System and method for identifying surface collapse based on multispectral image

Abstract

The invention relates to the field of surface collapse recognition, in particular to a surface collapse recognition system and method based on a multispectral image, and the system comprises an image recognition module, a spectral information module, a penetration evaluation module, a region marking module and a geophysical prospecting analysis module. The spectrum information module is used for acquiring spectrum image sets under different precipitation amounts, the penetration evaluation module is used for estimating the penetration speed and the penetration depth, the area marking module is used for segmenting and marking a subsidence area, and the geophysical prospecting analysis module is used for determining a drilling point and performing collapse analysis. The method is convenient for large-scale general survey of the surface collapse, prevents geological disasters caused by the surface collapse, improves the recognition accuracy, analyzes historical collapse and long-term settlement trends, is suitable for regular general survey of mining areas, urban underground engineering and other areas, and realizes an efficient and low-cost land collapse screening scheme.

Description

Technical Field

[0001] This invention relates to the field of ground subsidence identification, specifically to a ground subsidence identification system and method based on multispectral imagery. Background Technology

[0002] Quaternary ground subsidence refers to the geological phenomenon where surface soil, after meeting the conditions for pore development, sinks downward under external disturbance, forming a subsidence pit on the ground. It is a common geological hazard in silty sedimentary areas. Causes of ground subsidence include underground mining activities, karstification, pipeline leaks and erosion, and excessive groundwater extraction.

[0003] When underground subsidence or cavitation occurs, it triggers a series of surface effects, altering the spectral reflectance of ground features. By capturing the spectral information of the same feature across multiple specific wavelength ranges, areas with developed fissures can be identified to some extent. While commonly used satellite monitoring methods for ground subsidence can identify abnormal soil masses, they are difficult to use in situations where vegetation obscures the surface or the area is buried by water. Multispectral imagery, on the other hand, utilizes light waves with varying transmittances for flexible spectral responses, offering wide coverage at a low cost.

[0004] However, multispectral image recognition methods are still immature, with limitations in spatial resolution and poor penetration, making it difficult to mark and verify subsidence areas. The probability of false alarms caused by non-subsidence factors such as climate, agriculture, and pests is relatively high, and there is a lack of data processing and recognition methods, which reduces the efficiency of subsidence prediction. Summary of the Invention

[0005] The purpose of this invention is to provide a ground subsidence identification system and method based on multispectral imagery to solve the problems mentioned in the background art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a ground subsidence identification system based on multispectral images, comprising: an image recognition module, a spectral information module, a permeability assessment module, a region marking module, and a geophysical analysis module; The image recognition module is used to monitor the ground using GIS remote sensing technology, select spectral waves with different penetration, obtain multispectral images, perform radiometric image correction based on sensors and GPS control points, and fuse panchromatic and multispectral images to obtain the land use distribution, plant types, plant growth status and surface thermal conductivity of the region. The spectral information module is used to associate remote sensing images with data from meteorological monitoring equipment deployed in the monitoring area to ensure that the remote sensing shooting time matches the time sequence of precipitation events, obtain precipitation data, including precipitation amount, precipitation duration, wind speed and temperature, continuously shoot ground images during precipitation to form a spectral atlas, and continuously monitor the ground until a spectral atlas is obtained under at least two precipitation conditions. The permeability assessment module is used to calculate the normalized differential moisture index using the shortwave infrared band and retrieve the surface temperature using the thermal infrared band, constructing a feature space for temperature and vegetation index. Based on the one-dimensional soil infiltration model, it estimates the infiltration rate and depth according to precipitation intensity, duration, surface temperature, soil type, and surface soil moisture change rate. Using the fused visible light vegetation image, it classifies land vegetation through machine learning algorithms, calculates the normalized vegetation index, uploads the vegetation type and vegetation index to the big data center, and estimates the range of vegetation root depth. The region marking module is used to identify closed regions that are different from the background values ​​of adjacent soil infiltration rates using edge detection or region growing algorithms, segment the infiltration rate map, and perform GIS spatial overlay analysis on the segmented regions with geological maps or known collapse landform maps. After overlapping and segmentation, collapse stress zones are obtained. When the infiltration depth in the collapse stress zone is less than the lower limit of the vegetation root depth or the infiltration depth of bare land is less than the background value of the segmented region and reaches a threshold, the collapse stress zone is marked as a collapse zone. The geophysical analysis module is used to select the high-density resistivity method, lay out geophysical lines along the edge of the collapse area to form a grid-like or cross-shaped survey network, determine the density of drilling core points on the geophysical lines according to the penetration depth, place the drilling points on the center and gradient zone of the interpreted collapse area, conduct standard penetration tests at the drilling points using the free-fall hammer method, collect soil and rock samples for collapse analysis, and classify the risk of the collapse area.

