CN112200349B - Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN - Google Patents

Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN Download PDF

Info

Publication number
CN112200349B
CN112200349B CN202010972149.1A CN202010972149A CN112200349B CN 112200349 B CN112200349 B CN 112200349B CN 202010972149 A CN202010972149 A CN 202010972149A CN 112200349 B CN112200349 B CN 112200349B
Authority
CN
China
Prior art keywords
picture
predrnn
formula
image
remote sensing
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
Application number
CN202010972149.1A
Other languages
Chinese (zh)
Other versions
CN112200349A (en
Inventor
罗培韬
吕新杰
姜峥超
李君�
王玲霞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Balance Machine Technology Shenzhen Co ltd
Original Assignee
Balance Machine Technology Shenzhen Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Balance Machine Technology Shenzhen Co ltd filed Critical Balance Machine Technology Shenzhen Co ltd
Priority to CN202010972149.1A priority Critical patent/CN112200349B/en
Publication of CN112200349A publication Critical patent/CN112200349A/en
Application granted granted Critical
Publication of CN112200349B publication Critical patent/CN112200349B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
    • 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
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • 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
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/13—Satellite images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
    • 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
    • Y02A30/00—Adapting or protecting infrastructure or their operation
    • Y02A30/60—Planning or developing urban green infrastructure

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Business, Economics & Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Operations Research (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Marketing (AREA)
  • Algebra (AREA)
  • Astronomy & Astrophysics (AREA)
  • Remote Sensing (AREA)
  • Multimedia (AREA)
  • Molecular Biology (AREA)
  • Development Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Game Theory and Decision Science (AREA)
  • Computational Linguistics (AREA)
  • Image Processing (AREA)
  • Computing Systems (AREA)
  • Quality & Reliability (AREA)
  • Biomedical Technology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)

Abstract

The invention discloses a remote sensing image heat island effect prediction method based on a single window algorithm and PredRNN, which comprises the following steps: downloading the image; image preprocessing; a single window algorithm; cutting the satellite remote sensing image processed by the single-window algorithm at each moment into a plurality of small pictures according to a certain size, marking the area and the time sequence number of each picture, and constructing a PredRNN model; processing the segmented picture set into a sequence set form, segmenting the picture sequence set according to the ratio of 8:2 of a training set to a testing set, training by using PredRNN, using the testing set as the evaluation of parameter tuning of the model, and finally determining the model to obtain the change relation of the model on the historical satellite remote sensing picture and the future satellite remote sensing picture. According to the method, on one hand, the earth surface temperature inversion result of the time point when the satellite image is missing is supplemented, and the regularity and continuity of time are guaranteed; on the other hand, the effect of predicting the future thermal environment of the city is realized.

