A method, device, medium and equipment for predicting urban expansion

Abstract

The present invention provides a method, device, medium and equipment for predicting urban expansion. The method comprises: obtaining data information for predicting urban expansion for a target area; the target area comprises a plurality of cells, and the data information comprises data of each cell in the plurality of cells; calculating the land use development probability of each cell according to the data of each cell; taking the cell as the center and taking a plurality of preset spatial neighborhood scales as the expansion range respectively, obtaining a plurality of expansion areas, and inputting the land use data of the plurality of expansion areas into a spatial neighborhood multi-scale feature fusion model for processing, so as to obtain the development probability of each cell under the joint action of the spatial neighborhood multi-scale; constructing the conversion constraint factor of the cell according to the land transfer ratio of the original urban land of each cell; respectively for each cell, calculating the total development probability of the cell based on the land use development probability of the cell, the conversion constraint factor of the cell and the development probability of the cell under the joint action of the spatial neighborhood multi-scale.

Description

Technical Field

[0001] The present invention relates to the field of geographic information science, and in particular to a method, device, medium and equipment for predicting urban expansion. Background Art

[0002] In the model development of urban expansion simulation, Cellular Automata (CA), as a discrete dynamic model that takes into account spatiotemporal characteristics, can effectively model various inherent complex spatiotemporal dynamic processes in the urbanization process (such as land use changes, etc.), and therefore has been widely improved to introduce the construction of urban expansion simulation methods. The CA model consists of five basic elements: cellular space, cellular time, cellular state, cellular neighborhood, and transition rules. Among them, cellular space, cellular time, and cellular state are inherent properties of cells, while cellular neighborhood and transition rules are the direct driving factors of cellular state changes.

[0003] In the process of modeling the existing urban expansion CA model, the calculation of the total probability of cellular development mainly includes three parts: land use development probability, development probability under the action of spatial neighborhood, and constraint factors. As an important component of the total probability of cellular development, the cellular neighborhood plays an indispensable role. In the past two decades, most traditional and modern CA model studies have used decision functions (i.e., the proportion of a certain land use in the neighborhood, representing its contribution to land use conversion) to express the effect of cellular space neighborhood at a single scale. Considering the difference in the performance of neighborhood effects on different land use types, existing CA models such as the FLUS model and the PLUS model have made targeted improvements to the decision function and added the weights of land use effects of different categories in different regions. However, this improvement relies heavily on expert experience and knowledge and has a certain degree of subjectivity. In addition, the heterogeneity of the scale effect of the cellular space neighborhood greatly affects the simulation results of the CA model. However, existing studies rarely conduct quantitative modeling of the scale heterogeneity of the cellular space neighborhood, resulting in certain applicability problems in the simulation results of CA.

[0004] In summary, the above-mentioned prior art has the following technical defects:

[0005] (1) It is difficult to extract the complex laws of spatial neighborhood effects by expressing them in the form of decision functions or their variants;

[0006] (2) It is difficult to uniformly and regularly express the changing rules of the spatial neighborhood scale effect and the optimal neighborhood scale. There is heterogeneity in the effects of different cellular spatial neighborhood scales, and there is a lack of a fusion modeling method for the multi-scale effects of spatial neighborhoods. Summary of the invention

[0007] The present invention provides a method, device, medium and equipment for predicting urban expansion, the purpose of which is to solve the problem that the existing urban expansion CA model is difficult to fully grasp the effect of spatial neighborhood action and the heterogeneity of cellular spatial neighborhood scales, and cannot fully and completely calculate the development probability under the action of spatial neighborhood in the urban expansion CA model.

[0008] In order to achieve the above object, the present invention provides a method for predicting urban expansion, comprising:

[0009] Step 1, obtaining data information for predicting urban expansion in a target area; the target area includes a plurality of cells, and the data information includes data of each of the plurality of cells;

[0010] Step 2, calculate the land development probability of each cell based on the data of each cell;

[0011] Step 3: For each cell, take the cell as the center and use multiple preset spatial neighborhood scales as the expansion range to obtain multiple expansion areas, and input the land use data of the multiple expansion areas into the spatial neighborhood multi-scale feature fusion model for processing to obtain the development probability of the cell under the joint action of the spatial neighborhood multi-scale;

[0012] Step 4: for each cell, construct the conversion constraint factor of the cell according to the land transfer ratio of the original urban land of the cell;

[0013] Step 5: For each cell, calculate the total development probability of the cell based on the land use development probability of the cell, the conversion constraint factor of the cell, and the development probability of the cell under the joint action of multiple scales in the spatial neighborhood.

