Water and soil loss intelligent monitoring system based on Internet of Things

Abstract

The invention relates to the technical field of water and soil monitoring, in particular to a water and soil loss intelligent monitoring system based on the Internet of Things, which comprises an earth surface remote sensing image data acquisition module for acquiring remote sensing image data covering a monitoring area through a remote sensing satellite; the perturbation perception compensation module is used for repairing data by adopting a compensation mechanism aiming at abnormal fluctuation of the remote sensing image caused by landslide geological disturbance, and specifically comprises a disturbance detection unit and a disturbance area compensation unit; and the disaster tracing module is used for reconstructing a loss path based on the remote sensing image data restored by the perturbation perception compensation module, simulating a diffusion path of water and soil loss and tracing the loss based on the diffusion path. According to the method, water and soil loss prevention and control guidance of the key area is provided, the accuracy and efficiency of water and soil loss tracing are improved, and powerful technical support is provided for source identification and prevention and control decision of water and soil loss.

Description

Technical Field

[0001] The present invention relates to the technical field of soil and water monitoring, and in particular to an intelligent soil and water loss monitoring system based on the Internet of Things. Background Art

[0002] Soil erosion not only threatens agricultural production and the ecological environment, but also easily causes secondary disasters. Therefore, how to monitor the process and impact of soil erosion in real time and accurately, assess its risks, and take effective prevention and control measures in a timely manner has become an important issue in the current field of environmental management and disaster prevention.

[0003] Existing soil and water loss monitoring methods mainly include ground monitoring, remote sensing monitoring, meteorological monitoring and other means. Among them, although ground monitoring has high accuracy, it has limitations in practical applications due to its limited spatial coverage, high monitoring cost and inability to reflect the dynamic changes of soil and water loss in a large area in real time. Remote sensing monitoring, as an efficient and non-destructive monitoring technology, can provide large-scale and periodic surface information and has become the main means of soil and water loss monitoring.

[0004] At present, remote sensing technology is mainly used in soil and water loss monitoring for surface erosion morphology analysis, water flow distribution monitoring, vegetation coverage assessment, etc. There are still some problems with traditional remote sensing image analysis methods: First, affected by geological disasters such as landslides and geological subsidence, remote sensing image data may be missing or fluctuate abnormally, resulting in reduced data continuity and accuracy; second, existing soil and water loss path simulation methods mostly rely on simple GIS analysis, ignoring the complex spatial characteristics and temporal variation laws in the dynamic process of soil and water loss, and are unable to fully capture the diffusion path and evolution trend of soil and water loss; and traditional risk assessment methods are mostly limited to static analysis, lacking risk prediction based on the dynamic changes of real-time remote sensing data and comprehensive consideration of spatiotemporal evolution. Summary of the invention

[0005] The invention provides an intelligent monitoring system for soil and water loss based on the Internet of Things.

[0006] An intelligent monitoring system for soil and water loss based on the Internet of Things, comprising:

[0007] Surface remote sensing image data acquisition module: remote sensing image data covering the monitoring area is acquired through remote sensing satellites. The remote sensing image data includes surface erosion morphology, water flow distribution, vegetation coverage and soil exposed area characteristics;

[0008] Perturbation perception compensation module: For abnormal fluctuations in remote sensing images caused by landslide geological disturbances, a compensation mechanism is used to repair data, including:

[0009] Disturbance detection unit: uses image differential analysis and edge detection algorithms to identify and locate surface change areas in remote sensing images, and detect image abnormal areas caused by landslide geological disturbances;

[0010] Disturbance area compensation unit: Combined with the image features of the area not disturbed by landslide geology, the remote sensing image disturbed by landslide geology is reconstructed at the pixel level through the image restoration algorithm of the generative adversarial network to compensate the remote sensing image data;

[0011] Disaster source tracing module: Based on the remote sensing image data repaired by the perturbation perception compensation module, the loss path is reconstructed. The loss path reconstruction uses a deep learning algorithm to model the compensated remote sensing image data, simulate the diffusion path of soil erosion, and trace the loss based on the diffusion path.

[0012] Optionally, the surface remote sensing image data acquisition module acquires remote sensing image data covering the monitoring area, including:

[0013] Using multispectral remote sensing technology, we collected multi-band reflectance data of the surface to analyze the erosion morphology, and identified the morphological characteristics of the erosion area through different band combinations and normalized difference erosion index;

[0014] Monitor water flow paths based on surface elevation model (DEM) data combined with water flow index;

[0015] The vegetation coverage rate is extracted using near-infrared and visible light band data, and the normalized difference vegetation index (NDVI) is used to calculate the vegetation coverage status of the monitored area and observe the vegetation change trend;

[0016] For areas with exposed soil, the soil-adjusted vegetation index is used to identify the areas with exposed soil.

