Multi-source fusion method, device and equipment for satellite surface temperature data and medium

Abstract

The invention provides a multi-source fusion method, device and equipment for satellite surface temperature data and a medium, and relates to the technical field of surface temperature, and the method comprises the steps: obtaining a first surface temperature and a second surface temperature of a research region at an initial moment, and a first surface temperature of the research region at a target moment; wherein the spatial resolution of the first surface temperature is lower than that of the second surface temperature; generating a first surface temperature downscaling result of the research area at the target moment based on the first surface temperature and the second surface temperature at the initial moment and the first surface temperature at the target moment through a pre-constructed target downscaling model; and based on the second surface temperature of the research area at the target moment, performing deviation correction on the first surface temperature downscaling result to obtain the target surface temperature of the research area at the target moment. The method can significantly improve the quality of the surface temperature.

Description

Technical Field

[0001] The present invention relates to the field of land surface temperature technology, and in particular to a multi-source fusion method, device, equipment and medium for satellite land surface temperature data. Background Art

[0002] Land surface temperature (LST) is an important physical parameter that links the material exchange and energy balance between the earth's surface and the atmospheric boundary layer. Its spatial distribution and temporal dynamics play an important role in atmospheric circulation processes, soil moisture monitoring, vegetation evapotranspiration estimation, and urban heat island effect research. Traditionally, land surface temperature is measured using site data obtained from ground meteorological stations. However, meteorological site data cannot provide continuous real land surface temperature data over a large area. Interpolation and other methods are usually used to obtain surface data in the study area, which cannot obtain accurate results. This has become a bottleneck hindering the development of land surface temperature research. In recent years, remote sensing technology has developed rapidly, and large-area, long-term series of remote sensing images can be obtained. Its thermal infrared band directly reflects the thermal radiation intensity of the surface and provides continuous data over a large range in real time. Using thermal infrared data to invert land surface temperature has become the mainstream direction of land surface temperature research. However, due to the complexity of the process of satellite receiving surface radiation energy and the direct influence of the atmospheric environment, the temperature reflected by satellite remote sensing images is different from the real surface temperature. The temperature directly measured by the thermal infrared band of remote sensing satellites is called brightness temperature, which represents the heat received by the sensor after being affected by the atmosphere. For the study of the real surface, it is necessary to establish a temperature field model to obtain the real surface temperature. For the study of surface temperature in urban areas, the existing satellite data cannot obtain the surface temperature distribution that satisfies its high temporal and high spatial resolution at the same time. For example, images such as FY3D can provide daily data, but the spatial resolution is 1000 meters, which is not enough to reflect the heat difference in the local structure of the city; Landsat-8 can provide surface temperature data with a spatial resolution of up to 30 meters, but its 16-day re-entry cycle and acquisition are subject to the influence of cloud and rain weather, so it cannot provide continuous surface temperature change data in time, and the timeliness is low. Therefore, the fusion of different remote sensing data to obtain surface temperature data with high temporal and spatial resolution plays an important role in the study of thermal environment in urban space.

[0003] Usually, Landsat-8 and MODIS data are used for spatial downscaling data fusion. Some scholars have also demonstrated the application of ASTER and AVHRR in data fusion. However, they all fuse the thermal infrared channels first and then perform surface temperature inversion. The fused channel data has a certain degree of distortion, which has a great impact on the subsequent surface temperature inversion. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a multi-source fusion method, device, equipment and medium for satellite surface temperature data, which can significantly improve the quality of surface temperature.

[0005] In a first aspect, an embodiment of the present invention provides a multi-source fusion method for satellite surface temperature data, comprising:

[0006] Obtaining a first surface temperature and a second surface temperature of the study area at an initial time, and the first surface temperature of the study area at a target time; wherein the spatial resolution of the first surface temperature is lower than the spatial resolution of the second surface temperature;

[0007] Through the pre-constructed target downscaling model, based on the first surface temperature and the second surface temperature at the initial moment, and the first surface temperature at the target moment, the downscaling result of the first surface temperature of the study area at the target moment is generated;

[0008] Based on the second surface temperature of the study area at the target time, the deviation of the first surface temperature downscaling result is corrected to obtain the target surface temperature of the study area at the target time.

[0009] In one embodiment, a downscaling result of the first land surface temperature of the study area at the target time is generated based on the first land surface temperature and the second land surface temperature at the initial time and the first land surface temperature at the target time through a pre-constructed target downscaling model, including:

[0010] Set a moving window for the study area;

[0011] For any pixel point in the current moving window, determine the difference between the first surface temperature of the pixel point at the initial time and the first surface temperature of the pixel point at the target time;

[0012] The difference values ​​corresponding to each pixel point in the moving window are spatially merged using the conversion coefficient corresponding to the land use coverage type to which the pixel point belongs and the target weight coefficient corresponding to the pixel point;

[0013] The spatial merging result is merged with the second surface temperature of the central pixel point in the moving window at the initial time, and the first surface temperature downscaling result of the central pixel point at the target time is obtained;

[0014] The moving window is controlled to move until the first surface temperature downscaling result of each central pixel point in the study area at the target time is determined.

