Training method and device for deep learning-based rural tri-cellular space evolution simulation

Abstract

The present disclosure provides a kind of training method and equipment based on deep learning of rural trinborn space evolution simulation, which comprises: determining the spatial vector data of the trinborn space corresponding to each historical year according to the basic data of the target rural area in multiple historical years;Determine the multi-dimensional auxiliary features of the trinborn space according to the spatial vector data corresponding to each historical year;Based on spatial vector data and multi-dimensional auxiliary features, the preset double-layer deep learning model is trained to obtain the trinborn space evolution simulation model, to predict the evolution simulation result of the trinborn space in the target rural area using the trinborn space evolution simulation model.The method and equipment proposed in the present disclosure can improve the precision of the evolution simulation of trinborn space, and ensure the accuracy of prediction.

Description

Technical Field

[0001] This disclosure relates to the field of land space analysis technology, and in particular to training methods and equipment for simulating the evolution of rural three-life space based on deep learning. Background Technology

[0002] With the significant improvement of socio-economic level and the accelerated process of industrialization and urbanization, research and practice on the three-life space have mostly focused on urban areas, while there are still obvious shortcomings in the analysis and regulation of the three-life space in rural areas.

[0003] Existing technologies mostly focus on land use needs, lacking a systematic consideration of the interrelationships between the three spaces (life, ecology, and environment) in rural areas. This results in evaluation results that cannot provide accurate guidance for the coordinated development of these spaces in rural areas. At the same time, in terms of evolutionary simulation and planning support, existing simulation models are unable to accurately identify the spatial change characteristics of the three spaces in rural areas, resulting in insufficient accuracy in the evolutionary simulation results to support planning decision-makers in quickly grasping the core contradictions and development trends.

[0004] Therefore, there is an urgent need to propose a method based on deep learning technology, which integrates multi-source spatial data and simulates the long-term dynamic evolution of the three-life space, so as to provide a scientific basis for the precise regulation and planning decision-making of the three-life space in rural areas. Summary of the Invention

[0005] To overcome the problems existing in related technologies, this disclosure provides a training method and equipment for simulating the evolution of rural three-life space based on deep learning, so as to accurately simulate the long-term dynamic evolution process of three-life space.

[0006] According to a first aspect of the present disclosure, a training method for simulating the evolution of rural three-life space based on deep learning is provided, the method comprising:

[0007] Based on the basic data of the target rural area in multiple historical years, determine the spatial vector data of the three spaces corresponding to each historical year; wherein, the basic data includes remote sensing image data, land use vector data and socio-economic statistics, and the three spaces include ecological space, production space and living space;

[0008] The multidimensional auxiliary features of the three-life space are determined based on the spatial vector data of the three-life space corresponding to each historical year.

[0009] Based on the spatial vector data and the multidimensional auxiliary features, a preset two-layer deep learning model is trained to obtain a three-life space evolution simulation model, which is then used to predict the evolution simulation results of the three-life space in the target rural area.

[0010] According to the method proposed in this disclosure, the two-layer deep learning model includes a spatial extraction sub-model and a temporal evolution sub-model; the step of training the preset two-layer deep learning model based on the spatial vector data and the multi-dimensional auxiliary features includes:

[0011] The spatial vector data is input into the spatial extraction sub-model to determine the spatial characteristics of the three-dimensional space;

[0012] The spatial features and the multidimensional auxiliary features are input into the temporal evolution sub-model to determine the temporal evolution features of the three-generation space;

[0013] A loss function is constructed based on the spatial features and the temporal evolution features, and the parameters of the spatial extraction sub-model and the temporal evolution sub-model are updated according to the loss function.

[0014] According to the method proposed in this disclosure, determining the spatial vector data of the three-dimensional space corresponding to each historical year based on the basic data of the target rural area in multiple historical years includes:

[0015] For each historical year, at least one of the following preprocessing operations is performed on the remote sensing image data of the target rural area in that historical year: radiometric normalization, geometric correction, coordinate transformation, and boundary clipping, to obtain preprocessed remote sensing image data.

[0016] Based on the land use vector data of the target rural area in the historical year and the preprocessed remote sensing image data, determine the regional boundaries corresponding to various spatial types in the target rural area;

[0017] Based on the socioeconomic statistics of the target rural area in the historical year, determine the corresponding socioeconomic information for various spatial types; wherein, the socioeconomic information includes at least one of economic output, economic investment and population information;

[0018] The spatial vector data is determined based on the regional boundaries corresponding to various spaces in the target rural area in different historical years and the corresponding socio-economic information.

[0019] According to the method proposed in the embodiments of this disclosure, the multidimensional auxiliary features include area features, temporal features, morphological features, and coupling coordination features;

[0020] The determination of the multidimensional auxiliary features of the three-generation space based on the spatial vector data corresponding to each historical year includes:

[0021] For each historical year, the regional boundaries and areas of various spaces are determined based on the spatial vector data corresponding to that historical year;

[0022] The area and morphological characteristics of each type of space are determined based on the perimeter and area of ​​the region boundary corresponding to each type of space.

[0023] The temporal characteristics of each type of space are determined based on the area corresponding to each type of space.

[0024] The coupling and coordination characteristics of the three-dimensional space are determined based on the values ​​of the preset evaluation indicators corresponding to each type of space.

[0025] According to the method proposed in this disclosure, determining the area and morphological characteristics of each type of space based on the perimeter and area of ​​the region boundaries corresponding to each type of space includes:

[0026] For each type of space, the area characteristics of the space are determined based on the annual rate of change of the space's area;

[0027] The ratio between the perimeter and the square root of the area corresponding to the boundary of the space is determined, and the product of the ratio and a first proportional threshold is determined as the morphological feature corresponding to the space.

[0028] According to the method proposed in this disclosure, determining the temporal characteristics of various types of spaces based on their corresponding areas includes:

[0029] For each historical year, determine the first region boundary corresponding to each space in the historical year, and determine the second region boundary corresponding to each space in the next year of the historical year.

[0030] For each first space, the transfer area of ​​the first space to the second space is determined based on the boundary of the first region corresponding to the first space and the boundary of the second region corresponding to the second space in the next year; the first space is any type of space among the three types of spaces in the historical year, and the second space is each type of space among the three types of spaces in the next year of the historical year;

[0031] The probability of the first space being converted into the second space is determined by the ratio of the area transferred from the first space to the second space in the historical year to the total area corresponding to the first space in the historical year.

[0032] The temporal features are determined based on the transition probabilities corresponding to each space.

[0033] According to the method proposed in this disclosure, determining the coupling and coordination characteristics of the three-dimensional spaces based on the values ​​of preset evaluation indicators corresponding to various types of spaces includes:

[0034] For each preset evaluation index corresponding to each type of space, determine the target weight of the preset evaluation index;

[0035] The functional score for each space is determined by the weighted sum of the target weights and the values ​​of the preset evaluation indicators for each space.

[0036] The degree of coupling and coordination between all spaces is determined based on the functional scores corresponding to each type of space.

[0037] The corresponding coupling stages and coupling types are determined based on the coupling coordination degree; wherein, the coupling stages include a first coupling stage, an antagonistic stage, a break-in stage, and a second coupling stage, and the coupling types include a disordered decline type and a coordinated development type;

[0038] The coupling coordination characteristics of the three-dimensional space are determined based on the coupling coordination degree, coupling stage, and coupling type among all spaces.

[0039] According to the method proposed in this disclosure, determining the target weight of each preset evaluation index for each type of space includes:

[0040] For each preset evaluation index corresponding to each type of space, determine the standard value corresponding to the preset evaluation index;

[0041] The ratio between the standard value and the sum of the standard values ​​corresponding to the preset evaluation index in all historical years is determined as the index weight; the entropy redundancy corresponding to the preset evaluation index is determined based on the index weight; the first weight of the preset evaluation index is determined based on the ratio between the entropy redundancy corresponding to the preset evaluation index and the sum of the entropy redundancy corresponding to the preset evaluation index in all historical years.

