Water conservation function evaluation method and device based on multi-source heterogeneous data fusion

Abstract

The present invention discloses a water conservation function evaluation method and device based on multi-source heterogeneous data fusion, which relates to the field of geographical ecological technology. The method comprises: obtaining multi-source heterogeneous data, and converting the multi-source heterogeneous data into unified raster data according to preset rules; obtaining an undirected graph of the raster data and a corresponding adjacency matrix, and determining an attribute enhancement matrix; establishing a neural network including a convolution module, inputting the undirected graph and the corresponding adjacency matrix and the attribute enhancement matrix into the graph convolution module for convolution operation, and learning the parameters that the neural network needs to learn to obtain a water conservation function evaluation model; when it is necessary to evaluate the water conservation function, obtaining the multi-source heterogeneous data to be evaluated, and inputting the multi-source heterogeneous data to be evaluated into the water conservation function evaluation model to obtain an evaluation result. The present invention solves the problem of low accuracy in the prior art when evaluating the water conservation function.

Description

Technical Field

[0001] The present invention relates to the field of geographical ecological technology, and in particular to a method and device for evaluating water conservation function based on multi-source heterogeneous data fusion. Background Art

[0002] Water conservation is considered one of the most important service functions of an ecosystem. It refers to the process and ability of an ecosystem to maintain water in the system under specific time and space conditions. It is often used as a key indicator to measure the health of an ecosystem. At present, the main water conservation capacity and measurement methods can be divided into two categories according to the differences in construction ideas: single model evaluation method and multi-model integrated evaluation method.

[0003] However, for single model evaluation methods, the comprehensive water storage capacity method and the canopy interception residual method are suitable for areas with abundant precipitation, but require a large amount of measured data; the annual flow method and the groundwater flow increment method evaluate water conservation by considering the total runoff of the basin and the groundwater flow increment in the vegetated area. The characteristics are that the required data are relatively simple, but the evaporation and transpiration of different vegetation types are not fully considered; and the precipitation collection method has low data requirements, simple calculations and easy to master, but its applicability and popularity are limited; due to the limitations of monitoring points and observation tools, these methods may encounter large errors when estimating water conservation in a large range. For multi-model integrated evaluation methods, compared with single model evaluation methods, multi-model integrated methods have more systematic and comprehensive characteristics, can more accurately estimate water conservation values, and can be applied to research areas with large temporal and spatial scales. However, some key parameters in the multi-model integrated evaluation method vary with the changes in the ecosystem. Over-reliance on the determination of these parameters by experts in the field can also easily lead to the problem of low accuracy of the evaluation results.

[0004] Therefore, both the single model evaluation method and the multi-model integrated evaluation method are prone to the problem of low accuracy of evaluation results. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a water conservation function evaluation method and device based on multi-source heterogeneous data fusion, aiming to solve the problem of inaccuracy in water conservation function evaluation in the prior art.

[0006] The embodiment of the present invention is implemented as follows:

[0007] A water conservation function evaluation method based on multi-source heterogeneous data fusion, the method comprising:

[0008] Acquire multi-source heterogeneous data composed of different types of evaluation index factors, and convert the multi-source heterogeneous data into unified raster data according to preset rules;

[0009] Obtaining an undirected graph of the raster data and a corresponding adjacency matrix, and determining an attribute enhancement matrix obtained by a feature matrix and an attribute matrix containing time series information;

[0010] Establishing a neural network including a convolution module, inputting the undirected graph and the corresponding adjacency matrix and attribute enhancement matrix into the graph convolution module for convolution operation, and learning the parameters that the neural network needs to learn to obtain a water conservation function evaluation model;

[0011] When it is necessary to evaluate the water conservation function, obtain multi-source heterogeneous data to be evaluated, and input the multi-source heterogeneous data to be evaluated into the water conservation function evaluation model to obtain the evaluation result;

[0012] The step of obtaining the undirected graph of the raster data and the corresponding adjacency matrix comprises:

[0013] The sampling points in the raster data are used as nodes in the graph structure, and the edges of the graph structure are established by measuring the similarity between the features of the sampling points to obtain the undirected graph;

[0014] The center coordinates of the sampling points are defined according to the spatial coordinates of the raster data, and then an adjacency matrix is ​​established through the Euclidean distance, and the final adjacency matrix is ​​obtained by normalizing with a Gaussian kernel function;

[0015] The expression of the convolution operation is:

[0016]

[0017]

[0018]

[0019] ：

[0020] in, express t-1 The output of the moment, gc(.) is the graph convolution process, 、 yes time t The update gate and reset gate of express t The output of the moment, 、 、 、 、 、 are all learnable parameters. is the attribute enhancement matrix, A is the adjacency matrix.