[0017] Optionally, the disturbance detection and positioning unit specifically includes:

[0018] Image differential analysis: register the multi-temporal remote sensing images of the monitored area to ensure that the remote sensing images collected at different times have a consistent geographic reference system. By calculating the pixel differences between adjacent time-series remote sensing images, the differential image ΔI(x,y) is obtained.

[0019] Edge detection algorithm: The Canny edge detection algorithm is used to extract the edge features of surface changes from the differential image ΔI(x,y). The gradient amplitude G(x,y) and gradient direction of the image are calculated, and the non-maximum suppression and double threshold algorithms are used to locate the edge.

[0020] Optionally, the disturbance detection unit further includes a threshold segmentation method, combining image differential analysis with an edge detection algorithm to obtain a differential edge image, performing threshold segmentation on the differential edge image based on the threshold segmentation method, identifying areas with significant differences, and thereby obtaining a landslide geological disturbance area:

[0021]

[0022] Among them, T diff is the threshold of the differential image, which is used to determine whether the pixel difference exceeds the preset significant difference standard. edge It is the threshold of the edge detection image, which is used to determine whether the edge is obvious and indicate the edge of the surface disturbance. diff and T edge ,The areas with large differences and clear edges are marked as landslide geological disturbance areas.

[0023] Optionally, the disturbance area compensation unit specifically includes:

[0024] Image feature extraction and region selection: Based on the landslide geological disturbance area in the obtained differential image, the unaffected image area around the landslide geological disturbance area is extracted from the remote sensing image as the reference image area. The specific operations are as follows:

[0025] The disturbance area is marked in the difference image by threshold segmentation method;

[0026] Extract the image area around the disturbance area that is not affected by the landslide, which is considered to be relatively stable before and after the surface changes;

[0027] Using a generative adversarial network to perform pixel-level restoration of the area disturbed by landslide geology, including using a generator in the generative adversarial network to accept a remote sensing image area of ​​the area disturbed by landslide geology as input, and using the image features of the reference image area to generate a compensated restoration image. The goal of the generator is to generate image information consistent with the surrounding area based on the texture, spectral characteristics and spatial structure of the surrounding reference image area, and to repair missing or abnormal data in the disturbance area;

[0028] The discriminator in the generative adversarial network is used to judge the authenticity of the restored image by comparing the generated restored image with the image of the real reference image area. The goal of the discriminator is to ensure that the structure and texture of the generated restored image are consistent with the actual remote sensing image, thereby improving the restoration effect;

[0029] The generator and discriminator are continuously optimized through adversarial training until the generator can generate repaired images consistent with the real data. The loss function includes content loss and adversarial loss.

[0030] After training with the generative adversarial network, the generator generates high-quality repaired images based on the image features of the surrounding reference image areas. These repaired images will replace the missing or abnormal parts of the original remote sensing images caused by geological disturbances such as landslides, and achieve pixel-level reconstruction of the disturbed areas.

[0031] Optionally, the loss path reconstruction specifically includes:

[0032] Time series processing: Process the compensated remote sensing image data in time series to obtain the changes in surface features at different time points;

[0033] The CNN-TCN hybrid model (convolutional neural network and temporal convolutional network) is used to model the processed time series remote sensing images;

[0034] Simulation of the diffusion path of soil erosion: Based on the trained CNN-TCN hybrid model, the diffusion path of soil erosion is predicted, the evolution process of soil erosion from the source to the diffusion path is simulated, the diffusion direction and range at each time node are obtained, the source of soil erosion is traced based on the prediction results, the starting point of the diffusion path is traced back, and the source of soil erosion is identified.

[0035] Optionally, the CNN-TCN hybrid model includes a CNN convolutional neural network layer and a TCN time series modeling layer, wherein:

[0036] The CNN convolutional neural network layer extracts spatial features from remote sensing images and identifies areas of surface change, such as water flow paths, soil erosion, and vegetation coverage;

[0037] The TCN time series modeling layer captures the dynamic changes in the time dimension and analyzes the evolution of soil erosion areas over time and their diffusion patterns.