[0014] More specifically, step 2 includes:

[0015] Preprocess the data of each cell and extract the driving factor of each cell;

[0016] Process the driving factor of each cell to obtain the characteristic vector of each cell;

[0017] The land use development probability of each cell is calculated based on the eigenvector of each cell.

[0018] Furthermore, the spatial neighborhood multi-scale feature fusion model includes a convolutional neural network, a cascade module and a SE module. The convolutional neural network includes a flattening layer, a fully connected layer and multiple parallel connected convolutional modules. Each convolutional module includes a first convolutional layer and a second convolutional layer. The output of the first convolutional layer is connected to the input of the second convolutional layer. The output of the second convolutional layer of each convolutional module is connected to the input of the cascade module, the output of the cascade module is connected to the input of the SE module, the output of the SE module is connected to the input of the flattening layer, and the output of the flattening layer is connected to the input of the fully connected layer. The land use data of multiple extended areas are input into the first convolutional layers of multiple convolutional modules one by one.

[0019] More specifically, step 3 includes:

[0020] For each cell, the cell is taken as the center, and multiple preset spatial neighborhood scales are used as the expansion range to obtain multiple expansion areas. The land use data of the multiple expansion areas are input into the first convolution layer of the convolutional neural network for feature extraction to obtain the CNN spatial neighborhood effect features of each cell at multiple scales.

[0021] Input the CNN spatial neighborhood effect features of each cell at multiple scales into the cascade module for feature cascading to obtain the multi-scale spatial neighborhood effect features of the cell;

[0022] The multi-scale spatial neighborhood effect features of the cells are input into the SE module for attention weighting to obtain the attention-weighted multi-scale spatial neighborhood effect features of the cells;

[0023] Based on the attention-weighted multi-scale spatial neighborhood action characteristics of cells, the encoding of land use conversion types is mapped through the flattening layer and fully connected layer of the convolutional neural network to obtain the development probability Ω of cells under the multi-scale joint action of spatial neighborhoods.

[0024] Further, the SE module includes a Squeeze submodule for performing feature aggregation on the multi-scale spatial neighborhood effect features of cells, an Excitation submodule for calculating the attention value of the multi-scale spatial neighborhood effect features of the aggregated cells, and a Scale submodule for performing attention weighting on the multi-scale spatial neighborhood effect features of the aggregated cells, which are connected in sequence. The input end of the Squeeze submodule is connected to the output end of the cascade module, the input end of the Scale submodule is connected to the output end of the cascade module, and the output end of the Scale submodule is connected to the input end of the flattening layer.

[0025] Furthermore, for each cell, the cell transformation constraint factor is

[0026]

[0027] Where Z represents the constraint factor coefficient, and type=urban represents the original urban land of the cell.

[0028] More specifically, step 5 includes:

[0029] For each cell, the total development probability of the cell is calculated by the land development probability of the cell, the conversion constraint factor of the cell, and the development probability of the cell under the joint action of multi-scale spatial neighborhood:

[0030] OP=P×Z×Ω

[0031] Where OP is the total development probability, P is the land use development probability of the cell, Z is the conversion constraint factor of the cell, and Ω is the development probability of the cell under the joint action of multiple scales in the spatial neighborhood.

[0032] The present invention also provides a city expansion prediction device, comprising:

[0033] A data acquisition module is used to acquire data information for predicting urban expansion in a target area; the target area includes a plurality of cells, and the data information includes data of each of the plurality of cells;

[0034] The first calculation module is used to calculate the land development probability of each cell according to the data of each cell;

[0035] The spatial neighborhood multi-scale feature fusion model is used to obtain multiple extended areas for each cell, taking the cell as the center and taking multiple preset spatial neighborhood scales as the extension range. The land use data of the multiple extended areas are input into the spatial neighborhood multi-scale feature fusion model for processing to obtain the development probability of the cell under the joint action of the spatial neighborhood multi-scale.

