Interpolation method, device and related equipment in spatial domain based on deep learning

Abstract

The present invention discloses an interpolation method in a spatial domain based on deep learning, which is applied to the field of machine learning and is used to improve the interpolation accuracy in the spatial domain based on deep learning. The method provided by the present invention comprises: obtaining a feature matrix and an inverse distance feature matrix of a space to be analyzed; constructing training data based on the feature matrix and the inverse distance feature matrix; inputting the training data into a pre-training model for training to obtain an inverse distance feature model; constructing an interpolation point feature matrix based on the features of the feature matrix; inputting the interpolation point feature matrix into the inverse distance feature model for calculation to obtain an inverse distance weighted value matrix; and performing spatial analysis on the space to be analyzed based on the inverse distance weighted value matrix to obtain an analysis result.

Description

Technical Field

[0001] The present invention relates to the field of machine learning, and in particular to an interpolation method, device, computer equipment and storage medium in a spatial field based on deep learning. Background Art

[0002] When performing spatial analysis in spatial fields such as geographic information and environmental monitoring and analysis, common interpolation methods include inverse distance weighted interpolation (IDW) and Kriging. Among them, the inverse distance weighted interpolation method is to obtain the interpolation unit by calculating the average value of each sampling point in the neighboring area of ​​the analyzed area, and the Kriging interpolation method refers to a method that can mathematically provide an optimal linear unbiased estimate of the object under study. The Kriging interpolation method refers to a method that can mathematically provide an optimal linear unbiased estimate of the object under study. This method is mainly an interpolation method that takes into account the various geometric characteristics and spatial structure of the sample.

[0003] When performing spatial analysis in a spatial domain, such as air monitoring and analysis, the inverse distance weighted interpolation method directly uses meteorological factors as factors affecting distance, and then interpolates on this basis, but it results in the loss of a lot of useful information, making the interpolation accuracy limited. At the same time, this method requires that discrete points are evenly distributed, and the density of sample points can have a variable amount in the result, and the inverse distance interpolation method only considers the distance between two points in the same plane, and does not consider the characteristics of spatial distribution, making the accuracy of spatial analysis low. Since the spatial distribution of observation points in the air pollution concentration monitored by the air monitoring station is very irregular, the spatial structure diagram of the Kriging interpolation method will have obvious concave and convex phenomena when the number of samples is small, which will affect the interpolation and make the interpolation accuracy limited. Therefore, the Kriging interpolation method is not suitable for the field of air monitoring and analysis, and the applicability of this method is low.

[0004] Therefore, when performing spatial interpolation on a spatial domain, the existing methods have the problem of low analysis accuracy. Summary of the invention

[0005] Embodiments of the present invention provide a deep learning-based interpolation method, apparatus, computer equipment, and storage medium in a spatial domain to improve the interpolation accuracy in a spatial domain based on deep learning.

[0006] A deep learning-based interpolation method in a spatial domain, comprising:

[0007] Obtain the characteristic matrix and inverse distance characteristic matrix of the space to be analyzed;

[0008] Constructing training data based on the feature matrix and the inverse distance feature matrix;

[0009] Inputting the training data into a pre-training model for training to obtain an inverse distance feature model;

[0010] Based on the characteristics of the characteristic matrix, construct an interpolation point characteristic matrix;

[0011] Inputting the interpolation point feature matrix into the inverse distance feature model for calculation to obtain an inverse distance weighted value matrix;

[0012] Based on the inverse distance weighted value matrix, spatial analysis is performed on the space to be analyzed to obtain an analysis result.

[0013] An interpolation device in a spatial domain based on deep learning, comprising:

[0014] A matrix acquisition module is used to obtain the characteristic matrix and inverse distance characteristic matrix of the space to be analyzed;

[0015] A training data acquisition module, used for constructing training data based on the feature matrix and the inverse distance feature matrix;

[0016] A feature model acquisition module is used to input the training data into a pre-training model for training to obtain a feature model;

[0017] An interpolation point feature matrix acquisition module is used to construct an interpolation point feature matrix based on the features of the feature matrix;

[0018] An inverse distance weighted value matrix acquisition module is used to input the interpolation point feature matrix into the inverse distance feature model for calculation to obtain an inverse distance weighted value matrix;

[0019] The spatial analysis module is used to perform spatial analysis on the space to be analyzed based on the inverse distance weighted value matrix to obtain an analysis result.