[0042] The absolute values ​​of the differences between the standard values ​​and the preset reference values ​​for the preset evaluation index in all historical years are determined. The maximum value among these absolute values ​​is determined as the first absolute value, and the minimum value among these absolute values ​​is determined as the second absolute value. The correlation coefficients corresponding to the preset evaluation index are determined based on the absolute values ​​of the differences between the standard values ​​and the preset reference values ​​for all historical years, the first absolute value, and the second absolute value. The correlation degree of the preset evaluation index is determined based on the correlation coefficients for each historical year. The second weight of the preset evaluation index is determined based on the ratio between the correlation degree of the preset evaluation index and the sum of the correlation degrees for each historical year.

[0043] The target weight of the preset evaluation index is determined based on the weighted sum of the first weight and the second weight.

[0044] According to the method proposed in this disclosure, determining the coupling coordination degree between all spaces based on the functional scores corresponding to each type of space includes:

[0045] The weighted sum of the functional scores corresponding to each type of space is determined as the comprehensive coordination index;

[0046] The first product value is determined by multiplying the functional scores of each type of space; the second product value is determined by multiplying the sum of the functional scores of every two types of spaces within each type of space; the coupling degree is determined by combining the first product value and the second product value.

[0047] The square root of the product of the comprehensive coordination index and the coupling degree is determined as the coupling coordination degree.

[0048] According to a second aspect of the present disclosure, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described training method for deep learning-based simulation of rural three-life space evolution.

[0049] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects:

[0050] By integrating historical remote sensing images from multiple periods, land use vectors, and socio-economic data to determine spatial vector data, compared to single data, it can more accurately reconstruct the historical state of the three-life space, providing realistic and multi-dimensional data support for subsequent training. At the same time, by mining multi-dimensional auxiliary features from the spatial vector data, and training a pre-set two-layer deep learning model based on the spatial vector data and multi-dimensional auxiliary features, the trained three-life space evolution simulation model can comprehensively learn the spatial change patterns of the three-life space, improving the modeling accuracy of the three-life space and the prediction accuracy of the evolution simulation results.

[0051] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0052] The accompanying drawings, which are incorporated in and form part of this disclosure, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0053] Figure 1 This disclosure is a flowchart illustrating a deep learning-based training method for simulating the evolution of rural three-dimensional space, according to an exemplary embodiment.

[0054] Figure 2 This is a flowchart illustrating a deep learning-based method for simulating the evolution of rural three-life space according to an exemplary embodiment of this disclosure.

[0055] Figure 3 This disclosure is a block diagram of a training apparatus for a deep learning-based simulation of the evolution of rural three-dimensional space, according to an exemplary embodiment.

[0056] Figure 4 This is a block diagram of a deep learning-based rural three-life space evolution simulation device according to an exemplary embodiment of the present disclosure.

[0057] Figure 5 This is a hardware structure diagram of a computer device shown in an embodiment of this disclosure. Detailed Implementation

[0058] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0059] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0060] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0061] The embodiments of this disclosure will now be described in detail.

[0062] like Figure 1 As shown, Figure 1 This disclosure is a flowchart illustrating a deep learning-based training method for simulating the evolution of rural three-dimensional space according to an exemplary embodiment, comprising the following steps:

[0063] Step 101: Determine the spatial vector data of the three-life space corresponding to each historical year based on the basic data of the target rural area in multiple historical years.

[0064] In this embodiment, basic data for each historical period in multiple historical years can be obtained for the target rural area. For example, the multiple historical periods may include at least three historical periods, and the basic data may include remote sensing image data, socio-economic statistics, and land use vector data for the corresponding time periods. The remote sensing image data may include TIFF (Tagged Image File Format) data acquired by a satellite remote sensing platform and multi-band image data with 3 bands and 8-bit depth, such as Landsat-8 (Land Satellite-8) multispectral data and Sentinel-2 (Sentinel Satellite-2) multispectral data. This remote sensing image data contains complete rural land use types.

[0065] In this embodiment, the spatial vector data of the three-dimensional spaces corresponding to each historical year can be determined based on the basic data, and the spatial vector data corresponding to each historical year can be used as the basic database for time series analysis. The three-dimensional spaces include three types of spaces: ecological space, production space, and living space. The spatial vector data can include the regional boundaries of the target rural area in each historical year corresponding to each type of space, such as the coordinates of the boundary points, the socio-economic information associated with each type of space, and the vector data composed of the spatial types to which each regional boundary belongs.

[0066] Through the above processing, spatial vector data of various spaces in the target rural area in different historical years can be obtained. These spatial vector data can provide data support for subsequent extraction of spatial vector information and training of the three-life spatial evolution simulation model.

[0067] Step 102: Determine the multidimensional auxiliary features of the three-life space based on the spatial vector data of the three-life space corresponding to each historical year.

[0068] In this embodiment, the multi-dimensional auxiliary features corresponding to each historical year can include various spatial features, including area features, temporal features, morphological features, and coupling coordination features corresponding to each historical year.

[0069] In this embodiment, for each type of space among production space, living space, and ecological space, the corresponding regional boundary of each type of space can be determined, and the morphological characteristics of the space can be determined based on the perimeter and area corresponding to the regional boundary; the area characteristics of the space can be determined based on the annual rate of change of the area corresponding to the regional boundary.

[0070] In this embodiment, by comparing the regional boundaries of various spaces in two adjacent historical periods, such as period t and period t+1, indicators such as the proportion of the area transferred from one type of space to another can be calculated. This allows for the quantification of the dynamic evolution of space over time and the acquisition of the temporal characteristics of various types of spaces, such as the proportion of the area converted from production space to living space.

[0071] In this embodiment, preset evaluation indicators can be determined based on the socio-economic information corresponding to various types of spaces in the spatial vector data. For production spaces, preset evaluation indicators may include data such as output value per unit area and production investment intensity, such as the annual grain output value per mu of farmland and the industrial output value per square meter of factory building; for living spaces, preset evaluation indicators may include per capita living area and infrastructure coverage rate; for ecological spaces, preset evaluation indicators may include ecological protection investment, vegetation coverage rate, and ecological value transformation output, thereby constructing evaluation indicators that reflect the economic functional attributes of the spaces.

[0072] After obtaining the preset evaluation indicators for each type of space, the functional scores of each type of space can be calculated based on the values ​​of these indicators. Then, based on these functional scores, the coupling coordination degree between all spaces can be calculated. The corresponding coupling stages, such as the first coupling stage and the second coupling stage, as well as the coupling types, such as imbalance and decline or coordinated development, can be determined by combining the range of this coupling coordination degree. This quantifies the overall coordination level of the three-dimensional spaces. The coupling coordination degree, coupling stage, and coupling type can be defined as coupling coordination characteristics. The coupling degree of the first coupling stage is less than that of the second coupling stage.

[0073] By combining multi-dimensional information, we can comprehensively depict the static attributes, dynamic evolution trends, and functional coordination status of the Three-Life Space in each historical year, providing complete information input for subsequent deep learning models to learn evolutionary patterns.

[0074] Step 103: Train the preset two-layer deep learning model based on spatial vector data and multi-dimensional auxiliary features to obtain the three-life space evolution simulation model.

[0075] In this embodiment, a preset two-layer deep learning model can be trained based on spatial vector data and multi-dimensional auxiliary features. This preset two-layer deep learning model can include a bottom-level spatial feature extraction sub-model and a top-level temporal evolution sub-model. The bottom-level spatial feature extraction sub-model can be used to extract features from the spatial vector data to obtain the spatial features of the three-dimensional space. These spatial features and multi-dimensional auxiliary features can be input into the top-level temporal evolution sub-model to capture the evolutionary dependencies of the three-dimensional space in the time dimension and output the corresponding temporal evolution features.

[0076] When using a trained model for prediction, spatial vector data and multi-dimensional auxiliary features of the target rural area can be input into the model. The model outputs spatial distribution information and temporal evolution information for future target years. Based on the spatial distribution information predicted by the model, the predicted areas of various spatial distributions and a preset evaluation index system can be further calculated to determine the coupling and coordination information for future target years. The evolution simulation results can be determined based on the spatial distribution information, temporal evolution information, and coupling and coordination information for multiple future years.