[0036] A construction module is used to construct a conversion constraint factor of each cell according to the land transfer ratio of the original urban land of the cell;

[0037] The second calculation module is used to calculate the total development probability of each cell based on the land development probability of the cell, the conversion constraint factor of the cell and the development probability of the cell under the joint action of multi-scale spatial neighborhood.

[0038] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the above-mentioned urban expansion prediction method is implemented.

[0039] The present invention also provides a city expansion prediction device, including a memory and a processor; the memory is used to store computer programs and intermediate data during program processing, and the processor is used to execute the computer program to implement the above-mentioned city expansion prediction method.

[0040] The above scheme of the present invention has the following beneficial effects:

[0041] The present invention calculates the land development probability of each cell through data information used for predicting urban expansion in a target area, and for each cell, takes the cell as the center and takes multiple preset spatial neighborhood scales as the expansion range to obtain multiple expansion areas, and inputs the land use data of the multiple expansion areas into a spatial neighborhood multi-scale feature fusion model for processing to obtain the development probability of the cell under the joint action of the spatial neighborhood multi-scale, and then constructs the conversion constraint factor of the cell according to the land transfer ratio of the original urban land of the cell, and combines the land development probability of the cell, the conversion constraint factor of the cell and the development probability under the joint action of the spatial neighborhood multi-scale to calculate the total development probability of the cell to predict the probability of urban expansion; the problem that the existing urban expansion CA model is difficult to fully grasp the spatial neighborhood effect and the heterogeneity of the spatial neighborhood scale of the cell, and the development probability under the joint action of the spatial neighborhood in the urban expansion CA model cannot be fully and completely calculated.

[0042] Other beneficial effects of the present invention will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0044] Figure 2 is a schematic diagram of a target area according to an embodiment of the present invention;

[0045] Figure 3 It is a structural diagram of a spatial neighborhood multi-scale feature fusion model in an embodiment of the present invention;

[0046] Figure 4 This is a comparison diagram of urban land use obtained by simulation using an embodiment of the present invention and actual urban land use. DETAILED DESCRIPTION

[0047] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0049] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installation", "connection" and "connection" should be understood in a broad sense, for example, it can be a locking connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0050] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0051] In view of the existing problems, the present invention provides a method, device, medium and equipment for predicting urban expansion.

[0052] like Figure 1 As shown, an embodiment of the present invention provides a method for predicting urban expansion, comprising:

[0053] Step 1, obtaining data information for predicting urban expansion in a target area; the target area includes a plurality of cells, and the data information includes data of each of the plurality of cells;

[0054] Step 2, calculate the land development probability of each cell based on the data of each cell;

[0055] Step 3: For each cell, take the cell as the center and use multiple preset spatial neighborhood scales as the expansion range to obtain multiple expansion areas, and input the land use data of the multiple expansion areas into the spatial neighborhood multi-scale feature fusion model for processing to obtain the development probability of the cell under the joint action of the spatial neighborhood multi-scale;

[0056] Step 4: for each cell, construct the conversion constraint factor of the cell according to the land transfer ratio of the original urban land of the cell;

[0057] Step 5: For each cell, calculate the total development probability of the cell based on the land use development probability of the cell, the conversion constraint factor of the cell, and the development probability of the cell under the joint action of multiple scales in the spatial neighborhood.

[0058] Specifically, step 2 includes:

[0059] The data of each cell is preprocessed to extract the driving factor of each cell; the driving factor of each cell is processed to obtain the characteristic vector of each cell; and the land use development probability of each cell is calculated according to the characteristic vector of each cell.

[0060] like Figure 2 As shown, the embodiment of the present invention is based on the 30m2 urban area of ​​Guangzhou (including Yuexiu District, Liwan District and Haizhu District) from 2005 to 2015. 2 Based on the land use data, the specific implementation process of the present invention is described:

[0061] In an embodiment of the present invention, firstly, land use data, DEM data, population density, GDP data and road network vector data of equally spaced years for predicting urban expansion in a target area are obtained; the target area includes a plurality of cells, and the data information includes data of each cell in the plurality of cells; secondly, the obtained data is vector-range clipped according to the target area, the land use data is recoded to adapt to the neural network training rules, and the total land use of each cell in each year is calculated; then, the driving factor of each cell is extracted, that is, the DEM data, population density and GDP data are recoded. Through operations such as reprojection, resampling, grid alignment and slope calculation, the spatial resolution was uniformly set to 30m, and four driving factors, namely elevation, slope, population density and GDP, were extracted. The main roads, secondary roads and residential roads were extracted through the road network vector data. The Euclidean distance field was generated and grid aligned with other driving factors. The distance to the main road, the distance to the secondary road and the distance to the residential road were extracted. A total of three driving factors, namely the distance to the main road, the distance to the secondary road and the distance to the residential road, were extracted. Among them, elevation and slope are the driving factors of natural environmental conditions, and population density, GDP, distance to the main road, distance to the secondary road and distance to the residential road are the driving factors of socio-economic conditions.

[0062] The seven driving factors extracted to characterize the natural environment and socioeconomic conditions of each cell are standardized to obtain the characteristic vector of each cell. For the kth driving factor, the standardized formula is:

[0063]

[0064] In the formula, x k (i) and x kn (i) are the attribute value of the kth driving factor corresponding to the i-th cell and its standardized eigenvalue, max(xk ) and min(x k ) represent the maximum and minimum attribute values ​​of the kth driving factor respectively.

[0065] The eigenvector of each cell can be expressed as x = [x kn (1),x kn (2)…x kn (7)] T Indicates that the set of feature vectors of all cells is constructed as the input feature sample set.

[0066] Specifically, step 2 includes:

[0067] The land conversion type of each cell is encoded as the sample label of the random forest algorithm. Based on the random forest algorithm, a regular mapping relationship between the land conversion type and the driving factor is established to calculate the land development probability of each cell.

[0068] In an embodiment of the present invention, the land use conversion type of each cell is encoded, and the land use conversion type of the original land of the cell includes converting non-urban land or water body into urban land and keeping urban land unchanged. The conversion of non-urban land or water body into urban land is coded as 1, and the land use remains unchanged as 0, which is constructed corresponding to the feature vector of each cell as an input feature sample set.

[0069] 20,000 land conversion types coded as 1 and 0 were equally selected as sample sets for the random forest algorithm, of which 80% were used as training sets and 20% were used as test sets. The training set in the sample set was trained based on the random forest algorithm to establish a regular relationship between land conversion types and driving factors, and the land development probability of each cell was calculated using the following formula (2):

[0070]

[0071] Among them, x is the cell characteristic vector, P is the cell’s own land development probability, and f i (x) is a single decision classifier, I(·) is the indicator function, y represents the code of the target land conversion type, and N represents the number of decision classifiers.

[0072] The random forest algorithm in the embodiment of the present invention uses CART tree (Classification And Regression Tree, CART) as a decision tree classifier, the number of decision trees is 50, and feature selection is performed based on the Gini coefficient.

[0073] Specifically, step 3 includes:

[0074] For each cell, the cell is taken as the center, and multiple preset spatial neighborhood scales are used as the expansion range to obtain multiple expansion areas. The land use data of the multiple expansion areas are input into the convolution layer of the convolutional neural network for feature extraction to obtain the CNN spatial neighborhood effect features of each cell at multiple scales.

[0075] The CNN spatial neighborhood effect features of each cell at multiple scales are input into the cascade module for feature cascading to obtain the multi-scale spatial neighborhood effect features of each cell;

[0076] The multi-scale spatial neighborhood effect feature of each cell is input into the SE module for attention weighting to obtain the attention-weighted multi-scale spatial neighborhood effect feature of each cell;

[0077] Based on the attention-weighted multi-scale spatial neighborhood action characteristics of each cell, the encoding of the land use conversion type is mapped through the flattening layer and the fully connected layer of the convolutional neural network to obtain the development probability Ω of each cell under the multi-scale joint action of the spatial neighborhood.

