A self-supervised method and system for downscaling remote sensing temperature products

Abstract

The present application relates to the technical field of design optimization, and provides a self-supervised downscaling method and system for remote sensing temperature products. The downscaling multiple is obtained based on the spatial resolution of the remote sensing temperature product data RS1 and the target resolution. The upscaled remote sensing raster data RS0 is obtained through upscaling. Then, the anomaly value raster data of RS1 and RS0 are calculated respectively. The original temperature data and the anomaly value raster data are used to express different data laws, and the extraction of self-supervised knowledge is carried out. According to the extracted self-supervised knowledge, the spatial resolution of the remote sensing temperature product data RS1 is downscaled to the target resolution. Finally, the high-precision surface modeling method is used to obtain the downscaled raster data RS 1α of the temperature simulation value. This method does not require data from other sources. Based on the characteristics of the remote sensing temperature product data itself, it fully explores the laws of the remote sensing data itself, and applies the extracted self-supervised knowledge to the downscaling process to achieve the conversion of the spatial resolution from low to high.

Description

Technical Field

[0001] The present application relates to the technical field of design optimization, and particularly relates to a self-supervised downscaling method, system, computer-readable storage medium, and electronic device for remote sensing temperature products. Background Art

[0002] Against the backdrop of global warming, extreme climate events occur frequently. For example, the frequent occurrence of disasters such as high-temperature heatwaves and floods has had a great impact on the economic development and physical health of human society. Especially in urban areas, high-temperature heatwaves have exacerbated the urban heat island, and the "heat"-related impacts have been further enhanced. Comprehensively promoting the construction of resilient and safe cities, and high-temperature resilience is an important part of the construction of resilient cities. If effective measures are taken to enhance the high-temperature resilience of cities, it is an important challenge currently faced. And an important prerequisite for solving this challenge is to have available high-precision and high-resolution environmental temperature-related data.

[0003] In order to better evaluate the high-temperature resilience of cities, generally, the land surface temperature products retrieved by satellite remote sensing are used for related research. However, the underlying surface in urban areas changes complexly, and the variability of the surface thermal environment is also very large. The land surface temperature data retrieved by satellite remote sensing usually has a relatively coarse spatial resolution and cannot meet the needs of urban high-temperature resilience research. Therefore, in order to obtain more scientific laws, it is often necessary to downscale the remote sensing products. However, current downscaling methods almost all establish corresponding statistical models using the regression relationship between certain explanatory factors and variables to improve the spatial resolution of remote sensing temperature products. However, the primary condition for such methods is to have corresponding observational data to establish an empirical model, that is, exogenous data must be introduced and cannot be used without the introduction of observational data, and the laws of the data itself are not fully explored.

[0004] Therefore, an improved technical solution is needed to address the above deficiencies in the prior art. Summary of the Invention

[0005] The purpose of the present application is to provide a self-supervised downscaling method, system, computer-readable storage medium, and electronic device for remote sensing temperature products. Based on the characteristics of remote sensing data itself without other exogenous data, this solution fully explores the laws of remote sensing data itself, combines with the High Accuracy Surface Modeling (HASM) technology to achieve self-supervised spatial downscaling of remote sensing temperature products, and finally obtains temperature data with high spatial resolution and high precision, which solves or alleviates to a certain extent the problems existing in the above prior art.

[0006] To achieve the above purpose, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a self-supervised downscaling method for remote sensing temperature products, including:

[0008] According to the spatial resolution of the remote sensing temperature product data RS1 and the target resolution to be achieved by downscaling, calculate the downscaling factor d;

[0009] Perform upscaling on the remote sensing temperature product data RS1 according to the downscaling factor d to obtain the upscaled remote sensing raster data RS0;

[0010] Subtract the mean value of each pixel in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 from their respective pixel values to obtain the first anomaly value raster data δ1 and the second anomaly value raster data δ0 corresponding to the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively;

[0011] Calculate the scale operator α using the spatial position mapping relationship and pixel values between the pixels in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0; calculate the scale operator β using the spatial position mapping relationship and pixel values between the pixels in the first anomaly value raster data δ1 and the second anomaly value raster data δ0;

[0012] Traverse all pixels of the remote sensing temperature product data RS1, and downscale each pixel in the remote sensing temperature product data RS1 to the target resolution according to the scale operator α to obtain the downscaled raster data RS 1α ; Traverse all pixels of the first anomaly value raster data δ1, and downscale each pixel in the first anomaly value raster data δ1 to the target resolution according to the scale operator β to obtain the downscaled anomaly value raster data δ 1β ;

[0013] Use the downscaled raster data RS 1α as the initial field of the high-accuracy surface modeling HASM method, and use the sum of the downscaled raster data RS 1α and the downscaled anomaly value raster data δ 1β as the optimization control condition of HASM, and perform HASM simulation to obtain the temperature simulation values of all pixels of the downscaled raster data RS 1α

[0014] Preferably, it further includes: constructing an outer constraint condition of HASM, where the outer constraint condition is used to constrain the external accuracy of each iteration in the HASM solution process, and the optimization control condition of HASM and the outer constraint condition of HASM constitute a pseudo-HASM with double constraints.

[0015] Preferably, constructing the outer constraint condition of HASM includes: ​

[0016] Calculate the quotient of the mean m0 of the upscaled remote sensing raster data RS0 and the mean m1 of the remote sensing temperature product data RS1 to obtain the first scale coefficient m; and use the result of dividing the mean m1 of the remote sensing temperature product data RS1 by the first scale coefficient m as the first outer constraint;

[0017] Correspondingly, expand the HASM simulation, including:

[0018] During the HASM solving process, calculate the mean m2 of the current simulation result for each iteration of the simulation result;

[0019] If the error between the mean m2 of the current simulation result and the first outer constraint is less than or equal to a preset first external error threshold, and the current simulation result reaches a preset HASM internal stop condition, then stop the iteration and use the current simulation result as the final temperature simulation result.

[0020] Preferably, construct the outer constraint conditions of HASM, including:

[0021] Calculate the quotient of the standard deviation s0 of the upscaled remote sensing raster data RS0 and the standard deviation s1 of the remote sensing temperature product data RS1 to obtain the second scale coefficient s; and use the result of dividing the standard deviation s1 of the remote sensing temperature product data RS1 by the second scale coefficient s as the second outer constraint;

[0022] Correspondingly, expand the HASM simulation, including:

[0023] During the HASM solving process, calculate the standard deviation s2 of the current simulation result for each iteration of the simulation result;

[0024] If the error between the standard deviation s2 of the current simulation result and the second outer constraint is less than or equal to a preset second external error threshold, and the current simulation result reaches a preset HASM internal stop condition, then stop the iteration and use the current simulation result as the final temperature simulation result.