[0007] Furthermore, the image recognition module includes: a GIS remote sensing unit, a spectral selection unit, and an image preprocessing unit; The GIS remote sensing unit is used to access GIS satellites or aerial equipment to acquire time-series spectral remote sensing data of the area to be monitored. The spectral selection unit is used to classify land cover using visible light band images, retrieve land surface temperature using thermal infrared band images, and measure soil moisture using near-infrared and short-wave infrared band images. The image preprocessing unit uses the FLASSH or 6S model for atmospheric correction to eliminate the effects of aerosols and water vapor, and converts the DN value of the image into surface reflectance.

[0008] Furthermore, the spectral information module includes: a spectral index unit and a meteorological atlas unit; The spectral index unit is used to build automated scripts to perform radiometric calibration, atmospheric correction and orthorectification on multispectral images, fuse panchromatic and multispectral bands, and repair image errors. The meteorological atlas unit is used to connect multispectral image capturing equipment to regional meteorological radar data, associate it with rain gauge data, and generate a continuous image set of soil moisture changes before and after precipitation.

[0009] Furthermore, the permeability assessment module includes: a water analysis unit and a vegetation assessment unit; The water analysis unit is used to combine the evaporation data provided by meteorological data to establish a simplified one-dimensional infiltration model and quantitatively solve the soil infiltration rate and infiltration depth. The vegetation assessment unit is used to identify plant types and vegetation distribution, determine vegetation ecological indicators, and estimate vegetation root depth.

[0010] Furthermore, the region marking module includes: an image segmentation unit and an overlap segmentation unit; The image segmentation unit is used to segment abnormal regions in the penetration rate map using a semantic segmentation model or OBIA image analysis method. The overlapping segmentation unit is used to spatially overlay and analyze the anomalous area with Quaternary geological maps, sinkhole databases, or InSAR surface deformation results to identify sinkhole areas.

[0011] Furthermore, the geophysical analysis module includes: a geophysical line unit, a drilling and coring unit, and a collapse analysis unit; The geophysical line unit is used to lay geophysical lines along the edge of the collapse area and inside it to form a gridded or encircling measurement network to obtain underground resistivity profiles. The core drilling unit is used to determine drilling points at resistivity anomaly points, collapse zone boundary points, and gradient zones, and to perform standard penetration tests using the free-fall hammer method. The collapse analysis unit is used to perform ERT anomaly analysis on the collected soil and rock samples and to perform three-dimensional spatial fusion comparison to confirm the source of the collapse.

[0012] A ground subsidence identification method based on multispectral imagery includes the following steps: Step S1. Using GIS remote sensing technology, the visible light band, thermal infrared band, near-infrared band and short-wave infrared band are selected respectively to monitor the ground and obtain multispectral images. Atmospheric correction, orthorectification correction and image fusion are performed on the multispectral images. Step S2. Associate the remote sensing images with the data from meteorological monitoring equipment deployed in the monitoring area to match the shooting time with the time sequence of precipitation events, obtain precipitation data, and continuously capture ground images during the precipitation process to form a spectral atlas; Step S3. Calculate the soil differential moisture index using shortwave infrared images from the spectral atlas, retrieve the surface temperature from thermal infrared images, and estimate the infiltration rate and depth based on a one-dimensional soil infiltration model, according to precipitation intensity, duration, surface temperature, soil type, and surface soil moisture change rate. Step S4. Identify closed areas that are different from the background values ​​of adjacent soil infiltration rates, segment the infiltration rate map, overlay the segmented areas with the geological map to obtain the collapse stress area, estimate the range of vegetation root depth based on the visible light band image, and mark the collapse stress area as a collapse area when the infiltration depth in the collapse stress area is less than the lower limit of vegetation root depth or the infiltration depth of bare land is less than the background value of the segmented area and reaches the threshold. Step S5. Lay out geophysical exploration lines along the edge of the collapse area to form a grid or cross-shaped survey network. Determine the density of drilling core points according to the penetration depth. Conduct standard penetration tests at the drilling points and collect soil and rock samples for collapse analysis.