[0015] In one embodiment, before generating a downscaling result of the first land surface temperature of the study area at the target time based on the first land surface temperature and the second land surface temperature at the initial time and the first land surface temperature at the target time by using a pre-constructed target downscaling model, the method further includes:

[0016] Obtaining a first surface temperature and a second surface temperature of the study area at a first historical moment, and a first surface temperature and a second surface temperature of the study area at a second historical moment;

[0017] For the pixel points corresponding to any land use coverage type in the study area, determine the difference between the first surface temperature of the pixel point at the first historical moment and its first surface temperature at the second historical moment, as well as the difference between the second surface temperature of the pixel point at the first historical moment and its second surface temperature at the second historical moment; determine the conversion coefficient corresponding to the land use coverage type based on the ratio of the two differences.

[0018] In one embodiment, before generating a downscaling result of the first land surface temperature of the study area at the target time based on the first land surface temperature and the second land surface temperature at the initial time and the first land surface temperature at the target time by using a pre-constructed target downscaling model, the method further includes:

[0019] For any pixel point in the current moving window, determine the surface temperature similarity and vegetation index similarity corresponding to the pixel point, and perform weighted summation of the surface temperature similarity and the vegetation index similarity to obtain the pixel similarity corresponding to the pixel point; and determine the distance between the pixel point and the center pixel point in the moving window;

[0020] Based on the pixel similarity and distance corresponding to the pixel point, the target weight coefficient corresponding to the pixel point is determined.

[0021] In one implementation, determining the surface temperature similarity and vegetation index similarity corresponding to the pixel point includes:

[0022] If the first surface temperature corresponding to the pixel point is greater than the second surface temperature corresponding to the pixel point, the ratio of the second surface temperature to the first surface temperature is used as the surface temperature similarity corresponding to the pixel point; or, if the first surface temperature corresponding to the pixel point is less than the second surface temperature corresponding to the pixel point, the ratio of the first surface temperature to the second surface temperature is used as the surface temperature similarity corresponding to the pixel point;

[0023] And, if the first vegetation index corresponding to the pixel point is greater than the second vegetation index corresponding to the pixel point, the ratio of the second vegetation index to the first vegetation index is used as the vegetation index similarity corresponding to the pixel point; or, if the first vegetation index corresponding to the pixel point is less than the second vegetation index corresponding to the pixel point, the ratio of the first vegetation index to the second vegetation index is used as the vegetation index similarity corresponding to the pixel point; wherein the spatial resolution of the first vegetation index is lower than the spatial resolution of the second vegetation index.

[0024] In one implementation, determining a target weight coefficient corresponding to the pixel point based on the pixel similarity and distance corresponding to the pixel point includes:

[0025] Based on the pixel similarity and distance corresponding to the pixel point, an initial weight coefficient corresponding to the pixel point is determined, and the initial weight coefficient is negatively correlated with the pixel similarity and positively correlated with the distance;

[0026] The ratio between the reciprocal of the initial weight coefficient corresponding to the pixel point and the sum of the reciprocals of the initial weight coefficients corresponding to each pixel point in the moving window is used as the target weight coefficient corresponding to the pixel point.

[0027] In one embodiment, based on the second surface temperature of the study area at the target time, the first surface temperature downscaling result is corrected for deviation to obtain the target surface temperature of the study area at the target time, including:

[0028] The pre-trained random forest machine learning model is used as the surface temperature correction model, and the second surface temperature of the study area at the target time, the first surface temperature downscaling results and the surface temperature influencing factors are input into the surface temperature correction model, so that the surface temperature correction model outputs the target surface temperature of the study area at the target time.

[0029] In a second aspect, an embodiment of the present invention further provides a multi-source fusion device for satellite surface temperature data, comprising:

[0030] A data acquisition module, used to acquire a first surface temperature and a second surface temperature of a study area at an initial time, and the first surface temperature of the study area at a target time; wherein a spatial resolution of the first surface temperature is lower than a spatial resolution of the second surface temperature;

[0031] A downscaling module is used to generate a downscaling result of the first surface temperature of the study area at the target time based on the first surface temperature and the second surface temperature at the initial time and the first surface temperature at the target time through a pre-constructed target downscaling model;

[0032] The deviation correction module is used to perform deviation correction on the first surface temperature downscaling result based on the second surface temperature of the study area at the target time to obtain the target surface temperature of the study area at the target time.

[0033] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement any one of the methods provided in the first aspect.

[0034] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement any one of the methods provided in the first aspect.

[0035] The embodiment of the present invention provides a multi-source fusion method, device, equipment and medium for satellite surface temperature data. First, a first surface temperature and a second surface temperature of a study area at an initial moment, and the first surface temperature of the study area at a target moment are obtained, and the spatial resolution of the first surface temperature is lower than the spatial resolution of the second surface temperature; then, a pre-constructed target downscaling model is used to generate a downscaling result of the first surface temperature of the study area at a target moment based on the first surface temperature and the second surface temperature at the initial moment and the first surface temperature at the target moment; finally, based on the second surface temperature of the study area at the target moment, the first surface temperature downscaling result is corrected for deviation, and the target surface temperature of the study area at the target moment is obtained. The above method generates a first surface temperature downscaling result based on the first surface temperature and the second surface temperature of the study area at the initial time and the first surface temperature at the target time through a target downscaling model, and uses the second surface temperature of the study area at the target time to correct the deviation of the first surface temperature downscaling result, thereby realizing the fusion of multi-source surface temperatures. The embodiment of the present invention performs fusion processing on the basis of surface temperature inversion to ensure the quality of single-source surface temperature, and provides a good basis for data fusion. The fusion result can provide support for global or regional meteorological climate, urban heat island effect and other research.