[0078] In the embodiment of the present invention, 11 spatial neighborhood scales are selected from the target area, which are respectively increased from Moore3×3, Moore5×5 to Moore23×23, that is, the range of the spatial neighborhood side length is 90m to 690m. Multiple preset spatial neighborhood scales are used as the expansion range to obtain multiple expansion areas, and the land use data of the multiple expansion areas are input into the spatial neighborhood multi-scale feature fusion model, such as Figure 3 As shown. The spatial neighborhood multi-scale feature fusion model includes a convolutional neural network, a cascade module and an SE module. The convolutional neural network includes a flattening layer, a fully connected layer and a plurality of parallel connected convolutional modules. Each convolutional module includes a first convolutional layer and a second convolutional layer. The output of the first convolutional layer is connected to the input of the second convolutional layer. The output of the second convolutional layer of each convolutional module is connected to the input of the cascade module. The output of the cascade module is connected to the input of the SE module. The output of the SE module is connected to the input of the flattening layer. The output of the flattening layer is connected to the input of the fully connected layer. The land use data of the multiple extended areas are input into the first convolutional layers of the multiple convolutional modules one by one.

[0079] In the embodiment of the present invention, the convolution kernel scales of the first convolution layer and the second convolution layer are 1 and ns respectively (ns represents the size of the spatial neighborhood scale). The convolution kernel scale of the first convolution layer is 1, which is used to increase the feature dimension of the cell itself; the convolution kernel scale of the second convolution layer is ns, which aims to explore the overall effect of the cells within the neighborhood range and output the spatial neighborhood features under the effect of the neighborhood scale. For each spatial neighborhood scale, 80% of the sample set is selected as the training set and 20% as the test set. The cross entropy is used as the loss function, the learning rate is 0.001, and the batch size is 32. 11 scale CNN spatial neighborhood effect features are extracted respectively.

[0080] In the embodiment of the present invention, the first convolutional layer in the convolutional neural network takes the cell state within the spatial neighborhood scale as input, and the feature value extracted by the convolutional layer in the convolutional neural network can be expressed as:

[0081] x j l+1 =f(∑ k∈N×N w kj l+1 × k l +b j l+1 ) (3)

[0082] In the formula, x j l+1 is the output feature value of the jth convolution kernel in the l+1th layer of the CNN convolutional neural network, w kj l+1 is the weight coefficient of the jth convolution kernel in the l+1th layer of the CNN convolutional neural network for the kth feature in the lth layer, x k l is the kth eigenvalue in the lth layer of CNN, b j l+1 is the bias term of the jth convolution kernel in the l+1th layer of CNN, N is the size of the convolution kernel, and f(·) is the activation function.

[0083] The CNN spatial neighborhood effect features of each cell at 11 scales are feature cascaded to obtain the multi-scale spatial neighborhood effect features of each cell; the multi-scale spatial neighborhood effect features are weighted by attention through the SE module to obtain the attention-weighted multi-scale spatial neighborhood effect features of each cell. The SE module includes a Squeeze submodule for performing feature aggregation on the multi-scale spatial neighborhood effect features of the cells, an Excitation submodule for calculating the attention value of the multi-scale spatial neighborhood effect features of the aggregated cells, and a Scale submodule for performing attention weighting on the multi-scale spatial neighborhood effect features of the aggregated cells, the input end of the Squeeze submodule is connected to the output end of the cascade module, the input end of the Scale submodule is connected to the output end of the cascade module, and the output end of the Scale submodule is connected to the input end of the flattening layer.

[0084] In an embodiment of the present invention, the Excitation submodule includes two fully connected layers with different feature dimensions, and the dimension reduction coefficient r is defined as 16, that is, the cascade feature dimension will first be reduced from 704 to 44, and then increased from 44 to 704, thereby obtaining the attention coefficient corresponding to the multi-scale feature, which is used as the weight coefficient of the scale space neighborhood effect feature of each cell; then, based on the attention-weighted multi-scale spatial neighborhood cascade feature, the encoding of the land use conversion type is mapped, and the development probability Ω of each cell under the joint action of the spatial neighborhood multi-scale is output.

[0085] In an embodiment of the present invention, the Squeeze submodule in the SE module realizes global feature aggregation through an adaptive average pooling function; the Excitation submodule is composed of two fully connected layers with different feature dimensions, the purpose of which is to learn the nonlinear dependency between features of different neighborhood scales, and then obtain the attention value corresponding to the multi-scale features; the Scale submodule operation is the weighted addition of the multi-scale spatial neighborhood cascade features, and the weight is the attention coefficient learned by the Excitation submodule.

