Water conservation function evaluation method and device based on convolutional auto-encoder

Abstract

The invention discloses a water conservation function evaluation method and device based on a convolutional auto-encoder, and the method and device are used for obtaining and analyzing multi-source heterogeneous data of a water conservation function and carrying out the evaluation and classification of the water conservation function. The method comprises the following steps: firstly, acquiring comprehensive environmental data through a data acquisition and preprocessing module, and preprocessing to ensure the consistency and quality of the data; then, a convolutional auto-encoder is combined with a joint training method, deep feature extraction is performed on the data set, and the accuracy of feature representation is enhanced; and finally, carrying out clustering analysis on the extracted features to obtain a classification result of the water conservation function. According to the method, the problems of inaccurate multi-source heterogeneous data feature extraction and inaccurate classification analysis are effectively solved, and reliable technical support is provided for water source management and decision making.

Description

Technical Field

[0001] The present invention relates to the fields of geographical ecological technology and artificial intelligence, and in particular to a water conservation function evaluation method and device based on a convolutional autoencoder. Background Art

[0002] Water conservation is an important component of ecosystem services and plays a key role in maintaining ecological balance and ensuring water resource security. Traditional methods for evaluating water conservation mainly rely on manual surveys and empirical judgments. These methods usually require a large amount of field data collection, which is cumbersome and costly. In addition, due to the limitations of data acquisition, traditional methods are difficult to provide comprehensive and accurate analysis in space and time.

[0003] As environmental changes intensify, the dynamic characteristics of water conservation functions become more and more obvious. Traditional evaluation methods are often unable to capture these changes in real time, resulting in lags and inaccuracies in decision support. Especially when facing complex ecosystems, the limitations of traditional methods are more prominent and cannot effectively deal with the interactive effects of multiple variables. Summary of the invention

[0004] In view of this, the object of the present invention is to provide a water conservation function evaluation method and device based on convolutional autoencoder, aiming to improve the accuracy and efficiency of water conservation function evaluation.

[0005] A first aspect of an embodiment of the present invention provides a water conservation function evaluation method based on a convolutional autoencoder, the method comprising:

[0006] Acquire multi-source heterogeneous data, and perform preprocessing and normalization processing, wherein the multi-source heterogeneous data includes vegetation data, terrain data, meteorological data, soil data, and human activity data, and the preprocessing is used to ensure the consistency of the multi-source heterogeneous data in space and time;

[0007] The preprocessed and normalized data is input into a convolutional autoencoder, and a joint training strategy is adopted to train a feature extraction model, wherein the feature extraction model includes a convolutional autoencoder composed of an encoder and a decoder, and a classifier. In the process of training using the joint training strategy, the encoder is shared and combined with the feedback of the classifier to optimize the expression of the features;

[0008] The feature representation extracted by the convolutional autoencoder is input into the evaluation model to perform cluster analysis to quantitatively analyze the water conservation function and complete the evaluation.

[0009] Furthermore, the pre-processing step includes:

[0010] The multi-source heterogeneous data are converted into raster data in a unified coordinate system and time format by ArcGIS software;

[0011] The grid data is further divided into grid units of 1 km×1 km in size using the fishing net technology, wherein each grid unit represents a specific geographical location and contains multi-source heterogeneous data of the geographical location.

[0012] Furthermore, Z-score standardization is used for normalization, and the expression is:

[0013]

[0014] Among them, x represents the original data value, μ is the mean of the data, and σ is the standard deviation of the data.

[0015] Furthermore, the convolutional layer in the convolutional autoencoder applies the convolution kernel to operate on the data. Specifically, assuming that the net input z of the lth layer is (l) is the activity value a of the l-1th layer (l-1) and the convolution kernel w (l) The convolution expression is:

[0016] z (l) =w (l) *a (l-1) +b (l)

[0017] Among them, the convolution kernel w (l) is the learnable weight vector, b (l) is a learnable bias;

[0018] The expression of the activation function is:

[0019] ReLU(x)=max(0,x).

[0020] Furthermore, the output expression of the convolutional layer in the convolutional autoencoder is:

[0021] a (l) =ReLU(z (l) )=ReLU(0,w (l) *a (l-1) +b (l) )

[0022] The output expression of the pooling layer is:

[0023] a (l) =MaxPool(a (l-1) )

[0024] The output expression of the fully connected layer is:

[0025] a(l) =σ(W (l) *a (l-1) +b (l) ).

[0026] Furthermore, the joint loss expression of the joint training strategy is:

[0027] JointLoss=α×MSE+β×CrossEntropy;

[0028] Among them, MSE is the mean square error, CrossEntropy is the cross entropy, α and β are the weight coefficients for balancing the two losses, and JointLoss is the joint loss;

[0029] The expression of mean square error is:

[0030]

[0031] Where n is the number of samples, y i is the original data, is the reconstructed data, MSE is the mean square error;

[0032] The expression of cross entropy loss is:

[0033] .

[0035] Furthermore, the square error expression of K-means in the evaluation model is:

[0036]

[0037] Where J is the total squared error, k is the number of clusters, and C i is the i-th cluster, x is a data point, μ i is the center or centroid of the ith cluster, ||x-μ i || 2 is the data point x and its cluster center μ i The square of the Euclidean distance between them.

