LCZ temporal-spatial resolution control method based on multi-source remote sensing information

Abstract

The invention discloses an LCZ temporal-spatial resolution control method based on multi-source remote sensing information, and relates to the technical field of image processing. The LCZ temporal-spatial resolution control method based on the multi-source remote sensing information comprises the following steps of scene weight evaluation, environment weight evaluation and temporal-spatial fusion feedback evaluation. According to the method, the initial weight is obtained through the multi-source remote sensing information identification scene, the remote sensing data and the scene characteristic data are obtained to obtain the scene weight evaluation coefficient, and the geographical environment data are obtained and combined with the remote sensing data to obtain the environment weight evaluation coefficient; then, whether dynamic weight redistribution is carried out or not is judged according to the scene weight evaluation coefficient and the environment weight evaluation coefficient, finally, a weight regulation and control evaluation coefficient is obtained according to the obtained space-time fusion feedback data to judge whether weight secondary distribution is carried out or not, and the effect of improving the space-time resolution control accuracy is achieved. The problem of inaccurate temporal-spatial resolution control in the prior art is solved.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular to a LCZ spatiotemporal resolution control method based on multi-source remote sensing information. Background Art

[0002] With the acceleration of urbanization and the intensification of climate change, local climate zones (LCZs) have attracted extensive attention as an important method for studying urban microclimate and thermal environment. However, the accuracy and spatiotemporal resolution of LCZ classification are often limited by the shortcomings of a single remote sensing data source, such as low spatial resolution and discontinuous temporal coverage.

[0003] Existing local climate zones (LCZ) research mainly relies on a single remote sensing data source, such as optical images or thermal infrared data. Although these data can provide information on land cover types, their spatial resolution is limited or the temporal observation frequency is insufficient, making it difficult to meet the high-precision requirements in dynamic and complex environments. Traditional methods mostly use fixed-resolution processing, ignoring the complementarity between different data sources, which easily leads to a decrease in classification accuracy. At the same time, the classification model based on machine learning has limited ability to integrate multi-scale and multi-temporal data, and lacks technology for dynamic optimization of spatiotemporal resolution, which restricts the promotion and application of LCZ.

[0004] For example, the patent application with publication number: CN117494765A discloses a twin network and method for change detection of remote sensing images with ultra-high spatial resolution, including: deep feature extraction through an asymmetric convolutional residual encoder, and then stacking difference features of different scales along the channel direction, and using the spatiotemporal attention mechanism to enhance the multi-scale difference features in the spatiotemporal domain; the multi-scale difference features after spatiotemporal domain enhancement are split according to the number of channels of different scales, and the enhanced difference features are fused with the dual-phase encoder features of the same scale through a feature fusion decoder.

[0005] For example, the invention patent with announcement number CN111695681B announces a high-resolution dynamic visual observation method and device, including: step 1, obtaining low-resolution event stream data, obtaining low spatial resolution event stream data from an event camera; step 2, performing data preprocessing, and calculating the multiple of the required resolution improvement according to the spatial resolution size of the input low-resolution event data and the spatial resolution size of the expected high-resolution event data; step 3, constructing a pulse neural network, using the Spike Response model as the neuron dynamics model, and constructing a fully connected pulse neural network; step 4, completing the high-resolution spatiotemporal pulse signal calculation; step 5, performing data post-processing to achieve high-resolution dynamic visual observation.

[0006] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:

[0007] In existing technologies, different scenes (such as cities, forests, and farmlands) have very different requirements for temporal and spatial resolution, while simple weighted methods use fixed weights or lack scene adaptation capabilities, resulting in inaccurate control of temporal and spatial resolution. Summary of the invention

[0008] The embodiment of the present application solves the problem of inaccurate spatiotemporal resolution control in the prior art by providing a LCZ spatiotemporal resolution control method based on multi-source remote sensing information, thereby improving the accuracy of spatiotemporal resolution control.

[0009] The embodiment of the present application provides a LCZ spatiotemporal resolution control method based on multi-source remote sensing information, comprising the following steps: S1, identifying the scene based on the multi-source remote sensing information to obtain the initial weight, obtaining the scene weight evaluation coefficient by acquiring the remote sensing data and the scene characteristic data within a preset time period, and the scene weight evaluation coefficient is used to quantify the degree of adaptation between the initial weight and the scene; S2, acquiring the geographic environment data and combining the remote sensing data to obtain the environment weight evaluation coefficient, judging whether to perform dynamic weight redistribution according to the scene weight evaluation coefficient and the environment weight evaluation coefficient, and the environment weight evaluation coefficient is used to quantify the degree of adaptation between the initial weight and the geographic environment; S3, obtaining the weight control evaluation coefficient according to the acquired spatiotemporal fusion feedback data, and judging whether to perform weight secondary distribution according to the weight control evaluation coefficient, and the weight control evaluation coefficient is used to feedback the rationality of the fusion after the time weight and the space weight are controlled.

[0010] Furthermore, the specific steps for obtaining the initial weights are as follows: A1, automatically identifying scenes and classifying scenes through image information of multi-source remote sensing information, wherein the scene classification represents dividing the automatically identified scenes according to the specific geographical conditions of the region; A2, constructing a mapping set of scene categories and initial weights, wherein the mapping set is constructed by preset staff based on historical data; A3, inputting the automatically identified scenes into the mapping set to obtain the corresponding initial weights.

[0011] Furthermore, the remote sensing data includes spatial resolution and temporal resolution; the scene characteristic data includes vegetation index, hydrological index, texture entropy and texture contrast; the geographic environment data includes surface roughness, temperature, rainfall and wind speed; the spatiotemporal fusion feedback data includes mean square error, structural similarity index, spatial residual, temporal residual and LCZ image clarity; the LCZ image clarity represents the clarity of the LCZ image generated after a preset time period.

