CN113654959B - Rapid inversion method and system for smoke cloud concentration space-time distribution - Google Patents
Rapid inversion method and system for smoke cloud concentration space-time distribution Download PDFInfo
- Publication number
- CN113654959B CN113654959B CN202110879755.3A CN202110879755A CN113654959B CN 113654959 B CN113654959 B CN 113654959B CN 202110879755 A CN202110879755 A CN 202110879755A CN 113654959 B CN113654959 B CN 113654959B
- Authority
- CN
- China
- Prior art keywords
- concentration
- smoke
- distribution
- cloud
- 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.)
- Active
Links
- 239000000779 smoke Substances 0.000 title claims abstract description 153
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000013528 artificial neural network Methods 0.000 claims abstract description 21
- 238000009792 diffusion process Methods 0.000 claims abstract description 16
- 238000004364 calculation method Methods 0.000 claims abstract description 15
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 9
- 230000008569 process Effects 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 21
- 238000005259 measurement Methods 0.000 claims description 17
- 238000005070 sampling Methods 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 4
- 230000008859 change Effects 0.000 abstract description 5
- 230000001419 dependent effect Effects 0.000 abstract description 3
- 230000001351 cycling effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume or surface-area of porous materials
- G01N15/06—Investigating concentration of particle suspensions
-
- 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
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/95—Lidar systems specially adapted for specific applications for meteorological use
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2132—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on discrimination criteria, e.g. discriminant analysis
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Chemical & Material Sciences (AREA)
- Electromagnetism (AREA)
- Pathology (AREA)
- Analytical Chemistry (AREA)
- Dispersion Chemistry (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Computer Networks & Wireless Communication (AREA)
- Evolutionary Biology (AREA)
- Immunology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biochemistry (AREA)
- Fire-Detection Mechanisms (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for quickly inverting the spatial-temporal distribution of smoke concentration, which comprise the following steps: obtaining measured smoke concentration data of a specific position, and obtaining rough distribution of the measured smoke concentration; the actually measured cloud concentration rough distribution is transmitted to a trained super-resolution neural network, and the reconstructed cloud concentration fine distribution at the initial moment is output; constructing a Gaussian diffusion model, and calculating to obtain the spatial fine distribution of the concentration of the smoke cloud under the steady-state condition; inputting the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the fine distribution of the concentration of the smoke cloud under the steady-state condition into a trained convolutional neural network, and outputting the fine distribution of the concentration of the smoke cloud at different moments by using step-by-step iterative inversion; the inversion method is not dependent on a finite element method, has low complexity, can be used in a scene of rapid inversion, has the characteristics of calculation accuracy and generalization, and can accurately reflect the change process of the concentration of the smoke cloud in a fixed area along with time.
Description
Technical Field
The invention relates to the technical field of smoke cloud concentration detection, in particular to a method and a system for quickly inverting smoke cloud concentration space-time distribution.
Background
Smoke cloud diffusion is a common physical phenomenon in the atmospheric environment. The space-time dynamics calculation model for developing the concentration distribution of the smoke cloud can effectively invert the space distribution and time variation of the concentration of the smoke cloud in the atmosphere, and has positive significance in the fields of weather forecast, military observation and the like.
At present, the developed cloud concentration distribution space-time dynamics model mainly comprises two types: firstly, developing an analytical model formed by starting from a first nature principle of diffusion and flow; second, empirical models developed based on experimental data. The analytic model is required to rely on a finite element method for calculation, has high complexity and is difficult to be used in a scene requiring rapid inversion; the application range of the empirical model is smaller, and the generalization is poorer.
Disclosure of Invention
Based on the technical problems in the background technology, the invention provides a rapid inversion method and a rapid inversion system for the concentration space-time distribution of a smoke cloud, and the rapid inversion method and the rapid inversion system have the characteristics of calculation accuracy and generalization.
