Unmanned aerial vehicle low-altitude remote sensing and ground measurement collaborative water and soil conservation monitoring method and system

Abstract

This application relates to a method and system for monitoring soil and water conservation using a combination of UAV low-altitude remote sensing and ground-based measurements. The method includes: acquiring remote sensing images, meteorological data, and ground-based measured data of a target area; fusing the features of the remote sensing images, meteorological data, and ground-based measured data to obtain comprehensive target features; filtering the comprehensive target features using a feature selector to obtain key comprehensive features; and analyzing the key comprehensive features based on a hybrid architecture integrating GCN and TCN to obtain soil erosion prediction results. Specifically, the GCN module in the hybrid architecture is used to mine spatial correlations between geographical areas, while the TCN module is used to capture the temporal nonlinear changes in soil erosion. This approach ensures both spatial accuracy of the prediction results and temporal continuity, improving the accuracy and reliability of soil erosion prediction and providing more scientific technical support for soil and water conservation planning and ecological restoration decisions.

Description

Technical Field

[0001] This application relates to the field of environmental monitoring and data analysis technology, and in particular to a method and system for monitoring soil and water conservation by combining low-altitude remote sensing by unmanned aerial vehicles with ground-based measurements. Background Technology

[0002] Soil erosion is a serious global ecological and environmental problem, leading not only to land degradation and decreased agricultural productivity, but also exacerbating floods and river siltation, severely threatening ecological security and sustainable development. Traditional soil erosion monitoring mainly relies on ground-based measurement methods (such as runoff plot observation and soil sampling), which, while highly accurate, suffer from drawbacks such as high cost, long cycle time, and limited spatial coverage, making it difficult to meet the needs of large-scale, dynamic, and high-frequency monitoring.

[0003] Traditional methods for predicting soil erosion (such as the Universal Soil Loss Equation (USLE) and the Modified Universal Soil Loss Equation (RUSLE)) rely on manually set empirical parameters, which cannot dynamically adapt to the differences in geographical environment across different regions. Furthermore, they have a weak ability to capture spatiotemporal dynamic changes, and their prediction accuracy is limited by the rationality of parameter calibration. With the application of machine learning and deep learning technologies, some models have begun to attempt to integrate multi-source data, but significant shortcomings still exist.

[0004] Therefore, there is an urgent need for a scientific and accurate method for monitoring soil and water conservation that combines low-altitude remote sensing by drones with ground-based measurements. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and system for monitoring soil and water conservation by combining UAV low-altitude remote sensing with ground-based measurements to address the aforementioned technical problems.

[0006] In a first aspect, this application provides a method for monitoring soil and water conservation by combining low-altitude remote sensing from unmanned aerial vehicles (UAVs) with ground-based measurements, the method comprising: Acquire remote sensing images, meteorological data, and ground-measured data of the target area; The remote sensing image, the meteorological data, and the ground-measured data are fused to obtain the comprehensive features of the target. The target composite features are filtered using a feature selector to obtain key composite features; Based on a hybrid architecture integrating GCN and TCN, the key comprehensive features are analyzed to obtain soil erosion prediction results; wherein, the GCN module in the hybrid architecture is used to explore the spatial correlation between geographical regions; and the TCN module in the hybrid architecture is used to capture the temporal nonlinear changes in soil erosion.

[0007] In one embodiment, the feature fusion of the remote sensing image, the meteorological data, and the ground-measured data to obtain the target comprehensive features includes: The spectral features in the remote sensing image and the soil features in the ground measurement data are copied and expanded according to the time step to obtain candidate data in the target time series format; The candidate data and the meteorological data are concatenated according to feature dimensions to obtain the comprehensive features of the target.

[0008] In one embodiment, the step of filtering the target composite features using a feature selector to obtain key composite features includes: The contribution of each target's comprehensive feature is iteratively evaluated through the recurrent neural network in the feature selector. The attention mechanism in the feature selector determines the association weight between each target comprehensive feature and the core factors of soil erosion in the target area; wherein, the core factors of soil erosion include soil type, rainfall intensity, rainfall erosivity R factor and / or soil erodibility K factor; The contribution and correlation weight of each target feature are fused to obtain the target feature weight; Based on the target feature weights, the key comprehensive features are selected from the target comprehensive features.