[0038] A second aspect of an embodiment of the present invention provides a water conservation function evaluation device based on a convolutional autoencoder, which is used to implement the water conservation function evaluation method based on a convolutional autoencoder described in the first aspect, and the system includes:

[0039] An acquisition module, used to acquire multi-source heterogeneous data, and perform preprocessing and normalization processing, wherein the multi-source heterogeneous data includes vegetation data, terrain data, meteorological data, soil data and human activity data, and the preprocessing is used to ensure the consistency of the multi-source heterogeneous data in space and time;

[0040] A training module, used for inputting the preprocessed and normalized data into the convolutional autoencoder, and adopting a joint training strategy to train a feature extraction model, wherein the feature extraction model includes a convolutional autoencoder composed of an encoder and a decoder, and a classifier. In the process of training using the joint training strategy, the encoder is shared and the feedback of the classifier is combined to optimize the expression of the features;

[0041] The input module is used to input the feature representation extracted by the convolutional autoencoder into the evaluation model to perform cluster analysis so as to quantitatively analyze the water conservation function and complete the evaluation.

[0042] A third aspect of an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the water conservation function evaluation method based on a convolutional autoencoder provided in the first aspect.

[0043] A fourth aspect of an embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the program, the water conservation function evaluation method based on convolutional autoencoder provided in the first aspect is implemented.

[0044] The present invention provides a water conservation function evaluation method and device based on a convolutional autoencoder, which is used to acquire and analyze multi-source heterogeneous data of water conservation functions and to evaluate and classify water conservation functions. First, through the data acquisition and preprocessing module, comprehensive environmental data is acquired, and preprocessing is performed to ensure the consistency and quality of the data. Then, the convolutional autoencoder is combined with the joint training method to perform deep feature extraction on the data set to enhance the accuracy of feature representation. Finally, the extracted features are clustered and analyzed to obtain the classification results of the water conservation function. The present invention effectively solves the problems of inaccurate feature extraction of multi-source heterogeneous data and inaccurate classification analysis, and provides reliable technical support for water management and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the framework of the water conservation function evaluation method based on the convolutional autoencoder in the first embodiment of the present invention;

[0046] Figure 2 Schematic diagram of the flow of the water conservation function evaluation method based on convolutional autoencoder in the second embodiment of the present invention;

[0047] Figure 3 This is a graph of vegetation coverage preprocessing results in the water conservation function evaluation method based on convolutional autoencoder in the second embodiment of the present invention;

[0048] Figure 4This is a graph of vegetation index preprocessing results in the water conservation function evaluation method based on convolutional autoencoder in the second embodiment of the present invention;

[0049] Figure 5 This is a graph of leaf area index preprocessing results in the water conservation function evaluation method based on convolutional autoencoder in Example 2 of the present invention;

[0050] Figure 6 This is a graph of slope preprocessing results in the water conservation function evaluation method based on convolutional autoencoder in the second embodiment of the present invention;

[0051] Figure 7 This is a graph of slope preprocessing results in the water conservation function evaluation method based on convolutional autoencoder in the second embodiment of the present invention;

[0052] Figure 8 This is a graph of altitude preprocessing results in the water conservation function evaluation method based on convolutional autoencoder in Example 2 of the present invention;

[0053] Fig. 9 This is a graph showing the soil texture preprocessing result in the water conservation function evaluation method based on the convolutional autoencoder in the second embodiment of the present invention;

[0054] Fig.10 This is a graph showing the soil type preprocessing result in the water conservation function evaluation method based on the convolutional autoencoder in the second embodiment of the present invention;

[0055] Fig.11 This is a graph of the preprocessing results of the available water content of plants in the water conservation function evaluation method based on the convolutional autoencoder in the second embodiment of the present invention;

[0056] Fig.12 This is a diagram showing the root restriction layer depth preprocessing result in the water conservation function evaluation method based on the convolutional autoencoder in the second embodiment of the present invention;

[0057] Fig.13 This is a graph of precipitation preprocessing results in the water conservation function evaluation method based on convolutional autoencoder in Example 2 of the present invention;

[0058] Fig.14 This is a graph of evapotranspiration preprocessing results in the water conservation function evaluation method based on convolutional autoencoder in Example 2 of the present invention;

[0059] Fig.15 This is a temperature preprocessing result diagram in the water conservation function evaluation method based on the convolutional autoencoder in the second embodiment of the present invention;

[0060] Fig.16 This is a graph of population density preprocessing results in the water conservation function evaluation method based on convolutional autoencoder in Example 2 of the present invention;

[0061] Fig.17 This is a graph of land use type preprocessing results in the water conservation function evaluation method based on convolutional autoencoder in the second embodiment of the present invention;

[0062] Fig.18 This is a module structure block diagram of a water conservation function evaluation device based on a convolutional autoencoder in Embodiment 3 of the present invention;

[0063] Fig.19 This is a structural block diagram of an electronic device provided in Embodiment 4 of the present invention.

[0064] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0065] In order to better understand the present invention, it will be described in detail below in conjunction with the relevant drawings. The drawings show several embodiments of the present invention. However, the present invention is not limited to these specific embodiments and can be implemented in many ways. The purpose of providing these embodiments is to more fully disclose the content of the present invention.

