System and method for monitoring hot melt collapse risk in permafrost region

Abstract

The invention relates to the technical field of hot melt collapse risk monitoring, and discloses a permafrost region hot melt collapse risk monitoring system and method, and the system comprises an on-site data collection module, a collected data transmission module, a transmission data storage module, an on-site data analysis and early warning module, a remote on-site monitoring module, and an autonomous power generation and supply module. According to the permafrost region hot melt collapse risk monitoring system and method, measures are taken in advance, damage to a permafrost ecological system is avoided or relieved, animal and plant habitats living depending on the permafrost environment are protected, traffic arteries such as railways and highways are monitored, the thermal condition and stability of permafrost under a roadbed can be mastered in time, hidden dangers are treated in advance, and the risk of permafrost region hot melt collapse is prevented. And for energy facilities such as oil and gas pipelines and the like, through monitoring, operation and maintenance personnel can find potential risks in advance, measures such as heat insulation and water drainage are taken, and the pipelines are prevented from being broken and leaked due to hot melt collapse.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal thaw landslide risk monitoring, and in particular to a thermal thaw landslide risk monitoring system and method in permafrost areas. Background Art

[0002] Permafrost zone refers to the soil layer area where the temperature below the surface is below 0 degrees Celsius for two consecutive years or more. Permafrost zone is mainly distributed in areas with cold climates, such as the Arctic region, high mountains and certain high-latitude areas. The formation of permafrost mainly depends on low temperature conditions. In areas with cold climates, the temperature below the surface is continuously below 0 degrees Celsius, causing the soil and rocks to freeze.

[0003] Permafrost degradation caused by climate warming and soil thermal changes often leads to active layer degradation and the development of thermokarst landforms. The melting of ice-rich permafrost is called thermokarst, which can profoundly affect local landscapes, ecosystems and global climate. Common thermokarst landforms include melt lakes, melt ponds, ice wedges, active layer separation and thermokarst landslides. In cold seasons, the permafrost layer on the surface or slopes freezes and provides a certain degree of stability. However, as the temperature rises, the permafrost begins to melt and the ice in it gradually turns into water. When the ice melts, it loses its original supporting function, resulting in a decrease in the strength of the soil or rock. In the continuous melting process, the soil or rock will The phenomenon of loosening and loss of shear strength, followed by sliding, collapse and subsidence, is called thermal thaw landslide. Disasters caused by permafrost melting due to temperature changes are collectively referred to as thermal thaw disasters, which have a significant impact on engineering construction, ecological environment and human safety. In alpine areas, due to extreme climatic conditions and complex geological structures, thermal thaw landslide disasters occur frequently, which not only cause slope collapse and retreat, making the construction of engineering structures in permafrost areas dangerous, but also change the hydrothermal process of the active layer, causing permafrost to lose its recovery ability and leading to the deterioration of the regional ecosystem. Existing technologies cannot carry out accurate monitoring and early warning, resulting in the inability to evacuate people in time when disasters occur.

[0004] Based on this, the present invention provides a system and method for monitoring the risk of thermal thaw and landslide in permafrost areas. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a system and method for monitoring the risk of thermal thaw and landslide in permafrost areas, which has the capabilities of real-time monitoring of temperature, displacement of the ground and soil, humidity and weather, and analysis and early warning of the monitoring data, thereby achieving the advantages of thermal thaw and landslide risk monitoring in permafrost areas and solving the problems raised in the background technology.

[0006] The present invention provides the following technical solution: a permafrost region thermal thaw landslide risk monitoring system, comprising a field data collection module, a data collection and transmission module, a data transmission and storage module, a field data analysis and early warning module, a remote field monitoring module, and an autonomous power generation and supply module:

[0007] Field data collection module: (1) Field temperature monitoring unit: permafrost areas are divided and numbered, and multiple holes with a diameter of 1m are drilled in each individual permafrost area. Temperature sensors are arranged inside the holes at different depths. The temperature sensors use thermocouples or thermistors to obtain temperature data of the permafrost layer and the active layer, and then the holes are filled. The depths of the multiple holes are 10cm, 20cm, 30cm, 40cm and 50cm respectively; (2) Field displacement monitoring unit: Synthetic aperture radar interferometry (SBAS-InSAR) technology is used, and Sentinel-1 satellite images are selected as SBAS-InSAR R deformation detection data, synthetic aperture radar interferometry technology is an InSAR time series method based on multiple main images. It only uses interference pairs with shorter time and space baselines to extract surface deformation information. First, the time and space baselines are calculated for multi-view SAR images covering different time periods in a certain area, and the appropriate time and space baseline threshold is selected to select interference pairs. Then, the selected interference pairs are subjected to differential interference processing and phase unwrapping. Finally, according to the formation of subsets of freely combined interference patterns, the least squares method or SVD method is used to estimate the deformation parameters of the phase equations composed of all interference patterns. In actual processing, the time and space filtering method is used to remove atmospheric The delayed image is used to separate the nonlinear deformation. The sum of the estimated low-frequency deformation and the nonlinear deformation is the deformation information of the entire study area. The displacement of the ground and the soil is monitored to determine whether there is a sliding trend. The slope of the landslide is divided into five levels: slightly inclined slope (0°-1°), gently inclined slope (1°-3°), slope (3°-8°), steep slope (8°-10°), and abrupt slope (10°-78.61°). The method for estimating the deformation time series is the singular value decomposition method. Assuming that N+1 SAR images covering the same study area are used, the time sequence is as follows: t = [t0, t1, t2, ..., t M-1 ] T #(2.2)

[0008] N single interferometer pairs are generated by M pairs of multi-view interferometer pairs obtained by the maximum spatiotemporal baseline threshold, and M satisfies:

[0009] The generated multi-parallax differential interference pattern is then filtered and phase unfolded, and the coherence is used to select appropriate reference points in the unfolded differential interference pattern. The generalized inverse matrix of the coefficient matrix can be obtained using the SVD method, thereby obtaining the minimum norm solution of the data vector. Finally, the speed of each time period gives the deformation of each time period; (3) Field humidity monitoring unit: New multi-point drilling is carried out in the area within 1m of the same location mentioned above. The different depths of the multi-point drilling are 10cm, 20cm, 30cm, 40cm and 50cm respectively, and humidity sensors are arranged inside the holes at different depths to obtain humidity data of the permafrost layer and active layer, and then the pit filling operation is carried out; (4) Field meteorological monitoring unit: A small meteorological station is installed on the ground in the area within 1m of the same location mentioned above to monitor the external temperature and precipitation data of the permafrost area in real time; (5) Vegetation cover (NDVI) data acquisition unit: Gaofen-2 (GF-2) images are used to calculate NDVI. GF-2 images mainly use the 2nd, 3rd, 4th, and 8th bands. The 2nd, 3rd, and 4th bands synthesize true color images, which are close to the real colors of the ground objects. The images are dull and the tones are dark. The calculation of NDVI requires the use of the 4th and 8th bands for combined calculation. NDVI is calculated based on the near-infrared band and red band of the remote sensing image. The NDVI index of the study area is calculated using the ArcGIS platform. The vegetation cover is classified according to the index as follows: no vegetation cover (NDVI < 0), low vegetation cover (0 < NDVI < 0.05), relatively low vegetation cover (0.05 < NDVI < 0.1), medium vegetation cover (0.1 < NDVI less than 0.2) and high vegetation cover (NDVI greater than 0.2). The calculation formula is: Among them, NIR represents the near infrared band, and R represents the infrared band;

[0010] Data collection and transmission module: using 4G / 5G wireless communication technology, remotely transmit the real-time monitoring data of the field temperature monitoring unit, field displacement monitoring unit, field humidity monitoring unit and field meteorological monitoring unit;

[0011] Transmission data storage module: uses the database to classify and store the data transmitted by the data transmission modules collected in different numbered permafrost areas;

[0012] Field data analysis and early warning module: It uses a large amount of normal data provided by the network information database as training samples and is based on a deep learning model U-Net for image segmentation tasks. The main feature of the U-Net model is that it combines a fully convolutional neural network (FCN) with an encoder-decoder structure to achieve semantic segmentation. Its architecture consists of three main parts: encoder, decoder, and skip connection. The U-Net training process usually uses the cross-entropy loss function to compare the difference between the predicted segmentation map and the actual segmentation map. The loss function expression is as follows:

[0013]

[0014] Where N represents the total number of image pixels, Represents the true segmentation label (0 or 1). By minimizing the loss function, the parameters of the U-Net model can be optimized during the training process. Appropriate data augmentation techniques, such as image rotation, are used to enable the model to better perform image segmentation tasks. The U-Net architecture is adopted, and the pre-configured ResNet101, ResNet152 and VGG16 neural networks are used as backbone models for transfer learning. The number of sample batches used in each iteration is limited to 8, and an automatic training stop mechanism is set in the model. Once the model accuracy no longer improves, the training can be stopped. The training results show that the ResNet101 model is the best choice for automatic identification of thermal melt collapse backbone models. When the U-Net architecture uses the ResNet101 model to monitor new data stored in the database, if the data point deviates from the normal distribution range, it is determined to be an abnormality with risk, realizing real-time monitoring and early warning of risks. When the monitoring data analysis results show that there is a high risk, the warning information is pushed to relevant departments and personnel via wireless transmission.

