A data processing method for long-term dynamic monitoring of water bodies

Abstract

This invention provides a data processing method for long-term dynamic monitoring of water bodies, relating to the field of computer data processing and analysis. It includes: acquiring and standardizing sensor data to obtain standardized sensor data; extracting features from the standardized sensor data to obtain water body feature data; modeling the dynamic evolution of the water body by performing a high-dimensional orthogonal projection transformation on the water body feature data to obtain an optimized water body feature matrix; designing and solving a Wasserstein variational loss function based on the optimized water body feature matrix to obtain an optimal water body feature matrix; and predicting water body changes based on the optimal water body feature matrix and minimizing the prediction loss function to obtain an optimized water body change prediction state matrix. This method solves the problems of difficulty in capturing long-term water body change trends, inability to effectively model the complex spatiotemporal dependencies of water body changes, and inability to meet the accuracy requirements of long-term water resource management and ecological monitoring.

Description

Technical Field

[0001] This invention relates to the field of computer data processing and analysis, and in particular to a data processing method for long-term dynamic monitoring of water bodies. Background Technology

[0002] With the intensification of global climate change and human activities, the distribution and dynamic changes of surface water bodies are becoming increasingly complex, making long-term monitoring an important foundation for achieving sustainable water resource utilization. However, as a dynamic natural entity, the area, boundaries, and morphology of water bodies are affected by multiple factors such as precipitation, evaporation, groundwater recharge, snowmelt, and human regulation, resulting in complex spatiotemporal variation patterns of water bodies.

[0003] Furthermore, the acquisition of sensor data is limited by sensor performance and environmental interference. Sensor data from different time steps may exhibit deviations, affecting data consistency and consequently the accuracy of water body dynamic monitoring. Therefore, a computational method is needed that can integrate time-series information, improve data processing accuracy, reduce noise interference, and stably track long-term water body change trends to ensure the scientific validity and reliability of water body monitoring. Summary of the Invention

[0004] This invention provides a data processing method for long-term dynamic monitoring of water bodies, addressing the problem that existing technologies struggle to accurately capture long-term trends in water body changes. It also solves the problem that traditional water monitoring methods, mostly based on simple statistical models or traditional regression methods, cannot effectively model the complex spatiotemporal dependencies of water body changes, resulting in low prediction accuracy and difficulty in achieving long-term stable monitoring. Furthermore, it addresses the deficiency of traditional water monitoring methods in computational instability when dealing with high-dimensional data, which affects the robustness and reliability of monitoring results and fails to meet the accuracy requirements of long-term water resource management and ecological monitoring.

[0005] The present invention provides a data processing method for long-term dynamic monitoring of water bodies, specifically comprising the following technical solutions:

[0006] A data processing method for long-term dynamic monitoring of water bodies includes the following steps:

[0007] S1. Acquire and standardize sensor data to obtain standardized sensor data; extract features from the standardized sensor data to obtain water body feature data; model the dynamic evolution of the water body, and perform high-dimensional orthogonal projection transformation on the water body feature data to obtain the optimized water body feature matrix.

[0008] S2. Based on the optimized water feature matrix, design and solve the Wasserstein variational loss function to obtain the optimal water feature matrix; based on the optimal water feature matrix, predict water changes and minimize the prediction loss function to obtain the optimized water change prediction state matrix.

[0009] Preferably, S1 specifically includes:

[0010] A time recursive model is established using a spatiotemporal multi-layer recursive variational optimization network to extract features from standardized sensor data, thereby obtaining water body characteristic data.

[0011] Preferably, S1 specifically includes:

[0012] In the process of establishing the time recursive model, the water body characteristic data is adaptively adjusted through the variational optimization mechanism. A spatiotemporal recursive calculation framework is introduced, and the dynamic evolution process of the water body is constructed by combining the water body state at the current moment and the water body characteristic data of the past time steps, so as to obtain the water body characteristic state matrix.

[0013] Preferably, S1 specifically includes:

[0014] Based on the water body feature state matrix, an orthogonal projection matrix is ​​constructed to project the water body feature data, so that the water body feature data is preserved after projection. An initial water body feature matrix is ​​set, and a regularization parameter is introduced to perform a high-dimensional orthogonal projection transformation on the water body feature data.