[0036] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0037] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A schematic flow chart of a multi-source fusion method for satellite surface temperature data provided by an embodiment of the present invention;

[0040] Figure 2 A schematic diagram of the overall process of a multi-source fusion method for satellite surface temperature data provided by an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of the structure of a multi-source fusion device for satellite surface temperature data provided by an embodiment of the present invention;

[0042] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described in combination with the embodiments below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] At present, the existing technology usually fuses the thermal infrared channels first and then performs surface temperature inversion. The fused channel data has a certain degree of distortion, which has a great impact on the subsequent surface temperature inversion. Based on this, the present invention implements a multi-source fusion method, device, equipment and medium for satellite surface temperature data, which can significantly improve the quality of surface temperature.

[0045] To facilitate understanding of this embodiment, a multi-source fusion method for satellite surface temperature data disclosed in an embodiment of the present invention is first described in detail. Figure 1 The flowchart of a multi-source fusion method of satellite surface temperature data is shown in FIG. 1 , and the method mainly includes the following steps S102 to S106:

[0046] Step S102, obtaining a first surface temperature and a second surface temperature of the study area at an initial time, and a first surface temperature of the study area at a target time.

[0047] Among them, the spatial resolution of the first surface temperature is lower than the spatial resolution of the second surface temperature. In one example, the first surface temperature is the FY3D surface temperature, and the second surface temperature is the Landsat-8 surface temperature. The FY3D surface temperature can be the surface temperature inverted based on the remote sensing image collected by the FY3D satellite, and the Landsat-8 surface temperature can be the surface temperature inverted based on the remote sensing image collected by the Landsat-8 satellite.

[0048] Step S104, using a pre-constructed target downscaling model, based on the first surface temperature and the second surface temperature at the initial moment, and the first surface temperature at the target moment, generates a downscaling result of the first surface temperature of the study area at the target moment.

[0049] Among them, the target downscaling model is constructed based on the principle that the ratio of the difference between the second surface temperature at the initial moment and the target moment and the difference between the first surface temperature at the initial moment and the target moment is fixed in the same land use and cover type. In one example, the conversion coefficient and weight coefficient of the target downscaling model can be pre-fitted, and then the first surface temperature at the initial moment, the second surface temperature at the initial moment, and the first surface temperature at the target moment are input into the target downscaling model to obtain the downscaling result of the first surface temperature at the target moment. The conversion coefficient is the ratio between the difference between the second surface temperature at two moments and the difference between the first surface temperature data at two moments; the weight coefficient is related to the pixel similarity of each pixel point in the moving window and its distance from the middle pixel point.

[0050] Step S106, based on the second surface temperature of the study area at the target time, the first surface temperature downscaling result is corrected for deviation, so as to obtain the target surface temperature of the study area at the target time.

[0051] In one example, a pre-trained random forest machine learning model can be used to generate a target surface temperature at a target time based on the second surface temperature at the target time, the downscaling results of the first surface temperature, and combined with land use data, meteorological station observation data (precipitation, wind speed, wind direction), vegetation index data, longitude, latitude, surface elevation, observation angle, slope, slope direction and other data.

[0052] The multi-source fusion method of satellite surface temperature data provided by the embodiment of the present invention generates a first surface temperature downscaling result based on the first surface temperature and the second surface temperature of the study area at the initial time and the first surface temperature at the target time through a target downscaling model, and uses the second surface temperature of the study area at the target time to correct the deviation of the first surface temperature downscaling result, thereby realizing the fusion of multi-source surface temperatures. The embodiment of the present invention performs fusion processing on the basis of surface temperature inversion to ensure the quality of single-source surface temperature, thereby providing a good basis for data fusion. The fusion result can provide support for global or regional meteorological climate, urban heat island effect and other research.

[0053] In one embodiment, the overall idea of ​​the embodiment of the present invention is: collect FY3D surface temperature and Landsat-8 surface temperature data in the overlapping period as sample data sets, the sample data sets should contain as many types of data as possible, such as time, land use type, vegetation index, satellite observation angle, weather conditions, etc., construct the conversion coefficient and weight coefficient of FY3D surface temperature under various land use types based on the sample data sets, calculate and obtain FY3D high spatial resolution surface temperature data (that is, the first surface temperature downscaling result), and then use the current Landsat-8 surface temperature as a reference value to analyze the deviation of FY3D high spatial resolution surface temperature data in the overlapping area, correct the deviation of FY3D high spatial resolution surface temperature data, and obtain the fused surface temperature data (that is, the target surface temperature). See Figure 2 The overall process diagram of a multi-source fusion method of satellite surface temperature data shown in the figure includes: performing data analysis and processing on the Landsat-8 surface temperature at time t1 and time t2, and performing data analysis and processing on the FY3D surface temperature at time t1 and time t2, where time t1 and time t2 are the first historical moment and the second historical moment respectively; calculating conversion coefficients and weight coefficients in combination with land use data and vegetation index data; performing high temporal and spatial resolution surface temperature calculation in combination with FY3D surface temperature and vegetation index to output FY3D high spatial resolution surface temperature data; performing Landsat surface temperature and FY3D high spatial resolution surface temperature deviation analysis in combination with Landsat-8 surface temperature, realizing FY3D high spatial resolution surface temperature deviation correction, and obtaining fused surface temperature data.