[0066] Unless otherwise specified, all technical and scientific terms used herein are consistent with the meanings commonly understood by those skilled in the art. The terms herein are only used to describe specific embodiments and are not intended to limit the present invention. The term "and / or" refers to any and all combinations of one or more related items.

[0067] The following embodiments can be applied to Figure 1 In the water conservation function evaluation method shown, the feature representation of multi-source heterogeneous data is obtained by training the convolutional autoencoder, and the water conservation function evaluation model is used to evaluate the water conservation function. Specifically, the convolutional autoencoder is used to extract features from the preprocessed data. The convolutional autoencoder can effectively capture the local features and spatial structure of the data through its convolutional layer, thereby generating high-quality feature representations. In order to further improve the feature extraction capability, the present invention adopts a joint training method to share the encoder part through multiple models to enhance the generalization ability and robustness of the feature representation. The joint training can utilize the complementary information of different models to improve the overall performance and prediction accuracy of the model. Finally, the extracted features are input into the K-means algorithm for cluster analysis. The K-means algorithm divides the data into multiple clusters, each cluster representing an evaluation and classification result of a water conservation function. In this way, complex environmental data is converted into intuitive classification information.

[0068] The present invention solves the problems of difficulty in integrating multi-source heterogeneous data, inaccurate feature extraction, and inaccurate classification analysis by combining convolutional autoencoders. The method can accurately extract and analyze key features related to water conservation and evaluate and analyze water conservation functions, providing scientific and reliable support for water management and decision-making.

[0069] How to improve the accuracy of water conservation function evaluation will be described in detail below with reference to specific embodiments and drawings.

[0070] Embodiment 1

[0071] See also Figure 1 , which is a schematic diagram of the framework of the water conservation function evaluation method based on the convolutional autoencoder in the first embodiment of the present invention, is described in detail as follows:

[0072] In the embodiment of the present invention, a convolutional autoencoder is first introduced. A convolutional autoencoder is a neural network model that is specifically used to automatically extract and reconstruct features of data. It effectively captures the local features and spatial structure of data through convolutional layers. The autoencoder consists of an encoder and a decoder. The encoder is responsible for converting input data into a low-dimensional feature representation, while the decoder is used to reconstruct the input data to ensure the validity and integrity of the features.

[0073] Next, the joint training strategy is introduced. In this framework, a multi-task learning framework is designed to train the convolutional autoencoder and the classifier simultaneously. The main task of the autoencoder is to reconstruct the input data to maintain the integrity of the features, while the classifier is responsible for predicting the label of the data. By sharing the encoder part, the model is able to optimize the reconstruction loss and the classification loss at the same time. This joint training strategy not only improves the accuracy of feature extraction, but also enhances the generalization ability and robustness of the model.

[0074] During training, the convolutional autoencoder and classifier are co-optimized through the shared encoder part. The reconstruction loss ensures that the autoencoder can effectively capture the key features of the data, while the classification loss guides the classifier to accurately predict the category label of the data. In this way, the model improves classification performance while maintaining high-quality feature representation.

[0075] Finally, the category features extracted by the trained convolutional autoencoder are input into the clustering algorithm for clustering analysis. By applying clustering algorithms such as K-means, the data is divided into multiple clusters, each cluster representing the evaluation and classification results of a water conservation function. This method effectively realizes the classification evaluation of water conservation functions, provides a scientific basis for management and decision-making, and solves the problems of difficulty in integrating multi-source heterogeneous data and inaccurate classification analysis.

[0076] Furthermore, the water conservation function evaluation method based on convolutional autoencoder proposed in the present invention can be summarized into the following three key steps:

[0077] First, train the autoencoder and classifier. At this stage, the present invention designs a multi-task learning framework to train the convolutional autoencoder and classifier at the same time. The task of the autoencoder is to convert the input data into a low-dimensional feature representation through the encoder part, and reconstruct the input data through the decoder part to ensure the integrity and validity of the features. The classifier is responsible for label prediction based on these features. By sharing the encoder part, the model can optimize the reconstruction loss and classification loss simultaneously during the training process. This joint training strategy not only improves the accuracy of feature extraction, but also enhances the generalization ability and robustness of the model, enabling it to better adapt to different types of data.

[0078] Secondly, feature extraction is the core step of the whole method. The trained convolutional autoencoder can extract high-quality feature representations from complex multi-source heterogeneous data. These feature representations retain the key information of the data and effectively remove noise and redundant information. Through the operation of the convolutional layer, the model can capture the local features and spatial structure of the data, thereby generating highly discriminative feature vectors. These features are not only crucial for classification tasks, but also provide a solid foundation for subsequent clustering analysis.

[0079] Finally, the extracted category features are input into the clustering algorithm for cluster analysis. By applying clustering algorithms such as K-means, the model divides the data into multiple clusters, each of which represents the evaluation and classification results of a water conservation function. Cluster analysis can effectively transform complex environmental data into intuitive classification information, helping to identify and evaluate the water conservation function of different regions. This method provides a scientific basis for water management and decision-making, solves the problems of difficulty in integrating multi-source heterogeneous data and inaccurate classification analysis, and ensures the efficiency and accuracy of water conservation function evaluation.