[0077] By fusing extracted spatial features with multidimensional auxiliary features to train the model, the trained model can accurately capture the nonlinear dependencies of the three-dimensional space in the spatiotemporal dimension, thus achieving accurate simulation of the future evolution process.

[0078] In one embodiment of this disclosure, the preprocessed data can be divided into a training set, a validation set, and a test set, which are used for model training, parameter tuning, and performance evaluation, respectively. The spatial extraction sub-model includes, but is not limited to, neural network structures with spatial feature extraction capabilities such as U-Net (Derived from U-shaped Convolutional Neural Network), FCN (Fully Convolutional Network), and SegNet (Derived from Semantic Segmentation Network). The underlying spatial feature extraction sub-model can employ a combination architecture of a bilateral deep convolutional network edge detector and a U-shaped network. During training, spatial vector data from each period can be transformed into corresponding feature matrices. The bilateral deep convolutional network edge detector extracts edge features of various spaces, and these features are fused with the deep semantic features extracted by the U-shaped network to obtain the spatial features of the three-dimensional space. These spatial features and multi-dimensional auxiliary features can be concatenated or fused and then input into the top-level temporal evolution sub-model. The temporal evolution prediction layer can employ a bidirectional LSTM (Long Short-Term Memory Network). This temporal evolution sub-model receives spatial features extracted from the lower layers and combines them with multi-dimensional auxiliary features determined through the aforementioned steps as joint input. Through the gating mechanism of the bidirectional LSTM, it simultaneously learns the forward and backward dependencies in the temporal dimension, thereby capturing the evolutionary dependencies of the three-life space in the temporal dimension and outputting the corresponding temporal evolution features. The weight parameters of the spatial extraction sub-model and the temporal evolution sub-model can be updated synchronously using the backpropagation algorithm until the model converges, resulting in a trained three-life space evolution simulation model.

[0079] During model training, a loss function can be constructed, and the parameters of the spatial extraction sub-model and the temporal evolution sub-model can be updated based on the loss function, thereby optimizing the prediction accuracy of spatial distribution information and temporal change information. This loss function can be composed of a weighted average of spatial loss and temporal loss, as shown in formula (1):

[0080] (1)

[0081] In formula (1), This represents the sum of spatial loss and temporal loss. Indicates space loss, Indicates timing loss, Spatial weights, For time series weights, such as =0.6、 =0.4, the specific value can be determined based on the actual situation. As an example, spatial loss can be obtained by comparing the spatial distribution information of the current three-life space obtained by the underlying spatial extraction sub-model based on the spatial features with the actual spatial distribution information of the three-life space in the current spatial vector data. For example, it can be determined based on category-balanced cross-entropy to ensure the accuracy of spatial distribution information extraction. Temporal loss can be obtained by comparing the spatial distribution information of the future three-life space predicted by the top-level temporal evolution sub-model based on the spatial features and multi-dimensional auxiliary features with the corresponding actual spatial distribution information of the three-life space in the spatial vector data. For example, it can be determined using mean squared error loss to ensure the balance between spatial prediction accuracy and temporal consistency. Among them, spatial distribution information can be used to characterize the attribute status or probability value of each basic spatial unit in the target rural area, such as the raster pixel or vector plot belonging to the categories of production space, living space and ecological space, respectively. Specifically, the spatial distribution information can be represented as a multi-channel probability distribution matrix, that is, recording the confidence or type label of each location belonging to each type of space. The geographical range and regional location of each type of space can be determined through this probability distribution matrix. Spatial and temporal losses can be normalized before weighting to eliminate the impact of dimensional differences on parameter updates.

[0082] During the training iterations of the three-life spatial evolution simulation model, the spatial vector data and multi-dimensional auxiliary features of the training set are input into the model. The network parameters are updated through the backpropagation algorithm, gradually reducing the sum of spatial loss and temporal loss. After a preset number of iterations, such as 50 iterations, the model performance can be evaluated using a validation set. If the prediction accuracy of spatial distribution information is greater than or equal to a preset accuracy threshold, such as 90%, and the prediction error of temporal evolution information is less than or equal to a preset error threshold, such as 5%, the parameters of the current three-life spatial evolution simulation model are recorded. If the validation set loss does not decrease for 10 consecutive iterations, an early stopping mechanism is triggered to avoid model overfitting.

[0083] After the simulation model of the evolution of the three-life space is trained, its generalization ability can be verified through a test set. If the average deviation between the spatial distribution information output by the model and the actual spatial distribution information is less than or equal to a preset error threshold, such as 5%, and the consistency between the temporal evolution information and the actual evolution information is greater than or equal to a preset consistency threshold, such as 85%, then the model is considered to have passed the training. The finally trained simulation model of the three-life space evolution can directly and automatically output the spatial distribution information and temporal evolution information of the three-life space based on the input spatial vector data, providing a quantitative decision-making basis for the planning and management of rural three-life spaces.

[0084] In one embodiment of this disclosure, spatial vector data of the three-dimensional space corresponding to each historical year is determined based on basic data of the target rural area in multiple historical years, including:

[0085] For each historical year, at least one of the following preprocessing operations is performed on the remote sensing image data of the target rural area in that historical year: radiometric normalization, geometric correction, coordinate transformation, and boundary clipping, to obtain preprocessed remote sensing image data. Based on the land use vector data of the target rural area in the historical year and the preprocessed remote sensing image data, the target rural area is divided according to each land use type to obtain the regional boundaries corresponding to each type of space in the three-dimensional space. Socioeconomic information corresponding to each type of space is determined based on the socioeconomic statistics of the target rural area in the historical year; wherein, socioeconomic information includes at least one of economic output, economic investment, and population information. Spatial vector data is determined based on the regional boundaries corresponding to each type of space of the target rural area in each historical year and the socioeconomic information corresponding to each type of space.

[0086] In this embodiment, for each historical year, at least one of the following preprocessing operations can be performed on the remote sensing image data of that historical year: radiometric normalization, geometric correction, coordinate transformation, and boundary clipping, to obtain preprocessed remote sensing image data. As an example, if there are inconsistencies in brightness in remote sensing image data due to differences in shooting time and sensors, radiometric normalization can be performed. For instance, typical features evenly distributed in the image, such as large areas of farmland or flat roofs, can be selected as references. The average brightness of these features in different bands can be calculated, and the brightness values ​​of each pixel can be unified to the same range through linear adjustment. This ensures that the brightness fluctuations of similar features are controlled within a preset range, such as a standard deviation ≤ 5, thus ensuring the consistency of the reflectivity characteristics of the features. If there are deformations in the remote sensing image data, such as a straight road appearing curved in the image, geometric correction can be performed to restore the deformed features to their true shape, ensuring that linear features such as roads and rivers remain continuous and without misalignment in the image. If the coordinate system used in the remote sensing image data is inconsistent with the research standard, coordinate transformation can be performed. For instance, the coordinates of the remote sensing image data can be uniformly converted to the target coordinate system, such as the Gauss-Kruger Zone. 18 (Gauss-Kruger projection zone 18) ensures that the coordinates of the same feature correspond completely in images from different years, avoiding the problem of different coordinates of the same location in different remote sensing image data; if the image range of the remote sensing image data is larger than the target rural area, boundary cropping can be performed. The administrative boundary of the target rural area can be used as a mask to retain the image data within the research range, reducing the amount of data for subsequent processing.

[0087] In this embodiment, after obtaining the preprocessed remote sensing image data, the regional boundaries corresponding to each land use type in the three-dimensional space can be determined by combining the land use vector data of the target rural area in the corresponding historical year. The land use vector data can include existing land ownership boundaries and land use type labels for that year, such as cultivated land, construction land, and forest land. By spatially overlaying the preprocessed remote sensing image data with the land use vector data, the regions corresponding to each land use type in the remote sensing image data can be extracted. Based on the spatial type corresponding to each land use type, the target rural area can be divided to obtain the regional boundaries corresponding to each type of space. The production space can include land use types such as farmland, factories, greenhouses, and roads outside the residential area; the living space can include land use types such as residences and roads within the residential area; and the ecological space can include land use types such as forest land, grassland shrubs, rivers, and lakes. As an example, deep learning segmentation algorithms, such as bidirectional convolutional neural network edge detector algorithms, U-shaped network algorithms, and multi-task road extractor algorithms, can be used to divide the target rural area into spatial types, generate preliminary spatial area boundaries, and then compare them with the plot boundaries in the land use vector data to correct the deviations and obtain the area boundaries of various spatial types.