The invention provides a rapid inversion method of a cloud concentration space-time distribution, which comprises the following steps:
obtaining measured smoke concentration data of a specific position, and obtaining rough distribution of the measured smoke concentration;
the actually measured cloud concentration rough distribution is transmitted to a trained super-resolution neural network, and the reconstructed cloud concentration fine distribution at the initial moment is output;
constructing a Gaussian diffusion model, and calculating to obtain the spatial fine distribution of the concentration of the smoke cloud under the steady-state condition;
and inputting the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the fine distribution of the concentration of the smoke cloud under the steady-state condition into a trained convolutional neural network, and outputting the fine distribution of the concentration of the smoke cloud at different moments by using step-by-step iterative inversion.
Further, the gaussian diffusion model is as follows:
wherein C (x, y, z) represents the concentration of the smoke cloud at the space (x, y, z); a (x) represents a concentration gain function, sigma y Represents the standard deviation, sigma, of the Gaussian distribution in the y-axis direction z The standard deviation of the gaussian distribution in the z-axis direction is shown.
Further, inputting the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the fine distribution of the concentration of the smoke cloud under the steady-state condition into a trained convolutional neural network, and inverting and outputting the fine distribution of the concentration of the smoke cloud at different moments by using step-and-step iteration, wherein the method specifically comprises the following steps:
constructing a second generation-countermeasure network structure, wherein the second generation-countermeasure network structure comprises a second generation network and a second discrimination network;
sending the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the spatial distribution of the concentration of the smoke cloud under the steady-state condition into a second generation network to generate the concentration distribution of the smoke cloud after a certain time step;
sending the actually measured cloud concentration rough distribution and the cloud concentration distribution after the certain time step into a second discrimination network, and outputting a judgment result of the cloud concentration distribution after the certain time step;
re-sending the judgment result of the cloud concentration distribution after the certain time step into a second generation network, and iteratively calculating the cloud concentration distribution at the next moment until the loss function of the discrimination similarity converges;
at this time, the cloud concentration calculated by the second generation network at all times is used as the required cloud concentration space-time distribution.
Further, the training process of the super-resolution neural network is as follows:
s21: constructing a first generation-countermeasure network structure and a smoke concentration training sample set, wherein the first generation-countermeasure network structure comprises a first generation network and a first discrimination network, and the smoke concentration training sample set comprises an actual measurement smoke concentration fine distribution sample and an actual measurement smoke concentration rough distribution sample;
s22: sampling each actually measured smoke cloud concentration fine distribution sample once for n times, sending the actually measured smoke cloud concentration rough distribution sample into a first generation network, outputting an initially moment smoke cloud concentration fine distribution sample, sending the initially moment smoke cloud concentration fine distribution sample and the actually measured smoke cloud concentration fine distribution sample into a first discrimination network together, and outputting a discrimination result;
s23: performing iterative training on the activation function parameters in the first generation network by using the output discrimination result of the first discrimination network;
s24: steps S22 to S23 are circulated until the loss function for judging the similarity is completely converged, and the downsampling multiple at the moment is output;
s25: and (3) finishing the super-resolution neural network training, taking the downsampling multiple at the moment as an actually measured sampling position for setting the smoke concentration, and taking the smoke concentration distribution output by the first generation network at the moment as the smoke concentration fine distribution at the initial moment of reconstruction.
A rapid inversion system for the concentration space-time distribution of a smoke cloud comprises an actual measurement acquisition module, a reconstruction module, a calculation module and an output module;
the actually measured acquisition module is used for acquiring actually measured smoke concentration data of a specific position to obtain roughly measured smoke concentration distribution;
the reconstruction module is used for transmitting the actually measured cloud concentration rough distribution to a trained super-resolution neural network and outputting a reconstructed initial time cloud concentration fine distribution;
the calculation module is used for constructing a Gaussian diffusion model and calculating to obtain the spatial fine distribution of the concentration of the smoke cloud under the steady-state condition;
the output module is used for inputting the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the fine distribution of the concentration of the smoke cloud under the steady-state condition into the trained convolutional neural network, and outputting the fine distribution of the concentration of the smoke cloud at different moments by means of step-by-step iterative inversion.