[0080] Embodiment 2

[0081] See also Figure 2 , which shows a water conservation function evaluation method based on a convolutional autoencoder in Embodiment 2 of the present invention, and the method includes steps S10 to S12.

[0082] Step S10, obtaining multi-source heterogeneous data, and performing preprocessing and normalization processing.

[0083] Among them, high-quality basic data provides solid support for subsequent analysis. Therefore, it is crucial to build a comprehensive and reasonable water conservation function evaluation index system. The establishment of this system can effectively integrate data from different sources and lay the foundation for the scientific evaluation of water conservation functions.

[0084] Specifically, the selection of evaluation index factors is the core of the construction of the entire index system. The evaluation index system includes four natural factors: vegetation factors, topographic factors, meteorological factors and soil factors, as well as human activity factors such as human activity factors. Natural factors play a decisive role in the water conservation function. Vegetation factors directly affect the water cycle and soil retention capacity by affecting transpiration and interception. Topographic factors determine the convergence and distribution characteristics of water sources by affecting the flow path and speed. Meteorological factors, such as precipitation and temperature, directly affect the recharge and evaporation process of water sources. Soil factors are closely related to the infiltration and storage of water. On the other hand, human activity factors cannot be ignored. Human activity factors include land use changes, agricultural activities and urbanization, which significantly affect the water conservation function by changing surface characteristics and hydrological processes. Reasonable assessment of the impact of human activities can help formulate more effective management strategies and reduce the adverse effects on water conservation functions.

[0085] In summary, the evaluation index system combining natural factors and human activity factors can not only comprehensively reflect the current status of water conservation function, but also provide a scientific basis for water source management and protection. This comprehensive evaluation method ensures the efficiency and accuracy of water conservation function evaluation and provides strong support for sustainable water resources management.

[0086] Further, in an embodiment of the present invention, various types of factors include some specific evaluation indicators. Vegetation factors mainly include indicators such as vegetation coverage. Vegetation coverage is an important indicator for evaluating regional ecological health and water conservation capacity, because vegetation affects water circulation through transpiration and interception, and enhances soil retention capacity. High vegetation coverage usually means better water conservation function. Topographic factors include indicators such as slope and aspect. Slope affects water flow velocity and soil erosion, while aspect affects light and water evaporation. These indicators are selected because they directly determine the convergence and distribution characteristics of water sources and have a significant impact on water conservation function. Meteorological factors cover indicators such as precipitation and evaporation. Precipitation is the main source of water replenishment, while evaporation affects water loss. Selecting these indicators can accurately reflect the impact of climatic conditions on water conservation and help evaluate the sustainability of water resources. Soil factors include indicators such as soil texture. Soil texture affects the infiltration and storage capacity of water and determines the water retention and water supply functions of soil. Selecting these indicators helps to evaluate the supporting role of soil for water conservation. Human activity factors such as population density and other indicators. Population density reflects the intensity of human activities and has a direct impact on land use change and water resource consumption. The selection of these indicators can help evaluate the pressure and impact of human activities on water conservation functions. These evaluation indicators are selected based on their direct impact and contribution to water conservation functions. By comprehensively analyzing these indicators, the water conservation capacity of the region can be comprehensively evaluated, providing a scientific basis and guidance for water resource management and protection.

[0087] Specifically, the evaluation indicators included in each evaluation factor are as follows:

[0088] (1) Vegetation factors: Among the vegetation factors, vegetation coverage, vegetation index (such as NDVI) and leaf area index (LAI) are key evaluation indicators. Vegetation coverage indicates the proportion of the surface covered by vegetation. High vegetation coverage usually means better water conservation function, because vegetation can reduce surface runoff by intercepting precipitation and increase soil infiltration and water retention capacity. The pre-processing results are as follows: Figure 3 The NDVI is an indicator of vegetation growth through remote sensing technology. A high NDVI value indicates healthy and dense vegetation, which helps to improve water retention and soil stability in the area, thereby enhancing water conservation capacity. The preprocessing results are shown in Figure 4 The leaf area index (LAI) is the total leaf area per unit surface area. A high LAI indicates dense vegetation, which can increase transpiration and photosynthesis, thereby promoting water circulation and soil moisture storage, and has a positive impact on water conservation function. The pretreatment results are shown in Figure 5 By comprehensively analyzing these vegetation factors, we can more accurately assess the water conservation capacity of a region and provide a scientific basis for ecological protection and water resources management.

[0089] (2) Topographic factors: Among the topographic factors, slope, aspect and altitude are important evaluation indicators. Slope refers to the degree of inclination of the surface. A steep slope will accelerate the flow of surface water, increase the risk of soil erosion, and thus affect the water conservation function. A gentler slope helps slow down the flow of water, promote water infiltration, and increase the water retention capacity of the soil. The pretreatment results are as follows: Figure 6 As shown in Figure 1. Slope aspect refers to the direction of the terrain surface relative to the sun's position. Different slope aspects affect light and temperature, thereby affecting vegetation growth and soil moisture. Sun-facing slopes usually receive more solar radiation, resulting in higher evaporation rates, which may reduce water retention capacity. The preprocessing results are shown in Figure 1. Figure 7 Altitude affects climate conditions such as temperature and precipitation. Higher altitudes are usually associated with lower temperatures and higher precipitation, which has a positive effect on water conservation. Altitude changes also affect vegetation types and soil properties, which in turn affect the water conservation capacity of the region. The preprocessing results are shown in Figure 8 By analyzing these topographic factors, we can better understand the impact of topography on water conservation function and provide a basis for effective water resources management.