[0038] Optionally, the compensated remote sensing image data is processed in time series to obtain an image data set {I1, I2, ..., I t}, where I t Represents the remote sensing image at time point t. Through this data set, the model captures the dynamic changes in the time series and analyzes the evolution of soil erosion;

[0039] The CNN convolutional neural network layer is used to extract spatial features in remote sensing images from remote sensing images, including input image data I t After the convolutional layer processing, the image feature representation F is generated t :

[0040] The TCN temporal modeling layer captures dynamic changes in the time dimension, and the input is the processed remote sensing image features F1, F2, ..., F t, through time series modeling, the prediction of future moments is generated, which is expressed as:

[0041] in, It is the model output, which represents the soil erosion diffusion path at time point t. LSTM is the long short-term memory network selected by the TCN time series modeling layer;

[0042] Optionally, the generated image feature representation F t Expressed as: F t =CNN(I t ), where CNN represents convolutional layer processing;

[0043] The convolution operation is expressed as: Among them, I t is the input image data, W i is the convolution kernel (filter), b i is the bias term, k is the number of convolution kernels, and * represents the convolution operation.

[0044] Optionally, the disaster tracing module also includes soil erosion risk assessment, based on the soil erosion diffusion path For each geographical unit (pre-divided into grid units) on the diffusion path, the risk of soil erosion it faces is assessed based on the weighted effect of terrain factors. The risk assessment is expressed as:

[0045] R t (i,j)=α·Slope(i,j)+β·Roughness(i,j)+γ·SoilType(i,j);

[0046] Among them, R t (i, j) represents the soil erosion risk value at coordinate (i, j) at time point t. Slope(i, j), Roughness(i, j), SoilType(i, j) are the slope, terrain undulation, and soil type factors of the corresponding geographical unit. α, β, and γ are the weights of each terrain factor, indicating the degree of influence on soil erosion.

[0047] By calculating each geographical unit, a comprehensive risk assessment value R is obtained. t (i,j), reflecting the soil and water loss risk faced by the geographical unit at the current moment.

[0048] According to R t (i, j), set the threshold T to delineate the high-risk area, R t The geographical unit where (i, j) exceeds the threshold T is a high-risk area.

[0049] Beneficial effects of the present invention:

[0050] The present invention integrates a perturbation perception compensation module and uses image differential analysis and generative adversarial network technology to effectively repair remote sensing image data anomalies caused by geological disasters such as landslides and subsidence. This compensation mechanism ensures the continuity and accuracy of time series remote sensing image data, avoids data breaks or discontinuities caused by emergencies, and thus provides high-quality, coherent remote sensing data. By repairing abnormal areas, it accurately reflects changes in surface features and provides reliable data support for disaster tracing, loss path reconstruction, and risk assessment. It improves the accuracy and credibility of soil and water loss monitoring, especially in post-disaster recovery and dynamic monitoring.

[0051] In the present invention, the disaster tracing module is combined with a deep learning algorithm, especially a CNN-TCN hybrid model, which successfully realizes the efficient simulation and source tracing of the diffusion path of soil and water loss. By performing deep modeling on the compensated time series remote sensing image data, it can automatically identify the diffusion path of soil and water loss and its impact range, and accurately simulate the evolution process of soil and water loss from source to diffusion. In addition, based on the reverse tracing function of the diffusion path, it can locate the source of soil and water loss, and clarify its impact on the surrounding areas, provide guidance on soil and water loss prevention and control in key areas, improve the accuracy and efficiency of soil and water loss tracing, and provide strong technical support for source identification and prevention and control decisions of soil and water loss.

[0052] The present invention, based on the reconstruction of the loss path, combines terrain factors (slope, terrain undulation, soil type) to conduct soil and water loss risk assessment. By performing a weighted assessment on each geographical unit on the diffusion path, the soil and water loss risk faced by each region is assessed, high-risk areas are identified in real time, and decision support is provided for soil and water loss management and prevention. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 A schematic diagram of system function modules according to an embodiment of the present invention;

[0055] Figure 2 Schematic diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0057] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0058] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0059] like Figure 1-Figure 2 As shown, an intelligent monitoring system for soil and water loss based on the Internet of Things includes:

[0060] Surface remote sensing image data acquisition module: remote sensing image data covering the monitoring area is acquired through remote sensing satellites. The remote sensing image data includes surface erosion morphology, water flow distribution, vegetation coverage and soil exposed area characteristics;