[0086] In this embodiment, the land transfer ratio of the main urban area of ​​Guangzhou City every 5 years from 2005 to 2015 is counted, and it is found that the land transfer ratio of urban land every 5 years does not exceed 0.4%. Therefore, it is reasonable to assume that the original urban land remains unchanged during the urban expansion of Guangzhou City. Based on this, the conversion constraint factor of each cell is constructed, and its specific expression is:

[0087]

[0088] Where Z represents the constraint factor coefficient, and type=urban represents the original urban land of the cell.

[0089] Specifically, step 5 includes:

[0090] For each cell, the total development probability of the cell is calculated by the land development probability of the cell, the conversion constraint factor of the cell, and the development probability of the cell under the joint action of multi-scale spatial neighborhood:

[0091] OP=P×Z×Ω (5)

[0092] Where OP is the total development probability, P is the land use development probability of the cell, Z is the conversion constraint factor of the cell, and Ω is the development probability of the cell under the joint action of multiple scales in the spatial neighborhood.

[0093] The MS-CA spatial neighborhood multi-scale feature fusion model adopted in the embodiment of the present invention has a time step of 1 year for each iteration process. The spatial pattern of urban land use in 2010 is taken as the initial state. Each iteration is probability sorted by the total probability of cellular development. According to the historical land use data of the main urban area of ​​Guangzhou, the demand for urban land in this iteration is calculated and converted into the corresponding number of cells. The overall spatial pattern of the city is updated through the roulette strategy, and then the next iteration is entered to calculate the total probability of cellular development until the total probability of cellular development in 2015 is calculated. The iteration is stopped to obtain the urban expansion simulation result in 2015, which is compared with the actual urban land use in 2015, as shown in the figure. Figure 4 shown.

[0094] Finally, the embodiment of the present invention selects overall accuracy (OA), kappa (Kappa Coeffcient) coefficient and FoM (Figure of Merit) coefficient as evaluation functions, and the calculation method thereof is as follows:

[0095]

[0096]

[0097] In the above formula, sum is the total number of cells in the target area, m is the number of land use categories, N kk N represents the number of cells that are actually correct for the k-th land use prediction. k represents the number of cells simulated as the kth type of land use, S k is the number of cells of the kth type, NAMQ represents expected consistency, A is the number of cells that have actually changed but are simulated unchanged, B is the number of cells that have actually changed but are correctly simulated, C is the number of cells that have actually changed but are incorrectly simulated, and D is the number of cells that have not actually changed but are simulated as other types.

[0098] The simulation results of the MS-CA spatial neighborhood multi-scale feature fusion model are compared with the decision function with the best simulation effect in 11 spatial neighborhood scales, a single CNN convolutional neural network, and the MS-CA model without the SE module through the evaluation function. The comparison of the simulation results of the MS-CA model with the evaluation indicators of other modeling strategies is shown in Table 1.

[0099] Modeling strategy OA Kappa FoM MS-CA Model 0.96492 0.91979 0.09016 Optimal scale of decision function 0.96421 0.91811 0.07788 CNN optimal scale 0.96481 0.91953 0.08786 MS-CA model without SE module 0.96460 0.91905 0.08522

[0100] Table 1

[0101] It can be seen from the above table that the MS-CA spatial neighborhood multi-scale feature fusion model performs best in all indicators, which proves the effectiveness of the strategy of the embodiment of the present invention.

[0102] The embodiment of the present invention obtains data information for urban expansion prediction of the target area, calculates the land development probability of each cell, takes the cell as the center, and takes multiple preset spatial neighborhood scales as the expansion range to obtain multiple expansion areas, and inputs the land use data of the multiple expansion areas into the spatial neighborhood multi-scale feature fusion model for feature extraction, feature cascading, and attention weighting, so as to output the development probability of each cell under the joint action of the spatial neighborhood multi-scale; then, according to the land transfer ratio of the original urban land of the cell, the conversion constraint factor of each cell is constructed, and the land development probability of the cell, the conversion constraint factor of the cell and the development probability of the cell under the joint action of the spatial neighborhood multi-scale are combined to calculate the total development probability of each cell; the problem that the existing urban expansion CA model is difficult to fully grasp the spatial neighborhood effect and the heterogeneity of the cellular spatial neighborhood scale, and the development probability under the action of the spatial neighborhood in the urban expansion CA model cannot be fully and completely calculated.