[0009] In one embodiment, the hybrid architecture based on the integration of GCN and TCN analyzes the key comprehensive features to obtain soil erosion prediction results, including: The GCN module performs neighborhood feature aggregation on the key comprehensive features to mine the spatial correlation between different geographical regions and outputs spatial correlation features. The TCN module performs causal dilated convolution on the key comprehensive features, and combines residual connections to expand the temporal receptive field, outputting temporal nonlinear features. The spatial correlation features and the temporal nonlinear features are fused through the cross-modal gated fusion layer in the hybrid architecture to obtain spatiotemporal fusion features; The spatiotemporal fusion features are input into a fully connected layer for feature mapping to obtain the soil erosion prediction results.

[0010] In one embodiment, the method further includes: Multi-band reflectance data are extracted from the remote sensing image, and each band is processed independently using max-min normalization to obtain the reflectance tensor. The soil erosion sensitive spectral index is calculated based on the reflectance tensor; wherein the soil erosion sensitive spectral index includes at least one of the vegetation index, soil-adjusted vegetation index and differential soil index. The spectral features are obtained by concatenating the soil erosion sensitive spectral index with the reflectance tensor.

[0011] In one embodiment, the method further includes: The ground-measured data is mapped to the remote sensing image using the inverse distance weighted interpolation method, and then normalized to obtain a gridded soil tensor. Based on the gridded soil tensor, the soil erodibility K factor is calculated; The soil characteristics are obtained by splicing the soil erodibility K factor with the gridded soil tensor.

[0012] In one embodiment, the remote sensing image includes multispectral reflectance and / or vegetation index; the meteorological data includes rainfall, wind speed, temperature, sunshine duration and / or rainfall erosivity R factor; and the ground-measured data includes soil moisture content, clay content, bulk density and / or soil erosibility K factor.

[0013] Secondly, this application also provides a soil and water conservation monitoring system that combines low-altitude remote sensing by unmanned aerial vehicles with ground-based measurements, the system comprising: The acquisition module is used to acquire remote sensing images, meteorological data, and ground-measured data of the target area; The feature fusion module is used to fuse the remote sensing image, the meteorological data, and the ground-measured data to obtain the comprehensive features of the target. The feature filtering module is used to filter the target comprehensive features through a feature selector to obtain key comprehensive features; The prediction module is used to analyze the key comprehensive features based on a hybrid architecture integrating GCN and TCN to obtain soil erosion prediction results; wherein, the GCN module in the hybrid architecture is used to explore the spatial correlation between geographical regions; and the TCN module in the hybrid architecture is used to capture the temporal nonlinear changes of soil erosion.

[0014] Thirdly, this application also provides an electronic device, including a processor and a memory; wherein the memory is used to store a computer program; and the processor is configured to, when executing the computer program, implement the steps of the method described in any embodiment of this application.

[0015] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the methods described in any embodiment of this application.

[0016] The aforementioned method for monitoring soil and water conservation through a combination of UAV low-altitude remote sensing and ground-based measurements comprehensively covers multiple influencing factors, including spatial characterization of soil erosion, dynamic meteorological drivers, and soil background properties, thus laying a solid data foundation for accurate prediction. By using a feature selector to filter comprehensive target features, redundant and invalid features can be accurately eliminated, significantly reducing model computational complexity and improving operational efficiency. Simultaneously, it focuses on core driving factors such as rainfall intensity, soil erodibility, and vegetation cover, effectively reducing the interference of irrelevant information on prediction results. Furthermore, based on a hybrid architecture integrating GCN and TCN, the analysis is conducted. The GCN module can deeply explore the spatial correlations between different geographical regions, accurately depicting the interactive influence of spatial elements such as topography, soil, and vegetation. The TCN module can efficiently capture the temporal nonlinear variation patterns of soil erosion, adapting to the cumulative effects of dynamic factors such as rainfall and soil moisture. The synergy of these two modules achieves integrated mining of the spatiotemporal characteristics of soil erosion, ensuring both spatial accuracy and temporal continuity in prediction results. Ultimately, this significantly improves the accuracy and reliability of soil erosion prediction, providing more scientific technical support for soil and water conservation planning and ecological restoration decisions. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method for monitoring soil and water conservation by combining low-altitude remote sensing from unmanned aerial vehicles (UAVs) with ground-based measurements, according to an exemplary embodiment. Figure 2 This is a flowchart illustrating a method for monitoring soil and water conservation by combining low-altitude remote sensing from unmanned aerial vehicles (UAVs) with ground-based measurements, according to an exemplary embodiment. Figure 3 This is a structural block diagram of a soil and water conservation monitoring system that combines low-altitude remote sensing by unmanned aerial vehicles with ground-based measurement, according to an exemplary embodiment. Figure 4 This is an internal structural diagram of an electronic device according to an exemplary embodiment. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] The terms "first," "second," and "third" used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or apparatus.