[0088] In this embodiment, after determining the regional boundaries of various spaces, the socio-economic information corresponding to each type of space can be determined by combining the socio-economic statistics of the target rural area for that historical year. These socio-economic statistics include township-level population, total GDP, and fixed asset investment, while the socio-economic information can include at least one of economic output, economic investment, and population information. For example, the economic output of a production space can be determined based on farmland grain yield and factory industrial output; economic investment in a production space can be determined based on productive investment projects such as factory construction and greenhouse renovation. Population information for a living space can be estimated based on the ratio of residential land area to per capita living area; economic investment in a living space can be determined based on living-related investments such as road paving and water supply facility renovation within the corresponding village. The economic value of an ecological space can be determined based on data such as the carbon sequestration value of forest land. Through this spatial allocation method, clear socio-economic information corresponds to the regional boundaries of each type of space.

[0089] In this embodiment, the regional boundaries corresponding to various spaces in the target rural area for each historical year, the regional boundaries corresponding to various land use types within each space, and the socio-economic information corresponding to each space can be combined to obtain spatial vector data for each historical year. This spatial vector data not only contains the geometric morphological information of the three spaces (production, living, and ecological spaces), but also associates the regional boundaries of various features within each space with the corresponding socio-economic information. It can fully reflect the distribution and functional characteristics of production, living, and ecological spaces within the target rural area in that historical year, providing standardized basic data for subsequent extraction of spatial vector information and training of the three-life space evolution simulation model.

[0090] In one embodiment of this disclosure, the spatial vector information of the three-life space is determined based on the spatial vector data corresponding to each historical year, including:

[0091] For each historical year, the regional boundaries and areas of various spaces are determined based on the spatial vector data corresponding to that historical year.

[0092] The area and morphological characteristics of each type of space are determined based on the perimeter and area of ​​the corresponding regional boundaries.

[0093] The temporal characteristics of each type of space are determined based on its corresponding area.

[0094] Based on the values ​​of the preset evaluation indicators corresponding to various types of spaces, the coupling and coordination characteristics of the three-life spaces are determined.

[0095] In this embodiment, data on the boundaries of various spaces, such as production space, living space, and ecological space, can be extracted from the spatial vector data corresponding to each historical year. This includes the latitude and longitude or Cartesian coordinates of each vertex. Using the polygon area calculation formula, the area of ​​this type of space in the current historical year can be obtained based on the position of each boundary vertex. The unit is uniformly set to square meters or acres to ensure dimensional consistency in subsequent feature calculations.

[0096] In this embodiment, after obtaining the regional boundaries and areas of various spaces, the area characteristics and morphological characteristics of each type of space are determined based on the perimeter and area of ​​its regional boundaries. Specifically, the area characteristics of each type of space can be determined based on the annual rate of change of its area between the current and previous historical years, and the shape characteristics can be determined based on the ratio of the perimeter of the regional boundary to the square root of its area.

[0097] In this embodiment, multiple historical years can be selected, such as the spatial vector data corresponding to period t and period t+1, respectively. The regional boundaries of the same type of space corresponding to two adjacent periods, such as period t and period t+1, are overlaid and analyzed to identify the transfer areas in period t where production space, living space, and ecological space are transformed into other spatial types in period t+1. The area of ​​the transfer area where each type of space in period t is transformed into other spaces in period t+1 can be determined. The spatial transfer probability matrix can be determined by the area of ​​each transfer area in each historical year, and this spatial transfer probability matrix can be determined as a temporal feature.

[0098] In this embodiment, preset evaluation indicators for various types of spaces can be determined based on the socio-economic information in the spatial vector data. These preset evaluation indicators can be set differently according to the type of each space. For example, the preset evaluation indicators for production spaces can be configured as output value per unit area, production investment intensity, etc., the preset evaluation indicators for living spaces can be configured as per capita living area, infrastructure coverage rate, etc., and the preset evaluation indicators for ecological spaces can be configured as vegetation coverage rate, ecological protection investment, etc. The specific configuration of the preset evaluation indicators can be determined based on the actual situation.

[0099] The values ​​of preset evaluation indicators corresponding to various spaces can be standardized to obtain standard values ​​for each preset evaluation indicator. Based on the values ​​of the preset evaluation indicators corresponding to various spaces in each period and the standard values, the coupling and coordination characteristics of all spaces in that period, i.e., the three-dimensional spaces, can be determined. These coupling and coordination characteristics can include coupling stages and coupling types.

[0100] In this embodiment, each historical year can be classified into production space, living space, and ecological space. The area characteristics, temporal characteristics, morphological characteristics, and coupling and coordination characteristics of each type of space in the target rural area in each historical year can be sorted out, and the feature vectors corresponding to each feature can be determined. These feature vectors are then integrated to obtain a complete feature matrix. This feature matrix contains comprehensive information on the scale, morphology, dynamic changes, economic attributes, and overall coordination level of the three spaces in the current historical year. It is the multi-dimensional auxiliary feature corresponding to the current historical year and can be used as the training input for the subsequent two-layer deep learning model to support the model's learning and prediction of the evolution law of the three spaces.

[0101] In one embodiment of this disclosure, the area and morphological characteristics of various spaces are determined based on the perimeter and area of ​​the region boundaries corresponding to each type of space, including:

[0102] For each type of space, the area characteristics of the space are determined based on the annual rate of change of the space's area.

[0103] The ratio between the perimeter and the square root of the area corresponding to the boundary of the region in space is determined, and the product of the ratio with the first proportional threshold is determined as the morphological feature corresponding to the space.

[0104] In this embodiment, for each type of space, the annual rate of change of the area of ​​that type of space can be determined based on formula (2), and the annual rate of change of the area of ​​that type of space can be determined as the area characteristic of that type of space. This area characteristic can be used to reflect the expansion or contraction trend of that type of space in terms of scale.

[0105] (2)

[0106] In the above formula (2), K This represents the annual rate of change of the area of ​​this type of space. This represents the area of ​​this type of space in the current year of calculation within a multi-period historical timeframe. This represents the area of ​​this type of space corresponding to the year preceding the current year in a multi-period historical calculation. This represents the year value corresponding to the current year being calculated. This represents the year value corresponding to the year preceding the year being calculated.

[0107] In this embodiment, for each type of space, the morphological characteristics of that type of space can be determined based on formula (3). These morphological characteristics can be used to reflect the complexity of the space in terms of its shape.

[0108] (3)

[0109] In the above formula (3), SI This represents the shape index corresponding to the space. This represents the first proportional threshold, which can be determined based on the actual situation; for example, it can be set to 0.25. P This represents the perimeter of the region boundary in the space. A This indicates the area occupied by the boundary of the region in the space.

[0110] Area characteristics can quantify the dynamic changes in spatial scale, while morphological characteristics can quantify the regularity of spatial form. Combining the two can comprehensively describe the static attributes and dynamic trends of this type of space in geographical distribution, providing basic characteristic support for the subsequent analysis of the evolutionary laws of three-dimensional spaces.

[0111] In one embodiment of this disclosure, determining the temporal characteristics of various spaces based on their corresponding areas includes:

[0112] For each historical year, the first region boundary corresponding to each space in the historical year is determined, as well as the second region boundary corresponding to each space in the following year. For each first space, the transfer area from the first space to the second space is determined based on the first region boundary corresponding to the first space and the second region boundary corresponding to the second space in the following year. The first space is any type of space among the three types of spaces in the historical year, and the second space is each type of space among the three types of spaces in the following year. The transfer probability from the first space to the second space is determined based on the ratio of the transfer area from the first space to the second space in the historical year to the total area corresponding to the first space in the historical year. The temporal characteristics are determined based on the transfer probabilities corresponding to each space.