[0015] Preferably, S1 specifically includes:

[0016] The standardized sensor data matrix is ​​optimized by using an orthogonal projection matrix to obtain an optimized water feature matrix.

[0017] Preferably, S2 specifically includes:

[0018] Water body change prediction is performed based on the optimal water body feature matrix to predict the future water body state. The water body feature data of the current time step is combined with the water body feature data of the past time steps and the historical optimal water body feature matrix, and an orthogonal projection matrix is ​​added to calculate the water body change prediction state matrix.

[0019] Preferably, S2 specifically includes:

[0020] Based on the water change prediction state matrix, a prediction loss function is introduced. The prediction state matrix is ​​compared with the actual water state matrix, the prediction error is calculated and minimized.

[0021] Preferably, S2 specifically includes:

[0022] The prediction loss function is minimized to obtain the optimized water change prediction state matrix, which serves as the prediction result for long-term water dynamic monitoring and provides the water state distribution at future times.

[0023] The beneficial effects of the technical solution of the present invention are:

[0024] 1. A time recursive model is established using a spatiotemporal multi-layer recursive variational optimization network to achieve high-precision extraction of water body feature data. By introducing time dependencies, the accuracy of water body classification is improved, misclassification caused by short-term environmental changes is reduced, and the continuity of water body features is enhanced to ensure the stability of dynamic water body monitoring.

[0025] 2. In the water body classification stage, this invention proposes a water body feature optimization method based on high-dimensional orthogonal projection. By constructing an orthogonal projection matrix, the water body feature data is projected onto a new feature space, which reduces the impact of background noise on water body classification, improves the robustness of water body classification, and reduces the error caused by water body boundary ambiguity.

[0026] 3. After water body classification is completed, this invention introduces the Wasserstein variational loss function constraint method to optimize the distribution of water body classification, so as to ensure that the water body classification results at different time steps are consistent in mathematical space, reduce water body classification bias caused by factors such as sensor performance and environmental interference, and improve the stability and interpretability of water body classification.

[0027] 4. In terms of long-term water body dynamic change prediction, this invention introduces an orthogonal projection matrix to constrain the range of water body characteristics, making the water body change prediction state matrix more consistent with the actual water body change trend, thus effectively improving the accuracy of water body change prediction. Attached Figure Description

[0028] Figure 1 This is a flowchart of a data processing method for long-term dynamic monitoring of water bodies as described in this invention. Detailed Implementation

[0029] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0031] The following description, in conjunction with the accompanying drawings, details a specific scheme for a data processing method for long-term dynamic monitoring of water bodies provided by the present invention.

[0032] See attached document Figure 1 The diagram illustrates a data processing method for long-term dynamic monitoring of water bodies according to an embodiment of the present invention. The method includes the following steps:

[0033] S1. Acquire and standardize sensor data to obtain standardized sensor data; extract features from the standardized sensor data to obtain water body feature data; model the dynamic evolution of the water body, and perform high-dimensional orthogonal projection transformation on the water body feature data to obtain the optimized water body feature matrix.

[0034] Long-term monitoring of water body information in the study area was conducted using hydrological sensors, current meters, and GNSS receivers to acquire sensor data. This data included water depth, flow velocity, water quality parameters (such as dissolved oxygen, turbidity, and pH), precipitation, and evaporation. During data acquisition, existing data cleaning algorithms were applied to remove outliers, ensuring the quality of the input data. All sensor data underwent calibration and regularization to ensure time alignment and consistency across different time steps, guaranteeing data comparability in subsequent dynamic water body monitoring.

[0035] Sensor data acquisition is influenced by multiple factors, including sensor noise, environmental interference, different sampling times, and changes in environmental conditions. These factors can lead to significant deviations in sensor data acquired from the same water body at different times. Therefore, it is necessary to standardize the sensor data to obtain standardized sensor data. This ensures that sensor data from different time steps have consistent numerical ranges and statistical characteristics, preventing differences in the numerical distribution of sensor data from affecting the accuracy of subsequent water body classification and prediction.

