Field monitoring and protecting equipment, method and system for water level on freezing layer in permafrost region

Abstract

The invention belongs to the field of field water level measuring and positioning, and relates to a field monitoring and protecting device, method and system for the water level on a freezing layer in a permafrost region, comprising a water level pipe, a nut, a steel wire rope, a temperature control system, a heating device group and a plurality of water level gauges; the nut is arranged at the top of the water level pipe; a plurality of water level gauges are uniformly bound on the steel wire rope, and one end of the steel wire rope is arranged at the central point of the nut, so that the steel wire rope and the water level gauges are suspended inside the water level pipe; the heating device set is arranged in the inner wall of the water level pipe and connected with the temperature control system through an electric lead. The temperature control system is used for controlling the temperature of the heating device; by arranging a plurality of water level gauges, water level change information can be accurately monitored.

Description

Technical Field

[0001] The present invention belongs to the field of field water level measurement and positioning, and particularly relates to a field monitoring and protection device, method and system for the water level above the frozen layer in permafrost regions. Background Art

[0002] Permafrost includes two parts: the frozen soil layer and the active layer. The frozen soil layer refers to the part of the soil below the permafrost where the soil temperature remains at or below 0°C for two years or more. The active layer refers to the part of the soil that undergoes freeze-thaw cycles with the changing seasons. The active layer is adjacent to the atmosphere above and the frozen soil layer below, and is the main channel for energy and water exchange between permafrost and the atmosphere. The active layer of permafrost is affected by factors such as seasonal changes, topography, and surface cover. Generally, the active layer is thinner in summer and thicker in winter; the undulation and slope of the topography can affect the flow and accumulation of water, thereby affecting the formation of the active layer. In areas with a large slope, water is likely to flow away, which may result in a thinner active layer; vegetation cover has a great impact on the active layer. Vegetation can reduce the surface temperature and slow down the melting rate of the soil. At the same time, the roots of vegetation can also affect the soil structure, thereby affecting the formation of the active layer.

[0003] In recent years, with the continuous intensification of climate warming, the phenomenon of permafrost degradation has emerged. The active layer has thickened, the infiltration of soil moisture has decreased, and ultimately the water supply to the overlying vegetation has become insufficient, leading to a large-scale degradation of the alpine ecosystem. Therefore, monitoring the changes in the water table and temperature in the permafrost is of great significance for the study of the local ecological environment. At present, the test instruments for detecting water levels are rarely used in alpine permafrost areas and do not have the function of recording time and water level simultaneously. When measuring the water level, the water level gauge needs to be in contact with water. In permafrost areas, the freezing and melting of soil water will cause extrusion damage to the water level gauge, affecting the normal use of the water level gauge. In view of the above problems, it is necessary to study a monitoring and protection device for water level gauges for long-term field monitoring. The altitude in permafrost areas is often relatively high, and it is difficult to find the locations of water level wells. The field positioning device mainly relies on positioning systems such as GPS and Beidou. However, GPS and Beidou can only find the approximate location of a point and it is difficult to carry out precise guidance. Some positioning uses GPS and total station for positioning, but the total station is expensive and heavy, and it is difficult to operate in permafrost areas and is not suitable for individual data collection work. If the above solutions are used and the field signal is lost, it is difficult to find the water level measurement point. At present, AI technology has gradually entered the mass market and has achieved remarkable achievements in various fields. Machine learning is an important branch of AI technology. It mainly learns patterns and rules through a large amount of data. Common applications of machine learning technology in the field of artificial intelligence include computer vision, natural language processing, robotics and automation. Therefore, this research invented a method that uses image recognition and processing technology based on deep learning and graph neural network models in combination with traditional GPS / Beidou navigation systems to solve the positioning problem of water level gauges in permafrost areas without signals in the wild. Summary of the Invention

[0004] To solve the problems of effectively measuring the changes in the groundwater level in permafrost areas and finding fixed points in the wild, the present invention provides a field monitoring and protection device for the water table above the frozen layer in permafrost areas. The device includes: a water level pipe, a nut, a steel wire rope, a temperature control system, a heating device group, and a plurality of water level gauges; the nut is arranged at the top of the water level pipe; a plurality of water level gauges are evenly tied to the steel wire rope, and one end of the steel wire rope is arranged at the center point of the nut, so that the steel wire rope and the water level gauges are suspended inside the water level pipe; the heating device group is arranged in the inner wall of the water level pipe and is connected to the temperature control system through a conducting wire; the temperature control system is used to control the temperature of the heating device.