[0013] Furthermore, step S1 includes: Step S11. Access GIS satellite or aerial identification to obtain time-series spectral remote sensing data of the area to be monitored, adjust the spectral illumination wavelength, select spectral waves with different penetration to monitor the ground, among which visible light band images are used for land cover classification, thermal infrared band images are used to retrieve surface temperature, and near-infrared and short-wave infrared band images are used to measure soil moisture. Step S12. Perform atmospheric correction using the FLASSH or 6S model to eliminate the effects of aerosols and water vapor, convert the DN values ​​of the image to surface reflectance, perform orthorectification based on the sensor and GPS control points, and fuse the panchromatic and multispectral images.

[0014] Furthermore, step S2 includes: Step S21. Connect the multispectral image capturing device to the regional meteorological radar data and associate it with the rain gauge data to match the remote sensing capture time with the precipitation event time sequence and obtain precipitation data, including precipitation amount, precipitation duration, wind speed and temperature, and continuous soil moisture change sequence before and after precipitation. Step S22. During precipitation, the remote sensing equipment takes multispectral images of the ground at fixed intervals, classifies the images according to the wavelength of the images, and forms a spectral atlas. The monitoring continues until at least two spectral atlases under precipitation conditions are obtained.

[0015] Furthermore, step S3 includes: Step S31. Use thermal infrared or shortwave infrared spectral characteristic spatial method to invert the surface soil moisture before and after precipitation, calculate the change in moisture, and use the surface energy balance algorithm or METRIC model to calculate the surface evapotranspiration rate. Step S32. Estimate the infiltration rate and depth based on precipitation intensity, duration, surface temperature, soil type, and surface soil moisture change rate, satisfying v = ΔS / [(P-ET)·t], where v is the infiltration rate, P is the precipitation amount in the interval, ET is the evaporation amount in the interval, ΔS is the water difference due to soil moisture change, and t is the time interval between two observations. Based on the Philip infiltration equation, with precipitation and soil type as input parameters, output the infiltration depth.

[0016] Furthermore, step S4 includes: Step S41. Using a semantic segmentation model or OBIA image analysis method, identify closed areas that are different from the background values ​​of adjacent soil infiltration rates, segment the infiltration rate map, and perform spatial overlay analysis on the abnormal areas with Quaternary geological maps, subsidence pit databases or InSAR surface deformation results to obtain the subsidence stress zone after segmentation. Step S42. Using the fused visible light vegetation image, land vegetation is classified through machine learning algorithms, the normalized vegetation index is calculated, the vegetation type and vegetation index are uploaded to the big data center, and the root depth range of vegetation is estimated. Step S43. When the infiltration depth in the collapse stress zone is less than the lower limit of the vegetation root depth or the infiltration depth of the bare land is less than the background of the segmented area and reaches the threshold, the collapse stress zone is marked as a collapse zone.

[0017] Furthermore, step S5 includes: Step S51. Select the high-density resistivity method, lay out geophysical lines along the edge of the collapse area and inside it to form a gridded or encircling measurement network, and obtain the underground resistivity profile; Step S52. Determine the density of core drilling points on the geophysical exploration line according to the penetration depth, set drilling points at resistivity anomaly points, collapse zone boundary points and gradient zones, conduct standard penetration tests using the free fall hammer method, collect soil and rock samples for collapse analysis, and classify the risk of the collapse zone.

[0018] Compared with the prior art, the beneficial effects achieved by the present invention are: 1. This invention uses GIS remote sensing technology to monitor the ground by selecting spectral waves with different penetration properties, performs radiometric correction based on sensors and GPS control points, and fuses panchromatic and multispectral images to obtain land parameters. By combining the advantages of spectral, spatial and temporal information, it significantly improves monitoring efficiency and accuracy, facilitates early warning and large-scale surveys of ground subsidence, and prevents geological disasters caused by ground subsidence.

[0019] 2. This invention links remote sensing images with meteorological monitoring equipment to continuously photograph the ground during rainfall. Based on the changes in shortwave infrared humidity of the land before and after precipitation and the environmental evaporation rate, the infiltration rate and depth of precipitation are calculated. Image segmentation technology is used to delineate closed areas that coincide with Quaternary geological subsidence landform areas to determine subsidence stress zones. This is suitable for regular surveys of mining areas, urban underground engineering projects, and other areas, constituting an efficient, economical, and scalable solution for monitoring ground subsidence.