[0025] Preferably, construct the outer constraint conditions of HASM, including:

[0026] Calculate the quotient of the mean m0 of the upscaled remote sensing raster data RS0 and the mean m1 of the remote sensing temperature product data RS1 to obtain the first scale coefficient m;

[0027] Calculate the quotient of the standard deviation s0 of the upscaled remote sensing raster data RS0 and the standard deviation s1 of the remote sensing temperature product data RS1 to obtain the second scale coefficient s;

[0028] Taking the result of dividing the mean value m1 of the remote sensing temperature product data RS1 by the first scale factor m as the first outer constraint, and taking the result of dividing the standard deviation s1 of the remote sensing temperature product data RS1 by the second scale factor s as the second outer constraint, and jointly constituting a two - conditional outer constraint by the first outer constraint and the second outer constraint;

[0029] Correspondingly, expanding the HASM simulation includes:

[0030] During the HASM solving process, for the simulation result of each iteration, calculate the mean value m2 and the standard deviation s2 of the current simulation result; if the error between the mean value m2 of the current simulation result and the first outer constraint is less than or equal to a preset first external error threshold, and the error between the standard deviation s2 of the current simulation result and the second outer constraint is less than or equal to a preset second external error threshold, and the current simulation result reaches a preset HASM internal stop condition, then stop the iteration and take the current simulation result as the final temperature simulation result.

[0031] Preferably, after upscaling the remote sensing temperature product data RS1 according to the downscaling multiple d to obtain the upscaled remote sensing raster data RS0, it further includes:

[0032] Performing outlier processing on the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively.

[0033] Preferably, the outlier processing includes the following steps:

[0034] Calculate the third quartile Q3 and the first quartile Q1 of the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively;

[0035] Calculate the difference between the third quartile and the first quartile to obtain the inter - quartile range IQR;

[0036] Traverse all the pixels in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively. If the value of the currently traversed pixel is greater than the sum of the third quartile Q3 and 1.5 times the inter - quartile range IQR, then determine that the value of this pixel is an outlier, and re - assign the value of the currently traversed pixel to the sum of the third quartile Q3 and 1.5 times the inter - quartile range IQR;

[0037] Or, if the value of the currently traversed pixel is less than the difference between the first quartile Q1 and 1.5 times the inter - quartile range IQR, then determine that the value of this pixel is an outlier, and re - assign the value of the currently traversed pixel to the difference between the first quartile Q1 and 1.5 times the inter - quartile range IQR.

[0038] In a second aspect, this embodiment provides a self-supervised remote sensing temperature product downscaling system, including:

[0039] A first calculation unit, configured to calculate a downscaling multiple d according to the spatial resolution of the remote sensing temperature product data RS1 and the target resolution required for downscaling;

[0040] An upscaling unit, configured to perform upscaling processing on the remote sensing temperature product data RS1 according to the downscaling multiple d to obtain upscaled remote sensing raster data RS0;

[0041] A second calculation unit, configured to subtract the mean value of each pixel value in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 from the pixel values in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively, to obtain the first anomaly value raster data δ1 and the second anomaly value raster data δ0 corresponding to the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively;

[0042] A scale operator calculation unit, configured to calculate a scale operator α by using the spatial position mapping relationship and pixel values between the pixels in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0; calculate a scale operator β by using the spatial position mapping relationship and pixel values between the pixels in the first anomaly value raster data δ1 and the second anomaly value raster data δ0;

[0043] A traversal unit, configured to traverse all pixels of the remote sensing temperature product data RS1, and downscale each pixel in the remote sensing temperature product data RS1 to the target resolution according to the scale operator α to obtain downscaled raster data RS 1α ; traverse all pixels of the first anomaly value raster data δ1, and downscale each pixel in the first anomaly value raster data δ1 to the target resolution according to the scale operator β to obtain downscaled anomaly value raster data δ 1β ;

[0044] A simulation unit, configured to use the downscaled raster data RS 1α as the initial field of the High Accuracy Surface Modeling (HASM) method, and use the sum of the downscaled raster data RS 1α and the downscaled anomaly value raster data δ 1β as the optimization control condition of HASM to perform HASM simulation to obtain the temperature simulation values of all pixels of the downscaled raster data RS 1α

[0045] ​In a third aspect, this embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the self-supervised remote sensing temperature product downscaling method described in any one of the above embodiments.

[0046] In a fourth aspect, this embodiment provides an electronic device, including: a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the self-supervised remote sensing temperature product downscaling method described in any one of the above embodiments.

[0047] The technical solution of the embodiment of this application has the following beneficial effects:

[0048] Using the spatial resolution of the remote sensing temperature product data RS1 and the target resolution, by upscaling and then calculating the anomaly value raster data of the original remote sensing temperature product data RS1 and the upscaled remote sensing raster data respectively to represent different data laws, and then extracting self-supervised knowledge of the remote sensing data itself by calculating the scale operator. On this basis, according to the extracted self-supervised knowledge, that is, the scale operator α and the scale operator β, perform scale operations to downscale the spatial resolution of the remote sensing temperature product data RS1 to the target resolution, and at the same time use the high-precision simulation advantage of the high-precision surface modeling HASM method to obtain the downscaled raster data RS 1α The temperature simulation values of all pixels. This method does not require data from other sources, only needs to collect the remote sensing temperature product data RS1, first performs upscaling for self-supervised knowledge extraction, and then applies the extracted self-supervised knowledge to the downscaling process to achieve the conversion of spatial resolution from low to high, and combines the HASM method to simulate the temperature values of each pixel at high resolution to meet the needs of urban high-temperature resilience research. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The specification drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. Among them:

[0050] Figure 1 is a flowchart of a self-supervised remote sensing temperature product downscaling method provided according to some embodiments of this application Figure 1 .

[0051] Figure 2 is a flowchart of a self-supervised remote sensing temperature product downscaling method provided according to some embodiments of this application Figure 2 .

[0052] Figure 3 is a flowchart of a self-supervised module provided according to some embodiments of this application.

[0053] Figure 4 A schematic flow chart of outlier processing provided according to some embodiments of the present application.

[0054] Figure 5 A schematic flow chart of calculating the scaling operator α provided according to some embodiments of the present application.

[0055] Figure 6 A schematic flow chart of HASM simulation provided according to some embodiments of the present application.

[0056] Figure 7 A schematic diagram of the process of downscaling using the scaling operator α provided according to some embodiments of the present application.

[0057] Figure 8 A schematic structural diagram of a self-supervised remote sensing temperature product downscaling system provided according to some embodiments of the present application.

[0058] Figure 9 A schematic structural diagram of an electronic device provided according to an embodiment of the present application.

[0059] Figure 10 A hardware structure diagram of an electronic device provided according to an embodiment of the present application. Detailed implementation manners

[0060] Exemplary embodiments of the present application include, but are not limited to, a self-supervised remote sensing temperature product downscaling method, system, computer-readable storage medium, and electronic device.

[0061] To more clearly understand the solutions in the embodiments of the present application, some related technologies involved in the embodiments of the present application are first explained below.

[0062] The research and evaluation of urban high-temperature resilience rely on temperature data support. Currently, there are mainly three methods to obtain surface temperature data:

[0063] (1) Meteorological station observations

[0064] Meteorological station observations are the primary way to obtain temperature data currently. Its advantages are high observation accuracy and the ability to obtain time-continuous observation data; the disadvantage is that the meteorological station observation method belongs to sparse observation, which is often discontinuous in space and cannot meet the needs of agriculture, hydrology, and other fields for spatially continuous data.

[0065] (2) Climate model simulation

[0066] Climate models are multidisciplinary "scientific crystallization" formed based on the operation and changes of the atmosphere and climate system, making full use of the development of computer technology on the basis of basic physical laws (such as Newton's laws of motion, laws of conservation of energy and mass). Climate models mainly include global climate models and regional climate models. Global Climate Models (GCMs) can simulate and predict the climate on a global scale. However, due to the complexity of the climate system itself and the differences in comprehensive conditions such as latitude, altitude, sea-land location, and underlying surface in different regions, the climate system exhibits different variation characteristics and intensities in different regions, making GCMs often have poor simulation effects on a regional scale and are difficult to represent local weather processes, especially extreme weather and climate events.