Description

Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN
Technical Field
The invention relates to a heat island effect prediction method, belongs to the technical field of earth science, and particularly relates to a remote sensing image heat island effect prediction method based on a single window algorithm and PredRNN.
Background
The surface temperature plays a crucial role in the energy exchange between the earth's surface and the atmosphere, and has a significant impact on climate, environmental changes, and human productivity. Any object whose temperature exceeds the thermodynamic temperature of 0K will constantly emit thermal infrared radiation outward. The thermal infrared remote sensing is used for identifying ground objects and quantitatively inverting ground surface parameter information, such as remote sensing inversion of ground surface temperature and ground surface emissivity, by acquiring thermal infrared information of two atmospheric windows of 3-5 microns and 8-14 microns of a target ground object based on a satellite-borne or airborne thermal infrared sensor.
The single window algorithm is suitable for remote sensing data with only one thermal wave band and is mainly used for performing surface temperature inversion on TM6 and ETM + data. The algorithm utilizes the radiation values observed by a single thermal infrared channel within an atmospheric window. In the prior art, the remote sensing image is subjected to temperature inversion only based on a single-window algorithm to obtain urban earth surface temperature data. Therefore, the earth surface temperature of the city over the years is monitored, and the development dynamics of the urban heat island effect is analyzed.
But the time continuity is poor due to the low time resolution of most freely acquired satellite images. Therefore, most researches only can carry out research and analysis on the existing satellite images, and the earth surface temperature is calculated by using a proper algorithm, so that the urban heat island effect is monitored and analyzed regularly.
Disclosure of Invention
In order to overcome the defects of the technology, the invention provides a remote sensing image heat island effect prediction method based on a single window algorithm and PredRNN.
In order to solve the technical problems, the invention adopts the technical scheme that: a remote sensing image heat island effect prediction method based on a single window algorithm and PredRNN comprises the following steps:
step 1, image downloading, namely downloading a remote sensing image of a designated research area over the years according to research requirements;
step 2, image preprocessing, namely preprocessing operations such as geometric correction, inlaying, cutting, radiometric calibration, atmospheric correction and the like are carried out on the image;
step 3, performing a single window algorithm, namely calculating a spectral radiance value of a DN value of an image thermal infrared band by using a formula 1, calculating a brightness temperature of the image thermal infrared band by using a formula 2, calculating an NDVI value by using the image infrared band and a near infrared band, calculating a ground surface emissivity by using the NDVI and a formula 3, estimating an atmospheric transmittance, and calculating a ground surface temperature by using the single window algorithm;
step 4, cutting the satellite remote sensing image processed by the single-window algorithm at each moment into a plurality of small pictures according to a certain size, marking the area and the time sequence number of each picture, and constructing a PredRNN model, wherein the PredRNN is connected to the first layer of the next sequence as a hidden layer input by establishing a channel at the topmost layer of the previous sequence, so that the effect that the network can combine and utilize the characteristics closest to an output layer is realized;
step 5, processing the picture set segmented in the step 4 into a sequence set form, segmenting the picture sequence set according to the ratio 8:2 of a training set to a testing set, training by using PredRNN, using the testing set as the evaluation of parameter tuning of the model, and finally determining the model to obtain the change relation of the model on the historical satellite remote sensing picture and the future satellite remote sensing picture;
and 6, carrying out extrapolation prediction on the future urban surface temperature through the trained PredRNN model.
Further, equation 1 is:
L10＝gain*DN+bias
in the formula: l is10Is a spectral radiance value; gain is the gain amount; bias is the offset.
Further, equation 2 is:
Figure GDA0002791660370000021
in the formula: t is10Representing the brightness temperature (K) of the thermal infrared band of the image; l is10Is the spectral radiance value W/(m)2*μm*sr)；K1K2A constant value preset before transmission.
Further, formula 3 includes formula 31 or formula 32;
equation 31 is:
ε＝PvRvεv+(1-Pv)Rmεm+dε
equation 32 is:
ε＝PvRvεv+(1-Pv)Rsεs+dε
formula 31 is used for calculating the surface emissivity of the town earth surface, and formula 32 is used for calculating the surface emissivity of the natural earth surface;
in the formula: epsilons，εm，εvRespectively the surface emissivity of bare soil, the surface of a building and the surface emissivity of vegetation pure pixel;
Figure GDA0002791660370000031
Pvis the proportion of the mixed pixels to be planted,
Figure GDA0002791660370000032
Rv＝0.9332+0.0585Pv
Rm＝0.9886+0.1287Pv
NDVIsa normalized vegetation index representing an area completely covered by bare soil or no vegetation; NDVIvIs the normalized vegetation index of the completely covered picture element; rv，RmThe temperature ratios of vegetation and buildings, respectively.