[0036] A spatiotemporal multi-layer recursive variational optimization network is used to establish a temporal recursive model. Features are extracted from standardized sensor data to obtain water body characteristic data. A variational optimization mechanism is used to adaptively adjust the water body characteristic data, capturing both long-term trends and short-term dynamics of water body changes, thus constructing a complete spatiotemporal evolution trajectory of the water body. By introducing a spatiotemporal recursive computation framework, the current water body state depends not only on the current sensor data but also on water body characteristic data from multiple past time steps, effectively modeling the dynamic evolution process of the water body and improving the accuracy and stability of classification. The recursive equation for the water body characteristic state is constructed as follows:

[0037]

[0038] in, yes The water body characteristic state matrix at time 1; It is the hyperbolic tangent activation function; It is a state recursive weight matrix used to calculate the time dependence of water body characteristics. It is learned through the existing gradient descent method and initialized with Xavier. yes The water body characteristic state matrix at time 1; It is the input data weight matrix, used to map the contribution of standardized sensor data to the current water body characteristics, and is learned through the existing gradient descent method; yes A standardized matrix of sensor data at any given time; This is a bias term used to adjust the calculation results. It is learned through the existing gradient descent method and its value ranges from [value range missing]. ; It is a recursive decay factor used to control the influence weight of historical sensor data. It is obtained through the existing gradient descent method, and its value range is [value range missing]. ; It is the historical time step, representing the number of past time steps; It was before Historical weight matrix at each moment; yes A standardized matrix of sensor data at any given time; It is a norm 2; It is a numerically stable term to prevent numerical overflow during calculations, and can take values ​​of [value missing]. ; yes Activation function; yes The weight parameters of the activation function are learned using the existing gradient descent method.

[0039] The above-mentioned recursive equation for the characteristic state of water bodies is an improved scheme that integrates existing technologies, organically combining the following three key technologies: Status update History regularization terms (similar to VAE / VRNN) Gating mechanism (GRU / LSTM-like) This integration approach significantly improves the accuracy and robustness of water body characteristic state modeling.

[0040] By calculating the water body characteristic state matrix at the current time step, while considering the previous time step... Standardized sensor data at each time step is used to establish a time recursive model to extract features from the standardized sensor data, thereby obtaining a water body feature state matrix and thus obtaining water body feature data, which enhances the temporal information of water body classification.

[0041] To further optimize the water body classification boundary, a high-dimensional orthogonal projection transformation is needed on the water body feature data to reduce interference from non-water body features and improve the accuracy of water body classification. An orthogonal projection matrix is ​​constructed to project the water body feature data into a new feature space, ensuring that the water body feature data is preserved to the greatest extent possible after projection, while effectively suppressing non-water body feature data to improve the separability of water body categories. The non-water body feature data refers to noisy data within the water body feature data. An initial water body feature matrix is ​​set. The initial water feature matrix contains sampling point information from all sensor data after water body classification processing. To ensure the stability of the transformed water feature data, orthogonal projection calculation is required. This ensures the independence of the water feature data in the projected space while reducing its similarity to sampling points from non-water feature data. The calculation of the orthogonal projection matrix is ​​based on the classical orthogonal complementary projection formula. Add regular expression The implementation is as follows:

[0042]

[0043] in, It is an orthogonal projection matrix; It is the identity matrix; This is a regularization parameter used to prevent the singularity problem of orthogonal projection matrices, set to... ; The initial water body feature matrix is ​​a static water body feature matrix obtained by standardizing and extracting features from the original sensor data. It is used to calculate the distribution characteristics of the sampling points of the sensor data after water body classification. The feature extraction can be performed using the existing principal component analysis method. These are adjustment coefficients for the water body characteristic state matrix, obtained through fitting calculations, with values ​​ranging from [value range missing]. ; Represents the transpose of a matrix; It is the square of the second norm.

[0044] By projecting water feature data into an orthogonal subspace, background information in the sampling points of the sensor data after water classification is removed, thus improving the robustness of water classification. Subsequently, an orthogonal projection matrix is ​​used... The standardized sensor data matrix is ​​optimized to obtain the optimized water feature matrix. :

[0045]

[0046] Optimizing water body classification through orthogonal projection transformation makes the water body feature matrix after orthogonal projection transformation more separable, reduces spectral confusion between water body and non-water body feature data sampling points, and thus improves classification accuracy.