[0090] (3) Soil factors: Among soil factors, soil texture, soil type, plant available water content and root restriction layer depth are key evaluation indicators. Soil texture refers to the ratio of sand, silt and clay particles in the soil. Texture affects the permeability and water retention capacity of the soil. Sandy soil has high permeability but poor water retention, while clay texture is the opposite. Moderate texture helps to effectively store and supply water. The pretreatment results are as follows: Fig. 9 Soil type involves the formation and classification characteristics of soil. Different types of soil have different nutrient and water retention capabilities. Some soil types are more suitable for vegetation growth, thereby enhancing the water conservation function of the region. The pretreatment results are shown in Fig.10 The plant available water content refers to the amount of water in the soil that can be absorbed and utilized by plant roots. A high content means that the soil can provide continuous water support for plants under drought conditions and enhance the stability of the ecosystem. The pretreatment results are shown in Fig.11 The root restriction layer depth refers to the depth of the layer in the soil that hinders the downward growth of plant roots. A deeper restriction layer allows plant roots to penetrate deeper into the soil, improving the absorption capacity of water and nutrients, which is beneficial to water conservation. The pretreatment results are shown in Fig.12 By comprehensively analyzing these soil factors, we can more accurately evaluate the supporting role of soil in water conservation and provide a scientific basis for ecological protection and water resources management.

[0091] (4) Meteorological factors: Among meteorological factors, precipitation, evapotranspiration and temperature are important evaluation indicators. Precipitation is the main source of water supply. Adequate precipitation helps to replenish surface water and groundwater resources and enhance the water conservation capacity of the region. The temporal and spatial distribution of precipitation also affects the sustainability and management strategy of water resources. The preprocessing results are as follows: Fig.13 As shown. Evapotranspiration includes the sum of evaporation and plant transpiration. It is a key process in the water cycle, affecting the availability of soil moisture and water sources. Higher evapotranspiration may lead to water loss and reduce water conservation function, so it is necessary to balance precipitation and evapotranspiration to maintain water resources stability. The pretreatment results are shown in Fig.14 Temperature affects evaporation rate and plant growth. Higher temperatures usually increase evaporation and transpiration, accelerating water cycle. Suitable temperatures help vegetation growth and improve the water conservation capacity of the region. The pretreatment results are shown in Fig.15 By analyzing these meteorological factors, we can better understand the impact of climate conditions on water conservation functions and provide a scientific basis for effective water resources management.

[0092] (5) Human activity factors: Among human activity factors, population density and land use type are key evaluation indicators. Population density reflects the intensity of human activities in a region. High population density is usually accompanied by higher water resource demand and pollution risks, which puts pressure on water conservation functions. Reasonable planning and management of population density can reduce the negative impact on water resources and promote sustainable development. The pre-processing results are as follows: Fig.16 As shown in Figure 1. Land use types include agriculture, urbanization, and forests. Changes in land use directly affect the water cycle and ecosystem health. Urbanization often increases the impervious area of ​​the surface, reducing water infiltration and conservation capacity, while forests and wetlands contribute to the natural conservation and protection of water sources. The pre-processing results are shown in Figure 1. Fig.17 By analyzing these human activity factors, we can better assess the impact of humans on water conservation functions and provide a scientific basis for formulating effective resource management and protection strategies.

[0093] It should be noted that when processing environmental data related to water conservation, that is, processing vegetation data, terrain data, meteorological data, soil data, and human activity data, a major challenge faced is that the data sources of various evaluation indicators are diverse and inconsistent in form. Therefore, it is necessary to unify these data in time and space to ensure the accuracy and consistency of subsequent analysis. The embodiment of the present invention achieves this goal through a series of steps.

[0094] First, all acquired data were preprocessed with the support of ArcGIS software. As a powerful geographic information system software, ArcGIS can effectively process and analyze spatial data. Through ArcGIS, multi-source heterogeneous data from different sources are converted into raster data in a unified coordinate system and time format. This process ensures the consistency of data in space and time, laying the foundation for subsequent analysis.

[0095] Next, the raster data is further divided into 1km×1km cells using the fishnet technique. Fishnet technology is a technique that divides spatial data into regular grids, which helps to analyze and process the data in a more detailed manner. Each raster cell represents a specific geographic location and contains all the relevant data for that location. This method not only improves the manageability of the data, but also makes subsequent data extraction and analysis more accurate.

[0096] After completing the spatial division of the data, the grid data is extracted. The purpose of this step is to extract information from each grid unit that is crucial to the evaluation of water conservation functions. This extracted information will serve as the input of the model and provide necessary data support for the comprehensive evaluation of water conservation functions.