[0054] For ease of understanding, an embodiment of the present invention provides a specific implementation of a multi-source fusion method for satellite surface temperature data.

[0055] Before executing step S104, the conversion coefficient and weight coefficient of the target downscaling model need to be determined in advance.

[0056] For satellite surface temperature, Landsat-8 and FY3D satellites have similar patterns. p The difference between the Landsat-8 surface temperature at the initial time t0 and the target time t p The ratio of the difference between the FY3D surface temperature at the initial time t0 and the FY3D surface temperature at the same land use type is fixed. p The FY3D surface temperature at the initial time t0 and the Landsat-8 surface temperature at the initial time t0 can be obtained. p The FY3D high spatial resolution surface temperature data, also known as the FY3D first surface temperature downscaling result T F→L (x,y,t p ), which can be used as the research area at the target time t p The predicted Landsat-8 surface temperature under. In the embodiment of the present invention, the size of the moving window is selected as 5 (the size of the moving window can be determined according to the actual situation), and the land use coverage type is the surface coverage type data of the corresponding study area. Based on this, the expression of the target downscaling model is as follows:

[0057]

[0058] Among them, T F→L (x,y,t o ) is the first surface temperature downscaling result of FY3D, T L is the Landsat-8 surface temperature, T F is the FY3D surface temperature, h k is the conversion coefficient between the two satellite surface temperature data for each land use cover type; W i The target weight coefficient for each pixel point similar to the central pixel point in the calculated moving window.

[0059] In one example, the conversion coefficient h can be calculated according to the following steps 1.1 to 1.2: k :

[0060] Step 1.1, obtaining the first surface temperature and the second surface temperature of the study area at the first historical moment, and the first surface temperature and the second surface temperature of the study area at the second historical moment;

[0061] Step 1.2, for any pixel point corresponding to a land use coverage type in the study area, determine the difference between the first surface temperature of the pixel point at the first historical moment and its first surface temperature at the second historical moment, as well as the difference between the second surface temperature of the pixel point at the first historical moment and its second surface temperature at the second historical moment; determine the conversion coefficient corresponding to the land use coverage type based on the ratio of the two differences.

[0062] Specifically, the conversion factor h k is the Landsat-8 surface temperature at the target time t p The difference between the initial time t0 and the FY3D surface temperature at the target time t p The ratio of the difference between the initial time t0 and the conversion coefficient h k The calculation formula is as follows:

[0063]

[0064] Among them, T KL (t2) is the surface temperature of the pixel point of the Kth land use and cover type of Landsat-8 at the second historical moment t2, T KL (t1) is the surface temperature of the pixel point of the Kth land use and cover type of Landsat-8 at the first historical moment t1. Similarly, T F (t2) is the surface temperature of FY3D at the same location at the second historical moment t2, T F (t1) is the surface temperature of FY3D at the same location at the first historical moment t1.

[0065] In one example, the target weight coefficient W i The calculation of is determined by two parts: pixel similarity and the distance between the center pixel point and other pixel points in the moving window. The weight coefficient W can be calculated according to the following steps 2.1 to 2.2 i :

[0066] Step 2.1, for any pixel point in the current moving window, determine the surface temperature similarity and vegetation index similarity corresponding to the pixel point, and perform weighted summation of the surface temperature similarity and vegetation index similarity to obtain the pixel similarity corresponding to the pixel point; and determine the distance between the pixel point and the center pixel point in the moving window.

[0067] The pixel similarity calculation is expressed by the ratio of the surface temperature and the vegetation index between Landsat-8 and FY3D.

[0068] Specifically, the process of determining the surface temperature similarity is as follows: if the first surface temperature corresponding to the pixel point is greater than the second surface temperature corresponding to the pixel point, the ratio of the second surface temperature to the first surface temperature is used as the surface temperature similarity corresponding to the pixel point; or, if the first surface temperature corresponding to the pixel point is less than the second surface temperature corresponding to the pixel point, the ratio of the first surface temperature to the second surface temperature is used as the surface temperature similarity corresponding to the pixel point. The calculation formula for surface temperature similarity is as follows:

[0069]

[0070] Among them, R i_T For the surface temperature similarity, T(L i ) is the surface temperature of Landsat-8, T(F i ) is the surface temperature of FY3D.