[0103] The embodiment of the present invention further provides a device for predicting urban expansion, including:

[0104] A data acquisition module is used to acquire data information for predicting urban expansion in a target area; the target area includes a plurality of cells, and the data information includes data of each of the plurality of cells;

[0105] The first calculation module is used to calculate the land development probability of each cell according to the data of each cell;

[0106] The spatial neighborhood multi-scale feature fusion model is used to obtain multiple extended areas for each cell, taking the cell as the center and taking multiple preset spatial neighborhood scales as the extension range. The land use data of the multiple extended areas are input into the spatial neighborhood multi-scale feature fusion model for processing to obtain the development probability of the cell under the joint action of the spatial neighborhood multi-scale.

[0107] A construction module is used to construct a conversion constraint factor of each cell according to the land transfer ratio of the original urban land of the cell;

[0108] The second calculation module is used to calculate the total development probability of each cell based on the land development probability of the cell, the conversion constraint factor of the cell and the development probability of the cell under the joint action of multi-scale spatial neighborhood.

[0109] The embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the above-mentioned urban expansion prediction method is implemented.

[0110] An embodiment of the present invention further provides a city expansion prediction device, including a memory and a processor; the memory is used to store computer programs and intermediate data during program processing, and the processor is used to execute the computer program to implement the above-mentioned city expansion prediction method.

[0111] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting urban expansion, characterized in that: include: Step 1, obtaining data information for predicting urban expansion in a target area; The target area includes a plurality of cells, and the data information includes data of each of the plurality of cells; Step 2, calculating the land development probability of each cell according to the data of each cell; Step 3, for each of the cells, taking the cell as the center and taking multiple preset spatial neighborhood scales as the expansion range, obtaining multiple expansion areas, and inputting the land use data of the multiple expansion areas into the spatial neighborhood multi-scale feature fusion model for processing, to obtain the development probability of the cell under the joint action of the spatial neighborhood multi-scale; The spatial neighborhood multi-scale feature fusion model includes a convolutional neural network, a cascade module and an SE module. The convolutional neural network includes a flattening layer, a fully connected layer and a plurality of parallel connected convolutional modules. Each of the convolutional modules includes a first convolutional layer and a second convolutional layer. The output end of the first convolutional layer is connected to the input end of the second convolutional layer. The output end of the second convolutional layer of each convolutional module is connected to the input end of the cascade module. The output end of the cascade module is connected to the input end of the SE module. The output end of the SE module is connected to the input end of the flattening layer. The output end of the flattening layer is connected to the input end of the fully connected layer. The land use data of the plurality of extended areas are input into the first convolutional layers of the plurality of convolutional modules in a one-to-one correspondence. For each of the cells, the cell is taken as the center, and multiple preset spatial neighborhood scales are used as the expansion range to obtain multiple expansion areas, and the land use data of the multiple expansion areas are input into the convolution layer of the convolutional neural network for feature extraction to obtain the CNN spatial neighborhood effect features of each cell at each spatial neighborhood scale; Inputting the CNN spatial neighborhood effect features of each cell at each spatial neighborhood scale into a cascade module for feature cascading to obtain multi-scale spatial neighborhood effect features of the cell; Inputting the multi-scale spatial neighborhood effect feature of the cell into the SE module for attention weighting to obtain the attention weighted multi-scale spatial neighborhood effect feature of the cell; Based on the attention-weighted multi-scale spatial neighborhood action characteristics of the cell, the encoding of the land conversion type is mapped through the flattening layer and the fully connected layer of the convolutional neural network to obtain the development probability Ω of the cell under the multi-scale joint action of the spatial neighborhood; Step 4, for each of the cells, construct a conversion constraint factor of the cell according to the land transfer ratio of the original urban land of the cell; Step 5, for each of the cells, based on the land development probability of the cell, the conversion constraint factor of the cell and the development probability of the cell under the multi-scale joint action of the spatial neighborhood, calculate the total development probability of the cell.

2. The urban expansion prediction method according to claim 1, characterized in that: The step 2 comprises: Preprocessing the data of each cell to extract the driving factor of each cell; Processing the driving factor of each cell to obtain a characteristic vector of each cell; The land use development probability of each cell is calculated according to the characteristic vector of each cell.