[0067] (3) Satellite remote sensing inversion

[0068] The development of space exploration technology has greatly increased the means for people to obtain surface data. Meteorological satellites can conduct meteorological observations of the Earth from space to obtain temperature data. Its advantage is that relatively accurate and spatially continuous temperature data can be obtained using meteorological satellites; the disadvantage is that it is vulnerable to weather effects, and although the spatial resolution has been improved compared to climate model simulations, it still does not meet the requirements for fine expression in urban areas.

[0069] In view of the discontinuity of meteorological station observation data and the limitations of climate model simulation data, currently the industry generally uses satellite remote sensing inversion surface temperature products (remote sensing temperature product data) to conduct research on urban heat resilience. However, the spatial resolution of remote sensing temperature product data is usually low, making it difficult to reflect the complex situation of the underlying surface in urban areas and still difficult to meet the requirements for urban heat resilience research. Therefore, the embodiments of this application provide a self-supervised method, system, computer-readable storage medium, and electronic device for downscaling remote sensing temperature products, which can perform self-supervised knowledge extraction by combining the characteristics of the data itself without introducing additional data, and achieve downscaling of remote sensing temperature product data.

[0070] The following will detail this application with reference to the accompanying drawings and in combination with embodiments. Each example is provided by way of explanation of this application rather than limiting this application. In fact, those skilled in the art will clearly understand that modifications and variations can be made to this application without departing from the scope or spirit of this application. For example, features shown or described as part of one embodiment can be used in another embodiment to yield yet another embodiment. Therefore, it is desirable that this application includes such modifications and variations that fall within the scope of the appended claims and their equivalents.

[0071] In the following description, the terms "first / second / third" involved only distinguish similar objects and do not represent a specific order of the objects. Understandably, "first / second / third" can be interchanged with a specific order or sequence when allowed, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0072] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this disclosure belongs. The terms used herein are only for the purpose of describing the embodiments of this disclosure and are not intended to limit this disclosure.

[0073] Method Embodiment

[0074] This embodiment provides a self-supervised remote sensing temperature product downscaling method. The execution subject of this method can be an electronic device such as a personal computer, a tablet, or a smart phone. As Figure 1 shown, this method includes:

[0075] Step S101: Calculate the downscaling factor d according to the spatial resolution of the remote sensing temperature product data RS1 and the target resolution to be achieved by downscaling.

[0076] It should be noted that downscaling is a technical method aimed at converting the low-resolution remote sensing temperature product data RS1 into high-resolution data applicable to urban heat resilience research.

[0077] In this embodiment, the remote sensing temperature product data RS1 (also referred to as the remote sensing temperature product) is the remote sensing temperature product of the target period in the study area. For example, it can be the MOSID LST product MOD11A1 provided by the MODIS satellite, or the surface parameters retrieved from the thermal infrared imaging data provided by the HotSat-1 satellite.

[0078] For the sake of convenience of description, the MOSID LST product MOD11A1 is taken as an example for the explanation of the technical solution hereinafter, and its spatial resolution is 1 km.

[0079] The target resolution to be achieved by downscaling can be determined according to the scope of the study area, research requirements, and surface characteristics. In urban heat resilience research, the target resolution can be as high as 100 meters, 50 meters, or even 1 meter, for example.

[0080] The purpose of step S101 is to determine the downscaling factor d, and its calculation formula is as follows:

[0081] d = r1 / r2 (1)

[0082] In the formula, r1 represents the spatial resolution of the original remote sensing product data, that is, the spatial resolution of the remote sensing temperature product data RS1 collected, and r2 represents the target resolution.

[0083] It should be understood that the value range of the downscaling multiple d is a positive integer.

[0084] For example, to downscale the remote sensing temperature product data RS1 (spatial resolution is 1 km) provided by the MODIS satellite to 100 m, then d = 10 (1000 / 100).

[0085] Step S102: Upscale the remote sensing temperature product data RS1 according to the downscaling multiple d to obtain the upscaled remote sensing raster data RS0.

[0086] In step S102, after obtaining the downscaling multiple d, first upscale the remote sensing temperature product data RS1 to obtain the upscaled remote sensing raster data RS0, so as to subsequently use the obtained upscaled remote sensing raster data RS0 for self-supervised knowledge extraction work, and express the law that the characteristics of remote sensing data remain consistent at different scales through self-supervised knowledge.

[0087] Among them, upscaling is a concept corresponding to downscaling, which refers to the process of converting high-spatial-resolution data into lower-spatial-resolution data. That is to say, after upscaling, due to the reduction of spatial resolution, the ground area represented by each pixel in the remote sensing data will be larger than the ground area represented by each pixel in the remote sensing data before upscaling, but the detailed features of the ground that can be expressed will also be greatly reduced.

[0088] In this embodiment, there are various ways to implement upscaling. For example, the spatial resolution of the remote sensing temperature product data RS1 can be upscaled by d times using the bilinear interpolation resampling method to obtain the upscaled remote sensing raster data RS0, or the upscaled remote sensing raster data RS0 can be inferred based on the downscaling multiple d using statistical models and probability distributions, or the upscaled remote sensing raster data RS0 can be generated using machine learning methods. The specific implementation method of upscaling is not limited in this embodiment.

[0089] Step S103: Subtract the mean value of each pixel value in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 from their respective mean values to obtain the first anomaly value raster data δ1 and the second anomaly value raster data δ0 corresponding to the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively.

[0090] The purpose of step S103 is to use the remote sensing temperature product data RS1 and the upscaled remote sensing grid data RS0 to obtain the corresponding anomaly value grid data, so as to provide a data basis for the subsequent extraction of self-supervised knowledge and the construction of constraints of the HASM method. In this step, the mean m1 of the remote sensing temperature product data RS1 and the mean m0 of the upscaled remote sensing grid data RS0 are firstly counted, and then the corresponding mean m1 and m0 are subtracted from the values of each pixel of the remote sensing temperature product data RS1 and the upscaled remote sensing grid data RS0 to obtain the corresponding anomaly value grid data. For the convenience of description, the anomaly value grid data corresponding to the remote sensing temperature product data RS1 is referred to as the first anomaly value grid data δ1, and the anomaly value grid data corresponding to the upscaled remote sensing grid data RS0 is referred to as the second anomaly value grid data δ0.

[0091] Step S104: Calculate the scale operator α using the spatial position mapping relationship and pixel values between the pixels in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0; calculate the scale operator β using the spatial position mapping relationship and pixel values between the pixels in the first anomaly value raster data δ1 and the second anomaly value raster data δ0.