[0113] In this embodiment, for each historical year, this historical year is used as the base period. Based on the spatial vector data corresponding to this historical year, the regional boundaries of production space, living space, and ecological space can be extracted and determined as the first regional boundary. Each regional boundary in the first regional boundary contains a sequence of coordinate points in a Cartesian coordinate system. The geographical scope of each space can be accurately defined through this coordinate point sequence. The following year can be used as the comparison period. From the spatial vector data of this year, the spatial regional boundaries of each type of space—namely, the regional boundaries of production space, living space, and ecological space—are extracted and uniformly determined as the second regional boundary.

[0114] After obtaining the boundaries of each first and second region, for each type of space, the area enclosed by each first region boundary and the area enclosed by each second region boundary can be determined. Based on the differences between the first and second region boundaries corresponding to the same space within each space, the change in the first region boundary corresponding to that space can be determined. If the change is positive, it indicates that the area of ​​the space expanded from the baseline period to the comparison period; if the change is negative, it indicates that the area of ​​the space contracted; if the change is 0, it indicates that the area of ​​the space did not change. By calculating the change, the trend of scale change of various types of spaces between adjacent years can be intuitively grasped, providing a reference for the subsequent determination of the transferred area.

[0115] The first space can be defined as any category of the three-life space in the current historical year, and the second space can be defined as the space of each category of the three-life space in the next year. If the first space is a production space, then the second space can be a production space, a living space, or an ecological space.

[0116] Taking the first space as the production space as an example, we can determine the overlapping parts of the first area boundary corresponding to the production space and the second area boundary corresponding to the production space, living space, and ecological space, respectively, and determine the overlapping boundary; then we can calculate the transfer area of ​​the production space retained as the production space, that is, the area that remains unchanged, the transfer area of ​​the production space converted into the living space, and the transfer area of ​​the production space converted into the ecological space.

[0117] Assume the total area of ​​production space in this historical year is calculated to be 100 mu. Calculations show that 15 mu is transferred from production space to living space, and 10 mu is transferred from production space to ecological space. Based on the total area correspondence, the area that remains unchanged after conversion from production space to production space is 75 mu. The above calculation results can then be logically verified by considering the change in area of ​​the first boundary corresponding to the production space. For example, if the boundary calculation shows that the change in area of ​​production space between this historical year and the next year is +20 mu, meaning the total area of ​​production space in the next year is 120 mu, according to the logic that the total area in the next year = retained area + transferred area, since the retained area of ​​production space in the above calculation is 75 mu, there must be a transferred area of ​​45 mu from other spaces into production space. If the sum of the transferred areas of other spaces, such as living space and ecological space, converted to production space is indeed 45 mu, then the above transfer area and probability calculations are directly confirmed to be accurate. If the transferred area does not match 45 mu, then the overlapping area calculation can be rechecked for errors, corrected, and verified again until the logic is consistent.

[0118] In the same manner, when the first space is a living space, determine the transfer area from living space to production space and from living space to ecological space; when the first space is an ecological space, determine the transfer area from ecological space to production space and from ecological space to living space.

[0119] After determining the transfer area of ​​all first spaces to second spaces, the transfer probability between the first space and each second space can be determined based on formula (4).

[0120] (4)

[0121] In the above formula (4), Indicates the first t The spatial type in the period is u Space transformation to the first t+1 The spatial type in the period is v The spatial transition probability, Indicates the first t The spatial type in the period is u Space transformation to the first t+1 The period space type is v The area of ​​spatial transfer, Indicates the first t The spatial type in the period is u The space is converted into various space types. v The sum of the spatial transfer areas, i.e., the spatial type is u The sum of the areas of all destinations within that year is equivalent to the area of ​​the first year. t The spatial type in the period is u The total area of ​​the space.

[0122] As an example, taking the first space as the production space, assume that in year t, the total area of ​​the production space is 100 mu. Among them, the area transferred to living space is 15 mu, the area transferred to ecological space is 5 mu, and the remaining 80 mu remains unchanged as production space. Then, the probability of transfer between production space and living space is 15 / 100=0.15, the probability of transfer between production space and ecological space is 5 / 100=0.05, and the probability of transfer between production spaces, that is, the probability of remaining unchanged, is 80 / 100=0.80. The calculation of the transfer probability of other first spaces is similar, and will not be elaborated in this disclosure.

[0123] In this embodiment, the transition probability matrix of the three-generation space can be determined based on the transition probability corresponding to each space, and the transition probability matrix can be determined as a temporal feature.

[0124] By determining the temporal features based on the transfer area between different spaces, the conversion probability between different spatial types can be intuitively displayed through the corresponding transfer probability matrix. This provides comprehensive and accurate temporal input features for the deep learning model of rural three-life space evolution simulation, helping the model to capture the patterns and dynamics of spatial type conversion.

[0125] In one embodiment of this disclosure, the coupling and coordination characteristics of the three-dimensional spaces are determined based on the values ​​of preset evaluation indicators corresponding to various types of spaces, including:

[0126] For each preset evaluation indicator corresponding to each type of space, the target weight of the preset evaluation indicator is determined. The functional score for each space is determined by the weighted sum of the target weights and the values ​​of the preset evaluation indicators. The coupling coordination degree between all spaces is determined based on the functional scores of each type of space. The corresponding coupling stages and coupling types are determined based on the coupling coordination degree. The coupling stages include the first coupling stage, the antagonistic stage, the adjustment stage, and the second coupling stage. The coupling types include the imbalance and decline type and the coordinated development type. The coupling degree in the first coupling stage is less than that in the antagonistic stage, and the coupling degree in the second coupling stage is greater than that in the adjustment stage. The coupling coordination characteristics of the three-dimensional spaces are determined based on the coupling coordination degree, coupling stages, and coupling types among all spaces.

[0127] In this embodiment, for each preset evaluation index corresponding to each type of space, the target weight of the preset evaluation index can be determined, and then the functional score corresponding to each type of space can be determined according to formula (5).

[0128] (5)

[0129] In the above formula (5), F This indicates the functional score corresponding to this type of space. m This indicates the number of preset evaluation indicators in this type of space. Indicates the first j The values ​​of the preset evaluation indicators Indicates the first j The target weights of each preset evaluation indicator.

[0130] In this embodiment, the coupling coordination degree between all spaces can be determined based on the functional scores corresponding to each type of space. This coupling coordination degree can be used to represent the intensity of interaction and the level of coordinated development among production space, living space, and ecological space during the functional operation process.

[0131] In this embodiment, the coupling stage may include a first coupling stage, an antagonistic stage, a break-in stage, and a second coupling stage. A corresponding preset range can be set for each stage. If the coupling coordination degree of the target rural area is within the first preset range, the coupling stage of the target rural area can be determined as the first coupling stage; if the coupling coordination degree of the target rural area is within the second preset range, the coupling stage of the target rural area can be determined as the antagonistic stage; if the coupling coordination degree of the target rural area is within the third preset range, the coupling stage of the target rural area can be determined as the break-in stage; and if the coupling coordination degree of the target rural area is within the fourth preset range, the coupling stage of the target rural area can be determined as the second coupling stage. The first preset range can be greater than or equal to 0 and less than 0.3; the second preset range can be greater than or equal to 0.3 and less than 0.5; the third preset range can be greater than or equal to 0.5 and less than 0.8; and the fourth preset range can be greater than or equal to 0.8 and less than 1. The specific preset ranges can be determined based on actual conditions, and this disclosure does not limit them.