[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] The method for monitoring soil and water conservation using a combination of UAV low-altitude remote sensing and ground-based measurements provided in this application embodiment can be applied to electronic devices or cloud servers. The electronic device can be any mobile terminal or fixed terminal. The terminal can be a device that provides voice and / or data connectivity to the user. For example, the terminal can be an IoT terminal, such as a sensor device, a mobile phone or so-called "cellular" phone, or a computer with an IoT terminal; for example, it can be a fixed, portable, pocket-sized, handheld, or computer-built-in system. The cloud server can be any virtualized computing resource or physical server cluster. The server can be a platform that provides on-demand, scalable computing, storage, networking, and application services to the user.

[0022] In some embodiments, such as Figure 1 As shown, a method for monitoring soil and water conservation by combining low-altitude remote sensing from unmanned aerial vehicles (UAVs) with ground-based measurements is provided. The method includes the following steps: S101: Acquire remote sensing images, meteorological data, and ground-measured data of the target area.

[0023] In some embodiments, the remote sensing image includes multispectral reflectance and / or vegetation index; the meteorological data includes rainfall, wind speed, temperature, sunshine duration and / or rainfall erosivity R factor; and the ground-measured data includes soil moisture content, clay content, bulk density and / or soil erosibility K factor.

[0024] In some embodiments, the original image of the target area is acquired by taking pictures with a drone, and the noise generated during the acquisition process is removed by Gaussian smoothing to obtain a remote sensing image.

[0025] In some embodiments, the method further includes: Multi-band reflectance data are extracted from the remote sensing image, and each band is processed independently using max-min normalization to obtain the reflectance tensor. The soil erosion sensitive spectral index is calculated based on the reflectance tensor; wherein the soil erosion sensitive spectral index includes at least one of the vegetation index, soil-adjusted vegetation index and differential soil index. The spectral features are obtained by concatenating the soil erosion sensitive spectral index with the reflectance tensor.

[0026] In one embodiment, the electronic device reads the raster data (reflectance data) of a remote sensing image using remote sensing compilation tools such as GDAL / ENVI, and splits and reassembles it into an initial feature tensor according to the band dimension; using max-min normalization, each band is processed independently to obtain the reflectance tensor. For example, one calculation method for the normalization process can be as follows: ; in, Indicates the original band emissivity; (i,j) indicates the pixel coordinates; c indicates the band index.

[0027] In this embodiment of the application, the vegetation index can be the Normalized Difference Vegetation Index (NDVI); the NDVI reflects vegetation coverage through the difference in reflectance between the near-infrared and red light bands.

[0028] In this embodiment, the Soil-Adjusted Vegetation Index (SAVI) is used to correct the interference of bare soil areas on the vegetation index by introducing an adjustment coefficient, making it more suitable for sparsely vegetated areas.

[0029] In this embodiment of the application, the differential soil index can be the Bare Soil Index (BSI), which identifies bare soil areas by the difference between short-wave infrared and red light bands.

[0030] In some embodiments, the electronic device can concatenate one or more soil erosion-sensitive spectral indices with a normalized reflectance tensor according to the feature dimension to obtain spectral features.

[0031] In this embodiment, by performing maximum-min normalization on the multi-band reflectance data of the remote sensing image band by band, the numerical differences in reflectance between different bands can be effectively eliminated, avoiding the dominance of high-value bands in feature expression and ensuring that the information of each band has equal weight in subsequent analysis. Based on the normalized reflectance tensor, vegetation indices, soil-adjusted vegetation indices, and other water and soil loss-sensitive spectral indices can be calculated, which can transform the original spectral information into features that are directly related to key water and soil loss influencing factors such as vegetation coverage and soil exposure, thereby enhancing the relevance and physical significance of the spectral data. Furthermore, by concatenating the water and soil loss-sensitive spectral indices with the reflectance tensor to form spectral features, the original spectral information of the remote sensing image is completely preserved, while water and soil loss-oriented enhancement features are added, enriching the feature dimensions and representation capabilities, and laying a solid foundation of high-quality spectral data for subsequent multi-source data fusion and accurate water and soil loss prediction.