[0005] A field monitoring method for the water table above the frozen layer in permafrost areas, which uses the above-mentioned field monitoring and protection device for water levels, includes:

[0006] S1. Select water level measurement points in different environments and obtain the surrounding environment images of the water level measurement points;

[0007] S2. Set the field monitoring and protection device for water levels at the corresponding water level measurement points;

[0008] S3. Obtain water level change information through the field monitoring and protection equipment for water level;

[0009] S4. According to the image features, obtain the patterns and their regular models suitable for learning image data, and conduct training and testing;

[0010] S5. Input the water level change information and the preprocessed images into the trained monitoring model to obtain the monitoring results, where the monitoring model uses the graph neural network model of deep learning.

[0011] A field monitoring system for the water level above the frozen layer in permafrost areas, which is used to execute the above-mentioned field monitoring method for the water level above the frozen layer in permafrost areas. The system includes: a surrounding environment image acquisition module, a field monitoring and protection equipment for water level, an image preprocessing module, a monitoring module, and an output module;

[0012] The surrounding environment image acquisition module is used to acquire the surrounding environment images of the water level measurement point;

[0013] The image preprocessing module is used to perform normalization processing and data enhancement on the surrounding environment images;

[0014] The field monitoring and protection equipment for water level is used to acquire the water level change information of the water level measurement point;

[0015] The monitoring module is used to process the enhanced images and the water level change information to obtain the monitoring results;

[0016] The output module is used to output the monitoring results.

[0017] Advantages of the present invention:

[0018] The field monitoring and protection equipment for water level designed by the present invention is provided with a plurality of round holes on the pipe wall of the water level pipe, which can accelerate the measurement of the water level, thereby improving the measurement efficiency; the present invention can accurately monitor the water level change information by setting a plurality of water level gauges; the method of the present invention can more accurately determine the location of the water level change through the environmental information and the water level change information, so as to achieve the purpose of accurate supervision. Description of the drawings

[0019] Figure 1 It is the water level well position map set by the present invention based on the influence of different elevations and vegetation on the water level;

[0020] Figure 2 It is the structural diagram of the field monitoring and protection equipment for water level of the present invention;

[0021] Figure 3 It is the schematic diagram of the side wall support of the water level pipe of the present invention;

[0022] Figure 4 Structural diagram of the temperature control system of the present invention;

[0023] Figure 5 Schematic diagram of the nut structure of the present invention;

[0024] Figure 6 Schematic diagram of the initial data collection of the present invention;

[0025] Figure 7 Flowchart of a field monitoring method for the water table above the frozen layer in permafrost regions of the present invention;

[0026] Among them, 1. Nut, 2. Steel wire rope, 3. Power supply, 4. Temperature control system, 5. First circuit, 6. Fixator, 7. Heating device, 8. First metal tripod, 9. Water level gauge; 10. Round hole, 11. Solar panel, 12. Second metal tripod, 13. Second circuit, 14. Pulley, 15. Third circuit, 16. Conductive wire, 17. Water storage tank, 18. Circuit connection switch. Detailed implementation manners

[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0028] A field monitoring and protection device for the water table above the frozen layer in permafrost regions, as Figure 2 shown, the device includes: a water level pipe, a nut 1, a steel wire rope 2, a temperature control system 4, a heating device group, and a plurality of water level gauges 9; the nut is arranged at the top of the water level pipe; a plurality of water level gauges are evenly tied to the steel wire rope, and one end of the steel wire rope is arranged at the center point of the nut, so that the steel wire rope and the water level gauge are suspended inside the water level pipe; the heating device group is arranged in the inner wall of the water level pipe and is connected to the temperature control system through a conductive wire; the temperature control system is used to control the temperature of the heating device.