[0012] Furthermore, the specific steps for obtaining the scene weight evaluation coefficient are as follows: B1, numbering the preset time periods and obtaining reference scene data from a preset database, wherein the reference scene data includes a preset vegetation index, a preset hydrological index, a reference texture entropy and a reference texture contrast; B2, performing data preprocessing on the remote sensing data, wherein the data preprocessing is used to de-unitize and normalize the remote sensing data; B3, obtaining a remote sensing base according to the remote sensing data and the initial weight, wherein the remote sensing base is used to quantify the current spatiotemporal resolution; B4, processing the remote sensing base, the scene characteristic data and the reference scene data to obtain the scene weight evaluation coefficient.

[0013] Furthermore, the specific limiting expression of the scene weight evaluation coefficient is as follows:

[0014] ;

[0015] Where t represents the number of the preset time period, , Indicates the total number of preset time periods. Represents the remote sensing base number of the t-th preset time period, represents the vegetation index of the t-th preset time period, represents the hydrological index of the t-th preset time period, represents the texture entropy of the t-th preset time period, represents the texture contrast of the t-th preset time period, Indicates the preset vegetation index. Indicates the preset hydrological index, represents the reference texture entropy, represents the reference texture contrast, Represents the scene weight evaluation coefficient for the tth preset time period.

[0016] Furthermore, the specific process of the environmental weight assessment coefficient is as follows: obtain reference geographic data from a preset database, the reference geographic data including reference surface roughness, reference temperature, reference rainfall and reference wind speed; obtain remote sensing base and process the relative deviation and absolute deviation of the geographic environment data and the corresponding reference geographic data to obtain the environmental weight assessment coefficient; the relative deviation includes relative temperature deviation, relative rainfall deviation and relative wind speed deviation; the relative temperature deviation represents the ratio of the absolute value of the difference between the temperature and the reference temperature to the reference temperature; the relative rainfall deviation represents the ratio of the absolute value of the difference between the rainfall and the reference rainfall to the reference rainfall; the relative wind speed deviation represents the ratio of the absolute value of the difference between the wind speed and the reference wind speed to the reference wind speed.

[0017] Furthermore, the specific process of determining whether to perform dynamic weight redistribution is as follows: obtaining the scene threshold and the environment threshold from a preset database, and comparing the scene weight evaluation coefficient and the environment weight evaluation coefficient with the scene threshold and the environment threshold respectively: if the scene weight evaluation coefficient is not less than the scene threshold and the environment weight evaluation coefficient is not less than the environment threshold, dynamic weight redistribution is not performed; if the scene weight evaluation coefficient is less than the scene threshold or the environment weight evaluation coefficient is less than the environment threshold, dynamic weight redistribution is performed; the dynamic weight redistribution includes dynamic adjustment of weights based on spatiotemporal change characteristics and spatiotemporal convolutional network optimization.

[0018] Furthermore, the specific content of the dynamic weight redistribution is as follows: the dynamic adjustment of weights based on spatiotemporal change characteristics is achieved through spatiotemporal block dynamic weighting and weight smoothing processing; the spatiotemporal block dynamic weighting means calculating the weights region by region after dividing the LCZ image generated based on multi-source remote sensing information.

[0019] Furthermore, the specific acquisition process of the weight control evaluation coefficient is as follows: obtaining the minimum LCZ image clarity value from a preset database, the minimum LCZ image clarity value represents the minimum limit of the clarity of the LCZ image generated based on multi-source remote sensing information; normalizing the mean square error, spatial residual and temporal residual, the normalization is used to unify the dimensions of spatiotemporal fusion data; obtaining the weight control evaluation coefficient by processing the sum of deviations, the structural similarity index and the LCZ image clarity and the corresponding minimum LCZ image clarity value respectively, the sum of deviations represents the result of adding the mean square error, the spatial residual and the temporal residual.

[0020] Furthermore, the specific process of determining whether to perform secondary weight distribution is as follows: obtaining a control threshold from a preset database, wherein the control threshold is used to determine whether to perform secondary weight distribution; comparing the weight control evaluation coefficient with the control threshold: if the weight control evaluation coefficient is not less than the control threshold, no secondary weight distribution is performed; if the weight control evaluation coefficient is less than the control threshold, secondary weight distribution is performed; the secondary weight distribution indicates that the initial weight after reallocation based on the dynamic weight in the next preset time period is used as the initial weight, and based on the initial weight, the scene weight evaluation coefficient, the environment weight evaluation coefficient and the weight control evaluation coefficient are continued to be obtained until the weight control evaluation coefficient is not less than the control threshold.

[0021] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0022] 1. The scene is identified through multi-source remote sensing information to obtain the initial weight and the remote sensing data and scene characteristic data within a preset time period are obtained to obtain the scene weight evaluation coefficient. At the same time, the geographic environment data is obtained and combined with the remote sensing data to obtain the environmental weight evaluation coefficient. Then, according to the scene weight evaluation coefficient and the environmental weight evaluation coefficient, it is determined whether to perform dynamic weight redistribution. Finally, according to the obtained spatiotemporal fusion feedback data, the weight control evaluation coefficient is obtained to determine whether to perform weight secondary distribution, thereby improving the effect of spatiotemporal fusion, and then achieving an improvement in the accuracy of spatiotemporal resolution control, which effectively solves the problem of inaccurate spatiotemporal resolution control in the prior art.

[0023] 2. By numbering the preset time periods and obtaining reference scene data from the preset database, the remote sensing data is preprocessed, and then the remote sensing base is obtained according to the remote sensing data and the initial weight. Finally, the remote sensing base, scene characteristic data and reference scene data are processed to obtain the scene weight evaluation coefficient, thereby more accurately evaluating the rationality of the initial weight in the corresponding scene, and then realizing dynamic balanced adjustment of the initial weight to improve the effect of space-time fusion.