[0032] In some embodiments, the method further includes: The ground-measured data is mapped to the remote sensing image using the inverse distance weighted interpolation method, and then normalized to obtain a gridded soil tensor. Based on the gridded soil tensor, the soil erodibility K factor is calculated; The soil characteristics are obtained by splicing the soil erodibility K factor with the gridded soil tensor.

[0033] In this embodiment of the application, the Inverse Distance Weighting (IDW) method is a method of estimation by calculating the inverse weighted average of the distances between the unknown point and the known sample points.

[0034] In this embodiment of the application, the soil erodibility K factor represents the amount of soil loss per unit area caused by the erosive force of a unit rainfall under standard conditions, and is used to quantify the soil's sensitivity to water erosion.

[0035] In some embodiments, electronic devices read soil data within a target time window via IoT gateways or sensors; the soil data is mapped to a remote sensing image grid using IDW interpolation to obtain a gridded soil tensor; the gridded soil tensor can be normalized to obtain a normalized soil tensor; based on the sand content, silt content, clay content, and organic matter content in the normalized soil tensor, the soil erodibility K-factor is calculated; where a larger K value indicates that the soil is more easily eroded, and a smaller K value indicates stronger erosion resistance; the soil erodibility K-factor is then concatenated with the normalized soil tensor to obtain soil characteristics.

[0036] In this embodiment of the application, the rainfall erosivity R factor represents the soil erosion and transport capacity per unit area caused by rainfall and runoff, reflecting the combined effect of rainfall kinetic energy and rainfall intensity.

[0037] In some embodiments, the electronic device can acquire time-series meteorological data of the target area within a target time window, such as rainfall, wind speed, temperature, and sunshine duration. The data is then split into time steps, and missing and outlier values ​​are removed to obtain meteorological data. The meteorological data for each time step can also be mapped to a remote sensing image grid using IDW interpolation to obtain a gridded meteorological time-series tensor. The gridded meteorological time-series tensor is normalized to obtain a normalized meteorological feature tensor. Based on the rainfall in the normalized meteorological feature tensor, the rainfall erosivity R-factor is determined; the higher the R-value, the stronger the scouring and splashing erosion of the soil by rainfall. The R-factor for each time step is mapped to the remote sensing image grid to form an R-factor tensor, which is then concatenated with the normalized meteorological feature tensor to output the meteorological features.

[0038] S102, the remote sensing image, the meteorological data and the ground measured data are fused to obtain the comprehensive features of the target.

[0039] In some embodiments, the step of fusing features from the remote sensing image, the meteorological data, and the ground-measured data to obtain comprehensive target features includes: The spectral features in the remote sensing image and the soil features in the ground measurement data are copied and expanded according to the time step to obtain candidate data in the target time series format; The candidate data and the meteorological data are concatenated according to feature dimensions to obtain the comprehensive features of the target.

[0040] In one embodiment, the electronic device reads the spectral features of the preprocessed remote sensing image and the soil features of the ground measurement data, wherein the spectral features and soil features are structures of a two-dimensional grid combined with the feature dimension corresponding to the spatial resolution of the target area; the time step corresponding to the meteorological data is extracted, and the spectral features and soil features are copied and extended along the time dimension, with the number of extensions consistent with the time step of the meteorological data, to obtain time-series spectral features and time-series soil features respectively, and the two are merged to obtain candidate data in the target time-series format; the electronic device reads the meteorological features in the preprocessed meteorological data, and performs a concatenation operation between the candidate data and the meteorological features along the feature dimension to obtain the target comprehensive features.

[0041] S103, the target comprehensive features are filtered by the feature selector to obtain key comprehensive features.

[0042] In some embodiments, such as Figure 2 As shown, the step of using a feature selector to filter the target composite features to obtain key composite features includes: S1031, The contribution of each target comprehensive feature is iteratively evaluated through the recurrent neural network in the feature selector; S1032, through the attention mechanism in the feature selector, determine the association weight between each target comprehensive feature and the core factors of soil erosion in the target area; wherein, the core factors of soil erosion include soil type, rainfall intensity, rainfall erosivity R factor and / or soil erodibility K factor; S1033, the contribution degree and the correlation weight of each target comprehensive feature are fused to obtain the target feature weight; S1034, Based on the target feature weights, the key comprehensive features are selected from the target comprehensive features.