[0020] 3. This invention utilizes the high-density resistivity method to determine the density of drilling core points according to the penetration depth. Standard penetration tests are conducted at the core points to collect soil and rock samples for subsidence analysis. By analyzing the surface environmental anomalies caused by subsidence, potential subsidence risks can be indirectly identified, which can greatly improve the accuracy of identification. It can also analyze historical subsidence and long-term settlement trends to achieve efficient and low-cost land subsidence screening. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the structure of the ground subsidence recognition system based on multispectral imagery of the present invention; Figure 2 This is a schematic diagram illustrating the steps of the ground subsidence identification method based on multispectral imagery of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 The present invention provides a technical solution: a ground subsidence identification system based on multispectral imagery, comprising: an image recognition module, a spectral information module, a permeability assessment module, a region marking module, and a geophysical analysis module; The image recognition module is used to monitor the ground using GIS remote sensing technology, select spectral waves with different penetration, obtain multispectral images, perform radiometric image correction based on sensors and GPS control points, and fuse panchromatic and multispectral images to obtain the land use distribution, plant types, plant growth status and surface thermal conductivity of the region. The image recognition module includes: a GIS remote sensing unit, a spectral selection unit, and an image preprocessing unit; The GIS remote sensing unit is used to access GIS satellites or aerial equipment to acquire time-series spectral remote sensing data of the area to be monitored. The spectral selection unit is used to classify land cover using visible light band images, retrieve land surface temperature using thermal infrared band images, and measure soil moisture using near-infrared and short-wave infrared band images. The image preprocessing unit uses the FLASSH or 6S model for atmospheric correction to eliminate the effects of aerosols and water vapor, and converts the DN value of the image into surface reflectance.

[0024] The spectral information module is used to associate remote sensing images with data from meteorological monitoring equipment deployed in the monitoring area to ensure that the remote sensing shooting time matches the time sequence of precipitation events, obtain precipitation data, including precipitation amount, precipitation duration, wind speed and temperature, continuously shoot ground images during precipitation to form a spectral atlas, and continuously monitor the ground until a spectral atlas is obtained under at least two precipitation conditions. The spectral information module includes: a spectral index unit and a meteorological atlas unit; The spectral index unit is used to build automated scripts to perform radiometric calibration, atmospheric correction and orthorectification on multispectral images, fuse panchromatic and multispectral bands, and repair image errors. The meteorological atlas unit is used to connect multispectral image capturing equipment to regional meteorological radar data, associate it with rain gauge data, and generate a continuous image set of soil moisture changes before and after precipitation.

[0025] The permeability assessment module is used to calculate the normalized differential moisture index using the shortwave infrared band and retrieve the surface temperature using the thermal infrared band, constructing a feature space for temperature and vegetation index. Based on the one-dimensional soil infiltration model, it estimates the infiltration rate and depth according to precipitation intensity, duration, surface temperature, soil type, and surface soil moisture change rate. Using the fused visible light vegetation image, it classifies land vegetation through machine learning algorithms, calculates the normalized vegetation index, uploads the vegetation type and vegetation index to the big data center, and estimates the range of vegetation root depth. The permeability assessment module includes: a water body analysis unit and a vegetation assessment unit; The water analysis unit is used to combine the evaporation data provided by meteorological data to establish a simplified one-dimensional infiltration model and quantitatively solve the soil infiltration rate and infiltration depth. The vegetation assessment unit is used to identify plant types and vegetation distribution, determine vegetation ecological indicators, and estimate vegetation root depth.

[0026] The region marking module is used to identify closed regions that are different from the background values ​​of adjacent soil infiltration rates using edge detection or region growing algorithms, segment the infiltration rate map, and perform GIS spatial overlay analysis on the segmented regions with geological maps or known collapse landform maps. After overlapping and segmentation, collapse stress zones are obtained. When the infiltration depth in the collapse stress zone is less than the lower limit of the vegetation root depth or the infiltration depth of bare land is less than the background value of the segmented region and reaches a threshold, the collapse stress zone is marked as a collapse zone. The region marking module includes: an image segmentation unit and an overlap segmentation unit; The image segmentation unit is used to segment abnormal regions in the penetration rate map using a semantic segmentation model or OBIA image analysis method. The overlapping segmentation unit is used to spatially overlay and analyze the anomalous area with Quaternary geological maps, sinkhole databases, or InSAR surface deformation results to identify sinkhole areas.