[0029] Use a drone to collect field image data, and a computer realizes image processing, model selection, model training, testing, model saving, and offline result prediction through Python language tools; a water level gauge testing and protection device.

[0030] In this embodiment, as Figure 1As shown in the figure, the water level gauge used in this application is a HoBo automatic water level recorder, with a diameter of 25 mm, a length of 150 mm, and a weight of 210 g. The water level measurement range is 0 - 4 meters. It is applicable to the measurement of temperature and water level in lakes, streams, wells, swamps, etc. The water level gauge obtains the height of the groundwater level through the water pressure of the upper water layer in the permafrost area. The device is equipped with a self-recording data collector with a storage function; the material of the water level gauge is titanium alloy, which has certain corrosion resistance. The sensor is a ceramic pressure sensor, and the pressure measurement range is 0 - 145 kPa. The working environment of the water level gauge is the field, so it has a lightning protection function. The interface is an optical special data download interface. When reading data, the Base station data transmission base is directly connected to the adapter, the adapter is connected to the water level gauge, and then inserted into the USB interface of the computer to start, set parameters, and download data. The air pressure and water level conversion formula is as follows:

[0031] P = ρgh

[0032] Where, P is the hydrostatic pressure, ρ is the density of water, g is the acceleration of gravity, and h is the water level depth.

[0033] The water level pipe is a galvanized cylindrical metal steel pipe, with a length of 1150 mm, an inner diameter of 45 mm, and a pipe wall thickness of 45 mm, which contains about 1 mm thick metallic zinc to ensure the normal progress of water level measurement work; a nut is added to the top of the water level pipe. The outer diameter of the nut is 60 mm and the height is 30 mm. An iron ring is welded to the top and connected to the water level gauge through a steel wire rope and a nylon rope. The side wall is provided with threads and is tightly connected to the water level pipe through the threads. When checking the inside or reading the data stored in the water level gauge, the nut can be opened with a pipe wrench; the water level gauge is connected to the nut through a steel wire rope and a nylon rope and is suspended in the center of the water level pipe and placed below the water level. When reading data, the Base station data transmission base is directly connected to the adapter, and the adapter is connected to the water level gauge. Then insert it into the USB interface of the computer to start, set parameters, and download data; the temperature control device is a self-made circuit closed caused by the freezing expansion of pure water and the circuit is disconnected after melting. The power supply is a device that contains electricity itself and can store electricity. The power supply system uses a solar panel 11 to provide power for the power supply. The heating device 7 is mainly composed of a heating pipe and a resistance wire.

[0034] In this embodiment, to enable the melted water in the active layer of permafrost to enter the piezometer tube, the present invention is designed with round holes with a diameter of 10 mm on the tube wall, with an upper and lower spacing of 50 mm, and staggered distribution left and right; to prevent fine sand and clay with finer particles from entering the inside of the piezometer tube along with the water during the melting process, a plastic filter screen is wrapped outside the piezometer tube; to ensure that the piezometer tube stands stably in the soil when the permafrost in the active layer melts, three metal tripods 8 are set at a distance of 40 cm from the top of the metal tube wall and inserted into the surrounding soil to ensure the stability of the piezometer tube. The total length of the metal frame is 300 mm, and the top 50 mm is set as a pointed foot, specifically as Figure 3 shown.

[0035] In this embodiment, to prevent the problem that the water level gauge cannot work properly due to damage caused by water freezing, as Figure 4 shown, a heating device is installed on the piezometer tube wall. The heating device is connected to the power supply through a temperature control system. The temperature control system controls the connection of the circuit according to the principle of thermal expansion and contraction. The main materials include a pulley 14 that can slide upward, a water storage tank 17 filled with distilled water, a buoyant insulating wire 18, a first wire 5 connected to a resistance wire and a wire 16, a second wire 13 connected to the power supply, and a third wire 15 connected to a wire 13 and a wire 16. The power supply uses a power supply with charging and power storage functions, and a solar panel can make up for the problem of insufficient power of the power supply.