[0061] Perturbation perception compensation module: For abnormal fluctuations in remote sensing images caused by landslide geological disturbances, a compensation mechanism is used to repair data, including:

[0062] Disturbance detection unit: uses image differential analysis and edge detection algorithms to identify and locate surface change areas in remote sensing images, and detect image abnormal areas caused by landslide geological disturbances;

[0063] Disturbance area compensation unit: Combined with the image features of the area not disturbed by landslide geology, the remote sensing image disturbed by landslide geology is reconstructed at the pixel level through the image restoration algorithm of the generative adversarial network to compensate the remote sensing image data;

[0064] Remote sensing images are part of time series data. Sudden events may cause data to be broken or discontinuous. The compensation mechanism ensures the continuity of the time series by reconstructing the images, thereby providing reliable dynamic data support for subsequent analysis. High-quality, continuous remote sensing image data is the basis for reconstructing the paths of landslides, subsidence and soil erosion. The compensation mechanism repairs abnormal areas in the images and provides accurate data for disaster tracing and risk assessment.

[0065] Disaster source tracing module: Based on the remote sensing image data repaired by the perturbation perception compensation module, the loss path is reconstructed. The loss path reconstruction uses a deep learning algorithm to model the compensated remote sensing image data, simulate the diffusion path of soil erosion, and trace the loss based on the diffusion path.

[0066] The surface remote sensing image data acquisition module acquires remote sensing image data covering the monitoring area, including:

[0067] Using multispectral remote sensing technology, we collected multi-band reflectance data of the surface to analyze the erosion morphology, and identified the morphological characteristics of the erosion area through different band combinations and normalized difference erosion index;

[0068] The normalized difference erosion index is expressed as: in;

[0069] R stable Represents the reflectance of stable areas, collected by the near-infrared band (NIR, 0.76–0.90 μm), which is used to characterize the vegetation or soil cover status of uneroded areas;

[0070] R erosion Represents the reflectance of the eroded area, collected by the shortwave infrared band (SWIR, 1.55–1.75 μm), which is used to characterize the spectral characteristics of the bare soil area;

[0071] When the NDEI value is large, it means that the area is highly eroded and is used to accurately locate the surface erosion morphology.

[0072] Monitor water flow paths based on surface elevation model (DEM) data combined with water flow index;

[0073] The water flow index is expressed as: Among them, Slope is the surface slope, which is obtained based on the surface elevation model and represents the potential downhill path of water flow, and Accum water It is the accumulated amount of water flow, and the regional intensity value of water flow convergence is detected by radar imaging. Area represents the unit area in the monitoring area and is used to standardize the water flow intensity. Among them, Δh is the elevation difference between adjacent pixels, and Δd is the horizontal distance between adjacent pixels. By calculating the water flow index, the water flow distribution path and change trend can be dynamically monitored.

[0074] The vegetation coverage rate is extracted using near-infrared and visible light band data, and the normalized difference vegetation index (NDVI) is used to calculate the vegetation coverage status of the monitored area and observe the vegetation change trend;

[0075] For areas with exposed soil, the soil-adjusted vegetation index is used to identify the areas with exposed soil.

[0076] The normalized difference vegetation index is expressed as: in;

[0077] R NIR It is the reflectance in the near-infrared band (0.76-0.90 μm), which represents the health status of vegetation;

[0078] R Red It is the reflectance in the red light band (0.63-0.69 μm), representing the photosynthesis signal absorbed by the surface;

[0079] NDVI values ​​in the range of 0.2-0.8 indicate the state of vegetation coverage, and the vegetation coverage and changes are evaluated based on the trend of the values;

[0080] The soil adjusted vegetation index is expressed as: in;

[0081] R NIR and R Red Consistent with the definition of NDVI, L is a correction factor with a value range of 0-1. It is adjusted based on the actual soil background in the monitored area and takes a value close to 1 when the vegetation is sparse.

[0082] The disturbance detection and positioning unit specifically includes:

[0083] Image differential analysis: The multi-temporal remote sensing images of the monitored area are registered to ensure that the remote sensing images collected at different times have a consistent geographic reference system. The difference image ΔI(x,y) is obtained by calculating the pixel difference between adjacent time-series remote sensing images. The calculation is: ΔI(x,y)=I t1 (x,y)-I t0 (x,y), where I t1 (x, y) represents the pixel value at position (x, y) in the remote sensing image at time t1, I t0 (x, y) represents the pixel value at position (x, y) in the remote sensing image at time t0. The calculated differential image reveals the changing area of ​​the surface, where the area with significant pixel value changes is the potential landslide area;

[0084] Edge detection algorithm: The Canny edge detection algorithm is used to extract the edge features of surface changes from the differential image ΔI(x,y). The gradient amplitude G(x,y) and gradient direction of the image are calculated, and the non-maximum suppression and double threshold algorithms are used for edge positioning. The formula is as follows:

[0085] Among them, I x (x,y) and I y (x, y) are the gradient values ​​of the image in the x and y directions respectively. Through this edge detection process, the significant change edges in the differential image are obtained to assist in identifying possible geological disturbance areas (landslides).