[0020] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned deep learning-based spatial domain interpolation method when executing the computer program.

[0021] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned deep learning-based spatial domain interpolation method.

[0022] The interpolation method, device, computer equipment and storage medium in the spatial domain based on deep learning provided by the embodiment of the present invention obtain the characteristic matrix and inverse distance characteristic matrix of the space to be analyzed; construct training data based on the characteristic matrix and the inverse distance characteristic matrix; input the training data into the pre-trained model for training to obtain the inverse distance characteristic model; construct an interpolation point characteristic matrix based on the characteristics of the characteristic matrix; input the interpolation point characteristic matrix into the inverse distance characteristic model for calculation to obtain the inverse distance weighted value matrix; perform spatial analysis on the space to be analyzed based on the inverse distance weighted value matrix to obtain the analysis result. By using the characteristic matrix and the inverse distance characteristic matrix of the space to be analyzed to construct training data, the training data can select multiple features according to the needs, thereby improving the adaptability of the model. At the same time, the model obtained by training the training data based on the constructed interpolation point characteristic matrix can be analyzed, which can effectively improve the interpolation accuracy of the spatial domain based on deep learning. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0024] Figure 1 is a schematic diagram of an application environment of an interpolation method in a spatial domain based on deep learning in one embodiment of the present invention;

[0025] Figure 2 is a flow chart of an interpolation method in a spatial domain based on deep learning in one embodiment of the present invention;

[0026] Figure 3 is a schematic structural diagram of an interpolation device in a spatial domain based on deep learning in one embodiment of the present invention;

[0027] Figure 4 is a schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] The spatial interpolation method based on deep learning provided in this application can be applied in Figure 1In an application environment, a computer device communicates with a server via a network. The computer device may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The server may be implemented as an independent server or a server cluster consisting of multiple servers.

[0030] In one embodiment, if Figure 2 As shown in the figure, a spatial interpolation method based on deep learning is provided. Figure 1 The server in the example is used to illustrate, including the following steps S10 to S60:

[0031] S10. Obtain a characteristic matrix and an inverse distance characteristic matrix of the space to be analyzed.

[0032] In step S10, the above-mentioned space to be analyzed refers to different spatial domains corresponding to different application scenarios. For example, when the application scenario is geographical analysis, the space to be analyzed includes but is not limited to mountains and rivers. When the application scenario is air quality analysis, the space to be analyzed includes but is not limited to random air collection points in a certain area.

[0033] The content of the above feature matrix includes but is not limited to the inherent features and associated features of the sampling points in the space to be analyzed.

[0034] It should be understood that the inherent characteristics of the sampling point here refer to the characteristics that are only related to the sampling point itself. For example, the sampling value corresponding to the sampling point. The associated characteristics of the sampling point here refer to the characteristics of the sampling point related to the application scenario. For example, when the application scenario is geographic analysis, the above-mentioned associated characteristics include but are not limited to distance characteristics and angle characteristics. When the application scenario is air quality analysis, the above-mentioned associated characteristics include but are not limited to wind speed characteristics, wind direction characteristics, and air humidity characteristics.

[0035] The above feature matrix has different feature representations according to different application scenarios, which is not limited here. In addition, the inherent features and associated features included in the feature matrix can be specifically set according to actual needs, which is not limited here.

[0036] The above inverse distance feature matrix is ​​obtained by calculating the inverse distance weighted value of each sampling point. The inverse distance weighted value is:

[0037]

[0038] in is the sampling value of the sample point x, The inverse distance weight between sampling points x and y is:

[0039]

[0040] in, is the Euclidean distance between points y and x; is the influence distance function.

[0041] It should be understood that the influence distance refers to a parameter set according to the influence of a sampling point on other sampling points or interpolation points. The expression of the influence distance parameter includes but is not limited to constants and functions. It can be used to describe the influence between sampling points for different application scenarios. For example, in describing the influence of terrain or atmosphere on the propagation of pollutants.

