite data of a preset channel of a target area; the determining module is used for determining a satellite data sequence with a preset frame number according to the first satellite data of each preset channel; and the processing module is used for generating and outputting target radar echo data of a target area by adopting an inversion model obtained by pre-training according to the satellite data sequence with the preset frame number.
The specific manner in which the respective modules perform operations has been described in detail in relation to the apparatus in this embodiment, and will not be elaborated upon here.
According to the data processing device for the satellite inversion radar, radar echo data of a target area are inverted by adopting a satellite data sequence with preset frame numbers in the target area, in the inversion process, the characteristics of current frame satellite data and historical frame satellite data can be combined, the accuracy of an inversion result can be effectively improved, the influence of data irrelevant to precipitation in the satellite data on the inversion result is reduced, and the problem that in the prior art, radar echo data of corresponding time is generated by only adopting satellite data of certain time, and the inversion result is often inaccurate is solved.
The device provided by the above embodiment is further described in an additional embodiment of the present invention.
As a practical way, on the basis of the above embodiment, optionally, the network architecture of the inversion model at least includes: the system comprises a 3D convolutional network module, an R (2+1) D residual error network module, a 3D non-local network module and a 3D transposition convolutional network module.
As another implementable manner, on the basis of the foregoing embodiment, optionally, the determining module is specifically configured to:
generating a preset number of channels according to a preset rule aiming at each frame of first satellite data of each preset channel, and superposing the channels according to the channel direction to form a frame of satellite input data;
and forming the satellite input data of each frame into a continuous satellite data sequence with a preset frame number.
Optionally, the determining module is further configured to normalize the satellite input data of each frame to obtain normalized satellite input data;
and the determining module is specifically used for forming the normalized satellite input data of each frame into a satellite data sequence with a continuous preset frame number.
As another implementable manner, on the basis of the foregoing embodiment, optionally, the obtaining module is further configured to combine each frame of normalized satellite input data into a continuous satellite data sequence with a preset frame number;
a processing module further configured to:
inputting a training satellite data sequence into a preset inversion network, inputting radar echo data output by the inversion network and corresponding training radar echo data into a first identification network, and performing countermeasure training on the inversion network;
and judging the end of training based on a preset loss function to obtain an inversion model.
Optionally, the processing module is specifically configured to:
the training satellite data sequence is input into a preset inversion network, radar echo data output by the inversion network and corresponding training radar echo data are input into a first identification network, the radar echo data output by the inversion network and a first amount of adjacent radar echo data form a radar echo data sequence, and the radar echo data sequence and the corresponding training radar echo data sequence are input into a second identification network for confrontation training.
As another implementable manner, on the basis of the foregoing embodiment, optionally, the obtaining module is specifically configured to:
acquiring original satellite data of a target area;
and preprocessing the original satellite data to obtain first satellite data of each preset channel, wherein the preprocessing comprises coordinate system conversion and interpolation processing.
As another implementable manner, on the basis of the foregoing embodiment, optionally, the processing module is further configured to:
performing inverse normalization processing on target radar echo data of a target area to obtain first radar echo data;
and generating a target radar echo map of the target area based on the first radar echo data.
Optionally, the processing module is further configured to determine a weather condition of the target area according to the target radar echo map.
The specific manner in which the respective modules perform operations has been described in detail in relation to the apparatus in this embodiment, and will not be elaborated upon here.
It should be noted that the respective implementable modes in the embodiment may be implemented individually, or may be implemented in combination in any combination without conflict, and the present invention is not limited thereto.
According to the data processing device for the satellite inversion radar, the satellite data sequence is used as input information, single-frame radar data are inverted, data irrelevant to the radar data in the satellite are filtered by capturing the characteristics of different change directions, speeds and trends of the satellite data and the radar data, and the inversion accuracy is improved. And an inversion network suitable for inputting multi-frame satellite data sequences to generate single-frame radar echo data is constructed by combining various neural network modules, and the Gate loss training combining batch GAN loss with WGAN-GP loss is adopted, so that although single-frame radar echo data is generated, the training stability and the performance of an inversion model are further improved by adopting a video discriminator, and meanwhile, the precision of the radar echo data generated by the inversion model in the actual inversion application is effectively improved by combining VGG performance loss and L1 loss.
Still another embodiment of the present invention provides an electronic device, configured to perform the method provided by the foregoing embodiment.
As shown in fig. 9, is a schematic structural diagram of the electronic device provided in this embodiment. The electronic device 50 includes: at least one processor 51 and memory 52;
the memory stores computer-executable instructions; the at least one processor executes computer-executable instructions stored by the memory to cause the at least one processor to perform a method as provided by any of the embodiments above.
According to the electronic equipment of the embodiment, the satellite data sequence is used as input information, single-frame radar data are inverted, data irrelevant to the radar data in the satellite are filtered by capturing the characteristics of different change directions, speeds and trends of the radar data in the satellite data, and the inversion accuracy is improved. And an inversion network suitable for inputting multi-frame satellite data sequences to generate single-frame radar echo data is constructed by combining various neural network modules, and the Gate loss training combining batch GAN loss with WGAN-GP loss is adopted, so that although single-frame radar echo data is generated, the training stability and the performance of an inversion model are further improved by adopting a video discriminator, and meanwhile, the precision of the radar echo data generated by the inversion model in the actual inversion application is effectively improved by combining VGG performance loss and L1 loss.