[0071] Specifically, the process of determining vegetation similarity is as follows: if the first vegetation index corresponding to the pixel point is greater than the second vegetation index corresponding to the pixel point, the ratio of the second vegetation index to the first vegetation index is used as the vegetation index similarity corresponding to the pixel point; or, if the first vegetation index corresponding to the pixel point is less than the second vegetation index corresponding to the pixel point, the ratio of the first vegetation index to the second vegetation index is used as the vegetation index similarity corresponding to the pixel point; wherein, the spatial resolution of the first vegetation index is lower than the spatial resolution of the second vegetation index, the first vegetation index is also the vegetation index of FY3D, and the second vegetation index is also the vegetation index of Landsat-8. The calculation formula for surface temperature similarity is as follows:

[0072]

[0073] Among them, R i_NDVl To compare vegetation index similarity, NDVI (L i ) is the vegetation index of Landsat-8, NDVI (F i ) is the vegetation index of FY3D.

[0074] Specifically, the calculation formula of pixel similarity is as follows:

[0075] R i =a1*R i_T +a2*R i_NDVI ;

[0076] Among them, R i is the calculated pixel similarity, a1 is the surface temperature similarity weight, a2 is the vegetation index similarity weight, R i_T Based on the surface temperature similarity, R i_NDVIThe vegetation index is similarity.

[0077] Specifically, the process of determining the distance between the pixel point and the center pixel point in the moving window is as follows: The distance between the center pixel in the moving window and the pixels similar to it is expressed as Euclidean distance, which is shown as follows:

[0078]

[0079] Among them, d i is the calculated Euclidean distance of the pixel point in the moving window, x w / 2 ,y w / 2 is the coordinate of the center pixel of the moving window, x i ,y i is the coordinate of any pixel point in the moving window, and w is the size of the moving window.

[0080] Step 2.2, based on the pixel similarity and distance corresponding to the pixel point, determine the target weight coefficient corresponding to the pixel point.

[0081] In one example, based on the pixel similarity and distance corresponding to the pixel point, the initial weight coefficient corresponding to the pixel point is determined, and the initial weight coefficient is negatively correlated with the pixel similarity and positively correlated with the distance; then the ratio of the reciprocal of the initial weight coefficient corresponding to the pixel point to the sum of the reciprocals of the initial weight coefficients corresponding to each pixel point in the moving window is used as the target weight coefficient corresponding to the pixel point.

[0082] Specifically, the process of determining the target weight coefficient is as follows:

[0083] The initial weight coefficient CW of the combination of pixel similarity and Euclidean distance can be expressed as:

[0084] CW=(1-R i )*d i ;

[0085] The larger the CW is, the smaller the similarity between the two images is. Therefore, the actual weight should be the reciprocal of CW. Considering that the range of the reciprocal may be too large, the reciprocal weight is used to define the target weight coefficient W. i The calculation formula is as follows:

[0086]

[0087] Based on the above target downscaling model, the embodiment of the present invention provides a specific implementation of the above step S104, including:

[0088] (a) Set a moving window for the study area.

[0089] (b) For any pixel point in the current moving window, determine the difference between the first surface temperature of the pixel point at the initial time and its first surface temperature at the target time. That is, determine the Landsat-8 surface temperature corresponding to the pixel point at the target time t p The difference between the initial time t0 and the FY3D surface temperature corresponding to the pixel point at the target time t p and the difference between the initial time t0.

[0090] (c) Using the conversion coefficient corresponding to the land use coverage type to which the pixel point belongs and the target weight coefficient corresponding to the pixel point, the difference corresponding to each pixel point in the moving window is spatially merged.

[0091] (d) The spatial merging result is merged with the second surface temperature of the central pixel point in the moving window at the initial time to obtain the first surface temperature downscaling result of the central pixel point at the target time.

[0092] In one example, the specific process from (b) to (d) is as shown in the expression of the aforementioned target downscaling model, and the embodiment of the present invention will not be further elaborated on this.

[0093] (e) Control the moving window to move until the first surface temperature downscaling result of each central pixel point in the study area at the target time is determined. In one example, (b) to (d) are repeated until the first surface temperature downscaling results of all central pixel points at the target time are obtained.

[0094] Furthermore, an embodiment of the present invention also provides a specific implementation method of the aforementioned step S106, including: using a pre-trained random forest machine learning model as a surface temperature correction model, and inputting the second surface temperature of the study area at the target time, the first surface temperature downscaling result and the surface temperature influencing factor into the surface temperature correction model, so that the surface temperature correction model outputs the target surface temperature of the study area at the target time.

[0095] The field of surface temperature deviation correction has made significant progress in recent years, mainly due to the development of remote sensing technology, the improvement of data acquisition capabilities, and the application of machine learning and artificial intelligence technologies. High-resolution satellites such as Landsat and FY series provide a large amount of surface temperature data, which provide an important basis for surface temperature deviation correction. In this study, based on the FY3D high spatial resolution surface temperature data and Landsat-8 surface temperature data obtained by downscaling, combined with relevant elements of surface temperature (land use type, weather conditions, vegetation index, longitude, latitude, surface elevation, observation angle, slope, and slope aspect), a machine learning method is used to construct a deviation correction model for FY3D high spatial resolution surface temperature and Landsat-8 surface temperature in the overlapping area, and the FY3D high spatial resolution surface temperature of the entire region is corrected to achieve multi-source fusion of surface temperature, providing data support for urban heat island effect, forest fire monitoring, agricultural production monitoring, etc.