Further, the gaussian diffusion model is as follows:
wherein C (x, y, z) represents the concentration of the smoke cloud at the space (x, y, z); a (x) represents a concentration gain function, sigma y Represents the standard deviation, sigma, of the Gaussian distribution in the y-axis direction z The standard deviation of the gaussian distribution in the z-axis direction is shown.
A computer readable storage medium having stored thereon a number of acquisition classification procedures for being invoked by a processor and performing a method of rapid inversion of a concentration spatiotemporal distribution of smoke as described above.
The rapid inversion method and system for the concentration space-time distribution of the smoke cloud provided by the invention have the advantages that: the inversion method and the system provided by the invention are not dependent on a finite element method, have low complexity, can be used in a fast inversion scene, have the characteristics of calculation accuracy and generalization, can accurately reflect the change process of the concentration of the smoke in a fixed area along with time, and obviously reduce the time cost of reconstruction through the parallel calculation of a neural network.
Drawings
FIG. 1 is a schematic diagram of the structure of the present invention;
Detailed Description
In the following detailed description of the present invention, numerous specific details are set forth in order to provide a thorough understanding of the present invention. The invention may be embodied in many other forms than described herein and similarly modified by those skilled in the art without departing from the spirit or scope of the invention, which is therefore not limited to the specific embodiments disclosed below.
As shown in fig. 1, the method for rapidly inverting the smoke concentration space-time distribution provided by the invention comprises the following steps S1 to S4:
s1: obtaining measured smoke concentration data of a specific position, and obtaining rough distribution of the measured smoke concentration;
actually measured smoke cloud concentration data of the characteristic position can be obtained through a laser radar, a concentration sensor and the like; the specific position is determined by the super-resolution neural network of step S2.
The actually measured smoke concentration rough distribution refers to the obtained smoke concentration rough distribution in a certain area, wherein a laser radar or a concentration sensor and the like are based on actually measured smoke concentration data acquired at a specific position in the area; the measured cloud concentration fine distribution means that in a certain area, a laser radar or a concentration sensor or the like is based on measured cloud concentration data acquired at all positions in the area, and the obtained cloud concentration fine distribution is obtained.
S2: the actually measured cloud concentration rough distribution is transmitted to a trained super-resolution neural network, and the reconstructed cloud concentration fine distribution at the initial moment is output;
and (3) processing the actually measured smoke cloud concentration data by the trained super-resolution neural network, outputting the final downsampling multiple of the actually measured smoke cloud concentration data, and taking the downsampling multiple as a sampling position for guiding the laser radar to realize, wherein the sampling position is the specific position in the step S1.
And meanwhile, the output smoke cloud concentration distribution is the reconstructed smoke cloud concentration fine distribution at the initial moment.
S3: constructing a Gaussian diffusion model, and calculating to obtain the spatial fine distribution of the concentration of the smoke cloud under the steady-state condition;
the gaussian diffusion model is as follows:
wherein C (x, y, z) represents the concentration of the smoke cloud at the space (x, y, z); a (x) represents a concentration gain function, sigma y Represents the standard deviation, sigma, of the Gaussian distribution in the y-axis direction z The standard deviation of the gaussian distribution in the z-axis direction is shown. And according to the Gaussian diffusion model, the steady-state data of the concentration of the smoke cloud under the steady-state condition can be obtained.
S4: and inputting the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the fine distribution of the concentration of the smoke cloud under the steady-state condition into a trained convolutional neural network, and outputting the fine distribution of the concentration of the smoke cloud at different moments by using step-by-step iterative inversion.