Further, the inversion algorithm of the single window algorithm is shown in the formula (3-5):
Figure GDA0002791660370000033
wherein LST is the surface temperature (K); t issRepresenting the BT value (K) received by the satellite; t isoRepresents the atmospheric mean temperature value (K): a. b is a reference coefficient, a-67.355351, b-0.458606: C. d is obtained as shown in formulas (3-6) and (3-7):
C＝ετ (3-6)
D＝(1-τ)[1+(1-ε)τ] (3-7)
where ε represents the surface emissivity and τ represents the atmospheric transmittance.
Further, the neural unit structure of PredRNN is shown in fig. 2:
wherein WtNeural elements at layer t for the PredRNN network, XtFor the input picture of the network at time t,
Figure GDA0002791660370000041
for the hidden state of the ith neuron in the network at time t,
Figure GDA0002791660370000042
for the output image of the i-th layer network at time t,
Figure GDA0002791660370000043
the picture is an output picture of the PredRNN network at the time t;
Figure GDA0002791660370000044
Figure GDA0002791660370000045
Figure GDA0002791660370000046
Figure GDA0002791660370000047
Figure GDA0002791660370000048
Figure GDA0002791660370000049
wherein
Figure GDA00027916603700000410
Pictures input to the first layer PredRNN network for time t, gtFor abstract extraction of the PredRNN neural unit for current time input and last time input information,
Figure GDA00027916603700000411
for the abstract information extraction layer to the convolution kernel of the input picture at the current moment,
Figure GDA00027916603700000412
outputting a convolution kernel of a picture at the last moment for the abstract information extraction layer; f. oftFor the forgetting gate of the PredRN network at time t,
Figure GDA00027916603700000413
Figure GDA00027916603700000414
respectively inputting a picture at the current moment, inputting a picture at the last moment and performing convolution kernel in a network hidden layer state at the last moment for the forgetting gate; i.e. itIs the input gate of the network at time t,
Figure GDA00027916603700000415
respectively inputting a picture at the current moment, inputting a picture at the last moment and performing convolution kernel of a network hidden layer state at the last moment for the input gate; otIs an output gate of the network and is,
Figure GDA00027916603700000416
respectively inputting pictures for the output gate at the current moment and the last momentInputting a picture, and performing convolution kernel of a network hidden layer state at the last moment; symbol indicates convolution operation, symbol indicates dot product operation.
The method carries out earth surface temperature inversion on the existing satellite images, and carries out model training by using the inversion result. On one hand, the model supplements the earth surface temperature inversion result of the time point when the satellite image is missing, and guarantees the regularity and continuity of time; on the other hand, the effect of predicting the future thermal environment of the city is realized.
Drawings
Fig. 1 is an overall flow chart of urban heat island effect prediction according to the invention.
FIG. 2 is a schematic diagram of the structure of the neural unit of PredRNN.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
FIG. 1 shows a method for predicting the heat island effect of a remote sensing image based on a single window algorithm and PredRNN, which is characterized in that: the method comprises the following steps:
step 1, downloading images: downloading remote sensing images of a designated research area over the years according to research requirements;
step 2, image preprocessing: carrying out preprocessing operations such as geometric correction, embedding, cutting, radiometric calibration, atmospheric correction and the like on the image;
step 3, single window algorithm:
a. calculating a spectral radiation value of the DN value of the thermal infrared band by using a formula 1, wherein the formula 1 is as follows:
L10＝gain*DN+bias
in the formula: l is10Is a spectral radiance value; gain is the gain amount; bias is offset;
b. calculating the brightness and temperature of the thermal infrared band of the image by using a formula 2, wherein the formula 2 is as follows:
Figure GDA0002791660370000051
in the formula: t is a unit of10Representing heat of imageAn infrared band brightness temperature (K); l is10Is the spectral radiance value W/(m)2*μm*sr)；K1K2A constant preset before transmission;
c. calculating the NDVI value by utilizing the infrared and near-infrared wave bands of the image;
d. calculating the surface emissivity by using the NDVI and the formula 3, wherein the formula 3 includes the formula 31 or the formula 32, and the formula 31 is:
ε＝PvRvεv+(1-Pv)Rmεm+dε
equation 32 is:
ε＝PvRvεv+(1-Pv)Rsεs+dε
formula 31 is used for calculating the surface emissivity of the town earth surface, and formula 32 is used for calculating the surface emissivity of the natural earth surface;
in the formula: epsilons，εm，εvRespectively the surface emissivity of bare soil, the surface of a building and the surface emissivity of vegetation pure pixel;
Figure GDA0002791660370000061
Pvis the proportion of the mixed pixels to be planted,
Figure GDA0002791660370000062
Rv＝0.9332+0.0585Pv
Rm＝0.9886+0.1287Pv
NDVIsa normalized vegetation index representing an area completely covered by bare soil or no vegetation; NDVIvIs the normalized vegetation index of the completely covered picture element; rv，RmTemperature ratios of vegetation and buildings respectively;