[0027] The geophysical analysis module is used to select the high-density resistivity method, lay out geophysical lines along the edge of the collapse area to form a grid-like or cross-shaped survey network, determine the density of drilling core points on the geophysical lines according to the penetration depth, place the drilling points on the center and gradient zone of the interpreted collapse area, conduct standard penetration tests at the drilling points using the free-fall hammer method, collect soil and rock samples for collapse analysis, and classify the risk of the collapse area.

[0028] The geophysical analysis module includes: a geophysical line unit, a drilling and coring unit, and a collapse analysis unit; The geophysical line unit is used to lay geophysical lines along the edge of the collapse area and inside it to form a gridded or encircling measurement network to obtain underground resistivity profiles. The core drilling unit is used to determine drilling points at resistivity anomaly points, collapse zone boundary points, and gradient zones, and to perform standard penetration tests using the free-fall hammer method. The collapse analysis unit is used to perform ERT anomaly analysis on the collected soil and rock samples and to perform three-dimensional spatial fusion comparison to confirm the source of the collapse.

[0029] like Figure 2 As shown, the ground subsidence identification method based on multispectral imagery includes the following steps: Step S1. Using GIS remote sensing technology, the visible light band, thermal infrared band, near-infrared band and short-wave infrared band are selected respectively to monitor the ground and obtain multispectral images. Atmospheric correction, orthorectification correction and image fusion are performed on the multispectral images. Step S1 includes: Step S11. Access GIS satellite or aerial identification to obtain time-series spectral remote sensing data of the area to be monitored, adjust the spectral illumination wavelength, select spectral waves with different penetration to monitor the ground, among which visible light band images are used for land cover classification, thermal infrared band images are used to retrieve surface temperature, and near-infrared and short-wave infrared band images are used to measure soil moisture. Step S12. Perform atmospheric correction using the FLASSH or 6S model to eliminate the effects of aerosols and water vapor, convert the DN values ​​of the image to surface reflectance, perform orthorectification based on the sensor and GPS control points, and fuse the panchromatic and multispectral images.

[0030] Step S2. Associate the remote sensing images with the data from meteorological monitoring equipment deployed in the monitoring area to match the shooting time with the time sequence of precipitation events, obtain precipitation data, and continuously capture ground images during the precipitation process to form a spectral atlas; Step S2 includes: Step S21. Connect the multispectral image capturing device to the regional meteorological radar data and associate it with the rain gauge data to match the remote sensing capture time with the precipitation event time sequence and obtain precipitation data, including precipitation amount, precipitation duration, wind speed and temperature, and continuous soil moisture change sequence before and after precipitation. Step S22. During precipitation, the remote sensing equipment takes multispectral images of the ground at fixed intervals, classifies the images according to the wavelength of the images, and forms a spectral atlas. The monitoring continues until at least two spectral atlases under precipitation conditions are obtained.

[0031] Step S3. Calculate the soil differential moisture index using shortwave infrared images from the spectral atlas, retrieve the surface temperature from thermal infrared images, and estimate the infiltration rate and depth based on a one-dimensional soil infiltration model, according to precipitation intensity, duration, surface temperature, soil type, and surface soil moisture change rate. Step S3 includes: Step S31. Use thermal infrared or shortwave infrared spectral characteristic spatial method to invert the surface soil moisture before and after precipitation, calculate the change in moisture, and use the surface energy balance algorithm or METRIC model to calculate the surface evapotranspiration rate. Step S32. Estimate the infiltration rate and depth based on precipitation intensity, duration, surface temperature, soil type, and surface soil moisture change rate, satisfying v = ΔS / [(P-ET)·t], where v is the infiltration rate, P is the precipitation amount in the interval, ET is the evaporation amount in the interval, ΔS is the water difference due to soil moisture change, and t is the time interval between two observations. Based on the Philip infiltration equation, with precipitation and soil type as input parameters, output the infiltration depth.