[0021] Furthermore, in the above-mentioned water conservation function evaluation method based on multi-source heterogeneous data fusion, the central coordinate expression of the sampling point is:

[0022] ;

[0023] The expression of the adjacency matrix is:

[0024] ;

[0025] in, c i and c j Respectively represent i , No. j The center coordinates of the sampling points, 、 are the coordinates of the boundary vertices of the preset resolution around the sampling point, 、 is the width and height of the bounding box of the area centered at the sampling point, The adjacency matrix i Line j Elements of the column, parameters is the scaling factor, τ is the preset threshold.

[0026] Furthermore, in the water conservation function evaluation method based on multi-source heterogeneous data fusion, the expression of the characteristic matrix is:

[0027] X∈ ；

[0028] The expression of the attribute matrix is:

[0029] ；

[0030] The expression of the attribute enhancement matrix is:

[0031]

[0032] Where X is the feature matrix, is the attribute matrix, is the attribute enhancement matrix, Nis the number of sampling points in the network, P represents the number of node time series features, t Represents time, k is the impact factor, m Represents a time window.

[0033] Furthermore, the water conservation function evaluation method based on multi-source heterogeneous data fusion, wherein the method also includes:

[0034] The output values ​​of the convolution operation are classified and evaluated through a fully connected module, and the calculated values ​​are calculated through an error function, and back propagation is performed to improve the accuracy of model training;

[0035] Among them, the output layer uses the Softmax function for classification, and the Softmax function is defined as follows:

[0036] ；

[0037] Loss Function Loss is defined as follows:

[0038] ;

[0039] in, Indicates i The sampling points belong to j The probability of the categories, j∈ {0, 1, 2, 3, 4, 5}, = [ ，…， ], Represents the dimensional vector output by the network of the i-th sampling point, d express The index of the element in N represents the total number of samples, i represents the index of the sample, = [ , … , ] ， Indicates i The code of the category to which the sample belongs.

[0040] Furthermore, in the water conservation function evaluation method based on multi-source heterogeneous data fusion, the step of converting the multi-source heterogeneous data into unified raster data according to preset rules includes:

[0041] The multi-source heterogeneous data are projected onto the same coordinate system and their representation forms are unified to obtain the raster data.

[0042] Another object of the present invention is to provide a water conservation function evaluation device based on multi-source heterogeneous data fusion, the device comprising:

[0043] An acquisition module is used to acquire multi-source heterogeneous data composed of different types of evaluation index factors, and convert the multi-source heterogeneous data into unified raster data according to preset rules;

[0044] A determination module is used to obtain an undirected graph of the raster data and a corresponding adjacency matrix, and determine an attribute enhancement matrix obtained by a feature matrix and an attribute matrix containing time series information;

[0045] A learning module is used to establish a neural network including a convolution module, input the undirected graph and the corresponding adjacency matrix and attribute enhancement matrix into the graph convolution module for convolution operation, and learn the parameters that the neural network needs to learn to obtain a water conservation function evaluation model;

[0046] An evaluation module, used for obtaining multi-source heterogeneous data to be evaluated when water conservation function evaluation is required, and inputting the multi-source heterogeneous data to be evaluated into the water conservation function evaluation model to obtain an evaluation result;

[0047] The step of obtaining the undirected graph of the raster data and the corresponding adjacency matrix comprises:

[0048] The sampling points in the raster data are used as nodes in the graph structure, and the edges of the graph structure are established by measuring the similarity between the features of the sampling points to obtain the undirected graph;

[0049] The center coordinates of the sampling points are defined according to the spatial coordinates of the raster data, and then an adjacency matrix is ​​established through the Euclidean distance, and the final adjacency matrix is ​​obtained by normalizing with a Gaussian kernel function;

[0050] The expression of the convolution operation is:

[0051]

[0052]

[0053]

[0054] ：

[0055] in, express t-1 The output of the moment, gc(.) is the graph convolution process, 、 yes time t The update gate and reset gate of express t The output of the moment, 、 、 、 、 、 are all learnable parameters. is the attribute enhancement matrix, A is the adjacency matrix.

[0056] Another object of the present invention is to provide a readable storage medium having a computer program stored thereon, wherein the program implements the steps of the above method when executed by a processor.

[0057] Another object of the present invention is to provide an electronic device, comprising 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 method when executing the program.