[0024] 3. By obtaining reference geographic data from the preset database, and then obtaining the remote sensing base and processing the geographic environment data and the relative deviation and absolute deviation of the corresponding reference geographic data, the environmental weight assessment coefficient is obtained, thereby more accurately evaluating the rationality of the initial weight in the corresponding geographic environment, and then realizing dynamic balanced adjustment of the initial weight to improve the effect of space-time fusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A flowchart of a method for controlling LCZ spatiotemporal resolution based on multi-source remote sensing information provided in an embodiment of the present application;

[0026] Figure 2 A schematic diagram of the weight control evaluation coefficient according to the mean square error provided in the embodiment of the present application;

[0027] Figure 3 A schematic diagram of the weight control evaluation coefficient according to the structural similarity index provided in the embodiment of the present application;

[0028] Figure 4 A schematic diagram of the weight control evaluation coefficient according to the spatial residual provided in the embodiment of the present application;

[0029] Figure 5 A schematic diagram of the change of the residual of the weight control evaluation coefficient over time provided in an embodiment of the present application;

[0030] Figure 6 A schematic diagram of the weight control evaluation coefficient provided in an embodiment of the present application changing with the clarity of the LCZ image. DETAILED DESCRIPTION

[0031] The embodiment of the present application solves the problem of inaccurate spatiotemporal resolution control in the prior art by providing an LCZ spatiotemporal resolution control method based on multi-source remote sensing information. The method solves the problem of inaccurate spatiotemporal resolution control in the prior art by numbering preset time periods and obtaining reference scene data from a preset database, performing data preprocessing on the remote sensing data, and obtaining a remote sensing base according to the remote sensing data and an initial weight. The remote sensing base, scene characteristic data, and reference scene data are then processed to obtain a scene weight evaluation coefficient. At the same time, reference geographic data is obtained from a preset database, and then the remote sensing base is obtained and the relative deviation and absolute deviation of the geographic environment data and the corresponding reference geographic data are processed to obtain an environmental weight evaluation coefficient. The method then determines whether to perform dynamic weight redistribution based on the scene weight evaluation coefficient and the environmental weight evaluation coefficient. Finally, a weight control evaluation coefficient is obtained based on the acquired spatiotemporal fusion feedback data to determine whether to perform a secondary weight distribution, thereby improving the accuracy of spatiotemporal resolution control.

[0032] The technical solution in the embodiment of the present application is to solve the above-mentioned problem of inaccurate spatiotemporal resolution control, and the overall idea is as follows:

[0033] The scene weight evaluation coefficient and the environment weight evaluation coefficient are obtained to determine whether dynamic weight redistribution is to be performed, and the weight control evaluation coefficient is obtained based on the acquired spatiotemporal fusion feedback data to determine whether secondary weight distribution is to be performed, thereby achieving the effect of improving the accuracy of spatiotemporal resolution control.

[0034] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0035] like Figure 1 As shown, it is a flowchart of a LCZ spatiotemporal resolution control method based on multi-source remote sensing information provided by an embodiment of the present application, the method comprising the following steps: S1, identifying the scene based on multi-source remote sensing information to obtain the initial weight, obtaining the scene weight evaluation coefficient by acquiring the remote sensing data and the scene characteristic data within a preset time period, the initial weight including the initial time weight and the initial space weight, the scene weight evaluation coefficient is used to quantify the degree of adaptation between the initial weight and the scene; S2, acquiring geographic environment data and combining the remote sensing data to obtain the environment weight evaluation coefficient, judging whether to perform dynamic weight redistribution according to the scene weight evaluation coefficient and the environment weight evaluation coefficient, the environment weight evaluation coefficient is used to quantify the degree of adaptation between the initial weight and the geographic environment; S3, obtaining the weight control evaluation coefficient according to the acquired spatiotemporal fusion feedback data, judging whether to perform weight secondary distribution according to the weight control evaluation coefficient, the weight control evaluation coefficient is used to feedback the rationality of the fusion after the time weight and the space weight are controlled.

[0036] In this embodiment, remote sensing data, scene characteristic data and geographic environment data are all relevant data of the same area, and LCZ images are generated after the start and end of the preset time period for comparative analysis to obtain spatiotemporal fusion feedback data; by integrating multi-source remote sensing information and combining spatiotemporal fusion feedback, more accurate allocation and dynamic adjustment of spatiotemporal weights are achieved, and the scientificity and rationality of weight allocation are improved; at the same time, this method not only comprehensively considers the impact of time, space and environmental factors on the scene, but also introduces weight evaluation coefficients and feedback mechanisms to ensure the rationality and effectiveness of weights in the fusion process. This comprehensive processing strategy not only enhances the adaptability and flexibility of the method, but also improves the accuracy of spatiotemporal resolution control.

[0037] It needs to be explained that remote sensing data includes spatial resolution and temporal resolution. The spatial resolution is obtained directly through satellite sensors, and the temporal resolution is extracted from satellite image data. The scene characteristic data include vegetation index, hydrological index, texture entropy and texture contrast. The geographic environment data include surface roughness, temperature, rainfall and wind speed. The spatiotemporal fusion feedback data include mean square error, structural similarity index, spatial residual, temporal residual and LCZ image clarity. The LCZ image clarity refers to the clarity of the LCZ image generated after the preset time period.