[0032] Step S4. Identify closed areas that are different from the background values ​​of adjacent soil infiltration rates, segment the infiltration rate map, overlay the segmented areas with the geological map to obtain the collapse stress area, estimate the range of vegetation root depth based on the visible light band image, and mark the collapse stress area as a collapse area when the infiltration depth in the collapse stress area is less than the lower limit of vegetation root depth or the infiltration depth of bare land is less than the background value of the segmented area and reaches the threshold. Step S4 includes: Step S41. Using a semantic segmentation model or OBIA image analysis method, identify closed areas that are different from the background values ​​of adjacent soil infiltration rates, segment the infiltration rate map, and perform spatial overlay analysis on the abnormal areas with Quaternary geological maps, subsidence pit databases or InSAR surface deformation results to obtain the subsidence stress zone after segmentation. Step S42. Using the fused visible light vegetation image, land vegetation is classified through machine learning algorithms, the normalized vegetation index is calculated, the vegetation type and vegetation index are uploaded to the big data center, and the root depth range of vegetation is estimated. Step S43. When the infiltration depth in the collapse stress zone is less than the lower limit of the vegetation root depth or the infiltration depth of the bare land is less than the background of the segmented area and reaches the threshold, the collapse stress zone is marked as a collapse zone.

[0033] Step S5. Lay out geophysical exploration lines along the edge of the collapse area to form a grid or cross-shaped survey network. Determine the density of drilling core points according to the penetration depth. Conduct standard penetration tests at the drilling points and collect soil and rock samples for collapse analysis.

[0034] Step S5 includes: Step S51. Select the high-density resistivity method, lay out geophysical lines along the edge of the collapse area and inside it to form a gridded or encircling measurement network, and obtain the underground resistivity profile; Step S52. Determine the density of core drilling points on the geophysical exploration line according to the penetration depth, set drilling points at resistivity anomaly points, collapse zone boundary points and gradient zones, conduct standard penetration tests using the free fall hammer method, collect soil and rock samples for collapse analysis, and classify the risk of the collapse zone.

[0035] Example: A certain area is located downstream of an alluvial plain. The surface soil type is loose rock sediments. The terrain is high in the west and low in the east, with the elevation gradually decreasing from 52.9 meters to 48.6 meters. The average annual precipitation is 619.43 mm. The landforms are mainly of three types: slightly sloping low plains, hills, and breached fan-shaped landforms. Among them, the slightly sloping low plains and breached fan-shaped landforms are subsidence landforms. During precipitation, multispectral images of the ground are taken every 10 minutes using remote sensing equipment. Infrared spectroscopy is used to determine that the soil moisture content of a plain area is lower than the background value. The calculated infiltration depth is 1 m, and the estimated plant root depth is 2-4 m. This area is then marked as a subsidence area. Geophysical exploration is carried out in the subsidence area to determine the subsidence risk level, and safety protection measures are implemented for the area.

[0036] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0037] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A ground subsidence identification method based on multispectral imagery, characterized in that, The method includes the following steps: Step S1. Using GIS remote sensing technology, the visible light band, thermal infrared band, near-infrared band and short-wave infrared band are selected respectively to monitor the ground and obtain multispectral images. Atmospheric correction, orthorectification correction and image fusion are performed on the multispectral images. Step S2. Associate the remote sensing images with the data from meteorological monitoring equipment deployed in the monitoring area to match the shooting time with the time sequence of precipitation events, obtain precipitation data, and continuously capture ground images during the precipitation process to form a spectral atlas; Step S3. Calculate the soil differential moisture index using shortwave infrared images from the spectral atlas, retrieve the surface temperature from thermal infrared images, and estimate the infiltration rate and depth based on a one-dimensional soil infiltration model, according to precipitation intensity, duration, surface temperature, soil type, and surface soil moisture change rate. Step S4. Identify closed areas that are different from the background values ​​of adjacent soil infiltration rates, segment the infiltration rate map, overlay the segmented areas with the geological map to obtain the collapse stress area, estimate the range of vegetation root depth based on the visible light band image, and mark the collapse stress area as a collapse area when the infiltration depth in the collapse stress area is less than the lower limit of vegetation root depth or the infiltration depth of bare land is less than the background value of the segmented area and reaches the threshold. Step S5. Lay out geophysical exploration lines along the edge of the collapse area to form a grid or cross-shaped survey network. Determine the density of drilling core points according to the penetration depth. Conduct standard penetration tests at the drilling points and collect soil and rock samples for collapse analysis.