[0042] In the space to be analyzed, the sampling points corresponding to the rows of the above feature matrix and the inverse distance feature matrix are the same.

[0043] By obtaining the characteristic matrix and inverse distance characteristic matrix of the space to be analyzed for different application scenarios, the characteristic matrix and inverse distance characteristic matrix can be adjusted according to the specific application scenarios, so that they have corresponding analysis adaptability to the spatial fields in different application scenarios, thereby improving the interpolation adaptability of the spatial field based on deep learning.

[0044] S20. Construct training data based on the feature matrix and the inverse distance feature matrix.

[0045] In step S20, the training data refers to data used to train the feature model.

[0046] The above construction methods include but are not limited to merging, summing, and weighted summing.

[0047] Preferably, a column-based merging method is adopted here.

[0048] Constructing training data through feature matrices and inverse distance feature matrices not only makes the source of training data reliable, but also makes the training data have diverse characteristics to adapt to the training of different models in different application scenarios, thereby improving the interpolation adaptability and analysis accuracy in the spatial field based on deep learning.

[0049] S30, inputting the training data into the pre-training model for training to obtain an inverse distance feature model.

[0050] In step S30, the method for establishing the pre-training model includes but is not limited to establishing a pre-training model based on an SVM algorithm and establishing a pre-training model based on a deep neural network algorithm.

[0051] The inverse distance feature model obtains the inverse distance weighted value of a point through the features of the point in the spatial domain.

[0052] Preferably, a pre-trained model is established based on a deep neural network algorithm. A deep neural network model is constructed through CNN, RNN, pooling layer, and fully connected layer.

[0053] By inputting the training data into the pre-trained model for training, a feature model is obtained to perform spatial analysis on the space to be analyzed. Since the training data has a reliable source and is diverse, the trained feature model can effectively adapt to different application scenarios, thereby improving the interpolation adaptability and analysis accuracy in the spatial field based on deep learning.

[0054] S40. Construct an interpolation point feature matrix based on the features of the feature matrix.

[0055] In step S40, the above-mentioned features refer to the features corresponding to each interpolation point in the feature matrix, including feature categories and feature numbers. It should be understood here that the feature category refers to the category information to which the column feature belongs, for example, wind speed features, wind direction features. The number of features refers to the total number of all feature categories.

[0056] The interpolation point characteristic matrix is ​​a characteristic matrix used for spatial interpolation of the space to be analyzed. The method for obtaining the value of the interpolation point characteristic matrix is ​​generally based on the measured value of the space.

[0057] By constructing the interpolation point feature matrix through the characteristics of the feature matrix, the interpolation point feature matrix can be interpolated with the corresponding feature model, saving the time of directly analyzing the characteristics of the spatial data to be analyzed and establishing the spatial classification map, and improving the interpolation efficiency and analysis accuracy in the spatial field based on deep learning.

[0058] S50, inputting the interpolation point feature matrix into the inverse distance feature model for interpolation calculation to obtain an inverse distance weighted value matrix.

[0059] In step S50, the interpolation point feature matrix is ​​input into the inverse distance feature model trained in step S30, and an inverse distance weighted value matrix of the interpolation points can be obtained.

[0060] S60. Based on the inverse distance weighted value matrix, perform spatial analysis on the space to be analyzed to obtain analysis results.

[0061] In step S60, the above-mentioned space analysis refers to the process of analyzing the distribution state of sampling points in the space to be analyzed.

[0062] The above analysis results refer to the interpolation results obtained by analyzing the spatial features to be analyzed.

[0063] For example, in air quality analysis, the quality analysis of the selected space to be analyzed is performed based on the inverse distance weighted value matrix, and the quality analysis may include small particle content analysis, PM2.5 analysis, etc. The obtained analysis results may be small particle content interpolation, PM2.5 interpolation, etc.

[0064] By performing spatial analysis on the analyzed space, the analysis results are obtained. During the analysis process, the interpolation efficiency and analysis accuracy in the spatial field based on deep learning are improved.