[0097] In order to ensure that data from different sources and scales can be effectively integrated in the model, the Z-score standardization method is used to normalize the data after input. Z-score standardization is a commonly used data preprocessing method. It eliminates the influence of different scales and units by converting the data into a standard normal distribution with a mean of zero and a standard deviation of one. This normalization process ensures the alignment of each indicator in the model, so that each indicator can contribute fairly to the evaluation of water conservation function. Its expression is:

[0098]

[0099] Among them, x represents the original data value, μ is the mean of the data, and σ is the standard deviation of the data. Through this standardization, the influence of different data scales is eliminated, making the data suitable for further model analysis.

[0100] Through the above series of steps, the embodiment of the present invention successfully transforms complex multi-source heterogeneous data into a unified analysis format, providing high-quality basic data for the scientific evaluation of water conservation functions. This process not only improves the efficiency and accuracy of data processing, but also provides solid support for water resources management and ecological protection.

[0101] Step S11, input the preprocessed and normalized data into the convolutional autoencoder, and adopt a joint training strategy to train a feature extraction model.

[0102] In an embodiment of the present invention, the standardized data is input into a convolutional autoencoder for training. A convolutional autoencoder is an unsupervised learning model that can effectively extract deep features from data. By using convolutional layers, the autoencoder can capture the spatial features of the data, which is particularly important for processing rasterized environmental data. The structure of the autoencoder includes two parts: an encoder and a decoder, wherein the encoder is responsible for converting the input data into a low-dimensional feature representation, while the decoder attempts to reconstruct the input data from these features.

[0103] It should be noted that the convolutional autoencoder consists of an encoder network and a decoder network. The encoder network consists of two convolutional layers, two pooling layers, and one fully connected layer. The role of the encoder is to map the input data to a low-dimensional latent space through a series of neural network layers. This process aims to extract the main features and patterns of the data and achieve data compression and dimensionality reduction. The decoder network contains one fully connected layer and two transposed convolutional layers. In the convolutional autoencoder, the task of the decoder is to restore the latent variables generated by the encoder to the high-dimensional space of the original data.

[0104] Specifically, the convolution layer applies the convolution kernel to operate on the data. Assume that the net input z of the lth layer is (l) is the activity value a of the l-1th layer (l-1) and the convolution kernel w (l) The convolution expression is:

[0105] z (l) =w (l) *a (l-1) +b (l)

[0106] Among them, the convolution kernel w (l) is the learnable weight vector, b (l) is a learnable bias;

[0107] The expression of the activation function is:

[0108] ReLU(x)=max(0,x)

[0109] The output expression of the convolutional layer is:

[0110] a (l) =ReLU(z (l) )=ReLU(0,w (l) *a (l-1) +b (l) )

[0111] The output expression of the pooling layer is:

[0112] a (l) =MaxPool(a(l-1) )

[0113] The output expression of the fully connected layer is:

[0114] a (l) =σ(W (l) *a (l-1) +b (l) ).

[0115] In order to further improve the feature extraction capability, the present invention adopts a joint training strategy, that is, while training the convolutional autoencoder, a classifier is also trained. The task of the classifier is to perform specific classification tasks based on the extracted features, such as evaluating the level of water conservation function. By sharing the encoder part, the classifier and the autoencoder can work together. The features extracted by the encoder are not only used to reconstruct the input data, but also to support the classification task. This sharing mechanism enables the encoder to learn richer and more useful feature representations.

[0116] Among them, the joint loss expression of the joint training strategy is:

[0117] JointLoss=α×MSE+β×CrossEntropy;

[0118] Among them, MSE is the mean square error, CrossEntropy is the cross entropy, α and β are the weight coefficients for balancing the two losses, and JointLoss is the joint loss;

[0119] The expression of mean squared error (MSE) is:

[0120]

[0121] Where n is the number of samples, y i is the original data, It is to reconstruct the data;

[0122] The expression of cross entropy loss is:

[0123]

[0124] Where C is the number of categories, y i is the true label, is the probability predicted by the model.

[0125] The advantage of joint training is that it can guide the learning process of the encoder through feedback from the classification task. The classifier's error is back-propagated to the encoder part, prompting the encoder to optimize its feature extraction ability to better meet the needs of the classification task. This approach improves the overall performance of the model because the encoder takes into account the needs of multiple tasks when extracting features, not just data reconstruction.

[0126] Through this joint training strategy, the present invention achieves efficient feature extraction and classification capabilities. This not only improves the accuracy and robustness of the model in the evaluation of water conservation function, but also provides a reliable basis for subsequent water resources management and decision-making. The combination of convolutional autoencoders and classifiers fully taps the potential of deep learning models in processing complex environmental data and provides an innovative solution for the scientific evaluation of water conservation functions.

[0127] Step S12, inputting the feature representation extracted by the convolutional autoencoder into the evaluation model, performing cluster analysis, so as to quantitatively analyze the water conservation function and complete the evaluation.

[0128] Among them, the features extracted from multi-source heterogeneous data are input into the K-means algorithm for cluster analysis and divided into multiple clusters. Each cluster represents an evaluation classification result, which effectively transforms complex environmental data into intuitive classification information.

[0129] Specifically, the feature representation extracted from multi-source heterogeneous data is input into the K-means algorithm for clustering analysis. K-means is a classic unsupervised learning algorithm widely used in data mining and pattern recognition. Its core idea is to divide data points into k clusters, each cluster is represented by a centroid, and data points are assigned to the nearest cluster based on their distance to the centroid.