Yet another embodiment of the present invention provides a computer-readable storage medium, in which computer-executable instructions are stored, and when the processor executes the computer-executable instructions, the method provided in any one of the above embodiments is implemented.
According to the computer-readable storage medium of the embodiment, the satellite data sequence is used as input information, single-frame radar data are inverted, and data irrelevant to the radar data in the satellite are filtered by capturing the characteristics of different change directions, speeds and trends of the radar data in the satellite data, so that the inversion accuracy is improved. And an inversion network suitable for inputting multi-frame satellite data sequences to generate single-frame radar echo data is constructed by combining various neural network modules, and the Gate loss training combining batch GAN loss with WGAN-GP loss is adopted, so that although single-frame radar echo data is generated, the training stability and the performance of an inversion model are further improved by adopting a video discriminator, and meanwhile, the precision of the radar echo data generated by the inversion model in the actual inversion application is effectively improved by combining VGG performance loss and L1 loss.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (10)
1. A data processing method for a satellite inversion radar is characterized by comprising the following steps:
acquiring first satellite data of a preset channel of a target area;
determining a satellite data sequence with a preset frame number according to the first satellite data of each preset channel;
and generating and outputting target radar echo data of the target area by adopting an inversion model obtained by pre-training according to the satellite data sequence with the preset frame number.
2. The method of claim 1, wherein the network architecture of the inversion model comprises at least: the system comprises a 3D convolutional network module, an R (2+1) D residual error network module, a 3D non-local network module and a 3D transposition convolutional network module.
3. The method as claimed in claim 1, wherein the determining a satellite data sequence of a predetermined number of frames from the first satellite data of each predetermined channel comprises:
generating a preset number of channels according to a preset rule aiming at each frame of first satellite data of each preset channel, and superposing the channels according to the channel direction to form a frame of satellite input data;
and forming the satellite input data of each frame into a continuous satellite data sequence with a preset frame number.
4. The method of claim 3, further comprising:
normalizing each frame of satellite input data to obtain normalized satellite input data;
forming each frame of satellite input data into a continuous satellite data sequence with a preset frame number, wherein the satellite data sequence comprises:
and forming the normalized satellite input data of each frame into a continuous satellite data sequence with a preset frame number.
5. The method of claim 1, wherein the inverse model is obtained by training:
acquiring a training satellite data sequence and corresponding training radar echo data;
inputting the training satellite data sequence into a preset inversion network, inputting radar echo data output by the inversion network and corresponding training radar echo data into a first identification network, and performing countermeasure training on the inversion network;
and judging the end of training based on a preset loss function to obtain the inversion model.
6. The method of claim 5, wherein the inputting the training satellite data sequence into a preset inversion network, and inputting radar echo data output by the inversion network and the corresponding training radar echo data into a first identification network, and performing countermeasure training on the inversion network comprises:
and inputting the training satellite data sequence into a preset inversion network, inputting radar echo data output by the inversion network and corresponding training radar echo data into a first discrimination network, forming a radar echo data sequence by the radar echo data output by the inversion network and a first amount of adjacent radar echo data, and inputting the radar echo data sequence and the corresponding training radar echo data sequence into a second discrimination network for confrontation training.
7. The method according to any one of claims 1-6, further comprising:
performing inverse normalization processing on the target radar echo data of the target area to obtain first radar echo data;
and generating a target radar echo map of the target area based on the first radar echo data.
8. A data processing apparatus for a satellite inversion radar, comprising:
the acquisition module is used for acquiring first satellite data of a preset channel of a target area;
the determining module is used for determining a satellite data sequence with a preset frame number according to the first satellite data of each preset channel;
and the processing module is used for generating and outputting target radar echo data of the target area by adopting an inversion model obtained by pre-training according to the satellite data sequence with the preset frame number.
9. An electronic device, comprising: at least one processor and memory;
the memory stores computer-executable instructions;
the at least one processor executing the computer-executable instructions stored by the memory causes the at least one processor to perform the method of any one of claims 1-7.