[0058] The present invention obtains multi-source heterogeneous data composed of different types of evaluation index factors, and converts the multi-source heterogeneous data into unified raster data according to preset rules; obtains an undirected graph of raster data and a corresponding adjacency matrix, and determines the attribute enhancement matrix obtained by the feature matrix and attribute matrix containing time series information; establishes a neural network including a convolution module, inputs the undirected graph and the corresponding adjacency matrix and attribute enhancement matrix into the graph convolution module for convolution operation, and learns the parameters that the neural network needs to learn to obtain a water conservation function evaluation model; when it is necessary to evaluate the water conservation function, obtain the multi-source heterogeneous data to be evaluated, and input the multi-source heterogeneous data to be evaluated into the water conservation function evaluation model to obtain the evaluation result. By analyzing the comprehensive impact of various influencing factors on the water conservation function, and by characterizing and learning the features of massive data, a water conservation function evaluation model is constructed. The model combines the spatial feature extraction capability with the temporal feature extraction advantage, thereby improving the accuracy of the water conservation function evaluation, and solving the problem of low accuracy in the water conservation function evaluation in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] FIG1 is a schematic diagram of the structure of a preset neural network in a water conservation function evaluation method based on multi-source heterogeneous data fusion in one embodiment of the present invention;

[0060] Figure 2It is a flow chart of a water conservation function evaluation method based on multi-source heterogeneous data fusion in the first embodiment of the present invention;

[0061] Figure 3 It is a schematic diagram of the normalized vegetation index NDVI preprocessing results in the water conservation function evaluation method based on multi-source heterogeneous data fusion in the first embodiment of the present invention;

[0062] Figure 4 It is a schematic diagram of land use type preprocessing results in the water conservation function evaluation method based on multi-source heterogeneous data fusion in the first embodiment of the present invention;

[0063] Figure 5 It is a schematic diagram of the slope preprocessing results in the water conservation function evaluation method based on multi-source heterogeneous data fusion in the first embodiment of the present invention;

[0064] Figure 6 It is a schematic diagram of the slope preprocessing results in the water conservation function evaluation method based on multi-source heterogeneous data fusion in the first embodiment of the present invention;

[0065] Figure 7 It is a schematic diagram of precipitation preprocessing results in the water conservation function evaluation method based on multi-source heterogeneous data fusion in the first embodiment of the present invention;

[0066] Figure 8 It is a schematic diagram of evapotranspiration preprocessing results in the water conservation function evaluation method based on multi-source heterogeneous data fusion in the first embodiment of the present invention;

[0067] Fig. 9 It is a structural block diagram of a water conservation function evaluation device based on multi-source heterogeneous data fusion in the second embodiment of the present invention.

[0068] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0069] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0070] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0071] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0072] The following embodiments can be applied to Figure 1 In the neural network shown in FIG. 1 , a water conservation function evaluation model is obtained by training the neural network. The water conservation function evaluation model is used to evaluate the water conservation function. Specifically, the neural network mainly includes a graph convolution module (GUN), a spatiotemporal convolution module (GRU) and a fully connected module. h t-n is the hidden state at the previous time step, y t-n It is the predicted value of the model output at different time steps. Specifically, the graph convolution module is responsible for processing spatial information. This module uses the graph convolution neural network to process the spatial correlation in the remote sensing data, so as to effectively extract spatial features. The spatiotemporal convolution module processes time series data. The fully connected module integrates the outputs of the graph convolution module and the spatiotemporal convolution module, and finally generates the evaluation results of the water conservation value through the fully connected network. By comparing with the measured values, the back propagation algorithm is used to optimize the network parameters and improve the evaluation accuracy.

[0073] How to improve the accuracy of water conservation function evaluation will be described in detail below with reference to specific embodiments and drawings.

[0074] Embodiment 1

[0075] See also Figure 2 , which shows a water conservation function evaluation method based on multi-source heterogeneous data fusion in the first embodiment of the present invention, and the method includes steps S10 to S13.

[0076] Step S10, obtaining multi-source heterogeneous data composed of different types of evaluation index factors, and converting the multi-source heterogeneous data into unified raster data according to preset rules.

[0077] Among them, constructing a reasonable evaluation index system is crucial for accurately evaluating the water conservation function. Determining the evaluation index factors for the evaluation of water conservation function is the basis for constructing the evaluation system. Evaluation index factors include but are not limited to factors such as precipitation, topography, vegetation coverage, soil type and hydrogeological conditions. These factors directly affect the conservation and loss of water, and are key natural factors for evaluating the water conservation function. When selecting evaluation indicators, it is necessary to consider the feasibility of obtaining, operating and quantifying the indicators. This means that the selected indicators should be able to be obtained through existing technical means and data resources, and can maintain a certain degree of accuracy and repeatability during the operation process. In addition, these indicators should be able to be quantified, that is, to reflect the strength of the water conservation function through numerical values, which is crucial for the objectivity and accuracy of the evaluation. Finally, based on the selection of the above-mentioned natural factors and evaluation indicators, the constructed evaluation index system should be able to fully reflect the geographical environment characteristics of the target study area. In the embodiment of the present invention, the evaluation index factors for the evaluation of the water conservation function are established for the water conservation function area in the upper reaches of the Yellow River in Gannan, and the evaluation index system is further constructed according to the selected evaluation index factors.