[0096] The FY3D surface temperature, Landsat-8 surface temperature, land use data, meteorological station observation data (precipitation, wind speed, wind direction), vegetation index data, longitude, latitude, surface elevation, observation angle, slope, aspect and other data of overlapping years (such as 2020-2023) are used as samples to construct a training data set. The Random Forest (RF) machine learning model is used to establish the quantitative relationship between surface temperature and various factors. During the model training process, the root mean square error RMSE is used as the objective function, and the three key parameters in the RF model, namely n_estimators, max_depth, and max_features, are determined through grid search. The random forest machine learning model is shown below:

[0097] T′ F =

[0098] RF(T L ,T F→L ,LUC,TPW,U,V,NDVI,Lon,Lat,DEM,Angle,Slop,ASP);

[0099] Where T′ F is the revised FY3D high temporal and spatial resolution surface temperature data, that is, the target surface temperature, T L is the surface temperature at the closest time between Landsat-8 and FY3D, T F→LIt is the first surface temperature downscaling result of FY3D, LUC is land use data, TPW is atmospheric precipitation, U is wind direction, V is wind speed, NDVI is vegetation index, Lon is longitude, Lat is latitude, DEM is surface elevation, Angle is observation angle, Slop is slope, and ASP is slope aspect.

[0100] Based on the above steps, the conversion coefficient and weight coefficient of the FY3D and Landsat-8 surface temperatures are first calculated, and the FY3D surface temperature data with high temporal and spatial resolution is obtained through spatial downscaling. Then, combined with the Landsat-8 surface temperature data at the most recent moment, a surface temperature deviation model is constructed through surface temperature deviation analysis, and the deviation of the FY3D surface temperature with high temporal and spatial resolution is corrected. Finally, the multi-source fusion of medium and high-resolution satellite surface temperatures is realized, providing data reference for urban heat island effect, forest fire monitoring, agricultural production monitoring, etc.

[0101] In summary, the method provided by the embodiment of the present invention can realize multi-source fusion of land surface temperature, improve the accuracy, spatiotemporal coverage and application breadth of data, which is mainly reflected in the following aspects:

[0102] (1) Fusion of data from different sensors can combine the advantages of each sensor and reduce the systematic and random errors that may exist in a single data source.

[0103] (2) By utilizing the data details of high-resolution sensors and the coverage of low-resolution sensors, the overall accuracy of the data can be improved while ensuring spatial resolution.

[0104] (3) By integrating multi-source data, the consistency of surface temperature data can be improved and the coordination between different data sources can be ensured.

[0105] Based on the above-mentioned embodiment, the present invention provides a multi-source fusion device for satellite surface temperature data. Figure 3 The structure diagram of a multi-source fusion device for satellite surface temperature data is shown in FIG. 1 , and the device mainly includes the following parts:

[0106] The data acquisition module 302 is used to acquire a first surface temperature and a second surface temperature of the study area at an initial time, and the first surface temperature of the study area at a target time; wherein the spatial resolution of the first surface temperature is lower than the spatial resolution of the second surface temperature;

[0107] The downscaling module 304 is used to generate a downscaling result of the first land surface temperature of the study area at the target time based on the first land surface temperature and the second land surface temperature at the initial time and the first land surface temperature at the target time through a pre-constructed target downscaling model;

[0108] The deviation correction module 306 is used to perform deviation correction on the first surface temperature downscaling result based on the second surface temperature of the study area at the target time to obtain the target surface temperature of the study area at the target time.

[0109] The multi-source fusion device of satellite surface temperature data provided by the embodiment of the present invention generates a first surface temperature downscaling result based on the first surface temperature and the second surface temperature of the study area at the initial time and the first surface temperature at the target time through a target downscaling model, and uses the second surface temperature of the study area at the target time to correct the deviation of the first surface temperature downscaling result, thereby realizing the fusion of multi-source surface temperatures. The embodiment of the present invention performs fusion processing on the basis of surface temperature inversion to ensure the quality of single-source surface temperature, and provides a good basis for data fusion. The fusion result can provide support for global or regional meteorological climate, urban heat island effect and other research.

[0110] In one implementation, the downscaling module 304 is specifically configured to:

[0111] Set a moving window for the study area;

[0112] For any pixel point in the current moving window, determine the difference between the first surface temperature of the pixel point at the initial time and the first surface temperature of the pixel point at the target time;

[0113] The difference values ​​corresponding to each pixel point in the moving window are spatially merged using the conversion coefficient corresponding to the land use coverage type to which the pixel point belongs and the target weight coefficient corresponding to the pixel point;

[0114] The spatial merging result is merged with the second surface temperature of the central pixel point in the moving window at the initial time, and the first surface temperature downscaling result of the central pixel point at the target time is obtained;

[0115] The moving window is controlled to move until the first surface temperature downscaling result of each central pixel point in the study area at the target time is determined.