[0132] In this embodiment, the coupling type can include a misalignment / decline type and a coordinated development type. A corresponding coordination degree threshold can be set. If the coupling coordination degree is greater than or equal to the coordination degree threshold, the coupling type is determined to be a coordinated development type; if the coupling coordination degree is less than the coordination degree threshold, the coupling type is determined to be a misalignment / decline type. The coordination degree threshold can be determined according to actual conditions, such as setting it to 0.5. Furthermore, the misalignment / decline type and the coordinated development type can be further classified into different levels. For example, the misalignment / decline type can be divided into severe misalignment / decline, mild misalignment / decline, and near-misalignment / decline; the coordinated development type can be divided into barely coordinated development, moderate coordinated development, good coordinated development, and excellent coordinated development. As an example, if the coupling coordination degree of the target rural area is greater than or equal to 0 and less than 0.3, the coupling type of the target rural area is determined to be severely disordered and declining; if the coupling coordination degree of the target rural area is greater than or equal to 0.3 and less than 0.4, the coupling type of the target rural area is determined to be mildly disordered and declining; if the coupling coordination degree of the target rural area is greater than or equal to 0.4 and less than 0.5, the coupling type of the target rural area is determined to be on the verge of disorder and decline; if the coupling coordination degree of the target rural area is greater than or equal to 0.5 and less than 0.6, the coupling type of the target rural area is determined to be barely coordinated and developing; if the coupling coordination degree of the target rural area is greater than or equal to 0.6 and less than 0.7, the coupling type of the target rural area is determined to be moderately coordinated and developing; if the coupling coordination degree of the target rural area is greater than or equal to 0.7 and less than 0.8, the coupling type of the target rural area is determined to be good coordinated and developing; if the coupling coordination degree of the target rural area is greater than or equal to 0.8 and less than 1, the coupling type of the target rural area is determined to be excellent coordinated and developing.

[0133] By quantifying the importance of each preset evaluation index by target weight, then using weighted calculation to determine the functional score corresponding to the space, and finally combining the stages and types of coupling coordination degree, the evaluation results of the coupling coordination degree of the three-life space can be more accurate.

[0134] In one embodiment of this disclosure, the coupling coordination degree between all spaces is determined based on the values ​​of preset evaluation indicators corresponding to various types of spaces, including:

[0135] The weighted sum of the functional scores corresponding to each type of space is determined as the comprehensive coordination index; the first product value is determined by multiplying the functional scores of each type of space; the second product value is determined by multiplying the sum of the functional scores of every two types of spaces in each type of space; the coupling degree is determined by the first product value and the second product value; and the square root of the product of the comprehensive coordination index and the coupling degree is determined as the coupling coordination degree.

[0136] As an example, the comprehensive coordination index of the three-life space corresponding to the target rural area can be determined according to formula (6).

[0137] (6)

[0138] In the above formula (6), , and These represent the functional scores corresponding to production space, living space, and ecological space, respectively. , , These represent the weights corresponding to production space, living space, and ecological space, respectively, with T representing the comprehensive coordination index. , , This can be determined based on the actual situation; this disclosure does not impose any limitations. It can be 0.33, It can be 0.33, The value can be 0.34, with the sum of the weights for the production space, living space, and ecological space being 1. By assigning corresponding weights to the production space, living space, and ecological space to calculate the comprehensive coordination index, we can avoid the evaluation being dominated by a single space score, and adjust the weights according to actual needs, accurately reflecting the synergistic index of the overall functions of the three types of spaces.

[0139] The coupling degree of the three-life space corresponding to the target rural area can be determined according to formula (7).

[0140] (7)

[0141] In the above formula (7), , and These represent the functional scores corresponding to production space, living space, and ecological space, respectively. The corresponding product value can represent the first product value. The corresponding product value can represent the second product value, and C represents the coupling degree of the target rural area. By quantifying the coupling degree of the three types of spaces, the interdependence of the three types of spaces can be accurately captured, avoiding the limitation of traditional methods that ignore the correlation between spaces.

[0142] In this embodiment, the coupling coordination degree corresponding to the target rural area can be determined according to formula (8).

[0143] (8)

[0144] In the above formula (8), D represents the coupling coordination degree of the three-life space corresponding to the target rural area, C represents the coupling degree of the three-life space corresponding to the target rural area, and T represents the comprehensive coordination index of the three-life space corresponding to the target rural area. By integrating the overall level corresponding to the comprehensive coordination index with the interaction intensity corresponding to the coupling degree, it is possible to consider both the overall functional level of the three types of spaces and the interaction intensity between them, so that the result of the coupling coordination degree is more in line with the actual needs of the coordinated development of the three-life space.

[0145] In one embodiment of this disclosure, for each preset evaluation index corresponding to each type of space, determining the target weight of the preset evaluation index includes:

[0146] For each preset evaluation index corresponding to each type of space, determine the standard value of the preset evaluation index in each historical year.

[0147] The weight of the indicator for each historical year is determined by the ratio between the standard value corresponding to each historical year and the sum of the standard values ​​corresponding to the preset evaluation indicator for all historical years. The entropy redundancy of the preset evaluation indicator is determined based on the weight of each indicator. The first weight of the preset evaluation indicator is determined based on the ratio between the entropy redundancy of the preset evaluation indicator and the sum of the entropy redundancy of all preset evaluation indicators corresponding to this type of space.

[0148] The absolute values ​​of the differences between the standard values ​​and preset reference values ​​corresponding to the preset evaluation index in all historical years are determined. The maximum value among these absolute values ​​is determined as the first absolute value, and the minimum value among these absolute values ​​is determined as the second absolute value. The correlation coefficients corresponding to the preset evaluation index are determined based on the absolute values ​​of the differences between the standard values ​​and preset reference values ​​corresponding to the preset evaluation index in all historical years, the first absolute value, and the second absolute value. The correlation degree of the preset evaluation index is determined based on the correlation coefficients corresponding to the preset evaluation index in each historical year. The second weight of the preset evaluation index is determined based on the ratio between the correlation degree of the preset evaluation index and the sum of the correlation degrees of all preset evaluation indices corresponding to this type of space.

[0149] The target weight of the preset evaluation index is determined based on the weighted sum of the first weight and the second weight.

[0150] In this embodiment, the values ​​of preset evaluation indicators corresponding to various types of spaces in each historical year can be standardized to obtain standard values ​​for each preset evaluation indicator. For example, the maximum and minimum values ​​of each preset evaluation indicator in each historical year can be determined. If the preset evaluation indicator is a positive indicator, such as output per unit area, the first difference between the value of the preset evaluation indicator and the minimum value can be determined, and the second difference between the maximum and the minimum value can be determined. The ratio of the first difference to the second difference is determined as the standardized standard value of the preset evaluation indicator. If the preset evaluation indicator is a negative indicator, such as per capita living area, the third difference between the maximum value and the value of the preset indicator can be determined, and the ratio of the third difference to the second difference is determined as the standardized standard value of the preset evaluation indicator. Specifically, the positive indicator value is positively correlated with the quality of spatial function; that is, the higher the value of the positive indicator, the better the corresponding spatial function. The negative indicator value is negatively correlated with the quality of spatial function; that is, the lower the value of the negative indicator, the better the corresponding spatial function. By standardizing the data to distinguish between positive and negative indicators, we can avoid bias in the original data.

[0151] In this embodiment, the first weight of each preset evaluation index corresponding to each type of space can be determined according to formulas (9)-(12).

[0152] (9)

[0153] (10)

[0154] (11)

[0155] (12)

[0156] In the above formula (9), Indicates the first i The first in the historical years j The standard values ​​corresponding to each preset evaluation indicator; This represents the first period corresponding to all historical years. j The sum of the standard values ​​of each preset evaluation indicator; Indicates the first i The first in the historical years j The weight of each preset evaluation indicator. Indicates the ordinal index of the year. n This indicates the number of historical years.

[0157] In the above formula (10), n Indicates the number of historical years. This represents the information entropy corresponding to the j-th preset evaluation index. Represents the natural logarithm function. i The index number representing the year. Indicates the first i The first in the historical years j The weight of each preset evaluation indicator.

[0158] In the above formula (11), Indicates the first j The entropy redundancy corresponding to each preset evaluation index Indicates the first j Information entropy corresponding to each preset evaluation indicator.

[0159] In the above formula (12), Indicates the first j The first weight of each preset evaluation indicator This represents the entropy redundancy corresponding to the j-th preset evaluation index. m This indicates the number of preset evaluation indicators.

[0160] By calculating the information entropy and entropy redundancy of the preset evaluation indicators, the dispersion of the preset evaluation indicators in multiple historical data periods can be objectively quantified, avoiding interference from subjective judgment and accurately determining the preset evaluation indicators with high spatial evaluation discrimination.