2. The ground subsidence identification method based on multispectral imagery according to claim 1, characterized in that: Step S1 includes: Step S11. Access GIS satellite or aerial identification to obtain time-series spectral remote sensing data of the area to be monitored, adjust the spectral illumination wavelength, select spectral waves with different penetration to monitor the ground, among which visible light band images are used for land cover classification, thermal infrared band images are used to retrieve surface temperature, and near-infrared and short-wave infrared band images are used to measure soil moisture. Step S12. Perform atmospheric correction using the FLASSH or 6S model to eliminate the effects of aerosols and water vapor, convert the DN values ​​of the image to surface reflectance, perform orthorectification based on the sensor and GPS control points, and fuse the panchromatic and multispectral images. Step S2 includes: Step S21. Connect the multispectral image capturing device to the regional meteorological radar data and associate it with the rain gauge data to match the remote sensing capture time with the precipitation event time sequence and obtain precipitation data, including precipitation amount, precipitation duration, wind speed and temperature, and continuous soil moisture change sequence before and after precipitation. Step S22. During precipitation, the remote sensing equipment takes multispectral images of the ground at fixed intervals, classifies the images according to the wavelength of the images, and forms a spectral atlas. The monitoring continues until at least two spectral atlases under precipitation conditions are obtained.

3. The ground subsidence identification method based on multispectral imagery according to claim 2, characterized in that: Step S3 includes: Step S31. Use thermal infrared or shortwave infrared spectral characteristic spatial method to invert the surface soil moisture before and after precipitation, calculate the change in moisture, and use the surface energy balance algorithm or METRIC model to calculate the surface evapotranspiration rate. Step S32. Estimate the infiltration rate and depth based on precipitation intensity, duration, surface temperature, soil type, and surface soil moisture change rate, satisfying v = ΔS / [(P-ET)·t], where v is the infiltration rate, P is the precipitation amount in the interval, ET is the evaporation amount in the interval, ΔS is the water difference due to soil moisture change, and t is the time interval between two observations. Based on the Philip infiltration equation, with precipitation and soil type as input parameters, output the infiltration depth.

4. The ground subsidence identification method based on multispectral imagery according to claim 3, characterized in that: Step S4 includes: Step S41. Using a semantic segmentation model or OBIA image analysis method, identify closed areas that are different from the background values ​​of adjacent soil infiltration rates, segment the infiltration rate map, and perform spatial overlay analysis on the abnormal areas with Quaternary geological maps, subsidence pit databases or InSAR surface deformation results to obtain the subsidence stress zone after segmentation. Step S42. Using the fused visible light vegetation image, land vegetation is classified through machine learning algorithms, the normalized vegetation index is calculated, the vegetation type and vegetation index are uploaded to the big data center, and the root depth range of vegetation is estimated. Step S43. When the infiltration depth in the collapse stress zone is less than the lower limit of the vegetation root depth or the infiltration depth of the bare land is less than the background of the segmented area and reaches the threshold, the collapse stress zone is marked as a collapse zone.

5. The ground subsidence identification method based on multispectral imagery according to claim 4, characterized in that: Step S5 includes: Step S51. Select the high-density resistivity method, lay out geophysical lines along the edge of the collapse area and inside it to form a gridded or encircling measurement network, and obtain the underground resistivity profile; Step S52. Determine the density of core drilling points on the geophysical exploration line according to the penetration depth, set drilling points at resistivity anomaly points, collapse zone boundary points and gradient zones, conduct standard penetration tests using the free fall hammer method, collect soil and rock samples for collapse analysis, and classify the risk of the collapse zone.