[0078] Specifically, the geographical environment characteristics of the upper reaches of the Yellow River in Gannan are analyzed in detail, including climate conditions, topography, soil type, vegetation distribution, etc. These natural conditions are the basis of the water conservation function and are crucial to the selection of evaluation index factors. According to the evaluation principles of scientificity, operability, regionality and dynamicity, the evaluation index factors mainly selected by the embodiment of the present invention include four aspects: climate conditions, vegetation coverage, terrain conditions and human activities. Climate conditions are the basic factors that determine the water conservation function, including precipitation, temperature, evaporation, etc. Due to its unique geographical location and high altitude characteristics in Gannan, the impact of climate factors on the water conservation function is particularly significant. Precipitation directly determines the replenishment of regional water resources, while temperature and evaporation affect the rate of water loss. Vegetation coverage is an important factor affecting soil moisture evaporation and surface runoff. The vegetation types in Gannan are diverse, from alpine meadows to forests to river valley grasslands, and the contributions of various types of vegetation to water conservation are also different. Topographic conditions, such as slope, direction and height, directly affect the flow direction and water collection capacity of rainwater. The terrain in Gannan is complex, from plateaus to mountains to river valleys, and the water conservation capacity under different terrain conditions varies significantly. Human activities, including agricultural farming, urban construction, industrial production, and water conservancy facilities construction, have direct and indirect impacts on water conservation functions.

[0079] Furthermore, the evaluation index factors include:

[0080] (1) Climate factors: First of all, precipitation is the most important factor affecting surface runoff, thus affecting water conservation function. Evaporation and temperature have similar effects, but the influence of evaporation on water conservation is much greater than that of temperature. Finally, precipitation and evaporation are selected as the influencing factors of climate factors. Studies have shown that water conservation function has a high positive correlation with precipitation and a high negative correlation with evaporation.

[0081] (2) Vegetation factors: Vegetation plays a role in water conservation, which is specifically reflected in the redistribution of atmospheric precipitation by the vegetation canopy and litter layer. Studies have shown that different vegetation types have different characteristics of their canopy and litter layer, and their water interception capacity is also different. The greater the vegetation coverage, the higher the water interception capacity, that is, the higher the water conservation capacity. The vegetation index NDVI can show the factors behind the plant canopy, such as soil, wetlands, snow, dead leaves and plant surface roughness, and is associated with the coverage of vegetation. Finally, the NDVI index with comprehensive environmental expression was selected as the influencing factor in the vegetation factor.

[0082] (3) Topographic factors: A large number of research documents show that water conservation functions vary in space. The water conservation capacity increases with the increase of slope, and the water conservation capacity of shady slopes is also higher than that of sunny slopes. Slope and slope aspect are selected as the influencing factors of topographic factors.

[0083] (4) Human activity factors: Land use types actually reflect the impact of human activity factors. Human activities within the study area have an impact on water conservation functions. For example, woodlands, shrubs and grasslands have a strong ability to intercept and store precipitation due to their dense vegetation canopies and a large amount of litter, so they play a significant role in protecting and maintaining water sources. In contrast, the vegetation root system of farmland is relatively shallow, and its water conservation function is relatively weak. Bare land has poor water conduction and precipitation regulation capabilities because its surface is covered with rocks and gravel and lacks natural vegetation cover. In addition, bare land is often used for construction and population gathering areas, so it is greatly affected by human factors, further reducing its water conservation function.

[0084] In addition, due to the rich content and diverse data sources of the data composed of multiple evaluation index factors, the data standards, models, formats, precision, storage forms, etc. are quite different. It is necessary to merge various types of data with a large amount of data, establish dynamic detection data sets for different periods and different observation scales, and form a unified representation form of evaluation and monitoring data. Therefore, the data obtained by the embodiment of the present invention are all pre-processed under the technical support of ArcGIS software. After data pre-processing, multiple influencing factor data with different sources and different structures are projected into the same coordinate system, and their representation forms are unified into raster data. Then, ArcGIS is used to calibrate the sampling points of the pre-processed raster data through operations such as creating a fishing net, raster conversion, and calibrating the study area. According to the different pixel sizes of the fishing net, the number of sampling points will change accordingly. In order to unify the spatial coordinate system, the present invention unifies the pixel size to 500*500 in the process of calibrating the sampling points, providing support for the realization of the model.