[0065] The interpolation method in the spatial domain based on deep learning provided by the embodiment of the present invention obtains the characteristic matrix and the inverse distance characteristic matrix of the space to be analyzed; constructs training data based on the characteristic matrix and the inverse distance characteristic matrix; inputs the training data into the pre-trained model for training to obtain the inverse distance characteristic model; constructs the interpolation point characteristic matrix based on the characteristics of the characteristic matrix; inputs the interpolation point characteristic matrix into the inverse distance characteristic model for calculation to obtain the inverse distance weighted value matrix; performs spatial analysis on the space to be analyzed based on the inverse distance weighted value matrix to obtain the analysis result. By constructing the training data using the characteristic matrix and the inverse distance characteristic matrix of the space to be analyzed, the training data can select multiple features according to the needs, thereby improving the adaptability of the model. At the same time, the model obtained by training the training data based on the constructed interpolation point characteristic matrix is ​​analyzed, which can effectively improve the interpolation accuracy of the spatial domain based on deep learning.

[0066] In some optional implementations of this embodiment, step S10 further includes the following steps S101 to S104:

[0067] S101. Obtain all sampling points of the space to be analyzed.

[0068] S102: Based on a preset sampling factor, extract features from each sampling point to obtain a sampling matrix corresponding to each sampling point.

[0069] S103: All sampling matrices are combined to obtain a feature matrix.

[0070] S104: Perform inverse distance calculation on all sampling points according to a preset distance calculation method to obtain an inverse distance feature matrix.

[0071] In step S102, the preset sampling factors refer to the features corresponding to the sampling points.

[0072] For example, when the preset sampling factors are sampling value, wind speed feature, and wind direction feature, the corresponding sampling value, wind speed feature, and wind direction feature are extracted for the sampling point. It should be noted here that each sampling point needs to be calculated separately to obtain the sampling matrix corresponding to each sampling point.

[0073] In step S103, the above merging process refers to a method of merging the sampling matrices according to a certain merging order, for example, merging all sampling matrices according to the spatial distribution of the sampling points in the space to be analyzed.

[0074] In step S104, the above-mentioned preset distance calculation method includes but is not limited to Euclidean distance and inverse distance weighted sum method.

[0075] Through the above method, the feature matrix and the inverse distance feature matrix are obtained. For different application scenarios, the feature matrix and the inverse distance feature matrix can be adaptively adjusted according to the specific application scenarios, so that they have corresponding analysis adaptability to the spatial fields in different application scenarios, thereby improving the interpolation adaptability of the spatial field based on deep learning.

[0076] In some optional implementations of this embodiment, step S20 is specifically:

[0077] Based on the preset merging order, the feature matrix and the inverse distance feature matrix are merged to obtain training data.

[0078] The above merging refers to the process of merging the feature matrix and the inverse distance feature matrix in a preset order. Preferably, a splicing merging method is adopted here to merge the features corresponding to each sampling point in the feature matrix and the inverse distance feature matrix in a left-to-right order.

[0079] By constructing training data through the above-mentioned merging process, not only the source of training data is reliable, but also the training data has diverse characteristics to adapt to the training of different models in different application scenarios, thereby improving the interpolation adaptability and analysis accuracy of the spatial field based on deep learning.

[0080] In some optional implementations of this embodiment, step S30 further includes the following steps S301 to S305:

[0081] S301. Based on the deep neural network algorithm, a pre-training model is constructed, and the training data is input into the pre-training model.

[0082] S302: Calculate the training data using the pre-training model to obtain training data.

[0083] S303: Calculate the loss value of the training result according to the preset loss value calculation method to obtain the loss value.

[0084] S304: When the loss value is not lower than the preset loss value, the pre-trained model is optimized, and the step of returning to the pre-trained model to calculate the training data and obtain the training data is continued.

[0085] S305. When the loss value is lower than the preset loss value, the pre-trained model is used as an inverse distance feature model.

[0086] In step S301, the above-mentioned deep neural network algorithm includes but is not limited to CNN, RNN, pooling layer, and fully connected layer. The neural network model can be designed according to the number of features and interpolation accuracy.

[0087] In step S302, the above calculation is performed according to the network structure of the model.

[0088] In step S303, the above loss value calculation method includes but is not limited to a mean square error loss value calculation method and a maximum loss value calculation method.