[0036] In this embodiment, the piezometer tube is equipped with a nut. The side wall of the nut has a threaded groove, and a metal ring is welded on the top. Inside the piezometer tube, to fix the water level gauge at a specified height, the water level gauge is suspended in the piezometer tube through a steel wire rope and a nylon rope. The schematic diagram of the nut is as Figure 5 .

[0037] In this embodiment, three water level gauges are set at each well. The depths of the water level gauges at each position are 20 cm, 50 cm, and 80 cm respectively. The water level gauges are placed into the piezometer tube by connecting with the nut. The lower 1-meter part of the piezometer tube is buried in the active layer, and the upper 15-cm part is placed above the ground; after the water level gauges are buried, adjust the water level gauge recorder, record the burial time and date, adjust the parameters, set the monitoring period to 4 h, and obtain a more accurate water level change situation through the comparison of the measured values of the three groups of water levels.

[0038] A field monitoring method for the water level above the frozen layer in permafrost regions, as Figure 7 shown, this method includes:

[0039] S1. Select different environmental water level measurement points; respectively set the field monitoring and protection equipment for the water level at the corresponding water level measurement points;

[0040] S2. Obtain the surrounding environment images of the water level measurement points;

[0041] S3. Preprocess the surrounding environment images; the preprocessing of the images includes: normalizing the images and data augmentation;

[0042] In this embodiment, the processing of the images includes: initializing two empty lists features and labels, which are used to store the extracted image features and the corresponding labels respectively; traversing the image path list and performing the following operations on each image: reading the image, resizing the image to a size of (224, 224); calling the extract_features function to extract the image features and adding the feature vectors to the features list; calling the get_location_label function to obtain the location label according to the image path and adding it to the labels list; converting the features and labels lists into numpy arrays.

[0043] The processing of the images in the numpy array includes: calculating the color histogram of the images, where the parameters [0, 1, 2] indicate calculating the histograms of the three color channels (BGR), [8, 8, 8] indicates dividing each color channel into 8 intervals, and [0, 256, 0, 256, 0, 256] indicates that the value range of each channel is from 0 to 255. Normalize the color histogram, and then use flatten to flatten it into a one-dimensional array. Convert the image to a grayscale image, construct a gray-level co-occurrence matrix based on the grayscale image; calculate the edge histogram of the image; splice the color histogram, the gray-level co-occurrence matrix, and the edge histogram to form a one-dimensional feature vector.

[0044] Constructing the gray-level co-occurrence matrix includes defining a property list props, including contrast, dissimilarity, homogeneity, energy, and correlation; traversing the property list, using the graycoprops function to extract the value of each property and calculating the average value, and adding the average value to the features list; converting the features list into a numpy array.

[0045] S4. Divide the data features for model training and location result judgment, taking the image data features at different positions in the east direction as an example;

[0046]

[0047] S5. In order for the model to better obtain the data features, each environmental direction and its different distances should include sufficient picture information for the model to learn;

[0048] S6. The model uses a graph neural network model based on deep learning to improve the accuracy of sample feature extraction and the connection between nodes, enabling the model to still provide effective learning and prediction with few samples.

[0049] The training and evaluation of the model include: creating a CSV file containing all image paths and coordinates;

[0050] Loading the VGG16 model and pre-training it on a large-scale dataset (ImageNet) to improve the performance of model application and reduce training time;

[0051] The main structure of the VGG16 model is as follows:

[0052] Input(224×224×3): The input layer receives an RGB image (three channels) with a size of 224x224 pixels;

[0053] Conv3-64: The first convolutional layer is represented by 64 3x3 convolutional kernels, extracting the features of the input image, and the size of the feature map remains 224x224;

[0054] Max Pooling(2x2): The max pooling layer uses a 2x2 pooling window with a stride of 2, and the size of the output feature map is reduced to 112x112. This layer is used to reduce the size of the feature map, reduce computational complexity, and extract important features.

[0055] Conv3-128: The second convolutional layer uses 128 3x3 convolutional kernels, and the size of the output feature map is 112x112.