[0038] Among them, the vegetation index (i.e., normalized difference vegetation index) is generated through online platforms (such as Google Earth Engine) to measure the coverage and growth status of surface vegetation, and the hydrological index (i.e., normalized difference water index) is generated using Google Earth Engine. Engine is used to extract and measure the water coverage area, which is used to analyze the seasonal changes of hydrology. The texture entropy is calculated by the texture analysis tool in the ENVI software on the gray-level co-occurrence matrix, which is the entropy of the gray-level co-occurrence matrix of the LCZ image and describes the complexity of the gray-level value distribution in the image. The texture contrast is calculated by the contrast in the gray-level co-occurrence matrix. The surface roughness is obtained by ground radar measurement and is the contrast of the gray-level co-occurrence matrix of the LCZ image, which is used to measure the significance of the gray-level difference of pixels in the LCZ image. The temperature, rainfall and wind speed are all provided by satellite data. The mean square error is calculated by using the built-in functions of Python or MATLAB. The structural similarity index is calculated by using the built-in functions of Python (scikit-image library) or MATLAB. The spatial residual is calculated by analyzing the deviation of the spatial texture features of the image. The temporal residual is calculated by comparing the difference in the trend of the time series. The clarity of the LCZ image is obtained by remote sensing image processing software. The above data not only provide support for spatiotemporal fusion, but also provide a data basis for subsequent analysis.

[0039] Furthermore, the specific steps for obtaining the initial weights are as follows: A1, automatically identifying and classifying scenes through image information of multi-source remote sensing information, where scene classification means dividing the automatically identified scenes according to the specific geographical conditions of the region; A2, constructing a mapping set of scene categories and initial weights, where the mapping set is constructed by preset staff based on historical data; A3, inputting the automatically identified scenes into the mapping set to obtain the corresponding initial weights.

[0040] In this embodiment, automatic classification technology of remote sensing images is used, that is, a computer is used to identify and classify the attributes of information on the earth's surface and its environment in remote sensing images; and by constructing a mapping set of scene categories and initial weights, fast and accurate initial weight allocation is achieved. This process not only improves the level of automation and intelligence, reduces human intervention, but also ensures the rationality and reliability of weight allocation.

[0041] Furthermore, the specific steps for obtaining the scene weight evaluation coefficient are as follows: B1, numbering the preset time periods and obtaining reference scene data from a preset database, the reference scene data including a preset vegetation index, a preset hydrological index, a reference texture entropy and a reference texture contrast; B2, performing data preprocessing on the remote sensing data, the data preprocessing is used to de-unitize and normalize the remote sensing data; B3, obtaining the remote sensing base according to the remote sensing data and the initial weight, the remote sensing base is used to quantify the current spatiotemporal resolution; B4, processing the remote sensing base, the scene characteristic data and the reference scene data to obtain the scene weight evaluation coefficient.

[0042] In this embodiment, by comprehensively considering multi-source remote sensing data, an accurate evaluation of the initial weights in the corresponding scenarios is achieved, which not only improves the accuracy and reliability of the weight evaluation, but also provides strong support for subsequent analysis and decision-making; at the same time, the flexibility and scalability of the method also enable it to adapt to changes in different scenarios and needs, providing space for the in-depth application of remote sensing data.

[0043] Specifically, the specific limiting expression of the scene weight evaluation coefficient is as follows:

[0044] ;

[0045] ;

[0046] In the formula, t represents the number of the preset time period, , Indicates the total number of preset time periods. represents the spatial resolution of the t-th preset time period, represents the time resolution of the t-th preset time period, represents the initial time weight of the t-th preset time period, represents the initial spatial weight of the t-th preset time period, Represents the remote sensing base number of the t-th preset time period, represents the vegetation index of the t-th preset time period, represents the hydrological index of the t-th preset time period, represents the texture entropy of the t-th preset time period, represents the texture contrast of the t-th preset time period, Indicates the preset vegetation index. Indicates the preset hydrological index, represents the reference texture entropy, represents the reference texture contrast, Represents the scene weight evaluation coefficient for the tth preset time period.

[0047] In this embodiment, the algorithm combines the remote sensing base, scene characteristic data and reference scene data for comprehensive analysis to obtain the scene weight evaluation coefficient, wherein the remote sensing base is obtained by comprehensive processing according to the initial weight and the corresponding spatiotemporal resolution. When the remote sensing base is determined, as the scene characteristic data is closer to the corresponding reference scene data, the scene weight evaluation coefficient is larger. For example, when the vegetation index is closer to the preset vegetation index, the absolute deviation between the vegetation index and the preset vegetation index is smaller, and the corresponding scene weight evaluation coefficient is larger. The same is true for the hydrological index. For another example, when the texture entropy is closer to the reference texture entropy, the relative deviation between the texture entropy and the reference texture entropy is smaller, and the corresponding scene weight evaluation coefficient is larger. The same is true for the texture contrast. The larger the scene weight evaluation coefficient, the more balanced the fusion of spatiotemporal resolution is, and the more suitable the corresponding initial weight is for the current area.

[0048] The parameters involved in the processing of scene weight evaluation coefficients in this algorithm are interrelated and do not exist independently. Among them, vegetation growth is closely related to hydrological conditions. Vegetation affects water cycle through transpiration, and hydrological conditions in turn affect vegetation growth and distribution. Therefore, vegetation index and hydrological index are usually positively correlated, because vegetation growth depends on water resources, especially in wetlands or seasonal waters. In remote sensing image analysis, vegetation index and texture features are often used in combination to improve the accuracy of classification and recognition. Vegetation index can reflect information such as vegetation growth status and coverage. Vegetation coverage is uniform. Regions with complex hydrological conditions (such as interlaced water bodies and vegetation) usually have higher texture entropy, and texture features can provide additional information about surface cover type, soil moisture, etc. Texture entropy and texture contrast are two important parameters for describing image texture features. Regions with high texture contrast usually have higher corresponding texture entropy, especially in diverse terrain or ecological boundaries. Through the above analysis, the rationality of the initial weights in the corresponding scenes is more accurately evaluated, and then the dynamic balanced adjustment of the initial weights is achieved to improve the effect of spatiotemporal fusion.

[0049] Specifically, the reference scene data is obtained from a preset database. In a specific embodiment, the preset vegetation index and the preset hydrological index are the average values ​​of the vegetation index and the hydrological index of the corresponding season in the historical data, and the reference texture entropy and the reference texture contrast are set according to the LCZ image generation requirements, and the LCZ image generation requirements are specifically set by the preset staff according to the industry standards.