[0089] Preferably, a mean square error loss value calculation method is adopted here.

[0090] In step S304, the above-mentioned optimization process refers to a method for optimizing the pre-trained model.

[0091] The above optimization methods include but are not limited to Adam optimization and gradient descent optimization algorithm. Adam optimization refers to a method of adjusting different learning rates for each different parameter, updating frequently changing parameters with a smaller step size, and updating sparse parameters with a larger step size.

[0092] Through the above steps, the trained feature model performs spatial analysis on the space to be analyzed. Since the training data has a reliable source and is diverse, the trained feature model can effectively adapt to different application scenarios, improving the interpolation adaptability and analysis accuracy in the spatial field based on deep learning.

[0093] In some optional implementations of this embodiment, step S60 further includes:

[0094] Based on the inverse distance weighted value matrix, the inverse distance characteristic formula is used to perform reverse interpolation calculation on the analysis space to obtain the analysis result, wherein the analysis result is the interpolation result matrix.

[0095] The analysis results are calculated using the following formula:

[0096]

[0097] Among them, y (k) refers to the kth interpolation point, F(y (k) ) refers to the vector of the kth interpolation point in the interpolation result matrix, f(y (k) ) refers to the eigenvector of the kth interpolation point in the space to be analyzed, W i (x (i) ,y (k) ) refers to the inverse distance value between the i-th sampling point x and the k-th interpolation point y in the inverse distance weighted value matrix, the value range of k is (1, n), the value range of i is (1, m), n is the number of interpolation points, and m is the number of sampling points.

[0098] The interpolation point feature matrix is ​​interpolated through the feature model generated by training the training data to obtain the interpolation result matrix. It not only considers the distance between the sampling points in space, but also saves the time of directly analyzing the characteristics of the spatial data to be analyzed and establishing the spatial classification map, thereby improving the interpolation efficiency and analysis accuracy in the spatial field based on deep learning.

[0099] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0100] In one embodiment, a deep learning-based interpolation device in the spatial domain is provided, and the deep learning-based interpolation device in the spatial domain corresponds to the deep learning-based interpolation method in the spatial domain in the above-mentioned embodiment. Figure 3 As shown, the interpolation device in the space domain based on deep learning includes a matrix acquisition module 11, a training data acquisition module 12, a feature model acquisition module 13, an interpolation point feature matrix acquisition module 14, an inverse distance weighted value matrix acquisition module 15 and a space analysis module 16. The functional modules are described in detail as follows:

[0101] The matrix acquisition module 11 is used to acquire the characteristic matrix and the inverse distance characteristic matrix of the space to be analyzed.

[0102] The training data acquisition module 12 is used to construct training data based on the feature matrix and the inverse distance feature matrix.

[0103] The feature model acquisition module 13 is used to input the training data into the pre-training model for training to obtain the feature model.

[0104] The interpolation point feature matrix acquisition module 14 is used to construct an interpolation point feature matrix based on the features of the feature matrix.

[0105] The inverse distance weighted value matrix acquisition module 15 is used to input the interpolation point feature matrix into the inverse distance feature model for calculation to obtain the inverse distance weighted value matrix.

[0106] The spatial analysis module 16 is used to perform spatial analysis on the space to be analyzed based on the inverse distance weighted value matrix to obtain analysis results.

[0107] In one embodiment, the matrix acquisition module 11 further comprises:

[0108] The sampling point acquisition unit is used to acquire all sampling points of the space to be analyzed.

[0109] The sampling matrix acquisition unit is used to extract features from each sampling point based on a preset sampling factor to obtain a sampling matrix corresponding to each sampling point.

[0110] The feature matrix acquisition unit is used to merge all sampling matrices to obtain a feature matrix.

[0111] The inverse distance feature matrix acquisition unit is used to perform inverse distance calculation on all sampling points according to a preset distance calculation method to obtain an inverse distance feature matrix.

[0112] In one embodiment, the training data acquisition module 12 further includes:

[0113] The training data acquisition unit is used to merge the feature matrix and the inverse distance feature matrix based on a preset merging order to obtain training data.

[0114] In one embodiment, the feature model acquisition module 13 further includes:

[0115] The pre-training model building unit is used to build a pre-training model based on a deep neural network algorithm and input training data into the pre-training model.