[0092] In this embodiment, the spatial position mapping relationship between the pixels in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 refers to the corresponding relationship between RS1 and RS0 in terms of pixel positions. Figure 5 FIG. 1 is a schematic diagram of a calculation process of a scale operator α provided according to some embodiments of the present application. Figure 5 As shown in the figure, assuming that the downscaling factor d between the raster data RS1 and RS0 is 3 times, then a pixel at any position in the raster data RS0 (for example, the first pixel in the upper left corner) corresponds to the 3×3 pixels in the upper left corner of the raster data RS1, that is, any pixel in the raster data RS0 corresponds to the 9 pixels at the corresponding position of RS1, which is the spatial position mapping relationship between the two. This mapping relationship is determined in the upscaling / downscaling process and is information related to the spatial position. For example, in the upscaling process, the 3×3 pixels in the upper left corner of RS1 can be merged into the first pixel in the upper left corner of RS0, and the corresponding relationship between RS1 before merging and the first pixel in the upper left corner of RS0 after merging is recorded to obtain the spatial position mapping relationship between the two, and the spatial position mapping relationship of other pixels can be deduced by analogy.

[0093] After determining the spatial position mapping relationship between the pixels in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0, the scaling operator α can be obtained by dividing the pixel value in RS0 by the pixel value at the corresponding position in RS1. Figure 5Among them, the pixel value at the upper left corner of the raster data RS0 is 14, and the 3×3 pixel values at the corresponding position in RS1 are 14, 18, 22, 2, 5, 3, 19, 8, 32 in sequence. Then, the scaling operator α is exemplified as follows:

[0094]

[0095] It can be seen that the scaling operator α is obtained by dividing the pixel values at the corresponding positions in the remotely sensed temperature product data RS1 by the upscaled remotely sensed raster data RS0.

[0096] Using the same method, the scaling operator β is calculated by utilizing the spatial position mapping relationship and pixel values between the pixels in the first anomaly value raster data δ1 and the second anomaly value raster data δ0.

[0097] Step S105: Traverse all the pixels of the remotely sensed temperature product data RS1, and downscale each pixel in the remotely sensed temperature product data RS1 to the target resolution according to the scaling operator α, obtaining the downscaled raster data RS 1α ; traverse all the pixels of the first anomaly value raster data δ1, and downscale each pixel in the first anomaly value raster data δ1 to the target resolution according to the scaling operator β, obtaining the downscaled anomaly value raster data δ 1β .

[0098] The purpose of step S105 is to apply the scaling operator α and the scaling operator β to perform downscaling processing on the remotely sensed temperature product data RS1 and the first anomaly value raster data δ1 respectively, so as to obtain the raster data with the target spatial resolution required for the research. However, at this time, the pixel values in this raster data are only an initial value after downscaling, with low accuracy, and it is necessary to simulate high-precision temperature values by combining HASM.

[0099] Figure 7 FIG. is a schematic diagram of the process of downscaling by applying the scaling operator α according to some embodiments of the present application. As Figure 7 shown, the application process of the scaling operator α is essentially the reverse process of the calculation of the scaling operator α. In this process, each pixel of the remotely sensed temperature product data RS1 is divided by the scaling operator α corresponding to this pixel, obtaining 3×3 pixels. After traversing all the pixels, the downscaled raster data RS 1α is obtained. For example, the first pixel value at the upper left corner of the remotely sensed temperature product data RS1 is 14. After dividing by the scaling operator α calculated by formula (2), 9 pixels are obtained, and the values are in sequence: 5, 6.4, 7.8, 0.7, 1.7, 23.2, 6.7, 2.9, 11.4. Then, these 9 pixels are the pixels at the corresponding positions after downscaling.

[0100] Similarly, each pixel in the first anomaly value raster data δ1 is downscaled to the target resolution according to the scale operator β, and the downscaled anomaly value raster data δ is obtained. 1β .

[0101] Step S106: Use the downscaled raster data RS 1α as the initial field of the High Accuracy Surface Modeling (HASM) method, and use the sum of the downscaled raster data RS 1α and the downscaled anomaly value raster data δ 1β as the optimization control condition of HASM, and perform HASM simulation to obtain the temperature simulation values of all pixels in the downscaled raster data RS 1α .

[0102] Here, the HASM method is an ecological environment element spatial simulation method originally created by Chinese scholars. This technology has solved the error problem and multi-scale problem that have plagued the surface modeling process for half a century. Currently, HASM has been widely used in ecological environment-related simulations, such as upscaling, downscaling, multi-source data fusion, etc. However, the principle of HASM also determines that HASM must be used with additional information constraints. The method provided in this embodiment, based on the characteristics of remote sensing data, fully explores the laws of remote sensing data itself. Using the downscaled raster data RS 1α as the initial field of the High Accuracy Surface Modeling (HASM) method, and using the sum of the downscaled raster data RS 1α and the downscaled anomaly value raster data δ 1β as the optimization control condition (i.e., information constraint) of HASM, and combining the HASM technology to perform spatial downscaling of remote sensing temperature products in a self-supervised manner.

[0103] In this embodiment, use the downscaled raster data RS 1α as the initial field of the High Accuracy Surface Modeling (HASM) method, and use the sum of the downscaled raster data RS 1α and the downscaled anomaly value raster data δ 1β as the optimization control condition of HASM. Both the initial field and the optimization space condition are constructed based on the inherent characteristics and laws contained in the remote sensing temperature product data, without introducing external data. On this basis, perform HASM simulation, and solve the main equation of the HASM model through continuous iteration. If the simulation result obtained in the current iteration reaches the preset internal stop condition of HASM, stop the iteration and output the current simulation result as the final temperature simulation result.

[0104] In summary, the method provided in this embodiment utilizes the spatial resolution and target resolution of the remote sensing temperature product data RS1. After upscaling, the anomaly value raster data of the original remote sensing temperature product data RS1 and the upscaled remote sensing raster data are calculated respectively to express different data laws. Furthermore, the self-supervised knowledge extraction of the remote sensing data itself is realized by calculating the scale operator. On this basis, according to the extracted self-supervised knowledge, that is, the scale operator α and the scale operator β, the scale operation is performed to downscale the spatial resolution of the remote sensing temperature product data RS1 to the target resolution. At the same time, the high-precision simulation advantage of the High Accuracy Surface Modeling (HASM) method is utilized to obtain the downscaled raster data RS 1α The temperature simulation values of all pixels. This method does not require data from other sources. Only the remote sensing temperature product data RS1 needs to be collected, and by using the rule of scale law invariance, self-supervised knowledge extraction is first performed by upscaling, and then the extracted self-supervised knowledge is applied to the downscaling process to achieve the conversion of spatial resolution from low to high. Combined with the HASM method, the temperature values of each pixel at high resolution are simulated to meet the needs of urban high-temperature resilience research.

[0105] In order to further improve the accuracy of the downscaled remote sensing temperature product, in some embodiments, after upscaling the remote sensing temperature product data RS1 according to the downscaling factor d to obtain the upscaled remote sensing raster data RS0, it further includes: performing outlier processing on the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively.

[0106] In this embodiment, outliers, also known as extreme values, refer to those observation values that significantly deviate from the normal range or pattern. These outliers may be caused by factors such as noise in the data acquisition process, sensor failures, and environmental interference. The existence of outliers reduces the data quality and thus affects the accuracy of the downscaling result.

[0107] In this embodiment, outlier processing includes two steps: outlier identification and outlier handling. Among them, outlier identification can use the Z-Score method or the standard deviation method to identify outliers. For example, the Z-Score method calculates the standard score (Z-score) of each observation value, and considers the observation value with an absolute value greater than 3 as an outlier. The standard deviation method assumes that the data follows a normal distribution, and any observation value exceeding the mean ± 3 times the standard deviation is considered an outlier.