[0056] Max Pooling(2x2): Perform max pooling again, and the size of the output feature map is reduced to 56x56

[0057] Conv3-256: The third convolutional layer uses 256 3x3 convolutional kernels, and the size of the output feature map is 56x56;

[0058] Max Pooling(2x2): Perform max pooling, and the size of the output feature map is reduced to 28x28;

[0059] Conv3-512: The fourth convolutional layer uses 512 3x3 convolutional kernels, and the size of the output feature map is 28x28;

[0060] Max Pooling(2x2): Perform max pooling, and the size of the output feature map is reduced to 14x14;

[0061] Conv3-512: The fifth convolutional layer uses 512 3x3 convolutional kernels, and the size of the output feature map is 14x14;

[0062] Max Pooling(2x2): Perform max pooling, reducing the size of the output feature map to 7x7;

[0063] Fully Connected(4096): A fully connected layer that receives the flattened input and outputs 4096 nodes.

[0064] Output(Softmax)(Output layer): The last layer uses the Softmax activation function to convert the output of the fully connected layer into the image position.

[0065] Set the model to evaluation mode to ensure the consistency and stability of the model during inference or evaluation;

[0066] Use transforms for image preprocessing: Resize the input image to the specified size (224, 224) pixels; Convert the image to a PyTorch tensor and normalize the pixel values to the range [0, 1]; Normalize the image according to the mean and standard deviation of the ImageNet dataset; Extract the features of the image. In the VGG16 model, the image features are in the form of a four-dimensional tensor (batch_size, num_features, height, width); After conversion, the image features are represented as a two-dimensional NumPy array of [num_images, feature_dim], where num_images represents the number of images and feature_dim represents the dimension of each image; According to the feature extraction of the VGG16 model, the dimension of each feature vector can be 256, 512, or 1024. The higher the dimension, the richer the extracted features;

[0067] The NumPy array can be represented as: [[f11, f12,..., f1feature_dim], [f21, f22,..., f2feature_dim],..., [fn1, fn2,..., fnfeature_dim]]

[0068] Among them, [f11, f12,..., f1feature_dim] is the feature of the first image, [f21, f22,..., f2feature_dim] is the feature of the second image, and [fn1, fn2,..., fnfeature_dim] is the feature of the nth image.

[0069] Extract the position information of the images, and finally convert the extraction result into a two-dimensional array of [num_images, 2]. This array indicates that there are num_images images, and 2 means that each image is represented by two-dimensional coordinates; then this array can be expressed as: [[x1, y1], [x2, y2],... [xn, yn]], where [xn, yn] is the coordinate of the nth image. The images are regarded as nodes, the image features are node attributes, and the spatial features between nodes are connected by edges. The closer the edge distance is, the more similar the features between nodes are; build a graph structure with nodes and edges; use the k-nearest neighbor algorithm to construct other nodes (images) that are most similar to the node (image) features;

[0070] The algorithm is explained as follows:

[0071] If the k value is set to 5, it means that each node will be connected to 5 other nodes with the most similar features when constructing the graph structure;

[0072] To execute this algorithm, first search all the images. NearestNeighbors is a class in the sklearn.neighbors module, which is used to perform k-nearest neighbor search;

[0073] n_neighbors = k + 1: Set the number of neighbors to be searched as k + 1. Here, the search does not include the node itself, so one more is needed

[0074] algorithm = 'auto': The appropriate nearest neighbor search algorithm such as brute force search, KD tree or ball tree can be searched automatically through the algorithm. The specific choice depends on the characteristics of the data;

[0075] fit(features): Pass the extracted feature data (features) to the model for training. The model will learn the distance relationship between samples. The closer the distance is, the more similar the features are;

[0076] Return the sample information obtained by the model in the form of an array:

[0077] Distances: This command returns an array containing the distances from each sample to its k nearest neighbors;

[0078] Indices: This command returns an array containing the indices of the k nearest neighbors of each sample, which is used to provide the index information of the k nearest neighbors of each sample;

[0079] Suppose there are 3 samples, and the shape of the feature matrix is (3, feature_dim). If k is set to 5, the shapes of distances and indices will be as follows:

[0080] distances = [[0.0, 0.5, 0.7, 1.2, 1.5], [0.0, 0.6, 0.8, 1.0, 1.3], [0.0, 0.4, 0.9, 1.1, 1.4]], where [0.0, 0.5, 0.7, 1.2, 1.5] are the distances from the first sample to its neighbors, [0.0, 0.6, 0.8, 1.0, 1.3] are the distances from the second sample to its neighbors, and [0.0, 0.4, 0.9, 1.1, 1.4] are the distances from the third sample to its neighbors.