[0116] The preprocessing unit is used to calculate the training data through the pre-training model to obtain the training data.

[0117] The loss unit is used to calculate the loss value of the training result according to a preset loss value calculation method to obtain the loss value.

[0118] The optimization unit is used to optimize the pre-trained model when the loss value is not lower than the preset loss value, and return to the step of calculating the training data through the pre-trained model to continue execution.

[0119] The feature model acquisition unit is used to use the pre-trained model as the feature model when the loss value is lower than the preset loss value.

[0120] In one embodiment, the spatial analysis module 16 further includes:

[0121] The analysis result acquisition unit is used to perform inverse interpolation calculation on the analysis space based on the inverse distance weighted value matrix and using the inverse distance characteristic formula to obtain the analysis result, wherein the analysis result is an interpolation result matrix.

[0122] In one embodiment, the analysis result acquisition unit further comprises:

[0123] The analysis result calculation unit is used to calculate the analysis result using the following formula:

[0124]

[0125] Among them, y (k) refers to the kth interpolation point, F(y (k) ) refers to the vector of the kth interpolation point in the interpolation result matrix, f(y (k) ) refers to the eigenvector of the kth interpolation point in the space to be analyzed, W i (x (i) ,y (k) ) refers to the inverse distance value between the i-th sampling point x and the k-th interpolation point y in the inverse distance weighted value matrix, the value range of k is (1, n), the value range of i is (1, m), n is the number of interpolation points, and m is the number of sampling points.

[0126] The meaning of "first" and "second" in the above modules / units is only to distinguish different modules / units, and is not used to define which module / unit has a higher priority or other limiting meanings. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The division of modules in this application is only a logical division, and there may be other division methods when implemented in actual applications.

[0127] For the specific definition of the interpolation device in the spatial domain based on deep learning, please refer to the definition of the interpolation method in the spatial domain based on deep learning above, which will not be repeated here. Each module in the above-mentioned interpolation device in the spatial domain based on deep learning can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0128] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 4As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data involved in the interpolation method in the spatial field based on deep learning. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an interpolation method in the spatial field based on deep learning is implemented.

[0129] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the interpolation method in the spatial domain based on deep learning in the above embodiment are implemented, such as Figure 2 Alternatively, when the processor executes the computer program, the functions of each module / unit of the interpolation device in the spatial domain based on deep learning in the above-mentioned embodiment are realized, for example Figure 3 The functions of modules 11 to 16 are shown in Figure 1. To avoid repetition, they will not be described here.

[0130] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the computer device, and various interfaces and lines are used to connect various parts of the entire computer device.

[0131] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the computer device by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, video data, etc.), etc.

[0132] The memory may be integrated into the processor or may be arranged separately from the processor.

[0133] In one embodiment, a computer-readable storage medium is provided on which a computer program is stored. When the computer program is executed by a processor, the steps of the interpolation method in the spatial domain based on deep learning in the above embodiment are implemented, for example Figure 2 Alternatively, when the computer program is executed by a processor, the functions of each module / unit of the interpolation device in the spatial domain based on deep learning in the above-mentioned embodiment are realized, for example, Figure 3 The functions of modules 11 to 16 are shown in Figure 1. To avoid repetition, they will not be described here.