[0050] Furthermore, the specific process of the environmental weight assessment coefficient is as follows: obtain reference geographic data from a preset database, the reference geographic data including reference surface roughness, reference temperature, reference rainfall and reference wind speed; obtain remote sensing base and process the relative deviation and absolute deviation of the geographic environment data and the corresponding reference geographic data to obtain the environmental weight assessment coefficient; the absolute deviation represents the absolute value of the difference between the surface roughness and the reference surface roughness; the relative deviation includes the relative deviation of temperature, the relative deviation of rainfall and the relative deviation of wind speed; the relative deviation of temperature represents the ratio of the absolute value of the difference between the temperature and the reference temperature to the reference temperature; the relative deviation of rainfall represents the ratio of the absolute value of the difference between the rainfall and the reference rainfall to the reference rainfall; the relative deviation of wind speed represents the ratio of the absolute value of the difference between the wind speed and the reference wind speed to the reference wind speed.

[0051] The specific limiting expression of the environmental weight assessment coefficient is as follows:

[0052] ;

[0053] ;

[0054] In the formula, t represents the number of the preset time period, , Indicates the total number of preset time periods. represents the spatial resolution of the t-th preset time period, represents the time resolution of the t-th preset time period, represents the initial time weight of the t-th preset time period, represents the initial spatial weight of the t-th preset time period, Represents the remote sensing base number of the t-th preset time period, represents the surface roughness of the t-th preset time period, represents the temperature in the tth preset time period, represents the rainfall in the tth preset time period, represents the wind speed in the tth preset time period, represents the reference surface roughness, represents the reference temperature, represents the reference rainfall, represents the reference wind speed, Represents the environmental weight assessment coefficient for the tth preset time period.

[0055] In this embodiment, the remote sensing base, geographic environment data and corresponding reference geographic data are comprehensively analyzed to obtain the environmental weight evaluation coefficient, wherein the remote sensing base is obtained by comprehensive processing according to the initial weight and the corresponding spatiotemporal resolution. When the remote sensing base is determined, as the geographic environment data is closer to the corresponding reference geographic data, the environmental weight evaluation coefficient is larger. For example, when the surface roughness is closer to the reference surface roughness, the absolute deviation between the surface roughness and the reference surface roughness is smaller, and the corresponding environmental weight evaluation coefficient is larger. For another example, when the temperature is closer to the reference temperature, the relative deviation between the temperature and the reference temperature is smaller, and the corresponding environmental weight evaluation coefficient is larger. The same is true for rainfall and wind speed. The larger the environmental weight evaluation coefficient, the more balanced the fusion of spatiotemporal resolution is, and the corresponding initial weight is more suitable for the geographical environment of the current area.

[0056] The parameters involved in the processing of the environmental weight assessment coefficient in this algorithm are interrelated and do not exist independently. Among them, the air will be affected by ground obstacles during its flow. The greater the surface roughness, the greater the friction acting on the air, and the more the corresponding wind speed will decrease. At the same time, the surface roughness may affect the heat exchange process on the surface, thereby affecting the temperature. In addition, the wind speed may affect the distribution and change of the temperature. For example, in clear weather, the greater the wind speed, the faster the temperature may drop, because the wind will accelerate the heat loss of the surface. At the same time, the higher the wind speed, the more dispersed the rainfall distribution, which may lead to reduced precipitation in some areas. The long-term effect of rainfall may increase the surface roughness (promote vegetation growth or landform changes). At the same time, high roughness areas are usually more conducive to the formation of local rainfall. Through the above analysis, the rationality of the initial weight in the corresponding geographical environment is more accurately evaluated, and then the dynamic equilibrium adjustment of the initial weight is achieved to improve the effect of spatiotemporal fusion.

[0057] Specifically, the reference geographic data is obtained from a preset database. In a specific embodiment, the reference surface roughness is the average value of the surface roughness of the corresponding season in the historical data, and the reference temperature, reference rainfall and reference wind speed are the average values ​​of temperature, rainfall and wind speed provided by the Meteorological Bureau.

[0058] Furthermore, the specific process of determining whether to perform dynamic weight redistribution is as follows: obtain the scene threshold and the environment threshold from the preset database, and compare the scene weight evaluation coefficient and the environment weight evaluation coefficient with the scene threshold and the environment threshold respectively: if the scene weight evaluation coefficient is not less than the scene threshold and the environment weight evaluation coefficient is not less than the environment threshold, then no dynamic weight redistribution is performed; if the scene weight evaluation coefficient is less than the scene threshold or the environment weight evaluation coefficient is less than the environment threshold, then dynamic weight redistribution is performed; dynamic weight redistribution includes dynamic adjustment of weights based on spatiotemporal change characteristics and spatiotemporal convolutional network optimization.

[0059] In this embodiment, the dynamic weight redistribution process can flexibly and accurately deal with the weight distribution problems under different scenarios and environmental conditions, and can intelligently determine whether the current scene weight evaluation coefficient and the environmental weight evaluation coefficient meet the established standards through the scene threshold and the environmental threshold, thereby achieving flexible response and optimization of weight distribution under different scenarios and environmental conditions, thereby improving the adaptability and stability of the system.

[0060] Specifically, the scene threshold is obtained from a preset database. In a specific embodiment, the remote sensing data and scene characteristic data corresponding to the situation where the spatiotemporal fusion in the historical data does not meet the industry standards are substituted into the specific restriction expression of the scene weight evaluation coefficient to obtain the corresponding data set, and the data set is averaged to obtain the scene threshold.

[0061] Specifically, the environmental threshold is obtained from a preset database. In a specific embodiment, the remote sensing data and geographic environment data corresponding to the situation where the spatiotemporal fusion in the historical data does not meet the industry standards are substituted into the specific restriction expression of the environmental weight assessment coefficient to obtain the corresponding data set, and the data set is averaged to obtain the environmental threshold.