[0085] For example, the Normalized Difference Vegetation Index (NDVI) is derived from the MOD13A1 data of the MODIS product, which has a spatial resolution of 500m and a temporal resolution of 16 days. The NDVI raw data in the projection coordinate system is extracted, mosaicked, reprojected, and quantified to obtain the NDVI raster data file (TIF) in the specific study area. The results are shown in Figure 2. Figure 3 As shown;

[0086] The land use type data comes from the MCD12Q1 data of the MODIS product, with a spatial resolution of 500m and a temporal resolution of 1year. After comprehensively considering the water conservation function assessment needs and the characteristics of the ecosystem in the study area, the present invention selected the International Geo-Biosphere Program IGBP classification scheme to further screen the MCD12Q1 data, and obtained the initial land use data with a total of 17 land use types, including 11 natural vegetation types, 3 artificially altered types, and 3 non-vegetation types. The results are as follows Figure 4 As shown;

[0087] The slope and aspect data are all derived from the DEM elevation data of the geospatial data cloud, with a spatial resolution of 30m and a temporal resolution of 1year. After the DEM data in the study area is projected and the elevation is classified, the initial slope and aspect data required for the study can be calculated and analyzed based on the DEM data. The aspect data is divided into several different terrain slope directions, such as plane, north, northeast, east, east-west, etc. The results are as follows: Figure 5 , Figure 6 As shown;

[0088] The precipitation remote sensing data comes from the National Tibetan Plateau Data Center, with a spatial resolution of 1km and a temporal resolution of 1month. First, convert it into raster data, extract the precipitation values ​​by month according to the time field, and define its geographic coordinate system. Then project the raster data in the geographic coordinate system, and keep the projection coordinate system consistent with other data factors. Finally, extract the precipitation data in the study area with a time series of one month according to the mask. The results are as follows: Figure 7 As shown;

[0089] The evapotranspiration data comes from the MODIS product MOD16A2 data, with a spatial resolution of 500m and a temporal resolution of 8 days. Its preprocessing operation is similar to the NDVI index and land use type data mentioned above. The results are as follows Figure 8 shown.

[0090] Step S11, obtaining an undirected graph of the raster data and a corresponding adjacency matrix, and determining an attribute enhancement matrix obtained by a feature matrix and an attribute matrix containing time series information.

[0091] Among them, the input of the graph convolutional neural network is an undirected graph and its corresponding adjacency matrix. The sampling points in the raster data are used as nodes in the graph structure to represent the comprehensive geographical features at a specific spatial location; the edges of the graph structure are established by measuring the similarity between the features of the raster sampling points, representing the spatial relationship between these features. The water conservation function study area is composed of a variety of ecological environments, so it is very meaningful to define the adjacency matrix based on the geographical local correlation in the study area. The adjacency matrix is ​​constructed by Euclidean distance, and the dependency between sampling points is highly correlated with the distance between them. First, the central coordinates of the sampling points are defined according to the spatial coordinates of the raster data, and then the adjacency matrix is ​​established by Euclidean distance, and the final adjacency matrix is ​​obtained by normalization with the Gaussian kernel function.

[0092] Specifically, the center coordinate expression of the sampling point is:

[0093] ;

[0094] The expression of the adjacency matrix is:

[0095] ;

[0096] in, c i and c j Respectively represent i , No. j The center coordinates of the sampling points, 、 are the coordinates of the boundary vertices of the preset resolution around the sampling point, 、 is the width and height of the bounding box of the area centered at the sampling point, The adjacency matrix i Line j Elements of the column, parameters is the scaling factor, τ is the preset threshold.

[0097] The complex dependencies between nodes are modeled by gradually aggregating information in the neighborhood, and then a simple global maximum pooling strategy is adopted for graph features. GCN transfers information between neighboring nodes and updates each node according to a predefined adjacency matrix, enabling the model to effectively capture the spatial dependencies between objects.

[0098] Furthermore, it has a simpler structure, fewer parameters and faster convergence speed. Therefore, the present invention uses GRU to capture the time series dependency in remote sensing data. First, the time correlation needs to be represented. Define the feature matrix X∈ , which contains the time series information of the node. Where N is the number of sampling points in the network, and P represents the number of node time series features, such as the length of the historical time series. X∈ Represents the time series characteristics of the influencing factors of each sampling point at time t.

[0099] In the present invention, attribute-enhanced graph convolution operation is applied to the water conservation function evaluation model. The attribute enhancement unit integrates the characteristics of the time series of influencing factors and the characteristics of precipitation and evapotranspiration in the time series to characterize the different effects of different influencing factors on the water conservation function. The augmented matrix with the weighted adjacency matrix is ​​incorporated into the evaluation model to enhance the extraction of real-time and spatial dependencies.