According to the steps S1 to S4, reconstructing the actually measured smoke concentration data to obtain a reconstructed fine distribution of the smoke concentration at the initial moment, obtaining a spatial fine distribution of the smoke concentration under a steady state condition through a Gaussian diffusion model, and then combining a time relation to output a spatial-temporal distribution of the final smoke concentration; the inversion method is not dependent on a finite element method, has low complexity, can be used in a scene of rapid inversion, has the characteristics of calculation accuracy and generalization, can accurately reflect the dynamic change process of the concentration of the smoke cloud in a fixed area along with time, and obviously reduces the time cost of reconstruction through the parallel calculation of a neural network.
In this embodiment, the training process of the super-resolution neural network is as follows:
s21: constructing a first generation-countermeasure network structure and a smoke concentration training sample set, wherein the first generation-countermeasure network structure comprises a first generation network and a first discrimination network, and the smoke concentration training sample set comprises an actual measurement smoke concentration fine distribution sample and an actual measurement smoke concentration rough distribution sample;
the measured smoke cloud concentration fine distribution sample can obtain measured smoke cloud concentration data through all positions of a laser radar in a certain area, and the obtained smoke cloud concentration fine distribution sample is based on the measured smoke cloud concentration data; the actual measurement cloud concentration rough distribution sample can acquire actual measurement cloud concentration data through a specific position of the laser radar in a certain area, and the obtained cloud concentration rough distribution is based on the actual measurement cloud concentration data. The first generation network may be a srcn based super resolution neural network.
S22: sampling each actually measured smoke cloud concentration fine distribution sample once for n times, sending the actually measured smoke cloud concentration rough distribution sample into a first generation network, outputting an initially moment smoke cloud concentration fine distribution sample, sending the initially moment smoke cloud concentration fine distribution sample and the actually measured smoke cloud concentration fine distribution sample into a first discrimination network together, and outputting a discrimination result;
and (3) performing n times downsampling on each measured cloud concentration fine distribution sample once to simulate the final output result of the measured cloud concentration coarse distribution sample.
S23: performing iterative training on the activation function parameters in the first generation network by using the output discrimination result of the first discrimination network, and setting the maximum training step length to be 200;
s24: steps S22 to S23 are circulated until the loss function for judging the similarity is completely converged, and the downsampling multiple at the moment is output;
in the cycling steps S22 to S23, the measured smoke concentration sample is downsampled once m times in each cycle, and m and n may be the same or different.
S25: and (3) finishing the super-resolution neural network training, taking the downsampling multiple at the moment as an actual measurement sampling position for setting the smoke concentration in actual measurement use, and taking the smoke concentration distribution output by the first generation network at the moment as the smoke concentration fine distribution at the initial moment of reconstruction corresponding to the actual measurement smoke concentration data.
Through steps S21 to S25, the specific position and the reconstructed fine distribution of the concentration of the smoke and cloud at the initial moment can be output in the trained super-resolution neural network.
Step S4 includes the following steps S41 to S45:
s41: constructing a second generation-countermeasure network structure, wherein the second generation-countermeasure network structure comprises a second generation network and a second discrimination network;
s42: sending the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the spatial distribution of the concentration of the smoke cloud under the steady-state condition into a second generation network to generate the concentration distribution of the smoke cloud after a certain time step;
s43: sending the actually measured cloud concentration rough distribution and the cloud concentration distribution after the certain time step into a second discrimination network, and outputting a judgment result of the cloud concentration distribution after the certain time step;
the judgment result reflects the judgment of the calculation accuracy of the second judgment network on the calculated smoke concentration distribution after the certain time step, the output is 0 if the judgment network considers that the calculated smoke concentration distribution is inaccurate, and the output is 1 if the judgment network considers that the calculated smoke concentration distribution is accurate.
S44: re-sending the judgment result of the cloud concentration distribution after the certain time step into a second generation network, and iteratively calculating the cloud concentration distribution at the next moment until the loss function of the discrimination similarity converges;
s45: at this time, the cloud concentration calculated by the second generation network at all times is used as the required cloud concentration space-time distribution.