[0086] The image differential analysis results and edge detection results are combined to obtain a differential image with edge information. In this process, the potential change area is first located through differential analysis, and then the precise boundary of the change area is marked through edge detection. Combining the results of the two, a differential edge image is obtained, which shows areas with significant differences and clear edges. These areas are potential landslide geological disturbance areas.

[0087] The disturbance detection unit also includes a threshold segmentation method, which combines image differential analysis with an edge detection algorithm to obtain a differential edge image. Based on the threshold segmentation method, the differential edge image is threshold segmented to identify areas with significant differences, thereby obtaining the landslide geological disturbance area:

[0088]

[0089] Among them, T diff is the threshold of the differential image, which is used to determine whether the pixel difference exceeds the preset significant difference standard. edge It is the threshold of the edge detection image, which is used to determine whether the edge is obvious and indicate the edge of the surface disturbance. diff and T edge ,The areas with large differences and clear edges are marked as landslide geological disturbance areas.

[0090] The landslide geological disturbance area obtained by the threshold segmentation method is subjected to geometric morphological analysis to further locate the location and scope of the disturbance, as follows:

[0091] Region contour extraction: Use morphological operations (such as dilation and erosion) to extract the contour of the disturbed region and calculate the spatial characteristics (area, aspect ratio) of the disturbed region.

[0092] Calculation of morphological characteristics: By calculating the shape factor of the area (rectangularity, roundness, crack degree), the geological characteristics of the disturbed area are analyzed. For landslide or collapse areas, which usually have irregular and broken geometric shapes, morphological analysis can provide accurate information about the location and scale of the disturbance.

[0093] The disturbance area compensation unit specifically includes:

[0094] Image feature extraction and region selection: Based on the landslide geological disturbance area in the obtained differential image, the unaffected image area around the landslide geological disturbance area is extracted from the remote sensing image as the reference image area. The specific operations are as follows:

[0095] The disturbance area is marked in the difference image by threshold segmentation method;

[0096] Extract the image area around the disturbance area that is not affected by the landslide, which is considered to be relatively stable before and after the surface changes;

[0097] The use of a generative adversarial network to perform pixel-level repair on the area disturbed by landslide geology includes using the generator in the generative adversarial network to accept the remote sensing image area of ​​the area disturbed by landslide geology as input, and using the image features of the reference image area to generate a compensated repair image. The goal of the generator is to generate image information consistent with the surrounding area based on the texture, spectral characteristics and spatial structure of the surrounding reference image area, and repair the missing or abnormal data in the disturbed area, which is expressed as:

[0098] Among them, I context is the remote sensing image of the surrounding reference image area, G is the generator model, which generates the restored image by learning the image features of the surrounding reference image area;

[0099] Using the discriminator in the generative adversarial network, the discriminator judges the authenticity of the repaired image by comparing the generated repaired image with the image of the real reference image area. The goal of the discriminator is to ensure that the structure and texture of the generated repaired image are consistent with the actual remote sensing image, thereby improving the repair effect, which is expressed as:

[0100] D(I repaired )≈1,D(I real )≈1,D(I fake )≈0, where I repaired is the repaired image generated by the generator, I real is the real image of the unperturbed area, I fake is a generated fake image, and the discriminator tries to distinguish between true and false;

[0101] The generator and discriminator are continuously optimized through adversarial training until the generator can generate repaired images consistent with the real data. The loss function includes content loss and adversarial loss, which is expressed as:

[0102]

[0103] Among them, ContentLoss represents the content loss, which is used to minimize the pixel difference between the generated image and the real image, and ensure that the texture, color, and structure of the repaired area are consistent with the surrounding area. λ is the weight coefficient, which is used to balance the adversarial loss and content loss. is the total loss of the generator, which includes the combination of adversarial loss and content loss, is the expected value operation, It is part of the generator loss and is used to measure whether the generated image is close to the real image;

[0104] After training the generative adversarial network, the generator generates high-quality repaired images based on the image features of the surrounding reference image areas. These repaired images will replace the missing or abnormal parts of the original remote sensing images caused by geological disturbances such as landslides, and achieve pixel-level reconstruction of the disturbed areas. The specific process includes:

[0105] Repair each pixel in the disturbed area and generate pixel values ​​consistent with the surrounding environment through the generator;

[0106] Maintain the spectral, textural and spatial characteristics of the image so that the restored area blends seamlessly with the surrounding area, ensuring the naturalness and consistency of image restoration.