[0062] Furthermore, the specific contents of dynamic weight redistribution are as follows: the dynamic adjustment of weights based on spatiotemporal change characteristics is achieved through dynamic weighting of spatiotemporal blocks and weight smoothing processing; the dynamic weighting of spatiotemporal blocks means that the weights are calculated region by region after the LCZ image generated based on multi-source remote sensing information is divided into blocks; the specific contents of the spatiotemporal convolutional network optimization are as follows: the spatiotemporal weight map, NDVI (Normalized Difference Vegetation Index) change rate map, and edge complexity map of the previous preset time period are input into the spatiotemporal convolutional network, and the initial weights of the next preset time period are output. The spatiotemporal weight map is used to describe the weight distribution of spatial resolution and temporal resolution in the fusion process, the NDVI change rate map is used to reflect the change amplitude of the vegetation index within the preset time period, and the edge complexity map describes the density and complexity of the edges in the LCZ image generated based on multi-source remote sensing information.

[0063] In this embodiment, dynamic weight redistribution is achieved through dynamic weighting of spatiotemporal blocks, weight smoothing processing and spatiotemporal convolutional network optimization. This method can generate LCZ images based on multi-source remote sensing information, comprehensively consider factors such as spatial resolution, temporal resolution, vegetation index change range and edge complexity, thereby improving the accuracy and rationality of weight distribution.

[0064] Furthermore, the specific acquisition process of the weight control evaluation coefficient is as follows: the minimum value of the LCZ image clarity is obtained from the preset database, and the minimum value of the LCZ image clarity represents the minimum limit of the clarity of the LCZ image generated based on multi-source remote sensing information; the mean square error, spatial residual and temporal residual are normalized, and the normalization process is used to unify the dimensions of the spatiotemporal fusion data; the weight control evaluation coefficient is obtained by processing the sum of deviations, the structural similarity index, the LCZ image clarity and the corresponding minimum value of the LCZ image clarity respectively, and the sum of deviations represents the result of the addition operation of the mean square error, the spatial residual and the temporal residual.

[0065] The specific limiting expression of the weight control evaluation coefficient is as follows:

[0066] ;

[0067] In the formula, t represents the number of the preset time period, , Indicates the total number of preset time periods. represents the mean square error of the t-th preset time period, represents the structural similarity index of the t-th preset time period, represents the spatial residual of the t-th preset time period, represents the time residual of the t-th preset time period, represents the LCZ image clarity of the t-th preset time period, Indicates the minimum value of LCZ image clarity, Represents the weight control evaluation coefficient of the t-th preset time period.

[0068] In this embodiment, the algorithm combines the spatiotemporal fusion feedback data and the minimum value of the LCZ image clarity for comprehensive analysis to obtain the weight control evaluation coefficient. In the formula, as the mean square error, time residual and spatial residual decrease, the weight control evaluation coefficient increases accordingly, and as the structural similarity index increases, the weight control evaluation coefficient also increases accordingly. Similarly, as the LCZ image clarity increases, the weight control evaluation coefficient also increases accordingly. Among them, the mean square error, structural similarity index, spatial residual, time residual and LCZ image clarity are concepts and indicators in their respective fields. There is no direct connection between the data but they influence each other. For example, a higher mean square error usually corresponds to a lower Structural similarity index, because a larger pixel value error will lead to a decrease in image perceived quality, and the higher the mean square error, the lower the clarity of the LCZ image, because too large an error usually manifests as image blur or loss of details, and when the structural similarity index is lower, it usually indicates that the spatial structure or temporal consistency of the fused image is poor, and the corresponding spatial residual or temporal residual is larger. At the same time, the higher the structural similarity index, the higher the clarity of the LCZ image, because good structural similarity means that more image details and textures are retained; when the spatial residual and temporal residual are larger, because the spatial and temporal residuals will destroy the clarity and consistency of the image, the clarity of the LCZ image will be lower.

[0069] like Figure 2 As shown, it is a schematic diagram of the weight control evaluation coefficient provided in the embodiment of the present application changing with the mean square error. After the spatiotemporal fusion feedback data is de-unitized, assuming that the minimum LCZ image clarity is 0.1, when the structural similarity index is 0.8, the spatial residual is 0.5, the temporal residual is 0.5, and the LCZ image clarity is 1, as the mean square error increases, the image shows a downward trend, indicating that the mean square error is negatively correlated with the weight control evaluation coefficient; Figure 3 As shown, Figure 3The schematic diagram of the weight control evaluation coefficient changing with the structural similarity index provided in the embodiment of the present application, when the mean square error is 0.5, the spatial residual is 0.5, the temporal residual is 0.5, and the LCZ image clarity is 1, as the structural similarity index increases, the image shows an upward trend, indicating that the structural similarity index is positively correlated with the weight control evaluation coefficient; Figure 4 As shown, Figure 4 The schematic diagram of the weight control evaluation coefficient changing with the spatial residual provided in the embodiment of the present application, when the mean square error is 0.5, the structural similarity index is 0.8, the time residual is 0.5, and the LCZ image clarity is 1, as the spatial residual increases, the image shows a downward trend, indicating that the spatial residual and the weight control evaluation coefficient are negatively correlated; Figure 5 As shown, Figure 5 The schematic diagram of the weight control evaluation coefficient changing with the time residual provided in the embodiment of the present application, when the mean square error is 0.5, the structural similarity index is 0.8, the spatial residual is 0.5, and the LCZ image clarity is 1, as the time residual increases, the image shows a downward trend, indicating that the time residual is negatively correlated with the weight control evaluation coefficient; Figure 6 As shown, Figure 6 A schematic diagram of the weight control evaluation coefficient changing with the LCZ image clarity provided in the embodiment of the present application, when the mean square error is 0.5, the structural similarity index is 0.8, the spatial residual is 0.5, and the temporal residual is 0.5, as the LCZ image clarity increases, the image shows an upward trend, indicating that the LCZ image clarity is positively correlated with the weight control evaluation coefficient; the larger the weight control evaluation coefficient, the more suitable the initial weight after control is for the current scene and geographical environment. By analyzing the weight control evaluation coefficient, the effect of dynamic weight redistribution is more accurately evaluated, which is conducive to the continued improvement of weight distribution.