[0015] Remote field monitoring module: Cameras are installed in permafrost areas of different numbered areas to achieve real-time video monitoring of the on-site conditions in the permafrost areas. This assists the above-mentioned field data analysis and early warning module in pushing early warning information, enabling relevant departments and personnel to determine the occurrence of thermal thaw landslides;

[0016] Autonomous power supply module: Taking advantage of the high altitude of the long-term construction area, solar power generation facilities are installed. Solar power generation facilities are used to convert solar energy into electrical energy, and the electrical energy is stored in batteries. The batteries are connected to the above-mentioned monitoring equipment through cables to ensure that they are in a powered state, providing power support for the monitoring equipment of the remote field monitoring module.

[0017] Preferably, the permafrost zone is divided into zones and numbered in the form of English letters + numbers, with English letters using AZ as the first letter and numbers extending in sequence, for example: A1, A2, A3, A4, etc.

[0018] Preferably, the permafrost area is divided into zones and numbered, and the area of ​​each individual zone is measured using Beidou high-precision positioning equipment, and the division range of each individual zone is an interval of (90m-100m)*(90m-100m).

[0019] Preferably, the network information database uses web of science to provide data to ensure the authority and comprehensiveness of the data.

[0020] A method for monitoring the risk of thermal thaw landslides in permafrost areas comprises the following steps:

[0021] Step 1: The permafrost area is divided into zones and numbers using English and numerical numbers. The range of each zone is (90m-100m)*(90m-100m). Ten holes are randomly drilled in each zone with a diameter of 1m. The ten holes are divided into 10cm*2, 20cm*2, 30cm*2, 40cm*2 and 50cm*2. Temperature sensors and humidity sensors are placed in the two groups of holes with corresponding depths, and then the holes are filled.

[0022] Step 2: Use Synthetic Aperture Radar Interferometry (SBAS_InSAR) technology to monitor the displacement of the ground and soil to determine whether there is a sliding trend;

[0023] Step 3: Install a small weather station on the ground within 1m of the same location as in Step 1 to monitor the external temperature and precipitation data of the permafrost area in real time. Use the Gaofen-2 (GF-2) image to calculate the vegetation cover (NDVI) data.

[0024] Step 4: Install solar power generation facilities, use solar power generation facilities to convert solar energy into electrical energy, and store the electrical energy through batteries. The batteries are connected to the monitoring equipment through cables to ensure that they are powered on, providing power support for the monitoring equipment of the remote field monitoring module.

[0025] Step 5: Using 4G / 5G wireless communication technology, remotely transmit the real-time monitoring data of steps 1, 2, and 3;

[0026] Step 6: Use the database to classify and store the data transferred in step 4;

[0027] Step 7: Using the large amount of normal data provided by the network information database as training samples, the U-Net architecture uses the ResNet101 model to monitor new data stored in the database. If a data point deviates from the normal distribution range, it is determined to be abnormal and there is a risk, realizing real-time monitoring and early warning of risks. When the monitoring data analysis results show a high risk, early warning information is issued to relevant departments and personnel via text message push via wireless transmission.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The permafrost area thermal melt landslide risk monitoring system and method ensures the accuracy of monitoring data through temperature, ground and soil internal displacement, vegetation cover (NDVI) data obtained by calculating the Gaofen-2 (GF-2) image, humidity and meteorological real-time monitoring and real-time data transmission. A large amount of normal data provided by the network information database is used as training samples to learn the distribution characteristics of normal data. When monitoring new data stored in the database, if the data point deviates from the normal distribution range, it is determined to be abnormal and there is a risk, realizing real-time monitoring and early warning of the risk. When the monitoring data analysis results show that there is a high risk, SMS push is used to issue early warning information to relevant departments and personnel through wireless transmission, and cameras are installed to realize real-time video monitoring of the on-site situation in the permafrost area. When the field data analysis and early warning module pushes the early warning information, the relevant departments and personnel can judge the occurrence of thermal melt landslide. At the same time, the excess electricity is stored by solar power generation and used at night. The system is powered on rainy days to ensure the continuous operation of the monitoring equipment. This solution ensures the accuracy of data for thermal thaw landslide risk monitoring in permafrost areas through multi-faceted integration and real-time data monitoring and transmission. It can also conduct real-time analysis and early warning, making it easier for relevant departments to make preventive preparations, organize personnel evacuation and property transfer, and reduce casualties and property losses caused by thermal thaw landslides. Effective thermal thaw landslide risk monitoring in permafrost areas can timely detect thermal thaw landslide risks, take measures in advance, avoid or reduce damage to the permafrost ecosystem, and protect the habitats of animals and plants that rely on the permafrost environment for survival. In addition, monitoring of transportation arteries such as railways and highways can timely grasp the thermal conditions and stability of the permafrost under the roadbed, deal with hidden dangers in advance, prevent thermal thaw landslides from causing roadbed deformation and collapse, and ensure safe and smooth transportation. For energy facilities such as oil and gas pipelines, monitoring can enable operation and maintenance personnel to detect potential risks in advance and take measures such as insulation and drainage to avoid pipeline rupture and leakage due to thermal thaw landslides. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Schematic diagram of the system flow of the present invention;