[0161] In this embodiment, the second weight of each preset evaluation index corresponding to each type of space can be determined according to formulas (13)-(15).

[0162] (13)

[0163] (14)

[0164] (15)

[0165] In the above formula (13), Indicates the first j The pre-set evaluation indicators in the first i The correlation coefficients corresponding to historical years. Indicates the first j The pre-set evaluation indicators in the first i The absolute value of the difference between the standard value and the preset reference value for a given historical year; It represents the minimum absolute value among the differences between the standard value and the preset reference value for each corresponding standard value of the preset evaluation index in all historical years, and can also be represented as the second absolute value; It represents the maximum absolute value among the differences between the standard value and the preset reference value for each corresponding standard value of the preset evaluation index in all historical years, and can also be represented as the first absolute value; Represents the resolution coefficient. The preset reference values ​​for each preset evaluation index can be determined based on the average value of the corresponding values ​​for each historical year or based on the ideal expected value. The specific values ​​can be determined according to the actual situation, and this application does not impose any special restrictions.

[0166] In the above formula (14), Indicates the first j The correlation degree corresponding to each preset evaluation indicator n Indicates the number of historical years. i The index number representing the year.

[0167] In the above formula (15), Indicates the first j The second weight of each preset evaluation indicator This represents the correlation degree corresponding to the value of the j-th preset evaluation index. m This indicates the number of preset evaluation indicators. j This indicates the index number of the preset evaluation indicators.

[0168] By calculating correlation coefficients, correlation degrees, etc., the calculated weights can be made to better reflect actual functional requirements.

[0169] In this embodiment, the target weight of the preset evaluation index can be determined by the weighted sum of the first weight and the second weight. The weights corresponding to the first weight and the second weight can be determined according to the actual situation, and this disclosure does not limit them. By fusing the two weights, the one-sidedness of traditional single weighting or subjective weighting can be avoided, making the calculated target weight more in line with the actual contribution.

[0170] like Figure 2 As shown, Figure 2 This disclosure is a flowchart illustrating a deep learning-based method for simulating the evolution of rural three-dimensional space according to an exemplary embodiment, comprising the following steps:

[0171] Step 201: Determine the spatial vector data of the three-life space corresponding to each period based on the basic data of the target rural area for at least one period before the year of evolution.

[0172] In this embodiment, the starting year of the evolution can be determined, i.e., the starting year predicted in this simulation. If the predicted starting year is the current year, basic data from at least one historical year prior to the starting year can be retrieved. This basic data may include remote sensing image data, land use vector data, and socio-economic statistics for each year prior to the starting year. The spatial types of the three-dimensional space can include ecological space, production space, and living space. As an example, the remote sensing image data for each period can be preprocessed, including projecting the remote sensing images onto a preset coordinate system, performing radiometric normalization, geometric correction, and resolution unification. Then, the preprocessed remote sensing images are cropped according to the geographical extent of the target rural area to obtain remote sensing image data covering the target rural area.

[0173] By combining remote sensing imagery data, land use vector data, and socio-economic statistics from different years, we can identify areas that belong to production space, living space, and ecological space. By extracting information such as the boundary coordinates and shape parameters of these areas, we can convert them into vector format and finally form spatial vector data of the three spaces corresponding to each year. This data can accurately reflect the spatial distribution information of each type of space in each year.

[0174] The spatial vector data determined through the above method integrates the information advantages of multi-source data, improves the accuracy of spatial type identification, and provides reliable basic data support for obtaining the evolution simulation results of the three-life space using the trained three-life space evolution simulation model.

[0175] Step 202: Input the spatial vector data and the corresponding multidimensional auxiliary features into the three-life space evolution simulation model, and determine the evolution simulation results of the three-life space in the target rural area based on the prediction results of the three-life space evolution simulation model.

[0176] In this embodiment, based on the aforementioned method for determining multidimensional auxiliary features, the multidimensional auxiliary features of the three-life space can be determined according to the spatial vector data of each period in at least one period before the evolution start year. The spatial vector data and the corresponding multidimensional auxiliary features can be input into the trained three-life space evolution simulation model. Through the hierarchical calculation and analysis of the model, the spatial distribution information of the three-life space in the target rural area from the evolution start year and the preset years thereafter, such as 5-10 years, can be predicted. The spatial distribution information includes the geographical range and regional location of ecological space, production space, and living space in each year, which can intuitively reflect the future distribution changes of various types of space. On the other hand, the corresponding temporal evolution information can also be predicted, specifically including the probability of three-life space type transfer, the morphological evolution trend of each space type, and other information, to determine the dynamic change law of space over time.

[0177] After predicting the spatial distribution information and temporal evolution information, the coupling coordination information of the three spaces can be further calculated. For example, firstly, the functional scores of various types of spaces in each year are determined according to the preset evaluation indicators; then, the coupling coordination degree is obtained by calculating the coupling degree and the comprehensive coordination index; finally, the corresponding coupling stage and coupling type are determined according to the numerical range of the coupling coordination degree, and the complete coupling coordination information is obtained. The specific method for determining the coupling coordination information is the same as the corresponding method in the above embodiment.

[0178] In this embodiment, the evolution simulation results can be determined by spatial distribution information, temporal evolution information, and coupling coordination information. The evolution simulation results can include dynamic evolution maps of the three-dimensional space, such as heat maps of transition probabilities, trend curves of key indicators, such as line graphs of area changes, and evaluation reports of coupling coordination, so as to realize the visualization of the evolution simulation process.

[0179] Through the above process, the trained simulation model of the three-life space evolution can output prediction results from three dimensions: space, time, and evaluation. It not only achieves accurate prediction of the future layout of the three-life space, but also quantifies the spatial evolution law and coordinated development status. Compared with traditional single-dimensional prediction methods, the results are more comprehensive and reliable, and can provide a full-chain decision-making basis for the future layout of rural spatial planning, thereby improving the scientificity and foresight of the planning scheme.

[0180] Corresponding to the embodiments of the foregoing methods, this disclosure also provides embodiments of the apparatus and the terminal to which it is applied.

[0181] like Figure 3 As shown, Figure 3This is a block diagram of a training apparatus for simulating the evolution of rural three-dimensional space based on deep learning, according to an exemplary embodiment of this disclosure. The apparatus includes: a first determining module 310, a second determining module 320, and a model training module 330.

[0182] The first determining module 310 is used to determine the spatial vector data of the three-life space corresponding to each historical year based on the basic data of the target rural area in multiple historical years; wherein, the basic data includes remote sensing image data, land use vector data and socio-economic statistics, and the three-life space includes ecological space, production space and living space.

[0183] The second determining module 320 is used to determine the multi-dimensional auxiliary features of the three-life space based on the spatial vector data of the three-life space corresponding to each historical year.

[0184] The model training module 330 is used to train a preset two-layer deep learning model based on spatial vector data and multi-dimensional auxiliary features to obtain a three-life space evolution simulation model, so as to use the three-life space evolution simulation model to predict the evolution simulation results of the three-life space in the target rural area.

[0185] It should be noted that the training device for the deep learning-based rural three-life space evolution simulation in this embodiment is used to implement the corresponding deep learning-based rural three-life space evolution simulation training method in the aforementioned method embodiment, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0186] like Figure 4 As shown, Figure 4 This is a block diagram of a deep learning-based rural three-life space evolution simulation device according to an exemplary embodiment of the present disclosure. The device includes: a third determination module 410 and a prediction module 420.

[0187] The third determining module 410 is used to determine the spatial vector data of the three-life space corresponding to each period based on the basic data of the target rural area in at least one period before the year of evolution.

[0188] The prediction module 420 is used to input spatial vector data and the corresponding multidimensional auxiliary features into the three-life space evolution simulation model, and determine the evolution simulation results of the three-life space in the target rural area based on the prediction results of the three-life space evolution simulation model.