[0108] In order to obtain better outlier handling effects, in some embodiments, outlier handling includes the following steps: calculate the third quartile Q3 and the first quartile Q1 of the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively; calculate the difference between the third quartile and the first quartile to obtain the interquartile range IQR; traverse all the pixels in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively. If the value of the currently traversed pixel is greater than the sum of the third quartile Q3 and 1.5 times the interquartile range IQR, it is determined that the value of this pixel is an outlier, and the value of the currently traversed pixel is reassigned to the sum of the third quartile Q3 and 1.5 times the interquartile range IQR; or, if the value of the currently traversed pixel is less than the difference between the first quartile Q1 and 1.5 times the interquartile range IQR, it is determined that the value of this pixel is an outlier, and the value of the currently traversed pixel is reassigned to the difference between the first quartile Q1 and 1.5 times the interquartile range IQR.

[0109] Figure 4 FIG. is a schematic flowchart of outlier handling provided according to some embodiments of the present application. As Figure 4 shown, taking the calculation of the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 as the input raster data for outlier handling, calculate the Q1 quantile (first quartile) and Q3 quantile (third quartile) of the two respectively, then calculate the interquartile range according to IQR = Q3 - Q1, and compare the distance between the value x1 of each pixel in the input raster data and Q1 - 1.5×IQR and Q3 + 1.5×IQR respectively. If x1 > Q3 + 1.5×IQR or x1 < Q1 - 1.5×IQR, it is considered that the pixel value is an outlier, and the pixel value is reassigned to Q3 + 1.5×IQR or Q1 - 1.5×IQR, and the raster data after reassignment is output to achieve outlier handling.

[0110] Since the quartiles are not sensitive to the distribution form of the data, by using the above method to handle outliers, even if the data distribution is skewed, outliers can be effectively identified, and then the sum of the third quartile Q3 and 1.5 times the interquartile range IQR, or the difference between the first quartile Q1 and 1.5 times the interquartile range IQR is used to reassign the abnormal pixels, making the pixel values more in line with the actual situation, and further improving the rationality of urban heat resilience research.

[0111] During the HASM simulation process, in order to further improve the simulation accuracy, in some embodiments, HASM further includes an outer constraint condition, and the outer constraint condition is used to constrain the external accuracy of each iteration in the HASM solution process. The optimization control condition of HASM and the outer constraint condition of HASM constitute a pseudo-HASM with double constraints.

[0112] In the iterative solution process of the traditional HASM method, optimization control conditions are usually adopted to construct constraint equations to achieve internal constraints on HASM. However, considering the process of self-supervised downscaling, the optimization control conditions of HASM are the downscaled raster data RS 1α and the anomaly value raster data δ after downscaling 1β The sum of them still has a certain deviation from the data collected through observation stations. Therefore, a dual constraint combining outer-layer constraints and HASM internal constraints is adopted, and the HASM internal constraint conditions are used to ensure that the relationships between the internal parameters and variables of the HASM model remain reasonable and consistent, which can prevent drastic fluctuations in the internal parameter values during HASM simulation and is conducive to maintaining the stability and coherence of the simulation process; the outer-layer constraints are used to set reasonable boundary conditions to prevent the model from getting out of control under extreme conditions and ensure the stable operation of the model in a self-supervised scenario. The combination of internal and external constraints aims to further improve the accuracy of HASM simulation under self-supervised learning.

[0113] Furthermore, constructing the pseudo-HASM with dual constraints can include the following steps: The first scale coefficient m is obtained by calculating the quotient of the mean m0 of the upscaled remote sensing raster data RS0 and the mean m1 of the remote sensing temperature product data RS1; the result of dividing the mean m1 of the remote sensing temperature product data RS1 by the first scale coefficient m is used as the first outer-layer constraint. During the HASM solution process, the mean m2 of the current simulation result is calculated for each iteration of the simulation result; if the error between the mean m2 of the current simulation result and the first outer-layer constraint is less than or equal to a preset first external error threshold, and the current simulation result reaches the preset HASM internal stop condition, the iteration is stopped, and the current simulation result is used as the final temperature simulation result.

[0114] In this embodiment, the calculation formula of the first scale coefficient m is as follows:

[0115]

[0116] Assume that the process of HASM simulation is expressed as: y = HASM(x). Then, outer-layer constraints are constructed outside the iterative process of HASM to form a pseudo-HASM with dual constraints, and the expression is as follows:

[0117]

[0118] In the formula, y represents the result of each iteration of the HASM simulation process, mean(y) represents taking the mean of y to obtain m2, is the first outer-layer constraint.

[0119] In this embodiment, the result of dividing the mean value m1 of the remote sensing temperature product data RS1 by the first scale coefficient m is used as the first outer constraint to describe the theoretical value of the HASM simulation. If the mean value m2 of the actual simulation results obtained in each iteration is close to this theoretical value, that is, the error between the two is less than or equal to the preset first external error threshold, it can be considered that the accuracy of the HASM simulation reaches the outer constraint condition. At this time, if the internal stop condition of HASM can be satisfied simultaneously, the iteration can be stopped and the simulation result can be output.

[0120] Among them, the internal stop condition of HASM can be that the error inside the model is less than the preset internal error threshold, or it can be that the preset number of iterations is reached to satisfy the internal stop condition. The specific implementation method can refer to the prior art. For the sake of brevity of the text, it will not be elaborated here one by one.

[0121] In some other embodiments, constructing the dual-constrained pseudo-HASM may include the following steps: The second scale coefficient s is obtained by calculating the quotient of the standard deviation s0 of the upscaled remote sensing raster data RS0 and the standard deviation s1 of the remote sensing temperature product data RS1; the result of dividing the standard deviation s1 of the remote sensing temperature product data RS1 by the second scale coefficient s is used as the second outer constraint. During the HASM solution process, the standard deviation s2 of the current simulation result is calculated for the simulation result of each iteration; if the error between the standard deviation s2 of the current simulation result and the second outer constraint is less than or equal to the preset second external error threshold, and the current simulation result reaches the preset internal stop condition of HASM, the iteration is stopped, and the current simulation result is used as the final temperature simulation result.

[0122] In this embodiment, the calculation formula of the second scale coefficient s is as follows:

[0123]

[0124] Assume that the process of HASM simulation is expressed as: y = HASM(x). Then, an outer constraint is constructed outside the iteration process of HASM to form a dual-constrained pseudo-HASM, and the expression is as follows:

[0125]

[0126] In the formula, y represents the simulation result of each iteration of the HASM simulation process, std(y) represents calculating the standard deviation of y to obtain s2, is the second outer constraint.

[0127] In this embodiment, the result of dividing the standard deviation s1 of the remote sensing temperature product data RS1 by the second scale factor s is used as the second outer constraint to describe the theoretical value of the HASM simulation. If the standard deviation s2 of the actual simulation result obtained in each iteration is close to this theoretical value, that is, the error between the two is less than or equal to the preset second external error threshold (the value of the second external error threshold can be the same as or different from the value of the first external error threshold), it can be considered that the accuracy of the HASM simulation reaches the outer constraint condition. If the HASM internal stop condition is also met, the iteration can be stopped and the simulation result can be output.