[0100] Attribute matrix D∈ Represents k influencing factors at time t. At time t, the extended time window m+1 is used to express the attribute information instead of the original time window; that is, the attribute matrix is ​​expressed as: The attribute enhancement matrix S can be obtained by combining the feature matrix X and the attribute matrix Infer that:

[0101] .

[0102] Step S12, establish a neural network including a convolution module, input the undirected graph and the corresponding adjacency matrix and attribute enhancement matrix into the graph convolution module for convolution operation, and learn the parameters that need to be learned by the neural network to obtain a water conservation function evaluation model.

[0103] Among them, the expression of the convolution operation is:

[0104]

[0105]

[0106]

[0107] :

[0108] in, express t-1 The output of the moment, gc(.) is the graph convolution process, 、 yes time t The update gate and reset gate of express t The output of the moment, 、 、 、 、 、 are all learnable parameters. is the attribute enhancement matrix, A is the adjacency matrix.

[0109] The output values ​​of the convolution operation are classified and evaluated through a fully connected module, and the calculated values ​​are calculated through an error function, and back propagation is performed to improve the accuracy of model training;

[0110] Among them, the output layer uses the Softmax function for classification, and the Softmax function is defined as follows:

[0111] ；

[0112] Loss Function Loss is defined as follows:

[0113] ;

[0114] in, Indicates i The sampling points belong to j The probability of the categories, j∈ {0, 1, 2, 3, 4, 5}, = [ ，…， ], Represents the dimensional vector output by the network of the i-th sampling point, d express The index of the element in N represents the total number of samples, i represents the index of the sample, = [ , … , ] ， Indicates i The encoding of the category to which the sample belongs, specifically, one-hot encoding.

[0115] In summary, the water conservation function evaluation method based on multi-source heterogeneous data fusion in the above embodiment of the present invention obtains multi-source heterogeneous data composed of different types of evaluation index factors, and converts the multi-source heterogeneous data into unified raster data according to preset rules; obtains an undirected graph of the raster data and a corresponding adjacency matrix, and determines the attribute enhancement matrix obtained by the feature matrix and the attribute matrix containing time series information; establishes a neural network including a convolution module, inputs the undirected graph and the corresponding adjacency matrix and the attribute enhancement matrix into the graph convolution module for convolution operation, and learns the parameters that the neural network needs to learn to obtain a water conservation function evaluation model; when it is necessary to evaluate the water conservation function, obtain the multi-source heterogeneous data to be evaluated, and input the multi-source heterogeneous data to be evaluated into the water conservation function evaluation model to obtain the evaluation result. By analyzing the comprehensive impact of various influencing factors on water conservation function, and through feature representation and learning of massive data, a water conservation function evaluation model is constructed. The model combines the advantages of spatial feature extraction and temporal feature extraction, thereby improving the accuracy of water conservation function evaluation and solving the problem of low accuracy in water conservation function evaluation in existing technologies.

[0116] Embodiment 2

[0117] See also Fig. 9 , shown is a water conservation function evaluation device based on multi-source heterogeneous data fusion proposed in the second embodiment of the present invention, the device comprising:

[0118] The acquisition module 100 is used to acquire multi-source heterogeneous data composed of different types of evaluation index factors, and convert the multi-source heterogeneous data into unified raster data according to preset rules;

[0119] A determination module 200 is used to obtain an undirected graph of the raster data and a corresponding adjacency matrix, and determine an attribute enhancement matrix obtained by a feature matrix and an attribute matrix containing time series information;

[0120] The learning module 300 is used to establish a neural network including a convolution module, input the undirected graph and the corresponding adjacency matrix and attribute enhancement matrix into the graph convolution module for convolution operation, and learn the parameters required to be learned by the neural network to obtain a water conservation function evaluation model;

[0121] The evaluation module 400 is used to obtain multi-source heterogeneous data to be evaluated when water conservation function evaluation is required, and input the multi-source heterogeneous data to be evaluated into the water conservation function evaluation model to obtain an evaluation result.

[0122] Furthermore, in the above-mentioned water conservation function evaluation device based on multi-source heterogeneous data fusion, the step of obtaining the undirected graph of the raster data and the corresponding adjacency matrix includes:

[0123] The sampling points in the raster data are used as nodes in the graph structure, and the edges of the graph structure are established by measuring the similarity between the features of the sampling points to obtain the undirected graph;

[0124] The center coordinates of the sampling points are defined according to the spatial coordinates of the grid data, and then an adjacency matrix is ​​established through the Euclidean distance, and the final adjacency matrix is ​​obtained by normalizing it with a Gaussian kernel function.