[0081] indices = [[0, 1, 2, 3, 4], [1, 0, 2, 3, 4], [2, 0, 1, 3, 4]]; where [0, 1, 2, 3, 4] are the neighbor indices of the first sample, [1, 0, 2, 3, 4] are the neighbor indices of the second sample, and [2, 0, 1, 3, 4] are the neighbor indices of the third sample.

[0082] Get all images, create the indices of the samples, pass through the indices of each sample and its k nearest neighbors, and convert the array into a tensor to meet the requirements of PyTorch Geometric for edge indices and create a graph structure; import the PyTorch and PyTorch Geometric libraries, define the graph neural network model, extract the features of different convolutional layers and output the picture position information in the fully connected layer; set the graph convolutional layers to reduce the input feature dimension from 512 to 256, and then to 128 (self.conv1 and self.conv2) respectively, and use the fully connected layer to map the output of the convolutional layer to the final output space, where two values (such as (x, y) coordinates) are output.

[0083] The training, prediction, and evaluation of the model include: initialize the GNNModel model, use the Adam optimizer, and a reasonable learning rate setting helps to quickly find the optimal solution of the loss function and improve the model's performance. The initial learning rate is set to 0.01; select the loss function to evaluate the gap between the model output and the target (MSELoss); set the model to training mode (model.train), clear the gradients before each training step (optimizer.zero_grad), perform forward propagation; obtain the model output (output), calculate the model loss (loss = criterion), perform backward propagation, calculate the model gradients (loss.backward), update the model parameters (optimizer.step), set the model to evaluation mode (model.eva), then disable gradient calculation (with torch.no_grad), perform forward propagation to obtain the prediction results (predicted_positions), and output the predicted positions (print);

[0084] Input a new image according to the new image path, convert the image features into tensors, and then use a graph neural network for prediction;

[0085] Obtain a new image through the passed path (new_image_pathhuoqu), convert the new image into a tensor for feature extraction (new_features = extract_features), add a batch dimension to treat the feature tensor as a batch (new_features_tensor), set the graph neural network model to evaluation mode (model.eval) to ensure consistent output during inference, disable gradient calculation (with torch.no_grad) to reduce memory usage and improve speed; use the graph neural network model to perform forward propagation on the new features to obtain the predicted position (predicted_position), convert the prediction result into a NumPy array and print it to view the predicted position (print).

[0086] S7. Save the trained model;

[0087] S8. Input the water level change information and the preprocessed image into the trained monitoring model to obtain the monitoring result.

[0088] In this embodiment, the input examples of real-time image localization results include: loading a trained offline model; obtaining an image in real time, resizing the image to (224, 224); inputting the resized image into an ideal model for feature extraction; identifying the position in the image based on the extracted features. Inputting the resized image into the ideal model for feature extraction includes: first initializing an empty list features to store features; calculating color histogram features; converting the image to grayscale: using cv2.cvtColor to convert the image from the BGR color space to grayscale; calculating edge features; converting the features list to a numpy array. Actually

[0089] The specific process of calculating color histogram features includes: using cv2.calcHist to calculate the color histogram of the image, where the parameters [0, 1, 2] indicate calculating the histograms of the three color channels (BGR), [8, 8, 8] indicates dividing each color channel into 8 intervals, and [0, 256, 0, 256, 0, 256] indicates that the value range of each channel is 0 to 255; using cv2.normalize to normalize the color histogram, then using flatten to flatten it into a one-dimensional array, and adding it to the features list.