[0031] Figure 2 This is a schematic diagram of the field data collection module of the present invention;

[0032] Figure 3 Schematic diagram of the U-Net network structure of the present invention;

[0033] Figure 4 This is a schematic diagram of the thermal melt collapse training effect indicators of the present invention. DETAILED DESCRIPTION

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0035] See also Figure 1-4 A permafrost region thermal thaw landslide risk monitoring system includes a field data collection module, a data collection and transmission module, a data transmission and storage module, a field data analysis and early warning module, a remote field monitoring module, and an autonomous power generation and supply module.

[0036] Field data collection module: (1) Field temperature monitoring unit: permafrost areas are divided and numbered, and multiple holes with a diameter of 1m are drilled in each individual permafrost area. Temperature sensors are arranged inside the holes at different depths. The temperature sensors use thermocouples or thermistors to obtain temperature data of the permafrost layer and the active layer, and then the holes are filled. The depths of the multiple holes are 10cm, 20cm, 30cm, 40cm and 50cm respectively; (2) Field displacement monitoring unit: Synthetic aperture radar interferometry (SBAS-InSAR) technology is used, and Sentinel-1 satellite images are selected as SBAS-InSAR R deformation detection data, synthetic aperture radar interferometry technology is an InSAR time series method based on multiple main images. It only uses interference pairs with shorter time and space baselines to extract surface deformation information. First, the time and space baselines are calculated for multi-view SAR images covering different time periods in a certain area, and the appropriate time and space baseline threshold is selected to select interference pairs. Then, the selected interference pairs are subjected to differential interference processing and phase unwrapping. Finally, according to the formation of subsets of freely combined interference patterns, the least squares method or SVD method is used to estimate the deformation parameters of the phase equations composed of all interference patterns. In actual processing, the time and space filtering method is used to remove atmospheric The delayed image is used to separate the nonlinear deformation. The sum of the estimated low-frequency deformation and the nonlinear deformation is the deformation information of the entire study area. The displacement of the ground and the soil is monitored to determine whether there is a sliding trend. The slope of the landslide is divided into five levels: slightly inclined slope (0°-1°), gently inclined slope (1°-3°), slope (3°-8°), steep slope (8°-10°), and abrupt slope (10°-78.61°). The method for estimating the deformation time series is the singular value decomposition method. Assuming that N+1 SAR images covering the same study area are used, the time sequence is as follows: t = [t0, t1, t2, ..., t M-1 ] T #(2.2)

[0037] N single interferometer pairs are generated by M pairs of multi-view interferometer pairs obtained by the maximum spatiotemporal baseline threshold, and M satisfies:

[0038] The generated multi-parallax differential interference pattern is then filtered and phase unfolded, and the coherence is used to select appropriate reference points in the unfolded differential interference pattern. The generalized inverse matrix of the coefficient matrix can be obtained using the SVD method, thereby obtaining the minimum norm solution of the data vector. Finally, the speed of each time period gives the deformation of each time period; (3) Field humidity monitoring unit: New multi-point drilling is carried out in the area within 1m of the same location mentioned above. The different depths of the multi-point drilling are 10cm, 20cm, 30cm, 40cm and 50cm respectively, and humidity sensors are arranged inside the holes at different depths to obtain humidity data of the permafrost layer and active layer, and then the pit filling operation is carried out; (4) Field meteorological monitoring unit: A small meteorological station is installed on the ground in the area within 1m of the same location mentioned above to monitor the external temperature and precipitation data of the permafrost area in real time; (5) Vegetation cover (NDVI) data acquisition unit: Gaofen-2 (GF-2) images are used to calculate NDVI. GF-2 images mainly use the 2nd, 3rd, 4th, and 8th bands. The 2nd, 3rd, and 4th bands synthesize true color images, which are close to the real colors of the ground objects. The images are dull and the tones are dark. The calculation of NDVI requires the use of the 4th and 8th bands for combined calculation. NDVI is calculated based on the near-infrared band and red band of the remote sensing image. The NDVI index of the study area is calculated using the ArcGIS platform. The vegetation cover is classified according to the index as follows: no vegetation cover (NDVI < 0), low vegetation cover (0 < NDVI < 0.05), relatively low vegetation cover (0.05 < NDVI < 0.1), medium vegetation cover (0.1 < NDVI less than 0.2), and high vegetation cover (NDVI greater than 0.2). The calculation formula is: Among them, NIR represents the near infrared band, and R represents the infrared band;

[0039] Data collection and transmission module: using 4G / 5G wireless communication technology, remotely transmit the real-time monitoring data of the field temperature monitoring unit, field displacement monitoring unit, field humidity monitoring unit and field meteorological monitoring unit;

[0040] Transmission data storage module: using the database to classify and store the data transmitted by the data transmission modules collected in different numbered permafrost areas;

[0041] Field data analysis and early warning module: It uses a large amount of normal data provided by the network information database as training samples and is based on a deep learning model U-Net for image segmentation tasks. The main feature of the U-Net model is that it combines a fully convolutional neural network (FCN) with an encoder-decoder structure to achieve semantic segmentation. Its architecture consists of three main parts: encoder, decoder, and skip connection. The U-Net training process usually uses the cross-entropy loss function to compare the difference between the predicted segmentation map and the actual segmentation map. The loss function expression is as follows:

[0042]

[0043] Where N represents the total number of image pixels, Represents the true segmentation label (0 or 1). By minimizing the loss function, the parameters of the U-Net model can be optimized during the training process. Appropriate data augmentation techniques, such as image rotation, are used to enable the model to better perform image segmentation tasks. The U-Net architecture is adopted, and the pre-configured ResNet101, ResNet152 and VGG16 neural networks are used as backbone models for transfer learning. The number of sample batches used in each iteration is limited to 8, and an automatic training stop mechanism is set in the model. Once the model accuracy no longer improves, the training can be stopped. The training results show that the ResNet101 model is the best choice for automatic identification of thermal melt collapse backbone models. When the U-Net architecture uses the ResNet101 model to monitor new data stored in the database, if the data point deviates from the normal distribution range, it is determined to be an abnormality with risk, realizing real-time monitoring and early warning of risks. When the monitoring data analysis results show that there is a high risk, the warning information is pushed to relevant departments and personnel via wireless transmission.

[0044] Remote field monitoring module: Cameras are installed in permafrost areas of different numbered areas to achieve real-time video monitoring of the on-site conditions in the permafrost areas. This assists the above-mentioned field data analysis and early warning module in pushing early warning information, enabling relevant departments and personnel to determine the occurrence of thermal thaw landslides;

[0045] Autonomous power generation module: Taking advantage of the high altitude of the permafrost area, solar power generation facilities are installed. The solar power generation facilities are used to convert solar energy into electricity, and the electricity is stored in batteries. The batteries are connected to the above-mentioned monitoring equipment by cables to ensure that they are powered on, providing power support for the monitoring equipment of the remote field monitoring module. Through temperature, ground and soil displacement, and the use of Gaofen-2 (GF-2) image calculation to obtain vegetation cover (NDVI) data, humidity and meteorological real-time monitoring and real-time data transmission, the accuracy of the monitoring data is guaranteed. A large amount of normal data provided by the network information database is used as training samples to learn the distribution characteristics of normal data. When monitoring new data stored in the database, if the data point deviates from the normal distribution range, it is determined to be abnormal and there is a risk, realizing real-time monitoring and early warning of the risk. When the monitoring data analysis results show a high risk, the warning information is pushed to the relevant departments and personnel through SMS push via wireless transmission. Cameras are installed to realize real-time video monitoring of the on-site situation in the permafrost area, assisting the field data analysis and early warning module in early warning. When the warning information is pushed, the relevant departments and personnel can judge the occurrence of thermal melt landslides. At the same time, the excess electricity is stored by solar power generation to power the system at night or on rainy days to ensure the continuous operation of the monitoring equipment. In this solution, the accuracy of the data for thermal melt landslide risk monitoring in permafrost areas is guaranteed through multi-faceted combination and real-time data monitoring and real-time transmission. It can also conduct real-time analysis and early warning, making it convenient for relevant departments to make preventive preparations, organize personnel evacuation, property transfer, etc., reduce casualties and property losses caused by thermal melt landslides, and effectively carry out permafrost area Thermal thaw landslide risk monitoring can timely detect thermal thaw landslide risks, take measures in advance, avoid or reduce damage to permafrost ecosystems, and protect the habitats of animals and plants that rely on permafrost environments for survival. It also monitors transportation arteries such as railways and highways to timely grasp the thermal conditions and stability of the permafrost under the roadbed, deal with hidden dangers in advance, prevent thermal thaw landslides from causing roadbed deformation and collapse, and ensure safe and smooth transportation. For energy facilities such as oil and gas pipelines, monitoring can enable operation and maintenance personnel to discover potential risks in advance and take measures such as insulation and drainage to prevent pipelines from rupture and leakage due to thermal thaw landslides.