[0128] In some alternative embodiments, constructing the dual-constrained pseudo-HASM may include the following steps: The first scale factor m is obtained by calculating the quotient of the mean m0 of the upscaled remote sensing raster data RS0 and the mean m1 of the remote sensing temperature product data RS1; the second scale factor s is obtained by calculating the quotient of the standard deviation s0 of the upscaled remote sensing raster data RS0 and the standard deviation s1 of the remote sensing temperature product data RS1; the result of dividing the mean m1 of the remote sensing temperature product data RS1 by the first scale factor m is used as the first outer constraint, and the result of dividing the standard deviation s1 of the remote sensing temperature product data RS1 by the second scale factor s is used as the second outer constraint. The first outer constraint and the second outer constraint together constitute the total outer constraint. During the HASM solution process, the mean m2 and the standard deviation s2 of the current simulation result are calculated for each iteration result of the simulation; if the error between the mean m2 of the current simulation result and the first outer constraint is less than or equal to the preset first external error threshold, the error between the standard deviation s2 of the current simulation result and the second outer constraint is less than or equal to the preset second external error threshold, and moreover, the current simulation result reaches the preset HASM internal stop condition, then the iteration is stopped, and the current simulation result is used as the final temperature simulation result.

[0129] Assume that the process of HASM simulation is expressed as: y = HASM(x). Then, an outer constraint is constructed outside the iteration process of HASM to form a dual-constrained pseudo-HASM, and the expression is as follows:

[0130]

[0131] In the formula, y represents the result of each iteration of the HASM simulation process. mean(y) and std(y) represent calculating the mean and standard deviation of y respectively, and thus m2 and s2 can be obtained. Together, they constitute the constraint module of this embodiment. Then, the HASM simulation is run. By continuously solving the above constraint module until both the internal error and the outer error reach the preset error conditions or the preset number of iterations, then the operation of the program is stopped. At this time, the result obtained by the inner-layer HASM is the final downscaling result.

[0132] In this embodiment, by constructing a constraint module with two theoretical values , the two together constitute the outer constraint of the bicondition, making the outer constraint conditions further improved, which is beneficial to further improving the accuracy of HASM simulation.

[0133] As an example, as Figure 2 shown, the method of this embodiment can be implemented through a self-supervised module, a HASM module, and a constraint module. The overall process is described as follows: The remote sensing temperature product is used as input data and input into the self-supervised module to calculate the downscaling factor, upscaling process, and extraction of self-supervised knowledge, thereby constructing the initial field and optimized control conditions for HASM simulation, and then input into the HASM module to carry out HASM simulation to obtain the downscaling result. Among them, the self-supervised module can realize the extraction of self-supervised knowledge. Specifically, as Figure 3 shown, in the self-supervised module, the original remote sensing raster data (i.e., the remote sensing temperature product data RS1) is first upscaled to obtain the remote sensing raster data RS0 (i.e., the upscaled remote sensing raster data RS0), and then RS1 and RS0 are input into the outlier processing module, and the outlier processing is carried out in a quartile-based manner to correspondingly obtain the raster data RS1' and RS0'; the scale operator α is calculated according to the spatial position mapping relationship and pixel values between RS1' and RS0'; at the same time, the first anomaly value raster δ1 and the second anomaly value raster δ0 are calculated based on RS1' and RS0' respectively, and then the scale operator β is calculated using the spatial position mapping relationship and pixel values between δ1 and δ0. In addition, as Figure 3 shown, the self-supervised module further includes: respectively calculating the means m1, m0 and variances s1, s0 of RS1' and RS0', calculating the first scale coefficient m and the second scale coefficient s according to m1, m0 and variances s1, s0, and constructing the outer constraint conditions using the constraint module based on the mean m1, variance s1, first scale coefficient m and second scale coefficient s to further improve the effect of self-supervised downscaling. In the HASM module, as Figure 6 shown, first, the scale operator α and the scale operator β extracted by the self-supervised module are used to downscale the remote sensing temperature product data RS1 and the anomaly value raster δ1 respectively to obtain the raster data RS 1α with the target resolution and the anomaly value raster data δ 1β with the target resolution; then, using the raster data RS 1α as the initial field and the sum of the raster data RS 1α and the anomaly value raster data δ 1β as the optimized control condition, the HASM simulation is carried out to obtain the downscaling result.

[0134] In summary, the method provided in this embodiment extracts self-supervised knowledge from the original data of remote sensing temperature products, constructs the initial field and constraint control conditions of HASM, and through the pseudo-HASM with double constraints, while improving the spatial resolution of the data, fully explores the laws contained in the data itself to improve the accuracy of downscaling.

[0135] System embodiment

[0136] This embodiment provides a self-supervised remote sensing temperature product downscaling system, as Figure 8 shown. The system includes: a first calculation unit 801, an upscaling unit 802, a second calculation unit 803, a scale operator calculation unit 804, a traversal unit 805, and a simulation unit 806. Among them:

[0137] The first calculation unit 801 is configured to calculate the downscaling multiple d according to the spatial resolution of the remote sensing temperature product data RS1 and the target resolution required for downscaling.

[0138] The upscaling unit 802 is configured to perform upscaling processing on the remote sensing temperature product data RS1 according to the downscaling multiple d to obtain upscaled remote sensing raster data RS0.

[0139] The second calculation unit 803 is configured to subtract the mean value of each pixel in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 from the pixel values in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively, to obtain the first anomaly value raster data δ1 and the second anomaly value raster data δ0 corresponding to the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively.

[0140] The scale operator calculation unit 804 is configured to calculate the scale operator α by using the spatial position mapping relationship and pixel values between the pixels in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0; calculate the scale operator β by using the spatial position mapping relationship and pixel values between the pixels in the first anomaly value raster data δ1 and the second anomaly value raster data δ0.

[0141] The traversal unit 805 is configured to traverse all the pixels of the remote sensing temperature product data RS1, and downscale each pixel in the remote sensing temperature product data RS1 to the target resolution according to the scale operator α to obtain downscaled raster data RS 1α ; traverse all the pixels of the first anomaly value raster data δ1, and downscale each pixel in the first anomaly value raster data δ1 to the target resolution according to the scale operator β to obtain downscaled anomaly value raster data δ 1β ;

[0142] The simulation unit 806 is configured to use the downscaled raster data RS 1α as the initial field of the high-precision surface modeling HASM method, and use the downscaled raster data RS 1α and the sum of the downscaled anomaly value raster data δ 1β as the optimization control condition of HASM, and perform HASM simulation to obtain the temperature simulation values of all pixels of the downscaled raster data RS 1α

[0143] The self-supervised remote sensing temperature product downscaling system provided in this embodiment can implement the steps and processes of the self-supervised remote sensing temperature product downscaling method provided in any of the above embodiments, and achieve the same technical effects, which will not be elaborated here one by one.