[0047] S2. Based on the optimized water feature matrix, design and solve the Wasserstein variational loss function to obtain the optimal water feature matrix; based on the optimal water feature matrix, predict water changes and minimize the prediction loss function to obtain the optimized water change prediction state matrix.

[0048] Based on the optimized water body feature matrix, a Wasserstein variational loss function is designed to ensure the stability of water body classification and reduce classification bias between different time steps. Because sensor data is affected by factors such as sensor performance and environmental changes during acquisition at different time steps, even water bodies in the same area may exhibit significant fluctuations in their characteristics. Therefore, the Wasserstein variational loss function is used to constrain the distribution of water body classification, ensuring that the classification results at different time steps maintain consistency in the mathematical space, thereby improving classification stability. The Wasserstein variational loss function is defined as follows:

[0049]

[0050] in, It is the Wasserstein variational loss function; It is the Frobenius norm; The Wasserstein distance regularization factor is obtained through expert experience and its value ranges from [value range missing]. ; yes The target water body label data at any given time is obtained through manual annotation of historical sensor data.

[0051] The Wasserstein variational loss function mentioned above is the squared Frobenius norm. With Wasserstein regularization The combination of .

[0052] The Wasserstein variational loss function is solved using the existing gradient descent method to obtain the optimal water feature matrix. By using the Wasserstein variational loss function as a constraint, water body classification becomes more stable and the uncertainty of data classification is reduced.

[0053] Water body change prediction is based on the optimal water body feature matrix to predict future water body states. Since the spatiotemporal changes of water bodies are strongly correlated, the prediction process needs to consider not only the water body feature data at the current time step but also historical data. The model is constructed using water feature data from multiple time steps and the optimal water feature matrix from historical moments to enhance the temporal dependence of the prediction. Furthermore, to ensure the stationarity of the prediction and reduce the impact of error accumulation, an orthogonal projection matrix is ​​added to constrain the range of water feature variations, enabling it to more accurately reflect water change trends. Finally, the water change prediction formula establishes a recursive model by integrating information from multiple time steps to predict future water conditions. The water change prediction formula is as follows:

[0054]

[0055] in, yes The water change prediction state matrix at time step, where each element of the water change prediction state matrix represents the water change state value of a certain sampling point predicted at the current time step. These are the weighting parameters for predicting water body changes, obtained through expert experience and initialized according to Xavier or He. It is the number of historical time steps used to predict changes in water bodies, i.e., the size of the backtracking time window; , They are the first The dynamic adjustment coefficients for the water body characteristic state matrix and the optimal water body characteristic matrix at each historical time step are obtained through fitting calculations, and their values ​​range from [value range missing]. ; yes The water body characteristic state matrix at time t; yes The optimal water feature matrix at time t; It is the gradient adjustment factor, obtained through the existing gradient descent method, and its value range is [value range missing]. ; yes The optimal water feature matrix at time t; yes The optimal water feature matrix at time t; It is the time-weighted state input in a recurrent neural network (RNN), used for memorizing and modeling time series. It is a multi-source feature fusion in Transformer, which combines feature transformation, such as PCA or orthogonal projection, to orthogonally project the historical best water features from the historical best water feature matrix and introduce them into the prediction. These include gradient normalization (such as ResNet / LayerNorm) and smoothing trend detection (such as time difference normalization), used to capture short-term fluctuations and prevent gradient explosion / vanishing.

[0056] Because the dynamic changes of water bodies are time-dependent and influenced by various external factors, drastic fluctuations or cumulative drift may occur during the prediction process, affecting the stability of long-term predictions. Therefore, a prediction loss function is introduced. This function compares the predicted state matrix of water body changes with the actual state matrix, calculates the prediction error, and minimizes it to ensure the stability and rationality of long-term water body change trends. The mathematical expression of the prediction loss function is as follows:

[0057]