[0090] The process of calculating edge features includes: performing edge detection using cv2.Canny, where the parameters 100 and 200 are the thresholds for Canny edge detection and can be adjusted according to actual situations; calculating the histogram of the edge image using np.histogram, and adding the result of the histogram (here, taking the first element, i.e., the frequency of the histogram) to the features list.

[0091] In this embodiment, obtaining water level change information through the field monitoring and protection device for water level includes: calculating the average water level of the water level above the frozen layer according to the positions of each water level gauge; calculating the change in the water level above the frozen layer over a certain period of time, which can be any period from the start of recording to the last recording, and calculating the change rate of the water level above the frozen layer. If the water level gauge measures upward below the sea level water level, the input water level is positive, and if the water level gauge measures downward above the sea level water level, the input water level is negative; for more accurate measurement, before the experiment, the water level gauge is used to compensate for the error of the nearby air pressure.

[0092] The formula for calculating the average water level of the water level above the frozen layer is:

[0093]

[0094] where h1 is the position of the first water level gauge, h2 is the position of the second water level gauge, and h3 is the position of the third water level gauge.

[0095] The formula for the change in the water level above the frozen layer is:

[0096]

[0097] where H t1 is the average water level at the moment t1, and H t2 is the average water level at the moment t2.

[0098] The formula for calculating the change rate of the water level above the frozen layer is:

[0099]

[0100] where Δh is the height change amount and Δt is the change time.

[0101] In this embodiment, after the installation of the water level gauges is completed, a drone is used to collect information on the nearby environment. The information content includes the four directions of east, west, south, and north of the water level point, as well as different distances in different directions, as in the embodiment of the present invention Figure 6, the collected information is marked with direction and distance. The image information should be as detailed as possible as the basis for the judgment result of the positioning system. More samples usually can improve the generalization ability of the model, especially in complex environments, covering different environments, lighting conditions, weather conditions, shooting angles and object postures as much as possible. The drone can shoot videos from different heights and angles, which provides more perspectives and scene changes for the training set and helps the model learn more comprehensive features; the drone can cover a large area, especially in the wild or inaccessible places, which means that more diverse environmental data can be collected, including different terrains, plants and weather conditions; the videos collected from the drone can be data-augmented, such as by cropping, rotating, adjusting brightness, etc., to further increase the data diversity; the drone can usually shoot high-resolution videos, which can provide more details and information and help improve the accuracy of the model.

[0102] Organize the collected data, divide the image data features according to the direction and distance of the water level well, remove the missing values, outliers and noises of the data, and perform standardization processing and data augmentation (such as rotation, flipping, cropping, brightness adjustment, etc.) on the data to improve the robustness of the model; combining the advantages of deep learning and graph neural network, the present invention combines the advantages of deep learning and graph neural network, improves the accuracy of sample feature extraction and the connection between nodes, and enables the model to still provide effective learning and prediction under few samples. Combining the spatial position coordinates and geographical location coordinates of the geographical image feature target points makes the positioning more accurate and practical.

[0103] According to the azimuth and different distances of the image information, all the information is divided into a training set, a validation set and a test set through the corresponding language program and coding. The usual division ratio is 80% for the training set, 10% for the validation set, and 10% for the test set. Use the training set to train the model, ensure effective hyperparameter tuning of the model to obtain the best performance; use the validation set and the test set to evaluate the performance of the model. The evaluation basis indicators refer to the mean square error of the results and the direction and distance accuracy. Regularly evaluate the model performance. With the introduction of new data, the model needs to be retrained to maintain its accuracy; import the newly shot data on site, make the trained model predict the new data, and analyze the prediction results according to the requirements, such as exporting and visualizing the prediction results; the training, validation, testing, optimization and result output of the above model are all completed through the Python language, and the implementation process is as Figure 7 shown.

[0104] A field monitoring system for the water level above the frozen layer in permafrost regions, the system includes: a surrounding environment image acquisition module, a field monitoring protection device for the water level, an image preprocessing module, a monitoring module and an output module;

[0105] The surrounding environment image acquisition module is used to acquire the surrounding environment images of the water level measurement point;

[0106] The image preprocessing module is used to perform normalization processing and data enhancement on the surrounding environment images;

[0107] The water level field monitoring and protection device is used to acquire the water level change information of the water level measurement point;

[0108] The monitoring module is used to process the enhanced images and the water level change information to obtain the monitoring results;

[0109] The output module is used to output the monitoring results.