[0134] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0135] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0136] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A spatial domain interpolation method based on deep learning, characterized in that: include: Acquire a feature matrix and an inverse distance matrix of a space to be analyzed, wherein the space to be analyzed is a random air sampling point in a certain area, and the feature matrix includes inherent features and associated features of the sampling points in the space to be analyzed, wherein the inherent features of the sampling points refer to features only related to the sampling points themselves, and the associated features include wind speed features, wind direction features, and air humidity features; Constructing training data based on the feature matrix and the inverse distance matrix; The training data is input into a pre-training model for training to obtain an inverse distance feature model, wherein the inverse distance feature model obtains an inverse distance weighted value of a point in a spatial domain through the feature of the point; Based on the features of the feature matrix, construct an interpolation point feature matrix; Inputting the interpolation point feature matrix into the inverse distance feature model for calculation to obtain an inverse distance weighted value matrix; Based on the inverse distance weighted value matrix, spatial analysis is performed on the space to be analyzed to obtain an analysis result; The inverse distance matrix is ​​obtained by calculating the inverse distance weighted value of each sampling point, and the inverse distance weighted value is: in F(x) is the sampling value of the sample point x, w i (x,y) The inverse distance weight between sampling points x and y is: in, is the Euclidean distance between points y and x; It is the influence distance function, which refers to a function set according to the influence of the sampling point on other sampling points or interpolation points; The step of performing spatial analysis on the space to be analyzed based on the inverse distance weighted value matrix to obtain the analysis result includes: Based on the inverse distance weighted value matrix, the inverse distance characteristic formula is used to perform inverse interpolation calculation on the space to be analyzed to obtain an analysis result, wherein the analysis result is an interpolation result matrix; The step of performing inverse interpolation calculation on the space to be analyzed based on the inverse distance weighted value matrix and using the inverse distance characteristic formula to obtain the analysis result comprises: The analysis results are calculated by the following formula: Among them, y (k) refers to the kth interpolation point, F(y (k) ) refers to the vector of the kth interpolation point in the interpolation result matrix, f(y (k) ) refers to the eigenvector of the kth interpolation point in the space to be analyzed, W i (x (i) ,y (k) ) refers to the inverse distance value between the i-th sampling point x and the k-th interpolation point y in the inverse distance weighted value matrix, the value range of k is (1, n), the value range of i is (1, m), n is the number of interpolation points, and m is the number of sampling points.

2. The method according to claim 1, characterized in that The step of obtaining the characteristic matrix and the inverse distance matrix of the space to be analyzed comprises: Get all sampling points of the space to be analyzed; Based on the preset sampling factors, feature extraction is performed on each of the sampling points to obtain a sampling matrix corresponding to each of the sampling points; All the sampling matrices are combined to obtain a feature matrix; According to a preset distance calculation method, inverse distance calculation is performed on all the sampling points to obtain an inverse distance matrix.

3. The method according to claim 1, characterized in that: The step of constructing training data based on the feature matrix and the inverse distance matrix comprises: Based on a preset merging order, the feature matrix and the inverse distance matrix are merged to obtain training data.

4. The method according to claim 1, characterized in that The step of inputting the training data into the pre-training model for training to obtain the inverse distance feature model includes: Based on a deep neural network algorithm, a pre-training model is constructed, and the training data is input into the pre-training model; The training data is calculated by the pre-training model to obtain a training result; According to a preset loss value calculation method, the loss value is calculated for the training result to obtain a loss value; When the loss value is not lower than the preset loss value, the pre-trained model is optimized, and the step of returning to the pre-trained model to pre-process the training data to obtain the pre-processed data is continued; When the loss value is lower than a preset loss value, the pre-trained model is used as an inverse distance feature model.

5. A spatial domain interpolation device based on deep learning, characterized in that: The interpolation method in the spatial domain based on deep learning according to any one of claims 1 to 4 comprises: A matrix acquisition module is used to acquire a feature matrix and an inverse distance matrix of a space to be analyzed. The space to be analyzed is a random air sampling point in a certain area. The feature matrix includes inherent features and associated features of the sampling points in the space to be analyzed. The inherent features of the sampling points refer to features that are only related to the sampling points themselves. The associated features include wind speed features, wind direction features, and air humidity features. A training data acquisition module, used for constructing training data based on the feature matrix and the inverse distance matrix; A feature model acquisition module, used for inputting the training data into a pre-training model for training to obtain an inverse distance feature model; An interpolation point feature matrix acquisition module is used to construct an interpolation point feature matrix based on the features of the feature matrix; An inverse distance weighted value matrix acquisition module is used to input the interpolation point feature matrix into the inverse distance feature model for calculation to obtain an inverse distance weighted value matrix; The spatial analysis module is used to perform spatial analysis on the space to be analyzed based on the inverse distance weighted value matrix to obtain an analysis result.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the deep learning-based spatial domain interpolation method as described in any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the deep learning-based spatial domain interpolation method as claimed in any one of claims 1 to 4 are implemented.

Images