[0116] In one embodiment, a conversion coefficient determination module is further included, which is used to:

[0117] Obtaining a first surface temperature and a second surface temperature of the study area at a first historical moment, and a first surface temperature and a second surface temperature of the study area at a second historical moment;

[0118] For the pixel points corresponding to any land use coverage type in the study area, determine the difference between the first surface temperature of the pixel point at the first historical moment and its first surface temperature at the second historical moment, as well as the difference between the second surface temperature of the pixel point at the first historical moment and its second surface temperature at the second historical moment; determine the conversion coefficient corresponding to the land use coverage type based on the ratio of the two differences.

[0119] In one embodiment, a weight coefficient determination module is further included, which is used to:

[0120] For any pixel point in the current moving window, determine the surface temperature similarity and vegetation index similarity corresponding to the pixel point, and perform weighted summation of the surface temperature similarity and the vegetation index similarity to obtain the pixel similarity corresponding to the pixel point; and determine the distance between the pixel point and the center pixel point in the moving window;

[0121] Based on the pixel similarity and distance corresponding to the pixel point, the target weight coefficient corresponding to the pixel point is determined.

[0122] In one implementation, the weight coefficient determination module is specifically used to:

[0123] If the first surface temperature corresponding to the pixel point is greater than the second surface temperature corresponding to the pixel point, the ratio of the second surface temperature to the first surface temperature is used as the surface temperature similarity corresponding to the pixel point; or, if the first surface temperature corresponding to the pixel point is less than the second surface temperature corresponding to the pixel point, the ratio of the first surface temperature to the second surface temperature is used as the surface temperature similarity corresponding to the pixel point;

[0124] And, if the first vegetation index corresponding to the pixel point is greater than the second vegetation index corresponding to the pixel point, the ratio of the second vegetation index to the first vegetation index is used as the vegetation index similarity corresponding to the pixel point; or, if the first vegetation index corresponding to the pixel point is less than the second vegetation index corresponding to the pixel point, the ratio of the first vegetation index to the second vegetation index is used as the vegetation index similarity corresponding to the pixel point; wherein the spatial resolution of the first vegetation index is lower than the spatial resolution of the second vegetation index.

[0125] In one implementation, the weight coefficient determination module is specifically used to:

[0126] Based on the pixel similarity and distance corresponding to the pixel point, an initial weight coefficient corresponding to the pixel point is determined, and the initial weight coefficient is negatively correlated with the pixel similarity and positively correlated with the distance;

[0127] The ratio between the reciprocal of the initial weight coefficient corresponding to the pixel point and the sum of the reciprocals of the initial weight coefficients corresponding to each pixel point in the moving window is used as the target weight coefficient corresponding to the pixel point.

[0128] In one implementation, the deviation correction module 306 is specifically configured to:

[0129] The pre-trained random forest machine learning model is used as the surface temperature correction model, and the second surface temperature of the study area at the target time, the first surface temperature downscaling results and the surface temperature influencing factors are input into the surface temperature correction model, so that the surface temperature correction model outputs the target surface temperature of the study area at the target time.

[0130] The device provided in the embodiment of the present invention has the same implementation principle and technical effects as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0131] An embodiment of the present invention provides an electronic device. Specifically, the electronic device includes a processor and a storage device. The storage device stores a computer program, and when the computer program is executed by the processor, it executes the method described in any one of the above-mentioned implementation methods.

[0132] Figure 4 A structural diagram of an electronic device provided in an embodiment of the present invention, the electronic device 100 includes: a processor 40, a memory 41, a bus 42 and a communication interface 43, wherein the processor 40, the communication interface 43 and the memory 41 are connected via the bus 42; the processor 40 is used to execute an executable module stored in the memory 41, such as a computer program.

[0133] The memory 41 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 43 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0134] The bus 42 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0135] Among them, the memory 41 is used to store programs, and the processor 40 executes the program after receiving the execution instruction. The method executed by the device for flow process definition disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 40 or implemented by the processor 40.

[0136] The processor 40 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 40. The above processor 40 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor to execute, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 41, and the processor 40 reads the information in the memory 41 and completes the steps of the above method in combination with its hardware.

[0137] The computer program product of the readable storage medium provided in the embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be referred to the previous method embodiments, which will not be repeated here.

[0138] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0139] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A multi-source fusion method for satellite surface temperature data, characterized in that: include: Acquire a first surface temperature and a second surface temperature of a study area at an initial time, and the first surface temperature of the study area at a target time; wherein a spatial resolution of the first surface temperature is lower than a spatial resolution of the second surface temperature; Generate a downscaling result of the first land surface temperature of the study area at the target time based on the first land surface temperature and the second land surface temperature at the initial time and the first land surface temperature at the target time by using a pre-constructed target downscaling model; Based on the second surface temperature of the study area at the target time, the first surface temperature downscaling result is bias-corrected to obtain the target surface temperature of the study area at the target time.