[0130] Among them, the square error expression of K-means in the evaluation model is:

[0131]

[0132] Where J is the total squared error, k is the number of clusters, and C i is the i-th cluster, x is a data point, μ i is the center or centroid of the ith cluster, ||x-μ i || 2 is the data point x and its cluster center μ i The square of the Euclidean distance between them.

[0133] In the present invention, the application of K-means algorithm transforms complex environmental data into intuitive classification information. First, the features extracted by convolutional autoencoder and joint training classifier represent the deep structure and pattern of the data. K-means algorithm uses these low-dimensional features to aggregate data points into several clusters. Each cluster represents a specific evaluation classification result, indicating different water conservation capacity levels.

[0134] Through this cluster analysis, users can extract meaningful patterns and trends from large amounts of complex data. This classification method not only simplifies the complexity of the data, but also provides clear guidance for subsequent analysis and decision-making. The characteristics and representativeness of each cluster can be further used to identify key areas or issues and provide targeted strategies for water resources management and ecological protection.

[0135] In addition, the efficiency and scalability of the K-means algorithm make it suitable for processing large-scale data sets. When processing environmental data, the amount of data is often very large and the dimensions are high. K-means can converge quickly and provide stable clustering results to meet the needs of data of different scales.

[0136] By inputting features into K-means for cluster analysis, the present invention effectively transforms complex environmental data into intuitive classification information. This method not only enhances the interpretability of data analysis, but also provides strong support for scientific research and policy formulation, helping to achieve more accurate water conservation function evaluation and management.

[0137] In summary, the present invention proposes a water conservation function evaluation method based on convolutional autoencoder, and constructs a systematic water conservation function evaluation system. First, by obtaining multi-source heterogeneous data and preprocessing them, the consistency of the data in space and time is ensured. Using the preprocessing steps supported by ArcGIS software, various data are uniformly converted into raster format and divided into 1km×1km raster units by fishing net technology. This meticulous spatial division lays a solid foundation for subsequent feature extraction. Next, the standardized data is input into the convolutional autoencoder. The convolutional autoencoder can extract representative deep features from complex environmental data through its deep network structure. The joint training strategy further enhances the ability of feature extraction. By sharing the encoder part and combining the feedback of the classifier, the expression of the feature is optimized. This collaborative training not only improves the accuracy of the model, but also ensures the diversity and practicality of the extracted features. After completing the feature extraction, the data is input into the K-means algorithm for cluster analysis. The K-means algorithm converts complex environmental information into intuitive classification results by dividing the data into multiple clusters. Each cluster represents a specific water conservation function evaluation result, providing clear classification information. This method simplifies the complexity of the data, allowing users to identify and understand the water conservation capacity and ecological health status of different regions. Through this systematic method, the present invention not only achieves effective classification of water conservation functions, but also provides a scientific basis for water resources management and ecological protection. The innovation and practicality of this method provide new ideas for the analysis of complex environmental data, help achieve more accurate water conservation function evaluation, and promote the sustainable development of the ecological environment.

[0138] Embodiment 3

[0139] See also Fig.18 , which is a water conservation function evaluation device based on a convolutional autoencoder proposed in Embodiment 3 of the present invention, wherein the water conservation function evaluation device based on a convolutional autoencoder 200 comprises:

[0140] An acquisition module 21 is used to acquire multi-source heterogeneous data and perform preprocessing and normalization processing, wherein the multi-source heterogeneous data includes vegetation data, terrain data, meteorological data, soil data and human activity data, and the preprocessing is used to ensure the consistency of the multi-source heterogeneous data in space and time;

[0141] The training module 22 is used to input the preprocessed and normalized data into the convolutional autoencoder, and adopt a joint training strategy to train a feature extraction model, wherein the feature extraction model includes a convolutional autoencoder composed of an encoder and a decoder, and a classifier. In the process of training using the joint training strategy, the encoder is shared and the feedback of the classifier is combined to optimize the expression of the features;

[0142] The input module 23 is used to input the feature representation extracted by the convolutional autoencoder into the evaluation model to perform cluster analysis to quantitatively analyze the water conservation function and complete the evaluation.

[0143] 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.

[0144] Embodiment 4

[0145] Another aspect of the present invention provides an electronic device, see Fig.19 , shown is an electronic device in Embodiment 4 of the present invention, including a memory 20, a processor 10, and a computer program 30 stored in the memory and executable on the processor. When the processor 10 executes the computer program 30, the water conservation function evaluation method based on the convolutional autoencoder as described above is implemented.

[0146] In some embodiments, the processor 10 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 20, such as executing access restriction programs.

[0147] Among them, the memory 20 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. The memory 20 may be an internal storage unit of an electronic device in some embodiments, such as a hard disk of the electronic device. The memory 20 may also be an external storage device of an electronic device in other embodiments, such as a plug-in hard disk equipped on the electronic device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Further, the memory 20 may also include both an internal storage unit and an external storage device of the electronic device. The memory 20 may be used not only to store application software and various types of data of the electronic device, but also to temporarily store data that has been output or is to be output.