[0046] Among them, the permafrost zone is divided into zones using English + numbers, with AZ as the first English number and the numbers extending in sequence, for example: A1, A2, A3, A4, etc.

[0047] Among them, the permafrost area is divided into zones and numbered, and the area of ​​each individual area is measured using Beidou high-precision positioning equipment. The division range of each individual area is the interval of (90m-100m)*(90m-100m).

[0048] Among them, the online information database uses web of science to provide data to ensure the authority and comprehensiveness of the data.

[0049] A method for monitoring the risk of thermal thaw landslides in permafrost areas comprises the following steps:

[0050] Step 1: The permafrost area is divided into zones and numbers using English and numerical numbers. The range of each zone is (90m-100m)*(90m-100m). Ten holes are randomly drilled in each zone with a diameter of 1m. The ten holes are divided into 10cm*2, 20cm*2, 30cm*2, 40cm*2 and 50cm*2. Temperature sensors and humidity sensors are placed in the two groups of holes with corresponding depths, and then the holes are filled.

[0051] Step 2: Use Synthetic Aperture Radar Interferometry (SBAS_InSAR) technology to monitor the displacement of the ground and soil to determine whether there is a sliding trend;

[0052] Step 3: Install a small weather station on the ground within 1m of the same location as in Step 1 to monitor the external temperature and precipitation data of the permafrost area in real time. Use the Gaofen-2 (GF-2) image to calculate the vegetation cover (NDVI) data.

[0053] Step 4: Install solar power generation facilities, use solar power generation facilities to convert solar energy into electrical energy, and store the electrical energy through batteries. The batteries are connected to the monitoring equipment through cables to ensure that they are powered on, providing power support for the monitoring equipment of the remote field monitoring module.

[0054] Step 5: Using 4G / 5G wireless communication technology, remotely transmit the real-time monitoring data of steps 1, 2, and 3;

[0055] Step 6: Use the database to classify and store the data transferred in step 4;

[0056] Step 7: Using a large amount of normal data provided by the network information database as training samples, the U-Net architecture uses the ResNet101 model to monitor new data stored in the database. If the data point deviates from the normal distribution range, it is determined to be abnormal and there is a risk, thus achieving real-time monitoring and early warning of the risk. When the monitoring data analysis results show that there is a high risk, a text message is pushed to the relevant departments and personnel via wireless transmission to issue early warning information.

[0057] See also Figure 3,Each blue box in the U-Net network structure corresponds to a multi-channel feature map, the size of which is located at the lower left edge of the box. Different arrows represent different operations. The left part is feature extraction and the right part is upsampling, that is, the encoder-decoder structure.

[0058] See also Figure 4 The precision rate is also called the call accuracy rate, which indicates the proportion of samples predicted as positive that are actually positive samples, reflecting the model's ability to distinguish negative samples. The higher the precision rate, the stronger the model's ability to distinguish negative samples. The recall rate reflects the model's ability to identify positive samples. It measures the proportion of all positive samples that are correctly identified. The higher the recall rate, the stronger the model's ability to identify positive samples. The F1 Score is the weighted average of the precision rate and the recall rate, which is used to comprehensively evaluate the robustness and overall performance of the model. It can be clearly seen that among the three models, the ResNet101 backbone model has the best training effect on thermal melt collapse and has met the expected requirements, with an accuracy of more than 80%.