[0043] In some embodiments, a Long Short-Term Memory (LSTM) network is used to process the reshaped temporal features, leveraging its ability to capture long and short-term temporal dependencies to iteratively evaluate the contribution of each feature at each time step and each spatial node. LSTM quantifies the importance of each feature through dynamic adjustments of the input gate, forget gate, output gate, and cell state. The output contribution score ranges from 0 to 1, with higher scores indicating a greater impact of the feature on soil erosion.

[0044] In some embodiments, a multi-head attention mechanism is employed to further enhance the screening of known core drivers of soil erosion (such as rainfall erosivity R factor, soil erodibility K factor, and vegetation index NDVI). By constructing a query, key, and value matrix, the association weight between each feature and the core factor is calculated, and the weight corresponding to the core factor is increased, thereby achieving precise focusing on key features.

[0045] In some embodiments, the feature contribution score output by LSTM is fused with the association weight calculated by the attention mechanism, and the normalized result of the product of the two is used as the target feature weight; and the target comprehensive features with the highest weight ranking are selected as key comprehensive features.

[0046] In this embodiment, the contribution of each target comprehensive feature is iteratively evaluated through a recurrent neural network, which can fully explore the dynamic importance of features in the temporal or spatial dimensions and reduce static and one-sided judgments on feature importance. By using an attention mechanism to determine the correlation weight between features and core factors of soil erosion, the most critical feature dimensions affecting soil erosion can be accurately focused, strengthening the representation of core driving factors. The target feature weight obtained by fusing contribution and correlation weight takes into account both the inherent value of the feature itself and its targeted impact on soil erosion, making the weight assignment more scientific. Based on this weight, key comprehensive features are selected, which can effectively eliminate redundant and invalid features, reduce the computational complexity of the model, reduce the interference of irrelevant information, and improve the efficiency of subsequent GCN-TCN hybrid architecture analysis and the accuracy of soil erosion prediction results.

[0047] S104, Based on a hybrid architecture integrating GCN and TCN, the key comprehensive features are analyzed to obtain soil erosion prediction results; wherein, the GCN module in the hybrid architecture is used to mine the spatial correlation between geographical regions; the TCN module in the hybrid architecture is used to capture the temporal nonlinear changes in soil erosion.

[0048] In some embodiments, the hybrid architecture based on the integration of GCN and TCN analyzes the key comprehensive features to obtain soil erosion prediction results, including: The GCN module performs neighborhood feature aggregation on the key comprehensive features to mine the spatial correlation between different geographical regions and outputs spatial correlation features. The TCN module performs causal dilated convolution on the key comprehensive features, and combines residual connections to expand the temporal receptive field, outputting temporal nonlinear features. The spatial correlation features and the temporal nonlinear features are fused through the cross-modal gated fusion layer in the hybrid architecture to obtain spatiotemporal fusion features; The spatiotemporal fusion features are input into a fully connected layer for feature mapping to obtain the soil erosion prediction results.

[0049] In some embodiments, a Graph Convolutional Network (GCN) is used to process the filtered key comprehensive features, leveraging its ability to mine spatial topological relationships to aggregate the neighborhood features of each spatial node. The GCN first constructs a spatial adjacency matrix based on the geographical distance of the target region (nodes closer to each other have higher association weights, while those exceeding a threshold have weights of 0). Then, the adjacency matrix is ​​normalized, and matrix operations are used to aggregate the key comprehensive features of each node and its neighboring nodes, quantifying the spatial relationships between different geographical regions. The output spatial relationship features can accurately characterize the interactive influence of surrounding soil, vegetation, topography, and other factors on each location.

[0050] In some embodiments, electronic devices may employ a causal temporal convolutional network (TCN) to process the filtered key comprehensive features. Leveraging its ability to capture temporal nonlinear changes, the TCN performs causal dilated convolution operations on the features at each time step. The TCN expands the temporal receptive field by setting an increasing dilation coefficient, while introducing residual connection structures to preserve the original temporal feature information. This avoids gradient vanishing in deep networks and effectively captures the cumulative impact of long-term evolution of factors such as rainfall and soil moisture on soil erosion. The output temporal nonlinear features can reflect the dynamic changes of soil erosion driving factors at different time steps.