e. and (3) estimating atmospheric transmittance, and calculating the surface temperature by using a single-window algorithm, wherein the inversion algorithm of the single-window algorithm is shown as the formula (3-5):
Figure GDA0002791660370000063
wherein LST is the surface temperature (K); t issRepresents the BT value (K) received by the satellite; t isoRepresents the average action temperature value (K) of the atmosphere; a. b is a reference coefficient, a is-67.355351, b is 0.458606; C. d is obtained as shown in formulas (3-6) and (3-7):
C＝ετ (3-6)
D＝(1-τ)[1+(1-ε)τ] (3-7)
where ε represents the surface emissivity and τ represents the atmospheric transmittance. .
And 4, step 4:
a. cutting the satellite remote sensing image processed by the single-window algorithm at each moment into a plurality of small pictures according to a certain size, and marking the area and the time sequence number of each picture
b. Constructing a PredRNN model, wherein the earliest model for performing time-space prediction is a ConvLSTM model, and improving the model structure of the LSTM, adding a CNN convolution network layer in the model structure to ensure that the model has a better prediction effect on a radar echo diagram, and the model is used for extracting more complex features, wherein the extraction effect is usually realized by stacking a plurality of layers of networks, but the ConvLSTM model structure ignores the features extracted from the top layer of the last sequence, so that the PredRNN is structurally improved, and the nerve unit structure of the PredRNN is as shown in FIG. 2:
wherein WtNeural elements at layer t for the PredRNN network, XtFor the input picture of the network at time t,
Figure GDA0002791660370000071
for the hidden state of the ith neuron in the network at time t,
Figure GDA0002791660370000072
for the output image of the i-th layer network at time t,
Figure GDA0002791660370000073
is PredThe RNN outputs pictures at the time t;
Figure GDA0002791660370000074
Figure GDA0002791660370000075
Figure GDA0002791660370000076
Figure GDA0002791660370000077
Figure GDA0002791660370000078
Figure GDA0002791660370000079
wherein
Figure GDA00027916603700000710
Pictures input to the first layer PredRNN network for time t, gtFor abstract extraction of the PredRNN neural unit for current time input and last time input information,
Figure GDA00027916603700000711
for the abstract information extraction layer to the convolution kernel of the input picture at the current moment,
Figure GDA00027916603700000712
and outputting the convolution kernel of the picture at the last moment for the abstract information extraction layer. f. oftFor the forgetting gate of the PredRN network at time t,
Figure GDA00027916603700000713
Figure GDA00027916603700000714
and respectively inputting a picture at the current moment, inputting a picture at the last moment and performing convolution kernel in a network hidden layer state at the last moment for the forgetting gate. i.e. itIs the input gate of the network at time t,
Figure GDA00027916603700000715
and respectively inputting the picture at the current moment, the picture at the last moment and the convolution kernel in the network hidden layer state at the last moment for the input gate. otIs an output gate of the network and is,
Figure GDA00027916603700000716
respectively inputting a picture at the current moment, inputting a picture at the last moment and performing convolution kernel of a network hidden layer state at the last moment for the output gate; symbol represents convolution operation, symbol represents dot product operation;
PredRNN establishes a channel through the top layer of the last sequence to be connected to the first layer of the next sequence to be used as hidden layer input, and the effect that the network can combine and utilize the characteristics of the closest output layer is achieved;
step 5, processing the picture set segmented in the step 4 into a sequence set form, segmenting the picture sequence set according to the ratio 8:2 of a training set to a testing set, training by using PredRNN, using the testing set as the evaluation of parameter tuning of the model, and finally determining the model to obtain the change relation of the model on the historical satellite remote sensing picture and the future satellite remote sensing picture;
and 6, carrying out extrapolation prediction on the future urban surface temperature through the trained PredRNN model.
The method utilizes a deep learning model-PredRNN to simulate and predict the time point of satellite image loss while monitoring the urban heat island effect. On one hand, the time continuity of the heat island effect research of the city over the years can be enhanced, and on the other hand, the prediction analysis can be carried out on the future heat environment of the city. The method firstly utilizes the traditional algorithm to analyze the urban heat island effect capable of obtaining the satellite image, achieves the purpose of monitoring the historical result and simultaneously realizes the simulation and prediction of the urban heat island effect in the future. The method not only can carry out surface temperature inversion on the time points at which the satellite images can be obtained, but also utilizes the means of machine learning to carry out urban heat island inversion research on the time points at which the images are missing.
The above embodiments are not intended to limit the present invention, and the present invention is not limited to the above examples, and variations, modifications, additions and substitutions which may be made by those skilled in the art within the technical scope of the present invention are also within the protective scope of the present invention.