[0189] It should be noted that the deep learning-based rural three-life space evolution simulation device in this embodiment is used to implement the corresponding deep learning-based rural three-life space evolution simulation method in the aforementioned method embodiment, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0190] This disclosure discloses a training device or an embodiment of a deep learning-based rural three-life space evolution simulation device that can be applied to computer devices, such as servers or terminal devices. The device embodiment can be implemented in software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor reading the corresponding computer program instructions from non-volatile memory into memory and executing them. From a hardware perspective, such as... Figure 5 The diagram shown is a hardware structure diagram of a training device or a computer device for a deep learning-based rural three-life space evolution simulation according to an embodiment of this disclosure. Except for... Figure 5 In addition to the processor 510, memory 530, network interface 520, and non-volatile memory 540 shown, the server or electronic device on which the deep learning-based rural three-life space evolution simulation training device or the deep learning-based rural three-life space evolution simulation device is located in the embodiment may also include other hardware depending on the actual function of the computer device, which will not be described in detail here.

[0191] Accordingly, this disclosure also provides a training device or a simulation device for rural three-life space evolution based on deep learning, the device including a processor; a memory for storing processor-executable instructions; wherein the processor is configured to the above-mentioned training method or simulation method for rural three-life space evolution based on deep learning.

[0192] The specific implementation process of the functions and roles of each module in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0193] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0194] The foregoing has described specific embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0195] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention applied herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not claimed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0196] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

[0197] The above description is merely a preferred embodiment of this disclosure and is not intended to limit this disclosure. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A training method for simulating the evolution of rural three-life space based on deep learning, characterized in that, include: Based on the basic data of the target rural area in multiple historical years, determine the spatial vector data of the three spaces corresponding to each historical year; wherein, the basic data includes remote sensing image data, land use vector data and socio-economic statistics, and the three spaces include ecological space, production space and living space; The multidimensional auxiliary features of the three-life space are determined based on the spatial vector data of the three-life space corresponding to each historical year. Based on the spatial vector data and the multidimensional auxiliary features, a preset two-layer deep learning model is trained to obtain a three-life space evolution simulation model, so as to use the three-life space evolution simulation model to predict the evolution simulation results of the three-life space in the target rural area. The multidimensional auxiliary features include area features, temporal features, morphological features, and coupling coordination features; the determination of the multidimensional auxiliary features of the three-life space based on the spatial vector data of the three-life space corresponding to each historical year includes: For each historical year, the regional boundaries and areas of various spaces are determined based on the spatial vector data corresponding to that historical year; The area and morphological characteristics of each type of space are determined based on the perimeter and area of ​​the region boundary corresponding to each type of space. The temporal characteristics of each type of space are determined based on the area corresponding to each type of space. The coupling and coordination characteristics of the three-dimensional space are determined based on the values ​​of the preset evaluation indicators corresponding to each type of space.

2. The method according to claim 1, characterized in that, The two-layer deep learning model includes a spatial extraction sub-model and a temporal evolution sub-model; The step of training a pre-defined two-layer deep learning model based on the spatial vector data and the multi-dimensional auxiliary features includes: The spatial vector data is input into the spatial extraction sub-model to determine the spatial characteristics of the three-dimensional space; The spatial features and the multidimensional auxiliary features are input into the temporal evolution sub-model to determine the temporal evolution features of the three-generation space; A loss function is constructed based on the spatial features and the temporal evolution features, and the parameters of the spatial extraction sub-model and the temporal evolution sub-model are updated according to the loss function.

3. The method according to claim 1, characterized in that, The process of determining the spatial vector data of the three-dimensional space corresponding to each historical year based on the basic data of the target rural area in multiple historical years includes: For each historical year, at least one of the following preprocessing operations is performed on the remote sensing image data of the target rural area in that historical year: radiometric normalization, geometric correction, coordinate transformation, and boundary clipping, to obtain preprocessed remote sensing image data. Based on the land use vector data of the target rural area in the historical year and the preprocessed remote sensing image data, determine the regional boundaries corresponding to various spatial types in the target rural area; Based on the socioeconomic statistics of the target rural area in the historical year, determine the corresponding socioeconomic information for various spatial types; wherein, the socioeconomic information includes at least one of economic output, economic investment and population information; The spatial vector data is determined based on the regional boundaries corresponding to various spaces in the target rural area in different historical years and the corresponding socio-economic information.

4. The method according to claim 1, characterized in that, The process of determining the area and morphological characteristics of each type of space based on the perimeter and area of ​​the region boundaries corresponding to each type of space includes: For each type of space, the area characteristics of the space are determined based on the annual rate of change of the space's area; The ratio between the perimeter and the square root of the area corresponding to the boundary of the space is determined, and the product of the ratio and a first proportional threshold is determined as the morphological feature corresponding to the space.

5. The method according to claim 1, characterized in that, The determination of the temporal characteristics of various spaces based on their corresponding areas includes: For each historical year, determine the first region boundary corresponding to each space in the historical year, and determine the second region boundary corresponding to each space in the next year of the historical year. For each first space, the transfer area of ​​the first space to the second space is determined based on the boundary of the first region corresponding to the first space and the boundary of the second region corresponding to the second space in the next year; the first space is any type of space among the three types of spaces in the historical year, and the second space is each type of space among the three types of spaces in the next year of the historical year; The probability of the first space being converted into the second space is determined by the ratio of the area transferred from the first space to the second space in the historical year to the total area corresponding to the first space in the historical year. The temporal features are determined based on the transition probabilities corresponding to each space.

6. The method according to claim 1, characterized in that, The step of determining the coupling and coordination characteristics of the three-dimensional spaces based on the values ​​of preset evaluation indicators corresponding to various types of spaces includes: For each preset evaluation index corresponding to each type of space, determine the target weight of the preset evaluation index; The functional score for each space is determined by the weighted sum of the target weights and the values ​​of the preset evaluation indicators for each space. The degree of coupling and coordination between all spaces is determined based on the functional scores corresponding to each type of space. The corresponding coupling stages and coupling types are determined based on the coupling coordination degree; wherein, the coupling stages include a first coupling stage, an antagonistic stage, a break-in stage, and a second coupling stage, and the coupling types include a disordered decline type and a coordinated development type; The coupling coordination characteristics of the three-dimensional space are determined based on the coupling coordination degree, coupling stage, and coupling type among all spaces.

7. The method according to claim 6, characterized in that, The step of determining the target weight of each preset evaluation index for each type of space includes: For each preset evaluation index corresponding to each type of space, determine the standard value of the preset evaluation index in each historical year. The weight of the indicator for each historical year is determined by the ratio between the standard value corresponding to each historical year and the sum of the standard values ​​corresponding to the preset evaluation indicator for all historical years. The entropy redundancy of the preset evaluation indicator is determined based on the weight of each indicator. The first weight of the preset evaluation indicator is determined based on the ratio between the entropy redundancy of the preset evaluation indicator and the sum of the entropy redundancy of all preset evaluation indicators corresponding to this type of space. The absolute values ​​of the differences between the standard values ​​and preset reference values ​​corresponding to the preset evaluation index in all historical years are determined. The maximum value among these absolute values ​​is determined as the first absolute value, and the minimum value among these absolute values ​​is determined as the second absolute value. The correlation coefficients corresponding to the preset evaluation index are determined based on the absolute values ​​of the differences between the standard values ​​and preset reference values ​​corresponding to the preset evaluation index in all historical years, the first absolute value, and the second absolute value. The correlation degree of the preset evaluation index is determined based on the correlation coefficients corresponding to the preset evaluation index in each historical year. The second weight of the preset evaluation index is determined based on the ratio between the correlation degree of the preset evaluation index and the sum of the correlation degrees of all preset evaluation indices corresponding to this type of space. The target weight of the preset evaluation index is determined based on the weighted sum of the first weight and the second weight.

8. The method according to claim 6, characterized in that, The determination of the coupling coordination degree between all spaces based on the functional scores corresponding to each type of space includes: The weighted sum of the functional scores corresponding to each type of space is determined as the comprehensive coordination index; The first product value is determined by multiplying the functional scores of each type of space; the second product value is determined by multiplying the sum of the functional scores of every two types of spaces within each type of space; the coupling degree is determined by combining the first product value and the second product value. The square root of the product of the comprehensive coordination index and the coupling degree is determined as the coupling coordination degree.

9. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the training method for deep learning-based simulation of rural three-life space evolution as described in any one of claims 1-8.

Images