[0059] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0060] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A permafrost region thermal thaw landslide risk monitoring system, characterized in that: It includes field data collection module, data collection and storage module, data transmission and storage module, field data analysis and early warning module, remote field monitoring module and autonomous power generation module: Field data collection module: (1) Field temperature monitoring unit: permafrost areas are divided and numbered, and multiple holes with a diameter of 1m are drilled in each individual permafrost area. Temperature sensors are arranged inside the holes at different depths. The temperature sensors use thermocouples or thermistors to obtain temperature data of the permafrost layer and the active layer, and then the holes are filled. The depths of the multiple holes are 10cm, 20cm, 30cm, 40cm and 50cm respectively; (2) Field displacement monitoring unit: Synthetic aperture radar interferometry (SBAS-InSAR) technology is used, and Sentinel-1 satellite images are selected as SBAS-InSAR R deformation detection data, synthetic aperture radar interferometry technology is an InSAR time series method based on multiple main images. It only uses interference pairs with shorter time and space baselines to extract surface deformation information. First, the time and space baselines are calculated for multi-view SAR images covering different time periods in a certain area, and the appropriate time and space baseline threshold is selected to select interference pairs. Then, the selected interference pairs are subjected to differential interference processing and phase unwrapping. Finally, according to the formation of subsets of freely combined interference patterns, the least squares method or SVD method is used to estimate the deformation parameters of the phase equations composed of all interference patterns. In actual processing, time and space filtering is used. The method is used to remove the atmospheric delay from the image and separate the nonlinear deformation. The sum of the estimated low-frequency deformation and this nonlinear deformation is the deformation information of the entire study area. The displacement of the ground and the soil is monitored to determine whether there is a sliding trend. The slope of the landslide is divided into five levels: slightly inclined slope (0°-1°), gently inclined slope (1°-3°), slope (3°-8°), steep slope (8°-10°), and abrupt slope (10°-78.61°). The method for estimating the deformation time series is the singular value decomposition method. Assuming that N+1 SAR images covering the same study area are used, the time order is as follows: t=[t0,t1,t2,…,t M-1 ] T #(2.2) N single interferometer pairs are generated by M pairs of multi-view interferometer pairs obtained by the maximum spatiotemporal baseline threshold, and M satisfies: The generated multi-parallax differential interference pattern is then filtered and phase unfolded, and the coherence is used to select appropriate reference points in the unfolded differential interference pattern. The generalized inverse matrix of the coefficient matrix can be obtained using the SVD method, thereby obtaining the minimum norm solution of the data vector. Finally, the speed of each time period gives the deformation of each time period; (3) Field humidity monitoring unit: New multi-point drilling is carried out in the area within 1m of the same location mentioned above. The different depths of the multi-point drilling are 10cm, 20cm, 30cm, 40cm and 50cm respectively, and humidity sensors are arranged inside the holes at different depths to obtain humidity data of the permafrost layer and active layer, and then the pit filling operation is carried out; (4) Field meteorological monitoring unit: A small meteorological station is installed on the ground in the area within 1m of the same location mentioned above to monitor the external temperature and precipitation data of the permafrost area in real time; (5) Vegetation cover (NDVI) data acquisition unit: Gaofen-2 (GF-2) images are used to calculate NDVI. GF-2 images mainly use the 2nd, 3rd, 4th, and 8th bands. The 2nd, 3rd, and 4th bands synthesize true color images, which are close to the real colors of the ground objects. The images are dull and the tones are dark. The calculation of NDVI requires the use of the 4th and 8th bands for combined calculation. NDVI is calculated based on the near-infrared band and red band of the remote sensing image. The NDVI index of the study area is calculated using the ArcGIS platform. The vegetation cover is classified according to the index as follows: no vegetation cover (NDVI < 0), low vegetation cover (0 < NDVI < 0.05), relatively low vegetation cover (0.05 < NDVI < 0.1), medium vegetation cover (0.1 < NDVI less than 0.2), and high vegetation cover (NDVI greater than 0.2). The calculation formula is: Among them, NIR represents the near infrared band, and R represents the infrared band; Data collection and transmission module: using 4G / 5G wireless communication technology, remotely transmit the real-time monitoring data of the field temperature monitoring unit, field displacement monitoring unit, field humidity monitoring unit and field meteorological monitoring unit; Transmission data storage module: uses the database to classify and store the data transmitted by the data transmission modules collected in different numbered permafrost areas; Field data analysis and early warning module: It uses a large amount of normal data provided by the network information database as training samples and is based on a deep learning model U-Net for image segmentation tasks. The main feature of the U-Net model is that it combines a fully convolutional neural network (FCN) with an encoder-decoder structure to achieve semantic segmentation. Its architecture consists of three main parts: encoder, decoder, and skip connection. The U-Net training process usually uses the cross-entropy loss function to compare the difference between the predicted segmentation map and the actual segmentation map. The loss function expression is as follows: Where N represents the total number of image pixels, Represents the true segmentation label (0 or 1). By minimizing the loss function, the parameters of the U-Net model can be optimized during the training process. Appropriate data augmentation techniques, such as image rotation, are used to enable the model to better perform image segmentation tasks. The U-Net architecture is adopted, and the pre-configured ResNet101, ResNet152 and VGG16 neural networks are used as backbone models for transfer learning. The number of sample batches used in each iteration is limited to 8, and an automatic training stop mechanism is set in the model. Once the model accuracy no longer improves, the training can be stopped. The training results show that the ResNet101 model is the best choice for automatic identification of thermal melt collapse backbone models. When the U-Net architecture uses the ResNet101 model to monitor new data stored in the database, if the data point deviates from the normal distribution range, it is determined to be an abnormality with risk, realizing real-time monitoring and early warning of risks. When the monitoring data analysis results show that there is a high risk, the warning information is pushed to relevant departments and personnel via wireless transmission. Remote field monitoring module: Cameras are installed in permafrost areas of different numbered areas to achieve real-time video monitoring of the on-site conditions in the permafrost areas. This assists the above-mentioned field data analysis and early warning module in pushing early warning information, enabling relevant departments and personnel to determine the occurrence of thermal thaw landslides; Autonomous power supply module: Taking advantage of the high altitude of the long-term construction area, solar power generation facilities are installed. Solar power generation facilities are used to convert solar energy into electrical energy, and the electrical energy is stored in batteries. The batteries are connected to the above-mentioned monitoring equipment through cables to ensure that they are in a powered state, providing power support for the monitoring equipment of the remote field monitoring module.