[0070] Specifically, assuming that the minimum clarity of the LCZ image is 0.4, and after the spatiotemporal fusion feedback data is de-unitized, the data change table of the weight control evaluation coefficient can be obtained, as shown in Table 1:

[0071] Table 1 Data changes of weight control evaluation coefficients

[0072]

[0073] It can be seen from Table 1 that with the decrease of the mean square error, spatial residual and temporal residual, the increase of the structural similarity index and the LCZ image clarity, the weight control evaluation coefficient gradually increases. For example, when the mean square error decreases from 0.6 in the first row to 0.3 in the fifth row, the structural similarity index increases from 0.4 in the first row to 0.8 in the fifth row, the spatial residual decreases from 0.8 in the first row to 0.2 in the fifth row, the temporal residual decreases from 0.9 in the first row to 0.2 in the fifth row, the LCZ image clarity increases from 0.5 in the first row to 0.9 in the fifth row, and the corresponding weight control evaluation coefficient also increases from 1.01 in the first row to 1.61 in the fifth row. Through the specific analysis of the weight control evaluation coefficient, the effect of dynamic weight redistribution is quantified more accurately, which provides a basis for subsequent weight allocation.

[0074] Specifically, the minimum LCZ image clarity value is obtained from a preset database. In a specific embodiment, the minimum LCZ image clarity value is specifically set by a preset staff according to industry standards. For example, if the scene category is a city, and the LCZ image clarity of the city is required to be at least 0.6, then 0.6 is recorded as the minimum LCZ image clarity value.

[0075] Furthermore, the specific process of determining whether to perform secondary weight distribution is as follows: obtaining a control threshold from a preset database, and the control threshold is used to determine whether to perform secondary weight distribution; comparing the weight control evaluation coefficient with the control threshold: if the weight control evaluation coefficient is not less than the control threshold, no secondary weight distribution is performed; if the weight control evaluation coefficient is less than the control threshold, secondary weight distribution is performed; secondary weight distribution means that the initial weight after reallocation based on the dynamic weight in the next preset time period is used as the initial weight, and based on the initial weight, the scene weight evaluation coefficient, the environment weight evaluation coefficient and the weight control evaluation coefficient are continued to be obtained until the weight control evaluation coefficient is not less than the control threshold.

[0076] In this embodiment, taking the initial weights after dynamic weight redistribution as the starting point, through continuous iteration and optimization, the scene weight evaluation coefficient, the environment weight evaluation coefficient and the weight control evaluation coefficient are gradually adjusted until they reach or exceed the corresponding thresholds. This process not only improves the accuracy and rationality of weight allocation, but also ensures the stability and adaptability of the method in different scenarios and environments.

[0077] Specifically, the control threshold is obtained from a preset database. In a specific embodiment, the spatiotemporal fusion feedback data corresponding to the situation where the spatiotemporal fusion in the historical data does not meet the industry standards is substituted into the specific restriction expression of the weight control evaluation coefficient to obtain the corresponding data set, and the data set is averaged to obtain the control threshold.

[0078] To summarize, the embodiment of the present application obtains the initial weight through multi-source remote sensing information identification scene and obtains the remote sensing data and scene characteristic data within a preset time period to obtain the scene weight evaluation coefficient, and simultaneously obtains the geographic environment data and combines it with the remote sensing data to obtain the environmental weight evaluation coefficient, and then determines whether to perform dynamic weight redistribution based on the scene weight evaluation coefficient and the environmental weight evaluation coefficient, and finally obtains the weight control evaluation coefficient based on the obtained spatiotemporal fusion feedback data to determine whether to perform weight secondary distribution, thereby improving the effect of spatiotemporal fusion, and further achieving the improvement of the accuracy of spatiotemporal resolution control, effectively solving the problem of inaccurate spatiotemporal resolution control in the prior art.

[0079] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0080] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0081] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0082] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0083] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0084] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for controlling the spatiotemporal resolution of LCZ based on multi-source remote sensing information, characterized in that: The following steps are involved: S1, identifying the scene based on multi-source remote sensing information to obtain an initial weight, and obtaining a scene weight evaluation coefficient by acquiring remote sensing data and scene characteristic data within a preset time period, wherein the scene weight evaluation coefficient is used to quantify the degree of adaptation between the initial weight and the scene; S2, obtaining geographic environment data and combining it with remote sensing data to obtain an environmental weight assessment coefficient, and judging whether to perform dynamic weight redistribution according to the scene weight assessment coefficient and the environmental weight assessment coefficient, wherein the environmental weight assessment coefficient is used to quantify the degree of adaptation of the initial weight to the geographic environment; S3, obtaining a weight control evaluation coefficient based on the acquired spatiotemporal fusion feedback data, and judging whether to perform secondary weight distribution based on the weight control evaluation coefficient, wherein the weight control evaluation coefficient is used to feedback the rationality of the fusion after the time weight and space weight are regulated.

2. The LCZ spatiotemporal resolution control method based on multi-source remote sensing information as claimed in claim 1, characterized in that: The specific steps of obtaining the initial weights are as follows: A1, automatically identifying scenes and classifying scenes through image information of multi-source remote sensing information, wherein the scene classification means dividing the automatically identified scenes according to specific geographical conditions of the region; A2, constructing a mapping set of scene categories and initial weights, wherein the mapping set is constructed by a preset staff member based on historical data; A3, the automatically identified scene input mapping is concentrated to obtain the corresponding initial weights.