Through the steps S41 to S45, the change process of the concentration of the smoke cloud in a certain area along with the change process can be accurately obtained, and the spatial-temporal distribution of the concentration of the smoke cloud is obtained.
A rapid inversion system for the concentration space-time distribution of a smoke cloud comprises an actual measurement acquisition module, a reconstruction module, a calculation module and an output module;
the actually measured acquisition module is used for acquiring actually measured smoke concentration data of a specific position to obtain roughly measured smoke concentration distribution;
the reconstruction module is used for transmitting the actually measured cloud concentration rough distribution to a trained super-resolution neural network and outputting a reconstructed initial time cloud concentration fine distribution;
the calculation module is used for constructing a Gaussian diffusion model and calculating to obtain the spatial fine distribution of the concentration of the smoke cloud under the steady-state condition;
the output module is used for inputting the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the fine distribution of the concentration of the smoke cloud under the steady-state condition into the trained convolutional neural network, and outputting the fine distribution of the concentration of the smoke cloud at different moments by means of step-by-step iterative inversion.
A computer readable storage medium having stored thereon a plurality of acquisition classification procedures for being invoked by a processor and performing the above-described method of rapid inversion of a concentration spatiotemporal distribution of a smoke cloud.
Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware associated with program instructions, where the foregoing program may be stored in a computer readable storage medium, and when executed, the program performs steps including the above method embodiments; and the aforementioned storage medium includes: various media that can store program code, such as ROM, RAM, magnetic or optical disks.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should make equivalent substitutions or modifications according to the technical scheme of the present invention and the inventive concept thereof, and should be covered by the scope of the present invention.
Claims (6)
1. A method for rapidly inverting the spatial-temporal distribution of the concentration of a smoke cloud, comprising:
s1: obtaining measured smoke concentration data of a specific position, and obtaining rough distribution of the measured smoke concentration;
s2: the actually measured cloud concentration rough distribution is transmitted to a trained super-resolution neural network, the reconstructed cloud concentration fine distribution at the initial moment is output, and the training process of the super-resolution neural network is as follows:
s21: constructing a first generation-countermeasure network structure and a smoke concentration training sample set, wherein the first generation-countermeasure network structure comprises a first generation network and a first discrimination network, and the smoke concentration training sample set comprises an actual measurement smoke concentration fine distribution sample and an actual measurement smoke concentration rough distribution sample;
s22: sampling each actually measured smoke cloud concentration fine distribution sample once for n times, sending the actually measured smoke cloud concentration rough distribution sample into a first generation network, outputting an initially moment smoke cloud concentration fine distribution sample, sending the initially moment smoke cloud concentration fine distribution sample and the actually measured smoke cloud concentration fine distribution sample into a first discrimination network together, and outputting a discrimination result;
s23: performing iterative training on the activation function parameters in the first generation network by using the output discrimination result of the first discrimination network;
s24: steps S22 to S23 are circulated until the loss function for judging the similarity is completely converged, and the downsampling multiple at the moment is output;
s25: the super-resolution neural network training is completed, the downsampling multiple at the moment is used as an actually measured sampling position for setting the concentration of the smoke cloud, and the smoke cloud concentration distribution output by the first generation network at the moment is used as the fine smoke cloud concentration distribution at the initial moment of reconstruction;
s3: constructing a Gaussian diffusion model, and calculating to obtain the spatial fine distribution of the concentration of the smoke cloud under the steady-state condition;
s3: and inputting the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the fine distribution of the concentration of the smoke cloud under the steady-state condition into a trained convolutional neural network, and outputting the fine distribution of the concentration of the smoke cloud at different moments by using step-by-step iterative inversion.
2. The method of claim 1, wherein the gaussian diffusion model is as follows:
wherein,expressed in space->The concentration of the smoke cloud at the location; />Representing the concentration gain function, +.>Representing a Gaussian distributionyStandard deviation of axial direction>Representing a Gaussian distributionzStandard deviation in axial direction.