3. The urban expansion prediction method according to claim 1, characterized in that: The SE module includes a Squeeze submodule for performing feature aggregation on the multi-scale spatial neighborhood effect features of the cells, an Excitation submodule for calculating the attention value of the aggregated multi-scale spatial neighborhood effect features of the cells, and a Scale submodule for performing attention weighting on the aggregated multi-scale spatial neighborhood effect features of the cells, which are connected in sequence. The input end of the Squeeze submodule is connected to the output end of the cascade module, the input end of the Scale submodule is connected to the output end of the cascade module, and the output end of the Scale submodule is connected to the input end of the flattening layer.

4. The urban expansion prediction method according to claim 1, characterized in that: For each of the cells, the conversion constraint factor of each of the cells is Where Z represents the constraint factor coefficient, and type=urban represents the original urban land of the cell.

5. The urban expansion prediction method according to claim 1, characterized in that: The step 5 comprises: For each cell, the total development probability of the cell is calculated as OP = P × Z × Ω by using the land development probability of the cell, the conversion constraint factor of the cell and the development probability of the cell under the multi-scale joint action of the spatial neighborhood. Where OP is the total development probability, P is the land use development probability of the cell, Z is the conversion constraint factor of the cell, and Ω is the development probability of the cell under the joint action of multiple scales in the spatial neighborhood.

6. A device for predicting urban expansion, characterized in that: include: A data acquisition module, used to acquire data information for predicting urban expansion in a target area; The target area includes a plurality of cells, and the data information includes data of each of the plurality of cells; A first calculation module, used for calculating the land development probability of each cell according to the data of each cell; The spatial neighborhood multi-scale feature fusion model is used to obtain multiple extended areas for each of the cells, taking the cell as the center and taking multiple preset spatial neighborhood scales as the extension range, and inputting the land use data of the multiple extended areas into the spatial neighborhood multi-scale feature fusion model for processing to obtain the development probability of the cell under the joint action of the spatial neighborhood multi-scale; The spatial neighborhood multi-scale feature fusion model includes a convolutional neural network, a cascade module and an SE module. The convolutional neural network includes a flattening layer, a fully connected layer and a plurality of parallel connected convolutional modules. Each of the convolutional modules includes a first convolutional layer and a second convolutional layer. The output end of the first convolutional layer is connected to the input end of the second convolutional layer. The output end of the second convolutional layer of each convolutional module is connected to the input end of the cascade module. The output end of the cascade module is connected to the input end of the SE module. The output end of the SE module is connected to the input end of the flattening layer. The output end of the flattening layer is connected to the input end of the fully connected layer. The land use data of the plurality of extended areas are input into the first convolutional layers of the plurality of convolutional modules in a one-to-one correspondence. For each of the cells, the cell is taken as the center, and multiple preset spatial neighborhood scales are used as the expansion range to obtain multiple expansion areas, and the land use data of the multiple expansion areas are input into the convolution layer of the convolutional neural network for feature extraction to obtain the CNN spatial neighborhood effect features of each cell at each spatial neighborhood scale; Inputting the CNN spatial neighborhood effect features of each cell at each spatial neighborhood scale into a cascade module for feature cascading to obtain multi-scale spatial neighborhood effect features of the cell; Inputting the multi-scale spatial neighborhood effect feature of the cell into the SE module for attention weighting to obtain the attention weighted multi-scale spatial neighborhood effect feature of the cell; Based on the attention-weighted multi-scale spatial neighborhood action characteristics of the cell, the encoding of the land conversion type is mapped through the flattening layer and the fully connected layer of the convolutional neural network to obtain the development probability Ω of the cell under the multi-scale joint action of the spatial neighborhood; A construction module, for constructing a conversion constraint factor of each cell according to the land transfer ratio of the original urban land of the cell; The second calculation module is used to calculate the total development probability of the cell for each cell based on the land development probability of the cell, the conversion constraint factor of the cell and the development probability of the cell under the multi-scale joint action of the spatial neighborhood.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the urban expansion prediction method according to any one of claims 1 to 5 is implemented.

8. A city expansion prediction device, characterized in that: It comprises a memory and a processor; the memory is used to store computer programs and intermediate data during program processing, and the processor is used to execute the computer program to implement the urban expansion prediction method as described in any one of claims 1 to 5.

Images