[0051] In some embodiments, a cross-modal gating fusion layer in a hybrid architecture is used for dynamic fusion of spatial correlation features and temporal nonlinear features. This fusion layer first calculates dynamic gating coefficients (ranging from 0 to 1) through linear transformation and activation functions. These gating coefficients are used to dynamically adjust the contribution weights of the two types of features; the closer the coefficient is to 1, the higher the contribution of spatial correlation features; the closer it is to 0, the higher the contribution of temporal nonlinear features. Then, based on these coefficients, the two types of features are weighted and fused to obtain spatiotemporal fusion features that can take into account both spatial interaction and temporal evolution.

[0052] In some embodiments, the spatiotemporal fusion features are input into a fully connected layer for feature mapping. The fully connected layer maps the high-dimensional spatiotemporal fusion features into one-dimensional predicted values ​​that correspond one-to-one with the spatial grids in the remote sensing image of the target area through multi-layer linear transformation and activation functions. The final output predicted value directly corresponds to the soil erosion intensity at each spatial location and at each time step, i.e., the soil erosion prediction result.

[0053] Thus, by aggregating key comprehensive features using the GCN module to perform neighborhood feature aggregation, the spatial correlation between different geographical regions can be effectively explored, accurately depicting the interactive influence of surrounding areas on elements such as soil, vegetation, and topography in a particular region, thus overcoming the limitations of single-region feature analysis; through TCN... The module performs causal dilated convolution operations combined with residual connections. This expands the temporal receptive field to capture the cumulative effect of long-term evolution of factors such as rainfall and soil moisture on soil erosion, while avoiding the gradient vanishing problem of deep networks. It fully preserves the original temporal feature information and accurately restores the temporal nonlinear variation law of soil erosion. By fusing spatial correlation features and temporal nonlinear features through a cross-modal gating fusion layer, the contribution weights of the two types of features can be dynamically balanced, avoiding the drawback of single-modal features dominating the prediction results. This forms a spatiotemporal fusion feature that takes into account both spatial interaction and temporal dynamic changes. Finally, the spatiotemporal fusion feature is input into a fully connected layer for feature mapping, which can efficiently transform the high-dimensional fusion feature into a soil erosion prediction result that corresponds one-to-one with the spatial grid of the target area. This achieves a closed loop from multi-dimensional feature analysis to accurate prediction output, significantly improving the accuracy and reliability of soil erosion prediction.

[0054] The aforementioned method for monitoring soil and water conservation through a combination of UAV low-altitude remote sensing and ground-based measurements comprehensively covers multiple influencing factors, including spatial characterization of soil erosion, dynamic meteorological drivers, and soil background properties, thus laying a solid data foundation for accurate prediction. By using a feature selector to filter comprehensive target features, redundant and invalid features can be accurately eliminated, significantly reducing model computational complexity and improving operational efficiency. Simultaneously, it focuses on core driving factors such as rainfall intensity, soil erodibility, and vegetation cover, effectively reducing the interference of irrelevant information on prediction results. Furthermore, based on a hybrid architecture integrating GCN and TCN, the analysis is conducted. The GCN module can deeply explore the spatial correlations between different geographical regions, accurately depicting the interactive influence of spatial elements such as topography, soil, and vegetation. The TCN module can efficiently capture the temporal nonlinear variation patterns of soil erosion, adapting to the cumulative effects of dynamic factors such as rainfall and soil moisture. The synergy of these two modules achieves integrated mining of the spatiotemporal characteristics of soil erosion, ensuring both spatial accuracy and temporal continuity in prediction results. Ultimately, this significantly improves the accuracy and reliability of soil erosion prediction, providing more scientific technical support for soil and water conservation planning and ecological restoration decisions.

[0055] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0056] Based on the same inventive concept, this application also provides a UAV-based low-altitude remote sensing and ground-based measurement-based soil and water conservation monitoring system for implementing the aforementioned method of coordinating UAV low-altitude remote sensing and ground-based measurement in soil and water conservation monitoring. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more UAV-based low-altitude remote sensing and ground-based measurement-based soil and water conservation monitoring system embodiments provided below can be found in the limitations of the UAV-based low-altitude remote sensing and ground-based measurement-based soil and water conservation monitoring method described above, and will not be repeated here.