3. The method for rapidly inverting the cloud concentration space-time distribution according to claim 1, wherein the reconstructed fine distribution of the cloud concentration at the initial time and the fine distribution of the cloud concentration under the steady-state condition are input into a trained convolutional neural network, and the fine distribution of the cloud concentration at different time is outputted by using step-and-step iterative inversion, which comprises the following steps:
constructing a second generation-countermeasure network structure, wherein the second generation-countermeasure network structure comprises a second generation network and a second discrimination network;
sending the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the spatial distribution of the concentration of the smoke cloud under the steady-state condition into a second generation network to generate the concentration distribution of the smoke cloud after a certain time step;
sending the actually measured cloud concentration rough distribution and the cloud concentration distribution after the certain time step into a second discrimination network, and outputting a judgment result of the cloud concentration distribution after the certain time step;
re-sending the judgment result of the cloud concentration distribution after the certain time step into a second generation network, and iteratively calculating the cloud concentration distribution at the next moment until the loss function of the discrimination similarity converges;
at this time, the cloud concentration calculated by the second generation network at all times is used as the required cloud concentration space-time distribution.
4. The rapid inversion system for the smoke cloud concentration space-time distribution is characterized by comprising an actual measurement acquisition module, a reconstruction module, a calculation module and an output module;
the actually measured acquisition module is used for acquiring actually measured smoke concentration data of a specific position to obtain roughly measured smoke concentration distribution;
the reconstruction module is used for transmitting the actually measured cloud concentration rough distribution to a trained super-resolution neural network and outputting a reconstructed initial time cloud concentration fine distribution;
the calculation module is used for constructing a Gaussian diffusion model and calculating to obtain the spatial fine distribution of the concentration of the smoke cloud under the steady-state condition;
the output module is used for inputting the reconstructed fine distribution of the concentration of the smoke cloud at the initial moment and the fine distribution of the concentration of the smoke cloud under the steady-state condition into the trained convolutional neural network, and outputting the fine distribution of the concentration of the smoke cloud at different moments by using step-by-step iterative inversion;
the training process of the super-resolution neural network is as follows:
s21: constructing a first generation-countermeasure network structure and a smoke concentration training sample set, wherein the first generation-countermeasure network structure comprises a first generation network and a first discrimination network, and the smoke concentration training sample set comprises an actual measurement smoke concentration fine distribution sample and an actual measurement smoke concentration rough distribution sample;
s22: sampling each actually measured smoke cloud concentration fine distribution sample once for n times, sending the actually measured smoke cloud concentration rough distribution sample into a first generation network, outputting an initially moment smoke cloud concentration fine distribution sample, sending the initially moment smoke cloud concentration fine distribution sample and the actually measured smoke cloud concentration fine distribution sample into a first discrimination network together, and outputting a discrimination result;
s23: performing iterative training on the activation function parameters in the first generation network by using the output discrimination result of the first discrimination network;
s24: steps S22 to S23 are circulated until the loss function for judging the similarity is completely converged, and the downsampling multiple at the moment is output;
s25: and (3) finishing the super-resolution neural network training, taking the downsampling multiple at the moment as an actually measured sampling position for setting the smoke concentration, and taking the smoke concentration distribution output by the first generation network at the moment as the smoke concentration fine distribution at the initial moment of reconstruction.
5. The rapid inversion system of cloud concentration space-time distribution of claim 4, wherein said gaussian diffusion model is as follows:
wherein,expressed in space->The concentration of the smoke cloud at the location; />Representing the concentration gain function, +.>Representing a Gaussian distributionyStandard deviation of axial direction>Representing a Gaussian distributionzStandard deviation in axial direction.