[0107] Loss path reconstruction specifically includes:

[0108] Time series processing: The compensated remote sensing image data is processed in time series to obtain the changes in surface features at different time points. The remote sensing image at each time point is repaired by the perturbation perception compensation module and used as input data to capture the dynamic changes in the process of soil erosion through the time series method.

[0109] The CNN-TCN hybrid model (convolutional neural network and temporal convolutional network) is used to model the processed time series remote sensing images;

[0110] Simulation of the diffusion path of soil erosion: Based on the trained CNN-TCN hybrid model, the diffusion path of soil erosion is predicted, the evolution process of soil erosion from the source to the diffusion path is simulated, the diffusion direction and range at each time node are obtained, the source of soil erosion is traced based on the prediction results, the starting point of the diffusion path is traced back, and the source of soil erosion is identified.

[0111] Reverse tracing of the diffusion path can find the starting point or key area of ​​soil erosion, help identify the source area, reconstruct the path of soil erosion based on the source of erosion, and clarify which areas are the key areas of soil erosion.

[0112] The CNN-TCN hybrid model includes a CNN convolutional neural network layer and a TCN time series modeling layer, where:

[0113] The CNN convolutional neural network layer extracts spatial features from remote sensing images and identifies areas of surface change, such as water flow paths, soil erosion, and vegetation coverage;

[0114] The TCN time series modeling layer captures the dynamic changes in the time dimension and analyzes the evolution of soil erosion areas over time and their diffusion patterns.

[0115] After the compensated remote sensing image data is processed in time series, the image data set {I1, I2, ..., I t}, where I t Represents the remote sensing image at time point t. Through this data set, the model captures the dynamic changes in the time series and analyzes the evolution of soil erosion;

[0116] The CNN convolutional neural network layer is used to extract spatial features from remote sensing images, including input image data I t After the convolutional layer processing, the image feature representation F is generated t :

[0117] F t =CNN(I t ), where CNN represents convolutional layer processing;

[0118] The convolution operation is expressed as: Among them, I t is the input image data, W i is the convolution kernel (filter), b i is the bias term, k is the number of convolution kernels, and * represents the convolution operation;

[0119] The TCN temporal modeling layer captures dynamic changes in the time dimension, and the input is the processed remote sensing image features F1, F2, ..., F t , through time series modeling, the prediction of future moments is generated, which is expressed as:

[0120] in, It is the model output, which represents the soil erosion diffusion path at time point t. LSTM is the long short-term memory network selected by the TCN time series modeling layer;

[0121] The LSTM long short-term memory network is represented as follows:

[0122] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0123] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0124]

[0125] o t =σ(W o ·[h t-1 ,x t ]+b o );

[0126] h t =o t tanh(C t );

[0127] Among them, x t is the input feature at time point t, including F1, F2, ..., F t ,h t is the hidden state at the current moment, h t-1 is the hidden state at time t-1, C t is the cell state at the current moment, C t-1 is the cell state at time t-1, W f ,W i ,W C ,W o is the weight matrix, b f ,b i ,b C ,b o is the bias term, They are the calculation results of the forget gate, input gate, candidate cell state and output gate respectively.

[0128] Generate image feature representation F t Expressed as: F t =CNN(I t ), where CNN represents convolutional layer processing;

[0129] The convolution operation is expressed as: Among them, I t is the input image data, W i is the convolution kernel (filter), b iis the bias term, k is the number of convolution kernels, and * represents the convolution operation.

[0130] The disaster source tracing module also includes soil erosion risk assessment, based on the soil erosion diffusion path Y t For each geographical unit (pre-divided into grid units) on the diffusion path, the risk of soil erosion it faces is evaluated according to the weighted effect of terrain factors. The risk assessment is expressed as:

[0131] R t (i,j)=α·Slope(i,j)+β·Roughness(i,j)+γ·SoilType(i,j);

[0132] Among them, R t (i, j) represents the soil erosion risk value at coordinate (i, j) at time point t. Slope(i, j), Roughness(i, j), SoilType(i, j) are the slope, terrain undulation, and soil type factors of the corresponding geographical unit. α, β, and γ are the weights of each terrain factor, indicating the degree of influence on soil erosion.