[0110] The system implementation manner of the present invention is the same as the implementation manner of the method.

[0111] The above-mentioned embodiments further elaborate on the purpose, technical solution and advantages of the present invention. It should be understood that the above-mentioned embodiments are only the preferred implementation manners of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made to the present invention within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A field monitoring and protection device for the water level above the frozen layer in permafrost regions, characterized in that, Comprising: A water level pipe, nuts, steel wire ropes, a temperature control system, a heating device group, and multiple water level gauges; the nuts are arranged at the top of the water level pipe; the multiple water level gauges are evenly bound to the steel wire ropes, and one end of the steel wire rope is arranged at the center point of the nut, so that the steel wire rope and the water level gauges are suspended inside the water level pipe; the heating device group is arranged in the inner wall of the water level pipe and is connected to the temperature control system through conductive wires; the temperature control system is used to control the temperature of the heating device.

2. The field monitoring and protection device for the water table above the frozen layer in permafrost regions according to claim 1, characterized in that, The pipe wall of the water level pipe is provided with multiple round holes, and each round hole is evenly distributed on the pipe wall of the water level pipe for allowing the water to be monitored to flow into the water level pipe.

3. The field monitoring and protection device for the upper water level of the frozen layer in the permafrost region according to claim 2, characterized in that, The length of the water level pipe is 115 cm and the diameter is 45 cm.

4. The field monitoring and protection device for the upper water level of the frozen layer in the permafrost region according to claim 2, characterized in that, The outer pipe wall of the water level pipe is wrapped with a plastic filter screen to prevent fine sand and clay from entering the inside of the water level pipe along with the water during the melting process.

5. The field monitoring and protection device for the upper water level of the frozen layer in the permafrost region according to claim 2, characterized in that, The diameter of the round hole is 10 mm, and the distance between two adjacent round holes is 50 mm.

6. The field monitoring and protection equipment for the water table above the frozen layer in permafrost regions according to claim 1, characterized in that, The heating device group consists of multiple heating devices, and each heating device is connected in series through a wire, and the number of heating devices is the same as the number of water level gauges. The position of each heating device arranged on the inner wall of the water level pipe is at the same level as the position of the corresponding water level gauge inside the water level pipe.

7. The field monitoring and protection device for the upper water level of the frozen layer in the permafrost region according to claim 1, characterized in that, It further includes three metal tripods, and the metal tripods are evenly arranged on the outer wall of the water level pipe to stabilize the water level pipe.

8. A field monitoring method for the water table above the frozen layer in permafrost regions, which uses a field monitoring and protection device for the water table above the frozen layer in permafrost regions as described in any one of claims 1 to 7, characterized in that, Comprising: S1. Select different environmental water level measurement points; respectively set the water level field monitoring protection equipment at the corresponding water level measurement points; S2. Obtain the surrounding environment images of the water level measurement points; S3. Preprocess the surrounding environment images; the preprocessing of the images includes: performing normalization processing and data augmentation on the images; S4. Obtain the water level change information through the water level field monitoring protection equipment; S5. Input the water level change information and the preprocessed images into the trained monitoring model to obtain the monitoring result, where the monitoring model uses a graph neural network model based on deep learning.

9. A field monitoring system for the water table above the frozen layer in permafrost regions, which is used to implement the field monitoring method for the water table above the frozen layer in permafrost regions described in claim 8, characterized in that, The system includes: a surrounding environment image acquisition module, a water level field monitoring protection equipment, an image preprocessing module, a monitoring module, and an output module; The surrounding environment image acquisition module is used to acquire the surrounding environment images of the water level measurement points; The image preprocessing module is used to perform normalization processing and data augmentation on the surrounding environment images; The water level field monitoring protection equipment is used to acquire the water level change information of the water level measurement points; The monitoring module is used to process the enhanced images and the water level change information to obtain the monitoring result; The output module is used to output the monitoring result.

Images