6. A computer readable storage medium having stored thereon a plurality of acquisition classification procedures for being invoked by a processor and performing the rapid inversion method of the concentration-temporal profile of a cloud of smoke according to any one of claims 1-3.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110879755.3A CN113654959B (en) | 2021-07-29 | 2021-07-29 | Rapid inversion method and system for smoke cloud concentration space-time distribution |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110879755.3A CN113654959B (en) | 2021-07-29 | 2021-07-29 | Rapid inversion method and system for smoke cloud concentration space-time distribution |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113654959A CN113654959A (en) | 2021-11-16 |
CN113654959B true CN113654959B (en) | 2023-11-17 |
Family
ID=78479067
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110879755.3A Active CN113654959B (en) | 2021-07-29 | 2021-07-29 | Rapid inversion method and system for smoke cloud concentration space-time distribution |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113654959B (en) |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2013185939A (en) * | 2012-03-07 | 2013-09-19 | Toshiba Corp | Atmospheric diffusion material source survey device, atmospheric diffusion material source survey system and atmospheric diffusion material source survey method |
JP2014106203A (en) * | 2012-11-29 | 2014-06-09 | Mitsubishi Heavy Ind Ltd | Diffusate diffusion state prediction device, method of the same and program |
CN104792674A (en) * | 2015-04-01 | 2015-07-22 | 东南大学 | Method for measuring particle concentration |
CN106126909A (en) * | 2016-06-22 | 2016-11-16 | 重庆科技学院 | PM2.5 concentration prediction method based on Unscented kalman neutral net |
CN106404620A (en) * | 2015-07-30 | 2017-02-15 | 中国科学院遥感与数字地球研究所 | Method and system for inverting ground PM 2.5 through combination of geostatistical interpolation and satellite remote sensing |
CN108072597A (en) * | 2017-11-28 | 2018-05-25 | 南京大学 | Atmosphere particle concentration monitoring method based on support vector machines and leaf magnetics |
US10546243B1 (en) * | 2018-09-10 | 2020-01-28 | Merck Sharp & Dohme Corp. | Predicting particle size distribution and particle morphology |
CN110738354A (en) * | 2019-09-18 | 2020-01-31 | 北京建筑大学 | Method and device for predicting particulate matter concentration, storage medium and electronic equipment |
CN111859800A (en) * | 2020-07-15 | 2020-10-30 | 河海大学 | Method for spatio-temporal estimation and prediction of PM2.5 concentration distribution |
CN112378828A (en) * | 2020-12-11 | 2021-02-19 | 中科三清科技有限公司 | Method and device for inverting concentration of atmospheric fine particulate matters based on satellite remote sensing data |
CN112684471A (en) * | 2020-12-01 | 2021-04-20 | 兰州大学 | Aerosol micro-physical property inversion method based on laser radar |
-
2021
- 2021-07-29 CN CN202110879755.3A patent/CN113654959B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2013185939A (en) * | 2012-03-07 | 2013-09-19 | Toshiba Corp | Atmospheric diffusion material source survey device, atmospheric diffusion material source survey system and atmospheric diffusion material source survey method |
JP2014106203A (en) * | 2012-11-29 | 2014-06-09 | Mitsubishi Heavy Ind Ltd | Diffusate diffusion state prediction device, method of the same and program |
CN104792674A (en) * | 2015-04-01 | 2015-07-22 | 东南大学 | Method for measuring particle concentration |
CN106404620A (en) * | 2015-07-30 | 2017-02-15 | 中国科学院遥感与数字地球研究所 | Method and system for inverting ground PM 2.5 through combination of geostatistical interpolation and satellite remote sensing |
CN106126909A (en) * | 2016-06-22 | 2016-11-16 | 重庆科技学院 | PM2.5 concentration prediction method based on Unscented kalman neutral net |
CN108072597A (en) * | 2017-11-28 | 2018-05-25 | 南京大学 | Atmosphere particle concentration monitoring method based on support vector machines and leaf magnetics |