3. The LCZ spatiotemporal resolution control method based on multi-source remote sensing information as claimed in claim 1, characterized in that: The remote sensing data includes spatial resolution and temporal resolution; The scene characteristic data include vegetation index, hydrological index, texture entropy and texture contrast; The geographical environment data include surface roughness, temperature, rainfall and wind speed; The spatiotemporal fusion feedback data includes mean square error, structural similarity index, spatial residual, temporal residual and LCZ image clarity; The LCZ image clarity indicates the clarity of the LCZ image generated after a preset time period ends.

4. The LCZ spatiotemporal resolution control method based on multi-source remote sensing information as claimed in claim 3, characterized in that: The specific steps for obtaining the scene weight evaluation coefficient are as follows: B1, numbering the preset time periods and acquiring reference scene data from a preset database, wherein the reference scene data includes a preset vegetation index, a preset hydrological index, a reference texture entropy, and a reference texture contrast; B2, performing data preprocessing on the remote sensing data, wherein the data preprocessing is used to de-unitize and normalize the remote sensing data; B3, obtaining a remote sensing base number according to the remote sensing data and the initial weight, wherein the remote sensing base number is used to quantify the current spatiotemporal resolution; B4, the remote sensing base, scene characteristic data and reference scene data are processed to obtain the scene weight evaluation coefficient.

5. The LCZ spatiotemporal resolution control method based on multi-source remote sensing information as claimed in claim 4, characterized in that: The specific limiting expression of the scene weight evaluation coefficient is as follows: ； Where t represents the number of the preset time period, , Indicates the total number of preset time periods. Represents the remote sensing base number of the t-th preset time period, represents the vegetation index of the t-th preset time period, represents the hydrological index of the t-th preset time period, represents the texture entropy of the t-th preset time period, represents the texture contrast of the t-th preset time period, Indicates the preset vegetation index. Indicates the preset hydrological index, represents the reference texture entropy, represents the reference texture contrast, Represents the scene weight evaluation coefficient for the tth preset time period.

6. The LCZ spatiotemporal resolution control method based on multi-source remote sensing information as claimed in claim 3, characterized in that: The specific process of the environmental weight assessment coefficient is as follows: Acquire reference geographic data from a preset database, wherein the reference geographic data includes reference surface roughness, reference temperature, reference rainfall, and reference wind speed; Obtain remote sensing base data and process the relative deviation and absolute deviation of geographic environment data and corresponding reference geographic data to obtain environmental weight assessment coefficients; The relative deviation includes relative deviation of temperature, relative deviation of rainfall and relative deviation of wind speed; The relative temperature deviation represents the ratio of the absolute value of the difference between the temperature and the reference temperature to the reference temperature; The relative deviation of rainfall represents the ratio of the absolute value of the difference between rainfall and reference rainfall to the reference rainfall; The relative wind speed deviation represents the ratio of the absolute value of the difference between the wind speed and the reference wind speed to the reference wind speed.

7. The LCZ spatiotemporal resolution control method based on multi-source remote sensing information as claimed in claim 6, characterized in that: The specific process of determining whether to perform dynamic weight redistribution is as follows: Obtain the scene threshold and environment threshold from the preset database, and compare the scene weight evaluation coefficient and environment weight evaluation coefficient with the scene threshold and environment threshold respectively: If the scenario weight evaluation coefficient is not less than the scenario threshold and the environment weight evaluation coefficient is not less than the environment threshold, no dynamic weight redistribution is performed; If the scenario weight evaluation coefficient is less than the scenario threshold or the environment weight evaluation coefficient is less than the environment threshold, dynamic weight redistribution is performed; The dynamic weight redistribution includes dynamic weight adjustment based on spatiotemporal variation characteristics and spatiotemporal convolutional network optimization.

8. The LCZ spatiotemporal resolution control method based on multi-source remote sensing information as claimed in claim 7, characterized in that: The specific contents of the dynamic weight redistribution are as follows: The dynamic adjustment of weights based on spatiotemporal variation characteristics is achieved through spatiotemporal block dynamic weighting and weight smoothing processing; The spatiotemporal block dynamic weighting means that the weight is calculated region by region after the LCZ image generated based on multi-source remote sensing information is divided into blocks.

9. The LCZ spatiotemporal resolution control method based on multi-source remote sensing information as claimed in claim 1, characterized in that: The specific process of obtaining the weight control evaluation coefficient is as follows: Acquire a minimum value of LCZ image clarity from a preset database, wherein the minimum value of LCZ image clarity represents a minimum limit value of clarity of an LCZ image generated based on multi-source remote sensing information; Normalizing the mean square error, the spatial residual, and the temporal residual, wherein the normalization is used to unify the dimensions of the spatiotemporal fusion data; The weight control evaluation coefficient is obtained by processing the sum of deviations, the structural similarity index, the LCZ image clarity and the corresponding minimum LCZ image clarity respectively, wherein the sum of deviations represents the result of adding the mean square error, the spatial residual and the temporal residual.

10. The LCZ spatiotemporal resolution control method based on multi-source remote sensing information as claimed in claim 9, characterized in that: The specific process of determining whether to perform secondary weight distribution is as follows: Obtaining a control threshold from a preset database, where the control threshold is used to determine whether to perform secondary weight distribution; Compare the weight control evaluation coefficient with the control threshold: If the weight control evaluation coefficient is not less than the control threshold, no secondary weight distribution will be performed; If the weight control evaluation coefficient is less than the control threshold, the weight is redistributed; The secondary weight distribution means that the initial weight after dynamic weight redistribution in the next preset time period is used as the initial weight, and the scene weight evaluation coefficient, environment weight evaluation coefficient and weight control evaluation coefficient are continuously obtained based on the initial weight until the weight control evaluation coefficient is not less than the control threshold.

Images