2. The multi-source fusion method of satellite surface temperature data according to claim 1 is characterized in that: The method generates a downscaling result of the first land surface temperature of the study area at the target time based on the first land surface temperature and the second land surface temperature at the initial time and the first land surface temperature at the target time by using a pre-constructed target downscaling model, including: Setting a moving window for the research area; For any pixel point in the current moving window, determine the difference between the first ground surface temperature of the pixel point at the initial time and the first ground surface temperature of the pixel point at the target time; Using the conversion coefficient corresponding to the land use coverage type to which the pixel point belongs and the target weight coefficient corresponding to the pixel point, spatially merge the difference corresponding to each of the pixel points in the moving window; The spatial merging result is merged with the second surface temperature of the central pixel point in the moving window at the initial time, so as to obtain the downscaling result of the first surface temperature of the central pixel point at the target time; The moving window is controlled to move until the first surface temperature downscaling result of each central pixel point in the study area at the target time is determined.

3. The multi-source fusion method of satellite surface temperature data according to claim 2 is characterized in that: Before generating a downscaling result of the first land surface temperature of the study area at the target time based on the first land surface temperature and the second land surface temperature at the initial time and the first land surface temperature at the target time by using a pre-constructed target downscaling model, the method further includes: Acquire the first surface temperature and the second surface temperature of the study area at a first historical moment, and the first surface temperature and the second surface temperature of the study area at a second historical moment; For a pixel point corresponding to any land use coverage type in the study area, determine the difference between the first surface temperature of the pixel point at the first historical moment and the first surface temperature at the second historical moment, as well as the difference between the second surface temperature of the pixel point at the first historical moment and the second surface temperature at the second historical moment; and determine the conversion coefficient corresponding to the land use coverage type based on the ratio of the two differences.

4. The multi-source fusion method of satellite surface temperature data according to claim 2 is characterized in that: Before generating a downscaling result of the first land surface temperature of the study area at the target time based on the first land surface temperature and the second land surface temperature at the initial time and the first land surface temperature at the target time by using a pre-constructed target downscaling model, the method further includes: For any pixel point in the current moving window, determine the surface temperature similarity and vegetation index similarity corresponding to the pixel point, and perform weighted summation of the surface temperature similarity and the vegetation index similarity to obtain the pixel similarity corresponding to the pixel point; and determine the distance between the pixel point and the central pixel point in the moving window; Based on the pixel similarity and the distance corresponding to the pixel point, a target weight coefficient corresponding to the pixel point is determined.

5. The multi-source fusion method of satellite surface temperature data according to claim 4 is characterized in that: Determine the surface temperature similarity and vegetation index similarity corresponding to the pixel point, including: If the first ground surface temperature corresponding to the pixel point is greater than the second ground surface temperature corresponding to the pixel point, the ratio of the second ground surface temperature to the first ground surface temperature is used as the ground surface temperature similarity corresponding to the pixel point; or, if the first ground surface temperature corresponding to the pixel point is less than the second ground surface temperature corresponding to the pixel point, the ratio of the first ground surface temperature to the second ground surface temperature is used as the ground surface temperature similarity corresponding to the pixel point; And, if the first vegetation index corresponding to the pixel point is greater than the second vegetation index corresponding to the pixel point, the ratio of the second vegetation index to the first vegetation index is used as the vegetation index similarity corresponding to the pixel point; or, if the first vegetation index corresponding to the pixel point is less than the second vegetation index corresponding to the pixel point, the ratio of the first vegetation index to the second vegetation index is used as the vegetation index similarity corresponding to the pixel point; wherein the spatial resolution of the first vegetation index is lower than the spatial resolution of the second vegetation index.

6. The multi-source fusion method of satellite surface temperature data according to claim 4, characterized in that: Determining a target weight coefficient corresponding to the pixel point based on the pixel similarity and the distance corresponding to the pixel point includes: Based on the pixel similarity and the distance corresponding to the pixel point, determining an initial weight coefficient corresponding to the pixel point, wherein the initial weight coefficient is negatively correlated with the pixel similarity and positively correlated with the distance; The ratio between the reciprocal of the initial weight coefficient corresponding to the pixel point and the sum of the reciprocals of the initial weight coefficients corresponding to each pixel point in the moving window is used as the target weight coefficient corresponding to the pixel point.

7. The multi-source fusion method of satellite surface temperature data according to claim 1, characterized in that: Based on the second surface temperature of the study area at the target time, the first surface temperature downscaling result is corrected for deviation to obtain the target surface temperature of the study area at the target time, including: A pre-trained random forest machine learning model is used as a surface temperature correction model, and the second surface temperature of the study area at the target time, the first surface temperature downscaling result and the surface temperature influencing factor are input into the surface temperature correction model, so that the surface temperature correction model outputs the target surface temperature of the study area at the target time.

8. A multi-source fusion device for satellite surface temperature data, characterized in that: include: A data acquisition module, used to acquire a first surface temperature and a second surface temperature of a study area at an initial time, and the first surface temperature of the study area at a target time; wherein the spatial resolution of the first surface temperature is lower than the spatial resolution of the second surface temperature; A downscaling module, configured to generate a downscaling result of the first land surface temperature of the study area at the target time based on the first land surface temperature and the second land surface temperature at the initial time and the first land surface temperature at the target time through a pre-constructed target downscaling model; The deviation correction module is used to perform deviation correction on the first surface temperature downscaling result based on the second surface temperature of the study area at the target time, so as to obtain the target surface temperature of the study area at the target time.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the method according to any one of claims 1 to 7.

Images