[0148] It should be pointed out that Fig.19 The structure shown does not constitute a limitation on the electronic device. In other embodiments, the electronic device may include fewer or more components than those shown in the figure, or combine certain components, or arrange the components differently.

[0149] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the water conservation function evaluation method based on the convolutional autoencoder as described above.

[0150] 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 embodied in any computer-readable 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 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.

[0151] More specific examples of computer-readable media (a non-exhaustive list) 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 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.

[0152] 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, a plurality of 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.

[0153] 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.

[0154] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the attached claims.

Claims

1. A water conservation function evaluation method based on convolutional autoencoder, characterized in that: The method comprises: Acquire multi-source heterogeneous data, and perform preprocessing and normalization processing, wherein the multi-source heterogeneous data includes vegetation data, terrain data, meteorological data, soil data, and human activity data, and the preprocessing is used to ensure the consistency of the multi-source heterogeneous data in space and time; The preprocessed and normalized data is input into a convolutional autoencoder, and a joint training strategy is adopted to train a feature extraction model, wherein the feature extraction model includes a convolutional autoencoder composed of an encoder and a decoder, and a classifier. In the process of training using the joint training strategy, the encoder is shared and combined with the feedback of the classifier to optimize the expression of the features; The feature representation extracted by the convolutional autoencoder is input into the evaluation model to perform cluster analysis to quantitatively analyze the water conservation function and complete the evaluation.

2. The water conservation function evaluation method based on convolutional autoencoder according to claim 1 is characterized in that: The pre-processing steps include: The multi-source heterogeneous data are converted into raster data in a unified coordinate system and time format by ArcGIS software; The grid data is further divided into grid units of 1 km×1 km in size using the fishing net technology, wherein each grid unit represents a specific geographical location and contains multi-source heterogeneous data of the geographical location.

3. The water conservation function evaluation method based on convolutional autoencoder according to claim 1 is characterized in that: Z-score standardization is used for normalization, and the expression is: Among them, x represents the original data value, μ is the mean of the data, and σ is the standard deviation of the data.

4. The water conservation function evaluation method based on convolutional autoencoder according to claim 1 is characterized in that: The convolutional layer in the convolutional autoencoder applies the convolution kernel to operate on the data. Specifically, assuming that the net input z of the lth layer is (l) is the activity value a of the l-1th layer (l-1) and the convolution kernel w (l) The convolution expression is: z (l) =w (l) *a (l-1) +b (l) Among them, the convolution kernel w (l) is the learnable weight vector, b (l) is a learnable bias; The expression of the activation function is: ReLU(x)=max(0,x).

5. The water conservation function evaluation method based on convolutional autoencoder according to claim 4 is characterized in that: The output expression of the convolutional layer in the convolutional autoencoder is: a (l) =ReLU(z (l) )=ReLU(0,w (l) *a (l-1) +b (l) ) The output expression of the pooling layer is: a (l) =MaxPool(a (l-1) ) The output expression of the fully connected layer is: a (l) =σ(W (l) *a (l-1) +b (l) )。 6. The water conservation function evaluation method based on convolutional autoencoder according to claim 1 is characterized in that: The joint loss expression of the joint training strategy is: JointLoss=α×MSE+β×CrossEntropy; Among them, MSE is the mean square error, CrossEntropy is the cross entropy, α and β are the weight coefficients for balancing the two losses, and JointLoss is the joint loss; The expression of mean square error is: Where n is the number of samples, y i is the original data, is the reconstructed data, MSE is the mean square error; The expression of cross entropy loss is:

7. The water conservation function evaluation method based on convolutional autoencoder according to claim 1 is characterized in that: The square error expression of K-means in the evaluation model is: Where J is the total squared error, k is the number of clusters, and C i is the i-th cluster, x is a data point, μ i is the center or centroid of the ith cluster, ||x-μ i || 2 is the data point x and its cluster center μ i The square of the Euclidean distance between them.

8. A water conservation function evaluation device based on convolutional autoencoder, characterized in that: Used to implement the water conservation function evaluation method based on convolutional autoencoder as described in any one of claims 1 to 7, the device comprises: An acquisition module, used to acquire multi-source heterogeneous data, and perform preprocessing and normalization processing, wherein the multi-source heterogeneous data includes vegetation data, terrain data, meteorological data, soil data and human activity data, and the preprocessing is used to ensure the consistency of the multi-source heterogeneous data in space and time; A training module, used for inputting the preprocessed and normalized data into the convolutional autoencoder, and adopting a joint training strategy to train a feature extraction model, wherein the feature extraction model includes a convolutional autoencoder composed of an encoder and a decoder, and a classifier. In the process of training using the joint training strategy, the encoder is shared and the feedback of the classifier is combined to optimize the expression of the features; The input module is used to input the feature representation extracted by the convolutional autoencoder into the evaluation model to perform cluster analysis so as to quantitatively analyze the water conservation function and complete the evaluation.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the water conservation function evaluation method based on a convolutional autoencoder as described in any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: It comprises a memory, a processor and a computer program stored in the memory and running on the processor, and when the processor executes the program, it implements the water conservation function evaluation method based on convolutional autoencoder as described in any one of claims 1 to 7.

Images