10. A computer-readable storage medium having computer-executable instructions stored thereon which, when executed by a processor, implement the method of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011042984.1A CN112198483A (en) | 2020-09-28 | 2020-09-28 | Data processing method, device and equipment for satellite inversion radar and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011042984.1A CN112198483A (en) | 2020-09-28 | 2020-09-28 | Data processing method, device and equipment for satellite inversion radar and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112198483A true CN112198483A (en) | 2021-01-08 |
Family
ID=74006848
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011042984.1A Pending CN112198483A (en) | 2020-09-28 | 2020-09-28 | Data processing method, device and equipment for satellite inversion radar and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112198483A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115144835A (en) * | 2022-09-02 | 2022-10-04 | 南京信大气象科学技术研究院有限公司 | Method for inverting weather radar reflectivity by satellite based on neural network |
| CN115267786A (en) * | 2022-07-29 | 2022-11-01 | 北京彩彻区明科技有限公司 | Resunet-GAN global radar inversion method and device fusing satellite observation brightness and elevation |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108445464A (en) * | 2018-03-12 | 2018-08-24 | 南京恩瑞特实业有限公司 | Satellite radar inverting fusion methods of the NRIET based on machine learning |
| US20180260648A1 (en) * | 2017-03-09 | 2018-09-13 | Baidu Online Network Technology (Beijing) Co., Ltd | Area of interest boundary extracting method and apparatus, device and computer storage medium |
| CN109814175A (en) * | 2019-02-14 | 2019-05-28 | 浙江省气象台 | A kind of satellite-based strong convection monitoring method and its application |
| CN110221360A (en) * | 2019-07-25 | 2019-09-10 | 广东电网有限责任公司 | A kind of power circuit thunderstorm method for early warning and system |
| CN111474529A (en) * | 2020-06-10 | 2020-07-31 | 浙江省气象台 | Method for inverting radar echo by satellite, system for inverting radar echo and navigation radar |
| CN111507910A (en) * | 2020-03-18 | 2020-08-07 | 南方电网科学研究院有限责任公司 | Single image reflection removing method and device and storage medium |
-
2020
- 2020-09-28 CN CN202011042984.1A patent/CN112198483A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180260648A1 (en) * | 2017-03-09 | 2018-09-13 | Baidu Online Network Technology (Beijing) Co., Ltd | Area of interest boundary extracting method and apparatus, device and computer storage medium |
| CN108445464A (en) * | 2018-03-12 | 2018-08-24 | 南京恩瑞特实业有限公司 | Satellite radar inverting fusion methods of the NRIET based on machine learning |
| CN109814175A (en) * | 2019-02-14 | 2019-05-28 | 浙江省气象台 | A kind of satellite-based strong convection monitoring method and its application |
| CN110221360A (en) * | 2019-07-25 | 2019-09-10 | 广东电网有限责任公司 | A kind of power circuit thunderstorm method for early warning and system |
| CN111507910A (en) * | 2020-03-18 | 2020-08-07 | 南方电网科学研究院有限责任公司 | Single image reflection removing method and device and storage medium |
| CN111474529A (en) * | 2020-06-10 | 2020-07-31 | 浙江省气象台 | Method for inverting radar echo by satellite, system for inverting radar echo and navigation radar |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115267786A (en) * | 2022-07-29 | 2022-11-01 | 北京彩彻区明科技有限公司 | Resunet-GAN global radar inversion method and device fusing satellite observation brightness and elevation |
| CN115144835A (en) * | 2022-09-02 | 2022-10-04 | 南京信大气象科学技术研究院有限公司 | Method for inverting weather radar reflectivity by satellite based on neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11099275B1 (en) | LiDAR point cloud reflection intensity complementation method and system | |
| CN110427917B (en) | Method and device for detecting key points | |
| CN110660066B (en) | Training method of network, image processing method, network, terminal equipment and medium | |
| CN105335955B (en) | Method for checking object and object test equipment | |
| CN109636771B (en) | Flight target detection method and system based on image processing | |
| CN111369617B (en) | 3D target detection method of monocular view based on convolutional neural network | |
| CN107560592B (en) | Precise distance measurement method for photoelectric tracker linkage target | |
| CN111325107B (en) | Detection model training method, device, electronic equipment and readable storage medium | |
| CN109191513B (en) | Power equipment stereo matching method based on global optimization | |
| CN110825900A (en) | Training method of feature reconstruction layer, reconstruction method of image features and related device | |
| CN112198483A (en) | Data processing method, device and equipment for satellite inversion radar and storage medium | |
| CN113313047B (en) | Lane line detection method and system based on lane structure prior | |
| EP3998582A1 (en) | Three-dimensional model generation method and three-dimensional model generation device | |
| CN113192182A (en) | Multi-sensor-based live-action reconstruction method and system | |
| CN115115611B (en) | Vehicle damage identification method and device, electronic equipment and storage medium | |
| CN114415173A (en) | Fog-penetrating target identification method for high-robustness laser-vision fusion | |
| KR102238629B1 (en) | System and method for overcoming low visibility situation based on deep learning | |
| CN113597616A (en) | Pupil position determination method, device and system | |
| CN112115737B (en) | Vehicle orientation determining method and device and vehicle-mounted terminal | |
| CN112017243A (en) | Medium visibility identification method | |
| CN111611836A (en) | Ship detection model training and ship tracking method based on background elimination method | |
| CN116310899A (en) | YOLOv 5-based improved target detection method and device and training method | |
| CN116189140A (en) | Binocular vision-based vehicle three-dimensional target detection algorithm | |
| CN112069997B (en) | Unmanned aerial vehicle autonomous landing target extraction method and device based on DenseHR-Net | |
| CN114708544A (en) | Intelligent violation monitoring helmet based on edge calculation and monitoring method thereof |
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 |