[0133] By calculating each geographical unit, a comprehensive risk assessment value R is obtained. t (i,j), reflecting the soil and water loss risk faced by the geographical unit at the current moment.

[0134] According to R t (i, j), set the threshold T to delineate the high-risk area, R t The geographical unit where (i, j) exceeds the threshold T is a high-risk area.

[0135] Obtaining topographic data: Topographic data is a key factor in assessing soil erosion risk. Topographic data includes:

[0136] Slope: The steeper the slope, the stronger the energy of the water flow and the more easily the soil will erode.

[0137] Topographic undulation: The terrain is undulating, and the speed and path of water diffusion are more uneven, resulting in increased local soil erosion.

[0138] Soil type: Different soils have different resistance to erosion, for example sand is more susceptible to water erosion than clay.

[0139] The above data can be obtained through remote sensing images, geographic information systems (GIS) or existing terrain datasets.

[0140] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0141] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An intelligent monitoring system for soil erosion based on the Internet of Things, characterized in that: include: Surface remote sensing image data acquisition module: remote sensing image data covering the monitoring area is acquired through remote sensing satellites. The remote sensing image data includes surface erosion morphology, water flow distribution, vegetation coverage and soil exposed area characteristics; Perturbation perception compensation module: For abnormal fluctuations in remote sensing images caused by landslide geological disturbances, a compensation mechanism is used to repair data, including: Disturbance detection unit: using image differential analysis and edge detection algorithms to identify and locate the surface change areas in remote sensing images, and detect areas disturbed by landslide geology; Disturbance area compensation unit: Combined with the image features of the area not disturbed by landslide geology, the remote sensing image disturbed by landslide geology is reconstructed at the pixel level through the image restoration algorithm of the generative adversarial network to compensate the remote sensing image data; Disaster source tracing module: Based on the remote sensing image data repaired by the perturbation perception compensation module, the loss path is reconstructed. The loss path reconstruction uses a deep learning algorithm to model the compensated remote sensing image data, simulate the diffusion path of soil erosion, and trace the loss based on the diffusion path.

2. According to the intelligent soil and water loss monitoring system based on the Internet of Things in claim 1, it is characterized in that: The surface remote sensing image data acquisition module acquires remote sensing image data covering the monitoring area, including: Using multispectral remote sensing technology, we collected multi-band reflectance data of the surface to analyze the erosion morphology, and identified the morphological characteristics of the erosion area through different band combinations and normalized difference erosion index; Monitor water flow paths based on surface elevation model data combined with water flow indices; The vegetation coverage rate is extracted using near-infrared and visible light band data, and the normalized vegetation index is used to calculate the vegetation coverage status of the monitored area and observe the vegetation change trend; For areas with exposed soil, the soil-adjusted vegetation index is used to identify the areas with exposed soil.

3. According to the intelligent soil and water loss monitoring system based on the Internet of Things in claim 1, it is characterized in that: The disturbance detection and positioning unit specifically includes: Image differential analysis: register the multi-temporal remote sensing images of the monitored area to ensure that the remote sensing images collected at different times have a consistent geographic reference system. By calculating the pixel differences between adjacent time-series remote sensing images, the differential image ΔI(x,y) is obtained. Edge detection algorithm: The Canny edge detection algorithm is used to extract the edge features of surface changes from the differential image ΔI(x,y). The gradient amplitude G(x,y) and gradient direction of the image are calculated, and the non-maximum suppression and double threshold algorithms are used to locate the edge.

4. According to the intelligent soil and water loss monitoring system based on the Internet of Things in claim 3, it is characterized in that: The disturbance detection unit also includes a threshold segmentation method, which combines image difference analysis with an edge detection algorithm to obtain a differential edge image, performs threshold segmentation on the differential edge image based on the threshold segmentation method, identifies the difference area, and thus obtains the landslide geological disturbance area: Among them, T diff is the threshold of the differential image, which is used to determine whether the pixel difference exceeds the preset significant difference standard. edge It is the threshold of the edge detection image, which is used to determine whether the edge is obvious and indicates the edge of the surface disturbance.