[0058] in, It is a prediction loss function used to measure the performance of the predicted state matrix of water body changes. The prediction error at time point is calculated, and the regularization constraint on the water change prediction state matrix is ​​applied to enhance the stability of long-term water dynamic monitoring. yes The target water body label data at that time, i.e. The actual water state matrix at time t; It is the total number of time steps predicted, i.e., the length of the time series considered in long-term water body monitoring; This is the time decay parameter, used to determine the weighting effect of prediction errors in the time dimension. It is obtained through expert experience and its value ranges from [value missing]. ; This is a recursive regularization parameter used to control the smoothness of the water change prediction state matrix, ensuring the stability of the water change trend over time. It is obtained through expert experience and its value range is [value missing]. ; yes Predictive state matrix of water body changes at any given time; This is the time-weighted decay term for the prediction error, used to control the focus of prediction accuracy at different times. This indicates the degree of deviation from the future state of a water body (such as water area and water quality indicators) at the current prediction point in time. It is a time-focused weight, with the largest weight in the middle of the predicted sequence; This is a variation amplitude regularization term used to control the smoothness of the predicted sequence and prevent drastic changes. It is used to reflect whether the prediction results have good continuity over time.

[0059] By minimizing the prediction loss function, the resulting optimized water change prediction state matrix can be used as the prediction result for long-term water dynamic monitoring, providing the water state distribution at future times. Ultimately, the output optimized water change prediction state matrix can be used to plot water change trends, calculate water area change rates, analyze long-term trends, and provide data support for water resource management and ecological protection decision-making, thus realizing data processing for long-term water dynamic monitoring.

[0060] In summary, a data processing method for long-term dynamic monitoring of water bodies has been developed.

[0061] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0062] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0063] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A data processing method for long-term dynamic monitoring of water bodies, characterized in that, Includes the following steps: S1. Acquire and standardize sensor data to obtain standardized sensor data; based on the standardized sensor data, establish a time recursive model for feature extraction to obtain water body feature data; adaptively adjust the water body feature data through a variational optimization mechanism, introduce a spatiotemporal recursive calculation framework, combine the water body state at the current moment and the water body feature data at past time steps, and combine the historical regularization term of the standardized sensor data matrix based on historical moments and the gating mechanism of the standardized sensor data matrix based on the current moment to model the dynamic evolution process of the water body and obtain the water body feature state matrix; Based on the water body characteristic state matrix, an orthogonal projection matrix is ​​constructed to project the water body characteristic data, so that the water body characteristic data is preserved after projection. An initial water body characteristic matrix is ​​set, and a regularization parameter is introduced to perform a high-dimensional orthogonal projection transformation on the water body characteristic data. The standardized sensor data matrix is ​​optimized through the orthogonal projection matrix to obtain the optimized water body characteristic matrix. S2. Based on the optimized water feature matrix, design and solve the Wasserstein variational loss function to obtain the optimal water feature matrix; the Wasserstein variational loss function is: in, It is the Wasserstein variational loss function; It is the Frobenius norm; It is the optimized water body feature matrix; yes Target water body label data at any given time; The Wasserstein distance regularization factor; yes A standardized matrix of sensor data at any given time; Based on the optimal water body feature matrix, water body change prediction is performed, and the prediction loss function is minimized to obtain the optimized water body change prediction state matrix.

2. The data processing method for long-term dynamic monitoring of water bodies according to claim 1, characterized in that, S1 specifically includes: A time recursive model is established using a spatiotemporal multi-layer recursive variational optimization network to extract features from standardized sensor data, thereby obtaining water body characteristic data.

3. The data processing method for long-term dynamic monitoring of water bodies according to claim 1, characterized in that, S2 specifically includes: Water body change prediction is performed based on the optimal water body feature matrix to predict the future water body state. The water body feature data of the current time step is combined with the water body feature data of the past time steps and the historical optimal water body feature matrix, and an orthogonal projection matrix is ​​added to calculate the water body change prediction state matrix.

4. The data processing method for long-term dynamic monitoring of water bodies according to claim 3, characterized in that, S2 specifically includes: Based on the water change prediction state matrix, a prediction loss function is introduced. The prediction state matrix is ​​compared with the actual water state matrix, the prediction error is calculated and minimized.

5. The data processing method for long-term dynamic monitoring of water bodies according to claim 4, characterized in that, S2 specifically includes: The prediction loss function is minimized to obtain the optimized water change prediction state matrix, which serves as the prediction result for long-term water dynamic monitoring and provides the water state distribution at future times.

Images