Claims (3)

1. A remote sensing image heat island effect prediction method based on a single window algorithm and PredRNN is characterized by comprising the following steps: the method comprises the following steps:
step 1, downloading images: downloading a remote sensing image of a designated research area over the years according to research requirements;
step 2, image preprocessing: performing geometric correction, embedding, cutting, radiometric calibration and atmospheric correction on the image;
step 3, single window algorithm: calculating a spectral radiation value of DN value of the image thermal infrared band by using a formula 1, calculating brightness and temperature of the image thermal infrared band by using a formula 2, calculating an NDVI value by using the image infrared band and the near infrared band, calculating surface specific radiance by using NDVI and a formula 3, estimating atmospheric transmittance, and calculating surface temperature by using a single window algorithm;
step 4, cutting the satellite remote sensing image processed by the single-window algorithm at each moment into a plurality of small pictures according to a certain size, marking the area and the time sequence number of each picture, and constructing a PredRNN model, wherein the PredRNN is connected to the first layer of the next sequence as a hidden layer input by establishing a channel at the topmost layer of the previous sequence, so that the effect that the network can combine and utilize the characteristics closest to an output layer is realized;
step 5, processing the picture set segmented in the step 4 into a sequence set form, segmenting the picture sequence set according to the ratio 8:2 of a training set to a testing set, training by using PredRNN, using the testing set as the evaluation of parameter tuning of the model, and finally determining the model to obtain the change relation of the model on the historical satellite remote sensing picture and the future satellite remote sensing picture;
step 6, carrying out extrapolation prediction on the future urban ground surface temperature through the trained PredRNN model;
the formula 1 is:
L10＝gain*DN+bias
in the formula: l is a radical of an alcohol10Is a spectral radiance value; gain is the gain amount; bias is offset;
the formula 2 is:
Figure FDA0003572515170000011
in the formula: t is10Representing the brightness temperature (K) of the thermal infrared band of the image; l is10Is the spectral radiance value W/(m)2*um*sr)；K1、K2A constant preset before transmission;
the formula 3 includes a formula 31 or a formula 32;
the equation 31 is:
∈＝PvRvεv+(1-Pv)Rmεm+dε
the equation 32 is:
ε＝PvRvεv+(1-Pv)Rsεs+dε
formula 31 is used for calculating the surface emissivity of the town earth surface, and formula 32 is used for calculating the surface emissivity of the natural earth surface;
in the formula:
Figure FDA0003572515170000023
respectively the surface emissivity of bare soil, the surface of a building and the surface emissivity of vegetation pure pixel;
Figure FDA0003572515170000021
Pvis the proportion of the mixed pixels to be planted,
Figure FDA0003572515170000022
Rv＝0.9332+0.0585Pv
Rm＝0.9886+0.1287Pv
NDVIsa normalized vegetation index representing an area completely covered by bare soil or no vegetation; NDVIvIs the normalized vegetation index of the completely covered pixel; rvRmThe temperature ratios of vegetation and buildings, respectively.
2. The remote sensing image heat island effect prediction method based on the single window algorithm and PredRNN according to claim 1, characterized in that: the inversion algorithm of the single window algorithm is shown in the formula (3-5):
Figure FDA0003572515170000031
wherein LST is the surface temperature (K); t is a unit ofsRepresents the BT value (K) received by the satellite; t isaRepresents the atmospheric mean temperature value (K); a. b is a reference coefficient, a is-67.355351, b is 0.458606; C. d is obtained as shown in formulas (3-6) and (3-7):
C＝ετ (3-6)
D＝(1-τ)[1+(1-ε)τ] (3-7)
where ε represents the surface emissivity and τ represents the atmospheric transmittance.
3. The remote sensing image heat island effect prediction method based on the single window algorithm and PredRNN according to claim 1, characterized in that: PredRNN has a neural unit structure of: the pictures from t-1 to t +1 are marked as X, and the picture input at the time of t is marked as XtThe network of the model is denoted as W, and the neural units of the t-th layer network are denoted as WtAnd the output picture result is recorded as
Figure FDA0003572515170000032
The output at time t is recorded as
Figure FDA0003572515170000033
And H and M are the output image and hidden layer states for W,
Figure FDA0003572515170000034
is WtHidden state and output image at ith layer;
Figure FDA0003572515170000035
Figure FDA0003572515170000036
Figure FDA0003572515170000037
Figure FDA0003572515170000038
Figure FDA00035725151700000315
Figure FDA0003572515170000039
wherein
Figure FDA00035725151700000310
Pictures input to first layer PredRNN network for time t，gtFor abstract extraction of the PredRNN neural unit for current time input and last time input information,
Figure FDA00035725151700000311
for the abstract information extraction layer to the convolution kernel of the input picture at the current moment,
Figure FDA00035725151700000312
a convolution kernel of an image output at the last moment is taken as an abstract information extraction layer; f. oftFor the forgetting gate of the PredRN network at time t,
Figure FDA00035725151700000313
Figure FDA00035725151700000314
respectively inputting a picture at the current moment, inputting a picture at the last moment and performing convolution kernel in a network hidden layer state at the last moment for the forgetting gate; i.e. itIs the input gate of the network at time t,
Figure FDA0003572515170000041
respectively inputting a picture at the current moment, inputting a picture at the last moment and performing convolution kernel of a network hidden layer state at the last moment for the input gate; o istIs an output gate of the network and is,
Figure FDA0003572515170000042
respectively inputting a picture at the current moment, inputting a picture at the last moment and performing convolution kernel of a network hidden layer state at the last moment for the output gate; symbol indicates convolution operation, symbol indicates dot product operation.
CN202010972149.1A 2020-09-16 2020-09-16 Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN Active CN112200349B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010972149.1A CN112200349B (en) 2020-09-16 2020-09-16 Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010972149.1A CN112200349B (en) 2020-09-16 2020-09-16 Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN

Publications (2)

Publication Number Publication Date
CN112200349A CN112200349A (en) 2021-01-08
CN112200349B true CN112200349B (en) 2022-06-03

Family

ID=74015034

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010972149.1A Active CN112200349B (en) 2020-09-16 2020-09-16 Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN

Country Status (1)

Country Link
CN (1) CN112200349B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113743000B (en) * 2021-08-13 2023-04-18 电子科技大学 Method for generating all-weather surface temperature with high time resolution
CN115270638B (en) * 2022-08-11 2024-02-06 北华航天工业学院 Urban thermal environment downscaling space-time analysis and prediction method and system
CN116578662A (en) * 2023-05-23 2023-08-11 安徽建筑大学 Urban heat island effect monitoring system and method

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102103203A (en) * 2011-01-19 2011-06-22 环境保护部卫星环境应用中心 Environmental satellite 1-based surface temperature single-window inversion method
CN105678225A (en) * 2015-12-29 2016-06-15 中国科学院深圳先进技术研究院 Urban heat island effect space variation detection method and system
CN105930633A (en) * 2016-04-05 2016-09-07 中国科学院遥感与数字地球研究所 Method for forecasting urban heat island effect
CN107678075A (en) * 2017-11-13 2018-02-09 深圳先进技术研究院 A kind of urban heat land effect monitoring method and system based on domestic satellite
CN108168710A (en) * 2017-12-28 2018-06-15 福建农林大学 A kind of city tropical island effect appraisal procedure based on remote sensing technology
CN110967695A (en) * 2019-10-28 2020-04-07 兰州大方电子有限责任公司 Radar echo extrapolation short-term prediction method based on deep learning
CN111062410A (en) * 2019-11-05 2020-04-24 复旦大学 Star information bridge weather prediction method based on deep learning