5. According to the intelligent soil and water loss monitoring system based on the Internet of Things in claim 1, it is characterized in that: The disturbance area compensation unit specifically includes: Image feature extraction and region selection: Based on the landslide geological disturbance area in the obtained differential image, the unaffected image area around the landslide geological disturbance area is extracted from the remote sensing image as the reference image area; Using a generative adversarial network to perform pixel-level restoration of the area disturbed by landslide geology, including using a generator in the generative adversarial network to accept a remote sensing image area of ​​the area disturbed by landslide geology as input, and using the image features of the reference image area to generate a compensated restoration image. The goal of the generator is to generate image information consistent with the surrounding area based on the texture, spectral characteristics and spatial structure of the surrounding reference image area, and to repair missing or abnormal data in the disturbance area; Use the discriminator in the generative adversarial network to judge the authenticity of the restored image by comparing the generated restored image with the image of the real reference image area. The goal of the discriminator is to ensure that the generated restored image is consistent with the structure and texture of the actual remote sensing image; The generator and discriminator are continuously optimized through adversarial training until the generator can generate repaired images consistent with the real data. The loss function includes content loss and adversarial loss. After training with a generative adversarial network, the generator generates a restored image based on the image features of the surrounding reference image area.

6. The intelligent monitoring system for soil and water loss based on the Internet of Things according to claim 1 is characterized in that: The loss path reconstruction specifically includes: Time series processing: Process the compensated remote sensing image data in time series to obtain the changes in surface features at different time points; The CNN-TCN hybrid model is used to model the processed time series remote sensing images; Simulation of the diffusion path of soil erosion: Based on the trained CNN-TCN hybrid model, the diffusion path of soil erosion is predicted, and the source of soil erosion is traced based on the prediction results, the starting point of the diffusion path is traced back, and the source of soil erosion is identified.

7. The intelligent monitoring system for soil and water loss based on the Internet of Things according to claim 6 is characterized in that: The CNN-TCN hybrid model includes a CNN convolutional neural network layer and a TCN time series modeling layer, wherein: The CNN convolutional neural network layer extracts spatial features from remote sensing images and identifies areas of surface change; The TCN time series modeling layer captures the dynamic changes in the time dimension and analyzes the evolution of soil erosion areas over time and their diffusion patterns.

8. The intelligent monitoring system for soil and water loss based on the Internet of Things according to claim 7 is characterized in that: The compensated remote sensing image data is processed in time series to obtain an image data set {I1, I2, ..., I t }, where I t Represents the remote sensing image at time point t; The CNN convolutional neural network layer is used to extract spatial features in remote sensing images from remote sensing images, including input image data I t After the convolutional layer processing, the image feature representation F is generated t ; The TCN temporal modeling layer captures dynamic changes in the time dimension, and the input is the processed remote sensing image features F1, F2, ..., F t , through time series modeling, the prediction of future moments is generated, which is expressed as: in, It is the model output, which represents the soil erosion diffusion path at time point t. LSTM is the long short-term memory network selected by the TCN time series modeling layer.

9. The intelligent soil and water loss monitoring system based on the Internet of Things according to claim 8 is characterized in that: The generated image feature representation F t Expressed as: F t =CNN(I t ), where CNN represents convolutional layer processing; The convolution operation is expressed as: Among them, I t is the input image data, W i is the convolution kernel, b i is the bias term, k is the number of convolution kernels, and * represents the convolution operation.

10. The intelligent soil and water loss monitoring system based on the Internet of Things according to claim 8 is characterized in that: The disaster tracing module also includes soil erosion risk assessment, based on the soil erosion diffusion path For each geographical unit on the diffusion path, the soil and water loss risk it faces is assessed based on the weighted effect of terrain factors. The risk assessment is expressed as: R t (i,j)=α·Slope(i,j)+β·Roughness(i,j)+γ·SoilType(i,j); Among them, R t (i, j) represents the soil erosion risk value at coordinate (i, j) at time point t. Slope(i, j), Roughness(i, j), SoilType(i, j) are the slope, terrain undulation, and soil type factors of the corresponding geographical unit. α, β, and γ are the weights of each terrain factor, indicating the degree of influence on soil erosion. By calculating each geographical unit, a comprehensive risk assessment value R is obtained. t (i, j), reflects the soil and water loss risk faced by the geographical unit at the current moment; According to R t (i, j), set the threshold T to delineate the high-risk area, R t The geographical unit where (i, j) exceeds the threshold T is a high-risk area.

Images