2. The permafrost region thermal thaw landslide risk monitoring system according to claim 1, characterized in that: The permafrost zones are numbered in English + numbers, with AZ being the first letter and the numbers extending in sequence, for example: A1, A2, A3, A4, etc.

3. The permafrost region thermal thaw landslide risk monitoring system according to claim 1, characterized in that: The permafrost area is divided into zones and numbered, and the area of ​​each individual zone is measured using Beidou high-precision positioning equipment. The division range of each individual zone is an interval of (90m-100m)*(90m-100m).

4. The permafrost region thermal thaw landslide risk monitoring system according to claim 1, characterized in that: The network information database adopts the web of science to provide data, ensuring the authority and comprehensiveness of the data.

5. A method for monitoring the risk of thermal thaw landslides in permafrost areas according to claims 1-4, comprising the following steps: Step 1: The permafrost area is divided into zones and numbers using English and numerical numbers. The range of each zone is (90m-100m)*(90m-100m). Ten holes are randomly drilled in each zone with a diameter of 1m. The ten holes are divided into 10cm*2, 20cm*2, 30cm*2, 40cm*2 and 50cm*2. Temperature sensors and humidity sensors are placed in the two groups of holes with corresponding depths, and then the holes are filled. Step 2: Use Synthetic Aperture Radar Interferometry (SBAS_InSAR) technology to monitor the displacement of the ground and soil to determine whether there is a sliding trend; Step 3: Install a small weather station on the ground within 1m of the same location as in Step 1 to monitor the external temperature and precipitation data of the permafrost area in real time. Use the Gaofen-2 (GF-2) image to calculate the vegetation cover (NDVI) data. Step 4: Install solar power generation facilities, use solar power generation facilities to convert solar energy into electrical energy, and store the electrical energy through batteries. The batteries are connected to the monitoring equipment through cables to ensure that they are powered on, providing power support for the monitoring equipment of the remote field monitoring module. Step 5: Using 4G / 5G wireless communication technology, remotely transmit the real-time monitoring data of steps 1, 2, and 3; Step 6: Use the database to classify and store the data transferred in step 4; Step 7: Using the large amount of normal data provided by the network information database as training samples, the U-Net architecture uses the ResNet101 model to monitor new data stored in the database. If a data point deviates from the normal distribution range, it is determined to be abnormal and there is a risk, realizing real-time monitoring and early warning of risks. When the monitoring data analysis results show a high risk, early warning information is issued to relevant departments and personnel via text message push via wireless transmission.

Images