[0057] In one embodiment, such as Figure 3 As shown, a soil and water conservation monitoring system combining UAV low-altitude remote sensing and ground-based measurement is provided. The system includes: The acquisition module 10 is used to acquire remote sensing images, meteorological data and ground measurement data of the target area; Feature fusion module 20 is used to fuse the remote sensing image, the meteorological data and the ground measured data to obtain the target comprehensive features; Feature filtering module 30 is used to filter the target comprehensive features through a feature selector to obtain key comprehensive features; The prediction module 40 is used to analyze the key comprehensive features based on a hybrid architecture integrating GCN and TCN to obtain soil erosion prediction results; wherein, the GCN module in the hybrid architecture is used to explore the spatial correlation between geographical regions; and the TCN module in the hybrid architecture is used to capture the temporal nonlinear changes in soil erosion.

[0058] In one embodiment, the feature fusion module 20 is configured to perform the following steps: The spectral features in the remote sensing image and the soil features in the ground measurement data are copied and expanded according to the time step to obtain candidate data in the target time series format; The candidate data and the meteorological data are concatenated according to feature dimensions to obtain the comprehensive features of the target.

[0059] In one embodiment, the feature filtering module 30 is configured to perform the following steps: The contribution of each target's comprehensive feature is iteratively evaluated through the recurrent neural network in the feature selector. The attention mechanism in the feature selector determines the association weight between each target comprehensive feature and the core factors of soil erosion in the target area; wherein, the core factors of soil erosion include soil type, rainfall intensity, rainfall erosivity R factor and / or soil erodibility K factor; The contribution and correlation weight of each target feature are fused to obtain the target feature weight; Based on the target feature weights, the key comprehensive features are selected from the target comprehensive features.

[0060] In one embodiment, the prediction module 40 includes: The GCN module performs neighborhood feature aggregation on the key comprehensive features to mine the spatial correlation between different geographical regions and outputs spatial correlation features. The TCN module performs causal dilated convolution on the key comprehensive features, and combines residual connections to expand the temporal receptive field, outputting temporal nonlinear features. The spatial correlation features and the temporal nonlinear features are fused through the cross-modal gated fusion layer in the hybrid architecture to obtain spatiotemporal fusion features; The spatiotemporal fusion features are input into a fully connected layer for feature mapping to obtain the soil erosion prediction results.

[0061] In one embodiment, the system further includes: The extraction module is used to extract multi-band reflectance data from the remote sensing image, and to process each band independently through max-min normalization to obtain the reflectance tensor. The first calculation module is used to calculate the soil erosion sensitive spectral index based on the reflectance tensor; wherein the soil erosion sensitive spectral index includes at least one of the vegetation index, soil-adjusted vegetation index and differential soil index. The first splicing module is used to splice the soil erosion sensitive spectral index with the reflectance tensor to obtain the spectral features.

[0062] In one embodiment, the system further includes: The mapping module is used to map the ground measured data to the remote sensing image using the inverse distance weighted interpolation method, and perform normalization processing to obtain a gridded soil tensor. The second calculation module is used to calculate the soil erodibility K factor based on the gridded soil tensor. The second splicing module is used to splice the soil erodibility K factor with the gridded soil tensor to obtain the soil characteristics.

[0063] In one embodiment, the remote sensing image includes multispectral reflectance and / or vegetation index; the meteorological data includes rainfall, wind speed, temperature, sunshine duration and / or rainfall erosivity R factor; and the ground-measured data includes soil moisture content, clay content, bulk density and / or soil erosibility K factor.

[0064] The various modules in the aforementioned UAV low-altitude remote sensing and ground measurement collaborative soil and water conservation monitoring system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the electronic device in hardware form or independent of the processor, or stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0065] In one embodiment, an electronic device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4As shown, the electronic device includes a processor, memory, communication interface, display unit, and input system connected via a method bus. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores operating methods and computer programs. The internal memory provides an environment for the operation of the operating methods and computer programs stored in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a method for monitoring soil and water conservation using a combination of UAV low-altitude remote sensing and ground-based measurements. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the device's casing, or an external keyboard, touchpad, or mouse.

[0066] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0067] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0068] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps performed by the processor of the electronic device of any of the above.

[0069] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0070] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, compilable logic units, quantum computing-based data processing logic units, etc., and are not limited to these.