[0144] Device Embodiment

[0145] Figure 9 is a schematic structural diagram of an electronic device provided according to some embodiments of the present application; as Figure 9 shown, the electronic device includes:

[0146] One or more processors 901;

[0147] A computer-readable storage medium that can be configured to store one or more programs 902. When one or more processors 901 execute one or more programs 902, the following steps are implemented:

[0148] According to the spatial resolution of the remote sensing temperature product data RS1 and the target resolution to be achieved by downscaling, calculate the downscaling multiple d;

[0149] According to the downscaling multiple d, perform upscaling processing on the remote sensing temperature product data RS1 to obtain the upscaled remote sensing raster data RS0;

[0150] Subtract the mean value of each pixel in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 from their respective values to obtain the first anomaly value raster data δ1 and the second anomaly value raster data δ0 corresponding to the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively;

[0151] Use the spatial position mapping relationship and pixel values between the pixels in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 to calculate the scale operator α; use the spatial position mapping relationship and pixel values between the pixels in the first anomaly value raster data δ1 and the second anomaly value raster data δ0 to calculate the scale operator β;

[0152] ​Traverse all pixels of the remote sensing temperature product data RS1, and downscale each pixel in the remote sensing temperature product data RS1 to the target resolution according to the scale operator α to obtain the downscaled raster data RS. 1α Traverse all pixels of the first anomaly value raster data δ1, and downscale each pixel in the first anomaly value raster data δ1 to the target resolution according to the scale operator β to obtain the downscaled anomaly value raster data δ. 1β ;

[0153] Use the downscaled raster data RS 1α as the initial field of the high-precision surface modeling HASM method, and use the sum of the downscaled raster data RS 1α and the downscaled anomaly value raster data δ 1β as the optimization control condition of HASM, and perform HASM simulation to obtain the temperature simulation values of all pixels of the downscaled raster data RS 1α .

[0154] Figure 10 shows the hardware structure of an electronic device provided according to some embodiments of the present application; as Figure 10 shown, the hardware structure of the electronic device may include: a processor 1001, a communication interface 1002, a computer-readable storage medium (also referred to as a memory) 1003, and a communication bus 1004.

[0155] Among them, the processor 1001, the communication interface 1002, and the computer-readable storage medium 1003 complete mutual communication through the communication bus 1004.

[0156] The computer-readable storage medium 1003 can be configured to store one or more programs.

[0157] Optionally, the communication interface 1002 can be an interface of a communication module, such as an interface of a GSM module.

[0158] Among them, the processor 1001 executes one or more programs, and the program implements the following steps:

[0159] According to the spatial resolution of the remote sensing temperature product data RS1 and the target resolution to be achieved by downscaling, calculate the downscaling multiple d;

[0160] According to the downscaling multiple d, perform upscaling processing on the remote sensing temperature product data RS1 to obtain the upscaled remote sensing raster data RS0;

[0161] Subtract the value of each pixel in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 from their respective means to obtain the first anomaly value raster data δ1 and the second anomaly value raster data δ0 corresponding to the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0 respectively;

[0162] Calculate the scale operator α using the spatial position mapping relationship and pixel values between the pixels in the remote sensing temperature product data RS1 and the upscaled remote sensing raster data RS0; calculate the scale operator β using the spatial position mapping relationship and pixel values between the pixels in the first anomaly value raster data δ1 and the second anomaly value raster data δ0;

[0163] Traverse all pixels of the remote sensing temperature product data RS1, and downscale each pixel in the remote sensing temperature product data RS1 to the target resolution according to the scale operator α to obtain the downscaled raster data RS 1α ; Traverse all pixels of the first anomaly value raster data δ1, and downscale each pixel in the first anomaly value raster data δ1 to the target resolution according to the scale operator β to obtain the downscaled anomaly value raster data δ 1β ;

[0164] Use the downscaled raster data RS 1α as the initial field of the high-accuracy surface modeling HASM method, and use the sum of the downscaled raster data RS 1α and the downscaled anomaly value raster data δ 1β as the optimization control condition of HASM, and perform HASM simulation to obtain the temperature simulation values of all pixels of the downscaled raster data RS 1α

[0165] The processor 1001 can be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc., and 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 gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0166] The electronic device in the embodiments of the present application exists in various forms, including but not limited to:

[0167] ​(1) Mobile communication devices: These devices are characterized by having mobile communication functions and mainly aim to provide voice and data communications. Such terminals include: smart phones (e.g., iPhone), multimedia phones, functional phones, and low-end phones, etc.

[0168] (2) Ultra-mobile personal computer devices: These devices belong to the category of personal computers, have computing and processing functions, and generally also have the characteristic of mobile Internet access. Such terminals include: PDA, MID, and UMPC devices, etc., such as iPad.

[0169] (3) Portable entertainment devices: These devices can display and play multimedia content. Such devices include: audio and video players (e.g., iPod), handheld game consoles, e-books, and smart toys and portable in-vehicle navigation devices.

[0170] (4) Servers: Devices that provide computing services. The composition of a server includes a processor, hard disk, memory, system bus, etc. Servers are similar to general computer architectures, but due to the need to provide highly reliable services, they have higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability, etc.

[0171] (5) Other electronic devices with data interaction functions.

[0172] It should be noted that according to the needs of implementation, each component / step described in the embodiments of the present application can be split into more components / steps, or two or more components / steps or partial operations of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of the present application.

[0173] The methods according to the embodiments of the present application can be implemented in hardware, firmware, or be implemented as software or computer code that can be stored in a recording medium (such as CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or be implemented as computer code that is originally stored in a remote recording medium or a non-transitory machine storage medium and will be stored in a local recording medium and downloaded through a network. Thus, the methods described herein can be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as ASIC or FPGA). It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes storage components (such as RAM, ROM, flash memory, etc.) that can store or receive software or computer code. When the software or computer code is accessed and executed by the computer, the processor, or the hardware, the self-supervised remote sensing temperature product downscaling method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the method shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for executing the method shown herein.

[0174] Those of ordinary skill in the art can realize that the units and method steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application of the technical solution and the involved constraints. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the embodiments of this application.

[0175] It should be noted that the embodiments in this specification are all described in a progressive manner. For the same or similar parts between the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, reference can be made to the partial description of the method embodiments.

[0176] The device and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components referred to as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0177] The above are only the preferred embodiments of this application and are not used to limit this application. For those skilled in the art, various changes and modifications can be made to this application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims

1. A self-supervised method for downscaling remote sensing temperature products, characterized in that, including: According to the remote sensing temperature product data RS 1 Calculate the downscaling factor based on the spatial resolution of the remote sensing temperature product data and the target resolution required for downscaling d ; According to the downscaling multiple d , upscale the remote sensing temperature product data RS 1 to obtain upscaled remote sensing raster data RS 0 ; Subtract the respective mean values from the pixel values in the remote sensing temperature product data RS 1 and the upscaled remote sensing raster data RS 0 to obtain the first anomaly value raster data RS 1 corresponding to the remote sensing temperature product data RS 0 and the second anomaly value raster data δ 1 corresponding to the upscaled remote sensing raster data δ 0 ; Using the remote sensing temperature product data RS 1 and the spatial position mapping relationship and pixel values between pixels in the upscaled remote sensing raster data RS 0 calculate the scale operator α ; Using the first anomaly value raster data δ 1 and the spatial position mapping relationship and pixel values between pixels in the second anomaly value raster data δ 0 calculate the scale operator β ; The scale operator α and the scale operator β are used as self - supervised knowledge for downscaling operations, including: traversing all pixels of the remote sensing temperature product data RS 1 , and downscaling each pixel in the remote sensing temperature product data α to the target resolution according to the scale operator, obtaining the downscaled raster data RS 1 RS 1α ; traversing all pixels of the first anomaly value raster data δ 1 , and downscaling each pixel in the first anomaly value raster data β to the target resolution according to the scale operator, obtaining the downscaled anomaly value raster data δ 1 δ 1β ;​​ Using the downscaled raster data RS 1α as the initial field of the High-accuracy Surface Modeling (HASM) method, and using the downscaled raster data RS 1α and the sum of the downscaled anomaly value raster data δ 1β as the optimization control condition of HASM, conduct HASM simulation to obtain the temperature simulation values of all pixels of the downscaled raster data RS 1α ​ 2. The method according to claim 1, characterized in that, further including: constructing the outer constraint conditions of HASM, where the outer constraint conditions are used to constrain the external accuracy of each iteration in the HASM solution process, and the optimization control conditions of HASM and the outer constraint conditions of HASM constitute a pseudo-HASM with double constraints.