US10546243B1 (en) * | 2018-09-10 | 2020-01-28 | Merck Sharp & Dohme Corp. | Predicting particle size distribution and particle morphology |
CN110738354A (en) * | 2019-09-18 | 2020-01-31 | 北京建筑大学 | Method and device for predicting particulate matter concentration, storage medium and electronic equipment |
CN111859800A (en) * | 2020-07-15 | 2020-10-30 | 河海大学 | Method for spatio-temporal estimation and prediction of PM2.5 concentration distribution |
CN112684471A (en) * | 2020-12-01 | 2021-04-20 | 兰州大学 | Aerosol micro-physical property inversion method based on laser radar |
CN112378828A (en) * | 2020-12-11 | 2021-02-19 | 中科三清科技有限公司 | Method and device for inverting concentration of atmospheric fine particulate matters based on satellite remote sensing data |
Non-Patent Citations (5)
Title |
---|
Cloud removal in Sentinel-2 imagery using a deep residual neural network and SAR-optical data fusion;Andrea Meraner等;《ISPRS Journal of Photogrammetry and Remote Sensing》;第333-346页 * |
基于卫星遥感和气象再分析资料的北京市PM_(2.5)浓度反演研究;邵琦;陈云浩;李京;;地理与地理信息科学(03);第38-44页 * |
基于随机森林的高分辨率PM_(2.5)遥感反演――以广东省为例;申原;陈朝亮;钱静;刘军;;集成技术(03);第33-43页 * |
***烟云扩散的时空分布模型及特性;段中山;过惠平;冯孝杰;罗昆升;袁伟;;***与冲击(05);第84-91页 * |
空间烟云的扩散模型;陈宁;陈萍萍;孙胜利;;航天电子对抗(01);第26-28+52页 * |
Also Published As
Publication number | Publication date |
---|---|
CN113654959A (en) | 2021-11-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111862251B (en) | Method, device, storage medium and electronic equipment for medical image reconstruction technology | |
CN111811694B (en) | Temperature calibration method, device, equipment and storage medium | |
Mdrafi et al. | Joint learning of measurement matrix and signal reconstruction via deep learning | |
CN108896456B (en) | Aerosol extinction coefficient inversion method based on feedback type RBF neural network | |
CN113326851A (en) | Image feature extraction method and device, electronic equipment and storage medium | |
CN111832228A (en) | Vibration transmission system based on CNN-LSTM | |
Hao et al. | Displacement field denoising for high-temperature digital image correlation using principal component analysis | |
CN113487019A (en) | Circuit fault diagnosis method and device, computer equipment and storage medium | |
Daei et al. | Exploiting prior information in block-sparse signals | |
CN115457492A (en) | Target detection method and device, computer equipment and storage medium | |
Khlamov et al. | The astronomical object recognition and its near-zero motion detection in series of images by in situ modeling | |
CN113654959B (en) | Rapid inversion method and system for smoke cloud concentration space-time distribution | |
CN114519844A (en) | Crowd density estimation method and system based on visual converter | |
CN112711032B (en) | Radar target detection method and system based on graph data and GCN | |
CN112434935B (en) | Selectable PM2.5 concentration estimation method | |
Xie et al. | Multi-Frame Star Image Denoising Algorithm Based on Deep Reinforcement Learning and Mixed Poisson–Gaussian Likelihood | |
US20230401670A1 (en) | Multi-scale autoencoder generation method, electronic device and readable storage medium | |
Pu et al. | ORTP: A video SAR imaging algorithm based on low-tubal-rank tensor recovery | |
CN108492317B (en) | Infrared image reconstruction method based on Archimedes spiral as observation matrix | |
Watzenig et al. | Accelerated Markov chain Monte Carlo sampling in electrical capacitance tomography | |
CN116058852B (en) | Classification system, method, electronic device and storage medium for MI-EEG signals | |
Chertenkov et al. | Finite-size analysis in neural network classification of critical phenomena | |
CN117953029B (en) | General depth map completion method and device based on depth information propagation | |
CN113742911B (en) | High-precision self-adaptive blind estimation method of chaotic time sequence | |
WO2024071377A1 (en) | Information processing device, information processing method, and program |
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 |