6. A ground subsidence recognition system based on multispectral imagery, characterized in that, The system includes the following modules: image recognition module, spectral information module, penetration assessment module, area marking module, and geophysical analysis module; The image recognition module is used to monitor the ground using GIS remote sensing technology, select spectral waves with different penetration, obtain multispectral images, perform radiometric image correction based on sensors and GPS control points, and fuse panchromatic and multispectral images to obtain the land use distribution, plant types, plant growth status and surface thermal conductivity of the region. The spectral information module is used to associate remote sensing images with data from meteorological monitoring equipment deployed in the monitoring area to ensure that the remote sensing shooting time matches the time sequence of precipitation events, obtain precipitation data, including precipitation amount, precipitation duration, wind speed and temperature, continuously shoot ground images during precipitation to form a spectral atlas, and continuously monitor the ground until a spectral atlas is obtained under at least two precipitation conditions. The permeability assessment module is used to calculate the normalized differential moisture index using the shortwave infrared band and retrieve the surface temperature using the thermal infrared band, constructing a feature space for temperature and vegetation index. Based on the one-dimensional soil infiltration model, it estimates the infiltration rate and depth according to precipitation intensity, duration, surface temperature, soil type, and surface soil moisture change rate. Using the fused visible light vegetation image, it classifies land vegetation through machine learning algorithms, calculates the normalized vegetation index, uploads the vegetation type and vegetation index to the big data center, and estimates the range of vegetation root depth. The region marking module is used to identify closed regions that are different from the background values ​​of adjacent soil infiltration rates using edge detection or region growing algorithms, segment the infiltration rate map, and perform GIS spatial overlay analysis on the segmented regions with geological maps or known collapse landform maps. After overlapping and segmentation, collapse stress zones are obtained. When the infiltration depth in the collapse stress zone is less than the lower limit of the vegetation root depth or the infiltration depth of bare land is less than the background value of the segmented region and reaches a threshold, the collapse stress zone is marked as a collapse zone. The geophysical analysis module is used to select the high-density resistivity method, lay out geophysical lines along the edge of the collapse area to form a grid-like or cross-shaped survey network, determine the density of drilling core points on the geophysical lines according to the penetration depth, place the drilling points on the center and gradient zone of the interpreted collapse area, conduct standard penetration tests at the drilling points using the free-fall hammer method, collect soil and rock samples for collapse analysis, and classify the risk of the collapse area.

7. The ground subsidence recognition system based on multispectral imagery according to claim 6, characterized in that: The image recognition module includes: a GIS remote sensing unit, a spectral selection unit, and an image preprocessing unit; The GIS remote sensing unit is used to access GIS satellites or aerial equipment to acquire time-series spectral remote sensing data of the area to be monitored. The spectral selection unit is used to classify land cover using visible light band images, retrieve land surface temperature using thermal infrared band images, and measure soil moisture using near-infrared and short-wave infrared band images. The image preprocessing unit uses the FLASSH or 6S model for atmospheric correction to eliminate the effects of aerosols and water vapor, and converts the DN value of the image into surface reflectance.

8. The ground subsidence recognition system based on multispectral imagery according to claim 7, characterized in that: The spectral information module includes: a spectral index unit and a meteorological atlas unit; The spectral index unit is used to build automated scripts to perform radiometric calibration, atmospheric correction and orthorectification on multispectral images, fuse panchromatic and multispectral bands, and repair image errors. The meteorological atlas unit is used to connect multispectral image capturing equipment to regional meteorological radar data, associate it with rain gauge data, and generate a continuous image set of soil moisture changes before and after precipitation.

9. The ground subsidence recognition system based on multispectral imagery according to claim 8, characterized in that: The permeability assessment module includes: a water body analysis unit and a vegetation assessment unit; The water analysis unit is used to combine the evaporation data provided by meteorological data to establish a simplified one-dimensional infiltration model and quantitatively solve the soil infiltration rate and infiltration depth. The vegetation assessment unit is used to identify plant types and vegetation distribution, determine vegetation ecological indicators, and estimate vegetation root depth. The region marking module includes: an image segmentation unit and an overlap segmentation unit; The image segmentation unit is used to segment abnormal regions in the penetration rate map using a semantic segmentation model or OBIA image analysis method. The overlapping segmentation unit is used to spatially overlay and analyze the anomalous area with Quaternary geological maps, sinkhole databases, or InSAR surface deformation results to identify sinkhole areas.

10. The ground subsidence identification system based on multispectral imagery according to claim 9, characterized in that: The geophysical analysis module includes: a geophysical line unit, a drilling and coring unit, and a collapse analysis unit; The geophysical line unit is used to lay geophysical lines along the edge of the collapse area and inside it to form a gridded or encircling measurement network to obtain underground resistivity profiles. The core drilling unit is used to determine drilling points at resistivity anomaly points, collapse zone boundary points, and gradient zones, and to perform standard penetration tests using the free-fall hammer method. The collapse analysis unit is used to perform ERT anomaly analysis on the collected soil and rock samples and to perform three-dimensional spatial fusion comparison to confirm the source of the collapse.

Images