[0071] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0072] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for monitoring soil and water conservation by combining low-altitude remote sensing from unmanned aerial vehicles (UAVs) with ground-based measurements, characterized in that... The method includes: Acquire remote sensing images, meteorological data, and ground-measured data of the target area; The remote sensing image, the meteorological data, and the ground-measured data are fused to obtain the comprehensive features of the target. The target composite features are filtered using a feature selector to obtain key composite features; Based on a hybrid architecture integrating GCN and TCN, the key comprehensive features are analyzed to obtain soil erosion prediction results; wherein, the GCN module in the hybrid architecture is used to explore the spatial correlation between geographical regions; and the TCN module in the hybrid architecture is used to capture the temporal nonlinear changes in soil erosion.

2. The method according to claim 1, characterized in that, The step of fusing features from the remote sensing image, the meteorological data, and the ground-measured data to obtain the target's comprehensive features includes: The spectral features in the remote sensing image and the soil features in the ground measurement data are copied and expanded according to the time step to obtain candidate data in the target time series format; The candidate data and the meteorological data are concatenated according to feature dimensions to obtain the comprehensive features of the target.

3. The method according to claim 1, characterized in that, The step of filtering the target comprehensive features using a feature selector to obtain key comprehensive features includes: The contribution of each target's comprehensive feature is iteratively evaluated through the recurrent neural network in the feature selector. The attention mechanism in the feature selector determines the association weight between each target comprehensive feature and the core factors of soil erosion in the target area; wherein, the core factors of soil erosion include soil type, rainfall intensity, rainfall erosivity R factor and / or soil erodibility K factor; The contribution and correlation weight of each target feature are fused to obtain the target feature weight; Based on the target feature weights, the key comprehensive features are selected from the target comprehensive features.

4. The method according to claim 1, characterized in that, The hybrid architecture based on the integration of GCN and TCN analyzes the key comprehensive features to obtain soil erosion prediction results, including: The GCN module performs neighborhood feature aggregation on the key comprehensive features to mine the spatial correlation between different geographical regions and outputs spatial correlation features. The TCN module performs causal dilated convolution on the key comprehensive features, and combines residual connections to expand the temporal receptive field, outputting temporal nonlinear features. The spatial correlation features and the temporal nonlinear features are fused through the cross-modal gated fusion layer in the hybrid architecture to obtain spatiotemporal fusion features; The spatiotemporal fusion features are input into a fully connected layer for feature mapping to obtain the soil erosion prediction results.

5. The method according to claim 2, characterized in that, The method further includes: Multi-band reflectance data are extracted from the remote sensing image, and each band is processed independently using max-min normalization to obtain the reflectance tensor. The soil erosion sensitive spectral index is calculated based on the reflectance tensor; wherein the soil erosion sensitive spectral index includes at least one of the vegetation index, soil-adjusted vegetation index and differential soil index. The spectral features are obtained by concatenating the soil erosion sensitive spectral index with the reflectance tensor.

6. The method according to claim 2, characterized in that, The method further includes: The ground-measured data is mapped to the remote sensing image using the inverse distance weighted interpolation method, and then normalized to obtain a gridded soil tensor. Based on the gridded soil tensor, the soil erodibility K factor is calculated; The soil characteristics are obtained by splicing the soil erodibility K factor with the gridded soil tensor.

7. The method according to claim 1, characterized in that, The remote sensing images include multispectral reflectance and / or vegetation index; the meteorological data include rainfall, wind speed, temperature, sunshine duration and / or rainfall erosivity R factor; the ground-measured data include soil moisture content, clay content, bulk density and / or soil erosibility K factor.

8. A soil and water conservation monitoring system that combines low-altitude remote sensing from unmanned aerial vehicles (UAVs) with ground-based measurements, characterized in that... The system includes: The acquisition module is used to acquire remote sensing images, meteorological data, and ground-measured data of the target area; The feature fusion module is used to fuse the remote sensing image, the meteorological data, and the ground-measured data to obtain the comprehensive features of the target. The feature filtering module is used to filter the target comprehensive features through a feature selector to obtain key comprehensive features; The prediction module is used to analyze the key comprehensive features based on a hybrid architecture integrating GCN and TCN to obtain soil erosion prediction results; wherein, the GCN module in the hybrid architecture is used to explore the spatial correlation between geographical regions; and the TCN module in the hybrid architecture is used to capture the temporal nonlinear changes of soil erosion.

9. An electronic device, characterized in that, The system includes a processor and a memory; wherein the memory is used to store a computer program; and the processor is configured to, when executing the computer program, implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it performs the steps of the method according to any one of claims 1 to 7.

Images