[0125] Furthermore, in the above-mentioned water conservation function evaluation device based on multi-source heterogeneous data fusion, the central coordinate expression of the sampling point is:

[0126] ;

[0127] The expression of the adjacency matrix is:

[0128] ;

[0129] in, c i and c j Respectively represent i , No. j The center coordinates of the sampling points, 、 are the coordinates of the boundary vertices of the preset resolution around the sampling point, 、 is the width and height of the bounding box of the area centered at the sampling point, The adjacency matrix i Line j Elements of the column, parameters is the scaling factor, τ is the preset threshold.

[0130] Furthermore, in the above-mentioned water conservation function evaluation device based on multi-source heterogeneous data fusion, the expression of the characteristic matrix is:

[0131] X∈ ；

[0132] The expression of the attribute matrix is:

[0133] ；

[0134] The expression of the attribute enhancement matrix is:

[0135]

[0136] Where X is the feature matrix, is the attribute matrix, is the attribute enhancement matrix, N is the number of sampling points in the network, P represents the number of node time series features, t Represents time, k is the impact factor, m Represents a time window.

[0137] Furthermore, in the above-mentioned water conservation function evaluation device based on multi-source heterogeneous data fusion, the expression of the convolution operation is:

[0138]

[0139]

[0140]

[0141] ：

[0142] in, express t-1 The output of the moment, gc(.) is the graph convolution process, 、 yes time t The update gate and reset gate of express t The output of the moment, 、 、 、 、 、 are all learnable parameters. is the attribute enhancement matrix, A is the adjacency matrix.

[0143] Furthermore, in the above-mentioned water conservation function evaluation device based on multi-source heterogeneous data fusion, the method further comprises:

[0144] The output values ​​of the convolution operation are classified and evaluated through a fully connected module, and the calculated values ​​are calculated through an error function, and back propagation is performed to improve the accuracy of model training;

[0145] Among them, the output layer uses the Softmax function for classification, and the Softmax function is defined as follows:

[0146] ；

[0147] Loss Function Loss is defined as follows:

[0148] ;

[0149] in, Indicates i The sampling points belong to j The probability of the categories, j∈ {0, 1, 2, 3, 4, 5}, = [ ，…， ], Represents the dimensional vector output by the network of the i-th sampling point, d express The index of the element in N represents the total number of samples, i represents the index of the sample, = [ , … , ] ， Indicates i The code of the category to which the sample belongs.

[0150] Furthermore, in the above-mentioned water conservation function evaluation device based on multi-source heterogeneous data fusion, the step of converting the multi-source heterogeneous data into unified raster data according to preset rules includes:

[0151] The multi-source heterogeneous data are projected onto the same coordinate system and their representation forms are unified to obtain the raster data.

[0152] The functions or operation steps implemented when the above modules are executed are substantially the same as those in the above method embodiments, and will not be repeated here.

[0153] Embodiment 3

[0154] Another aspect of the present invention further provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of the method described in the first embodiment above.

[0155] Embodiment 4

[0156] Another aspect of the present invention provides an electronic device, which includes 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 method described in the first embodiment when executing the program.

[0157] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0158] Those skilled in the art will appreciate that the logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be specifically implemented in any computer-readable storage medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For purposes of this specification, a "computer-readable storage medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0159] More specific examples (a non-exhaustive list) of computer-readable storage media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable storage medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0160] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0161] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0162] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A water conservation function evaluation method based on multi-source heterogeneous data fusion, characterized in that: The method comprises: Acquire multi-source heterogeneous data composed of different types of evaluation index factors, and convert the multi-source heterogeneous data into unified raster data according to preset rules; Obtaining an undirected graph of the raster data and a corresponding adjacency matrix, and determining an attribute enhancement matrix obtained by a feature matrix and an attribute matrix containing time series information; Establishing a neural network including a convolution module, inputting the undirected graph and the corresponding adjacency matrix and attribute enhancement matrix into the graph convolution module for convolution operation, and learning the parameters that the neural network needs to learn to obtain a water conservation function evaluation model; When it is necessary to evaluate the water conservation function, obtain multi-source heterogeneous data to be evaluated, and input the multi-source heterogeneous data to be evaluated into the water conservation function evaluation model to obtain the evaluation result; The step of obtaining the undirected graph of the raster data and the corresponding adjacency matrix comprises: The sampling points in the raster data are used as nodes in the graph structure, and the edges of the graph structure are established by measuring the similarity between the features of the sampling points to obtain the undirected graph; The center coordinates of the sampling points are defined according to the spatial coordinates of the raster data, and then an adjacency matrix is ​​established through the Euclidean distance, and the final adjacency matrix is ​​obtained by normalizing with a Gaussian kernel function; The expression of the convolution operation is: ： in, express t-1 The output of the moment, gc(.) is the graph convolution process, 、 yes time t The update gate and reset gate of express t The output of the moment, 、 、 、 、 、 are all learnable parameters. is the attribute enhancement matrix, A is the adjacency matrix; Among them, multi-source heterogeneous data include climate factors, vegetation factors, and terrain factors. Climate factors include precipitation and evapotranspiration, vegetation factors include vegetation index, and terrain factors include slope and aspect.