3. The method according to claim 2, wherein Constructing the outer constraint conditions of HASM includes: Calculate the upscaled remote sensing raster data RS 0 for its mean value m 0 and the mean value of the remote sensing temperature product data RS 1 for its mean value m 1 to obtain the first scale coefficient m ; and use the result of dividing the mean value RS 1 of the remote sensing temperature product data m 1 by the first scale coefficient m as the first outer constraint; Correspondingly, expanding the HASM simulation includes: During the HASM solution process, the mean value of the current simulation results is calculated for the simulation results of each iteration. m 2 ; If the mean of the current simulation results m 2 has an error less than or equal to a preset first external error threshold with respect to the first outer constraint, and the current simulation results meet the preset HASM internal stop condition, then stop the iteration and use the current simulation results as the final temperature simulation results.

4. The method according to claim 2, wherein Constructing the outer constraint conditions of HASM includes: Calculate the upscaled remote sensing raster data RS 0 Standard deviation s 0 Of the remote sensing temperature product data RS 1 Standard deviation s 1 To obtain the second scaling coefficient s ; And using the standard deviation of the remote sensing temperature product data RS 1 Standard deviation s 1 Divided by the second scaling coefficient s The result is used as the second outer constraint; Correspondingly, expanding the HASM simulation includes: During the HASM solution process, the standard deviation of the current simulation result is calculated for the simulation results of each iteration. s 2 ; If the standard deviation of the current simulation result s 2 has an error less than or equal to a preset second external error threshold with respect to the second outer constraint, and the current simulation result meets the preset HASM internal stop condition, then stop the iteration and use the current simulation result as the final temperature simulation result.

5. The method according to claim 2, characterized in that, Constructing the outer constraint conditions of HASM includes: Calculate the upscaled remote sensing raster data RS 0 for its mean value m 0 and the mean value of the remote sensing temperature product data RS 1 for its mean value m 1 and obtain the first scale coefficient by taking the quotient m ; Calculate the upscaled remote sensing raster data RS 0 for the standard deviation s 0 and the standard deviation of the remote sensing temperature product data RS 1 to obtain the quotient, which is the second scaling coefficient s 1 ; s ; The remote sensing temperature product data RS 1 The mean m 1 Divide by the first scale factor m The result is used as the first outer constraint, with the remote sensing temperature product data RS 1 Standard Deviation s 1 Divide by the second scale factor s The result is used as the second outer constraint, and the first outer constraint and the second outer constraint together constitute a double-conditional outer constraint; Correspondingly, expanding the HASM simulation includes: During the HASM solution process, the mean value of the current simulation results is calculated for each iteration of the simulation results m 2 and the standard deviation s 2 ; if the error between the mean value of the current simulation results m 2 and the first outer constraint is less than or equal to a preset first external error threshold, and the standard deviation of the current simulation results s 2 and the error between the second outer constraint is less than or equal to a preset second external error threshold, and the current simulation results meet the preset HASM internal stop condition, then the iteration is stopped, and the current simulation results are used as the final temperature simulation results.

6. The method according to claim 1, wherein According to the upscaling factor d , the remote sensing temperature product data RS 1 is upscaled to obtain upscaled remote sensing raster data RS 0 After that, it further includes: Perform outlier processing on the remote sensing temperature product data respectively RS 1 and the upscaled remote sensing raster data RS 0 respectively.

7. The method according to claim 6, characterized in that, The outlier processing includes the following steps: Calculate the remote sensing temperature product data separately RS 1 and the upscaled remote sensing raster data RS 0 for the third quartile Q3 and the first quartile Q1; Calculating the difference between the third quartile and the first quartile to obtain the interquartile range IQR; Traverse the remote sensing temperature product data respectively RS 1 and all the pixels in the upscaled remote sensing raster data RS 0 If the value of the currently traversed pixel is greater than the sum of the third quartile Q3 and 1.5 times the interquartile range IQR, it is determined that the value of the pixel is an outlier, and the value of the currently traversed pixel is reassigned to the sum of the third quartile Q3 and 1.5 times the interquartile range IQR; Alternatively, if the value of the currently traversed pixel is less than the difference between the first quartile Q1 and 1.5 times the interquartile range IQR, then it is determined that the value of the pixel is an outlier, and the value of the currently traversed pixel is reassigned to the difference between the first quartile Q1 and 1.5 times the interquartile range IQR.

8. A self-supervised remote sensing temperature product downscaling system, characterized in that, including: The first computing unit is configured to calculate a downscaling factor according to the spatial resolution of remote sensing temperature product data RS 1 and the target resolution to be achieved by downscaling d ; An upscaling unit configured to perform upscaling on the remote sensing temperature product data according to the downscaling multiple d , to obtain upscaled remote sensing raster data RS 1 by performing upscaling processing on the remote sensing temperature product data RS 0 ; A second computing unit, configured to subtract the respective pixel values in the remotely sensed temperature product data RS 1 and the upscaled remotely sensed raster data RS 0 from the mean values of the remotely sensed temperature product data RS 1 and the upscaled remotely sensed raster data RS 0 respectively, to obtain the first anomaly value raster data RS 1 and the second anomaly value raster data RS 0 corresponding to the remotely sensed temperature product data δ 1 and the upscaled remotely sensed raster data δ 0 ; A scale operator calculation unit, configured to utilize the remote sensing temperature product data RS 1 and the spatial position mapping relationship and pixel values between pixels in the upscaled remote sensing raster data RS 0 to calculate a scale operator α ; utilize the first anomaly value raster data δ 1 and the spatial position mapping relationship and pixel values between pixels in the second anomaly value raster data δ 0 to calculate a scale operator β ; A traversal unit configured to perform a downscaling operation on a scale operator α and a scale operator β as self-supervised knowledge, including: traversing all pixels of the remote sensing temperature product data RS 1 , and downscaling each pixel in the remote sensing temperature product data α to the target resolution according to the scale operator, obtaining downscaled raster data RS 1 ; traversing all pixels of the first anomaly value raster data RS 1α , and downscaling each pixel in the first anomaly value raster data δ 1 to the target resolution according to the scale operator, obtaining downscaled anomaly value raster data β ; δ 1 ; δ 1β ; The simulation unit is configured to use the downscaled raster data RS 1α as the initial field of the High-accuracy Surface Modeling (HASM) method, and use the downscaled raster data RS 1α and the downscaled anomaly value raster data δ 1β The sum of them is used as the optimization control condition of HASM to carry out HASM simulation, and the temperature simulation values of all pixels of the downscaled raster data are obtained. RS 1α ​ 9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the self-supervised remote sensing temperature product downscaling method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, including: a memory, a processor, and a program stored in the memory and executable on the processor, and when the processor executes the program, it implements the self-supervised remote sensing temperature product downscaling method according to any one of claims 1 to 7.

Images