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102103203A (en) * 2011-01-19 2011-06-22 环境保护部卫星环境应用中心 Environmental satellite 1-based surface temperature single-window inversion method
CN105678225A (en) * 2015-12-29 2016-06-15 中国科学院深圳先进技术研究院 Urban heat island effect space variation detection method and system
CN105930633A (en) * 2016-04-05 2016-09-07 中国科学院遥感与数字地球研究所 Method for forecasting urban heat island effect
CN107678075A (en) * 2017-11-13 2018-02-09 深圳先进技术研究院 A kind of urban heat land effect monitoring method and system based on domestic satellite
CN108168710A (en) * 2017-12-28 2018-06-15 福建农林大学 A kind of city tropical island effect appraisal procedure based on remote sensing technology
CN110967695A (en) * 2019-10-28 2020-04-07 兰州大方电子有限责任公司 Radar echo extrapolation short-term prediction method based on deep learning
CN111062410A (en) * 2019-11-05 2020-04-24 复旦大学 Star information bridge weather prediction method based on deep learning

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
"时空序列预测之PredRNN（用ST-LSTM预测学习的循环神经网络";Miracle8070;《CSDNhttp://blog.csdn.net/abcdefg90876/article/details/104140563》;20200216;第1-16页 *
基于Landsat数据的昆明地区热岛效应分析;刘鑫等;《科学技术与工程》;20110508(第13期);第137-141页 *
应用单窗算法进行北京规划市区热岛效应研究;谢军飞等;《园林科技》;20111231;第7-11页 *

Also Published As

Publication number Publication date
CN112200349A (en) 2021-01-08

Similar Documents

Publication Publication Date Title
CN112200349B (en) Remote sensing image heat island effect prediction method based on single window algorithm and PredRNN
CN110969088B (en) Remote sensing image change detection method based on significance detection and deep twin neural network
Ma et al. Coupling urbanization analyses for studying urban thermal environment and its interplay with biophysical parameters based on TM/ETM+ imagery
CN108985959B (en) Wheat powdery mildew remote sensing monitoring method based on surface temperature inversion technology
CN102183237B (en) Device and method for measuring two-waveband cloud height of foundation
CN109598273B (en) City entity boundary identification method fusing earth surface temperature and building index
CN109115995B (en) Soil moisture information acquisition methods, device, computer equipment and storage medium
Li et al. Evaluation of 10 year AQUA/MODIS land surface temperature with SURFRAD observations
CN111582575B (en) Method for identifying urban thermal environment formation development leading factors under multiple space-time scales
Rahman et al. A comparison of four relative radiometric normalization (RRN) techniques for mosaicing H-res multi-temporal thermal infrared (TIR) flight-lines of a complex urban scene
CN115481368B (en) Vegetation coverage estimation method based on full remote sensing machine learning
CN105204024B (en) Method of the microwave remote sensing surface temperature to thermal infrared remote sensing skin face temperature transition
CN111323129A (en) Earth surface temperature inversion method based on broadband thermal infrared image
CN108051371A (en) A kind of shadow extraction method of ecology-oriented environment parameter remote-sensing inversion
CN113553697B (en) Long-time-sequence multi-source data-based vegetation disturbance analysis method for coal mining
CN112906310A (en) Method for optimizing BP neural network microwave remote sensing soil moisture inversion by considering firefly algorithm
Scolari et al. Local estimation of the global horizontal irradiance using an all-sky camera
CN111723711A (en) Plianes and object-oriented mulching film information extraction method and system
CN111178384A (en) Meteorological satellite dynamic convection layer top height inversion method and device, storage medium and computer equipment
CN102901563B (en) Method and device for determining land surface emissivity of narrow band and broad band simultaneously
CN112630160A (en) Unmanned aerial vehicle track planning soil humidity monitoring method and system based on image acquisition and readable storage medium
Khalil et al. Time series analysis of land surface temperature and drivers of urban heat island effect based on remotely sensed data to develop a prediction model
Chen et al. Sun glint correction with an inherent optical properties data processing system
Hu et al. Short-term tropical cyclone intensity forecasting from satellite imagery based on the deviation angle variance technique
CN113836490B (en) STARFM surface temperature fusion prediction method based on data linear regression

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