2. The water conservation function evaluation method based on multi-source heterogeneous data fusion according to claim 1 is characterized in that: The center coordinate expression of the sampling point is: ； The expression of the adjacency matrix is: ； in, c i and c j Respectively represent i , No. j The center coordinates of the sampling points, 、 are the coordinates of the boundary vertices of the preset resolution around the sampling point, 、 is the width and height of the bounding box of the area centered at the sampling point, is the adjacency matrix i Line j Elements of the column, parameters is the scaling factor, τ is the preset threshold.

3. The water conservation function evaluation method based on multi-source heterogeneous data fusion according to claim 1 is characterized in that: The expression of the feature matrix is: X∈ ； The expression of the attribute matrix is: ； The expression of the attribute enhancement matrix is: Where X is the feature matrix, is the attribute matrix, is the attribute enhancement matrix, N is the number of sampling points in the network, P represents the number of node time series features, t Represents time, k is the impact factor, m Represents a time window.

4. The water conservation function evaluation method based on multi-source heterogeneous data fusion according to claim 1 is characterized in that: The method further comprises: The output values ​​of the convolution operation are classified and evaluated through a fully connected module, and the calculated values ​​are calculated through an error function, and back propagation is performed to improve the accuracy of model training; Among them, the output layer uses the Softmax function for classification, and the Softmax function is defined as follows: ； Loss Function Loss is defined as follows: ； in, Indicates i The sampling points belong to j The probability of the categories, j∈ {0, 1, 2, 3, 4, 5}, = [ ，…， ], Represents the dimensional vector output by the network of the i-th sampling point, d express The index of the element in N represents the total number of samples, i represents the index of the sample, = [ , … , ] ， Indicates i The code of the category to which the sample belongs.

5. The water conservation function evaluation method based on multi-source heterogeneous data fusion according to claim 1 is characterized in that: The step of converting the multi-source heterogeneous data into unified raster data according to preset rules includes: The multi-source heterogeneous data are projected onto the same coordinate system and their representation forms are unified to obtain the raster data.

6. A water conservation function evaluation device based on multi-source heterogeneous data fusion, characterized in that: The device comprises: An acquisition module is used to acquire multi-source heterogeneous data composed of different types of evaluation index factors, and convert the multi-source heterogeneous data into unified raster data according to preset rules; A determination module is used to obtain an undirected graph of the raster data and a corresponding adjacency matrix, and determine an attribute enhancement matrix obtained by a feature matrix and an attribute matrix containing time series information; A learning module is used to establish a neural network including a convolution module, input the undirected graph and the corresponding adjacency matrix and attribute enhancement matrix into the graph convolution module for convolution operation, and learn the parameters that the neural network needs to learn to obtain a water conservation function evaluation model; An evaluation module, used for obtaining multi-source heterogeneous data to be evaluated when water conservation function evaluation is required, and inputting the multi-source heterogeneous data to be evaluated into the water conservation function evaluation model to obtain an evaluation result; The step of obtaining the undirected graph of the raster data and the corresponding adjacency matrix comprises: The sampling points in the raster data are used as nodes in the graph structure, and the edges of the graph structure are established by measuring the similarity between the features of the sampling points to obtain the undirected graph; The center coordinates of the sampling points are defined according to the spatial coordinates of the raster data, and then an adjacency matrix is ​​established through the Euclidean distance, and the final adjacency matrix is ​​obtained by normalizing with a Gaussian kernel function; The expression of the convolution operation is: ： in, express t-1 The output of the moment, gc(.) is the graph convolution process, 、 yes time t The update gate and reset gate of express t The output of the moment, 、 、 、 、 、 are all learnable parameters. is the attribute enhancement matrix, A is the adjacency matrix; Among them, multi-source heterogeneous data include climate factors, vegetation factors, and terrain factors. Climate factors include precipitation and evapotranspiration, vegetation factors include vegetation index, and terrain factors include slope and aspect.

7. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

8. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method according to any one of claims 1 to 5 are implemented.

Images