Ground surface matrix sensor based on hyperspectral technology and monitoring method

Abstract

The invention provides an earth surface matrix sensor based on a hyperspectral technology and a monitoring method, and the sensor comprises a sensor module which is used for extending into a stratum through a housing; the hyperspectral module extends into the stratum through the shell; the communication module is in communication connection with the sensor module and the hyperspectral module; the terminal module is in communication connection with the communication module and stores a plurality of association relationships between hyperspectral curve characteristics and surface matrix physicochemical property parameters (moisture, salinity, temperature, acidity and alkalinity (pH), oxidation-reduction potential (Eh) and the like), and each association relationship corresponds to a preset region type; the terminal module obtains hyperspectral information measured by the hyperspectral module; the terminal module determines an association relationship corresponding to the stratum according to a region type corresponding to the stratum; and the terminal module completes surface matrix monitoring of the stratum according to the hyperspectral information, the association relationship and the surface matrix physicochemical property parameters.

Description

Technical Field

[0001] This application relates to the technical field of surface matrix monitoring, and in particular, to a surface matrix sensor and a monitoring method based on hyperspectral technology. Background Art

[0002] As the basic material that nurtures and supports various natural resources such as soil, forests, grasslands, water, and wetlands on the earth's surface, the physical and chemical properties of the surface matrix directly control the spatial distribution pattern of surface agricultural production and vegetation ecology. It is the space where the multi-sphere interactions of the earth are most frequent, and it is also the material basis for the overall protection, systematic restoration, and comprehensive management of cultivated land and natural ecosystems. Traditional monitoring techniques focus on the fixed-point collection of shallow soil parameters within 0 - 2m. Data is obtained by manually excavating and burying single-point sensors such as temperature, humidity, and salinity. Such methods have significant drawbacks: First, the manual drilling and layout will damage the original structure of the formation, resulting in deviations in monitoring data due to soil disturbance; Second, the layout depth of the sensors is limited by the drilling process and it is difficult to penetrate dense layers or gravel layers to achieve deep matrix monitoring below 2m; Third, single-parameter monitoring cannot reflect the multi-parameter coupling mechanism. Especially under extremely arid conditions, the supporting role of the deep matrix in water transport and nutrient supply cannot be effectively characterized by traditional point data.

[0003] In recent years, research has shown that the surface matrix exhibits significant heterogeneity in the vertical profile. The water-salt migration law, the chemical composition of cementing substances, and the pore structure characteristics of the deep matrix below 2m have a decisive impact on plant root development and matrix water-holding capacity. However, due to technical limitations, existing monitoring equipment can neither achieve non-destructive in-situ burial nor has the ability to establish a correlation model between hyperspectral characteristics and multi-physical parameters, resulting in the long-term dependence on laboratory core analysis for deep matrix monitoring, with problems such as poor timeliness and low spatial coverage.

[0004] There is an urgent need for a sensor that can penetrate deep strata, synchronously obtain multi-dimensional parameters, and have intelligent analysis capabilities. Summary of the Invention

[0005] This application provides a surface matrix sensor and a monitoring method based on hyperspectral technology, aiming to solve the problems that existing monitoring equipment, due to technical limitations, can neither achieve non-destructive in-situ burial nor has the ability to establish a correlation model between hyperspectral characteristics and multi-physical parameters, resulting in the long-term dependence on laboratory core analysis for deep matrix monitoring, with problems such as poor timeliness and low spatial coverage.

[0006] In a first aspect, this application provides a surface matrix sensor based on hyperspectral technology, including: A sensor module, which is used to extend into the formation through the housing; A hyperspectral module, which extends into the formation through the housing; A communication module, which is communicatively connected to the sensor module and the hyperspectral module respectively; A terminal module, which is communicatively connected to the communication module. The terminal module stores a plurality of correlation relationships between hyperspectral curve features and surface matrix physical and chemical property parameters, and each correlation relationship corresponds to a preset regional type; the terminal module obtains the hyperspectral information measured by the hyperspectral module; the terminal module determines the correlation relationship corresponding to the formation according to the regional type corresponding to the formation; the terminal module completes the monitoring of the surface matrix of the formation according to the hyperspectral information, the correlation relationship and the surface matrix physical and chemical property parameters.

[0007] In a second aspect, the present application provides a method for monitoring surface matrix based on hyperspectral technology, which is characterized in that it is applied to the surface matrix sensor based on hyperspectral technology provided in any embodiment of the present application; the method includes: Obtaining a plurality of correlation relationships between hyperspectral curve features and surface matrix physical and chemical property parameters, and each correlation relationship corresponds to a preset regional type; Obtaining the hyperspectral information measured by the hyperspectral module; Determining the correlation relationship corresponding to the formation according to the regional type corresponding to the formation; Completing the monitoring of the surface matrix of the formation according to the hyperspectral information, the correlation relationship and the surface matrix physical and chemical property parameters.

[0008] The surface matrix sensor and monitoring method based on hyperspectral technology provided by the embodiments of the present application integrate a sensor module (measuring conventional surface matrix physical and chemical property parameters) and a hyperspectral module (obtaining formation spectral information), and cooperate with the communication module to upload data to the terminal module; a correlation relationship library between hyperspectral curve features and surface matrix physical and chemical property parameters under different regional types is pre-stored in the terminal module, and the corresponding correlation relationship is selected by matching the regional type of the target formation, and combined with real-time hyperspectral data and sensor parameters, multi-dimensional monitoring of the formation matrix is realized.

[0009] By embedding sensors in the formation in-situ, the destruction of the matrix structure caused by traditional manual drilling is avoided, significantly improving the representativeness and reliability of monitoring data, especially suitable for continuous monitoring of deep matrices below 2m; breaking through the limitation of traditional single-point sensors that can only obtain local parameters, using hyperspectral modules to capture the spectral characteristics of the entire formation, and combining with the correlation database of the terminal module to achieve spatial mapping of spectral data with multiple parameters such as moisture, salinity, and temperature, enhancing the monitoring dimension; pre-storing differential correlation relationships for the surface matrix characteristics in different regions (such as arid regions, wetlands) to solve the monitoring deviation problem caused by the "one-size-fits-all" model due to geological background differences, enhancing the technical universality; through cross-validation of hyperspectral inversion and sensor-measured data, providing data support with both depth and accuracy for land use and ecological restoration, especially accurately identifying the resource supply potential of deep matrices in extreme environments (such as drought). By deeply integrating hyperspectral technology with traditional sensors, a full-link monitoring system of "hardware embedding - data fusion - regional adaptation" is constructed, overcoming the technical bottlenecks of traditional methods in terms of destructiveness, monitoring depth, and parameter correlation, and filling the technical gap in intelligent monitoring of deep matrices.

[0010] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and do not limit this application. Brief Description of the Drawings

[0011] To more clearly illustrate the technical solutions of the embodiments of this application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0012] Figure 1 It is a schematic structural diagram of the first type of surface matrix sensor based on hyperspectral technology provided by an embodiment of this application; Figure 2 It is a bottom view of the first type of surface matrix sensor based on hyperspectral technology provided by an embodiment of this application; Figure 3 It is a schematic structural diagram of the second type of surface matrix sensor based on hyperspectral technology provided by an embodiment of this application; Figure 4 It is a schematic flow chart of the steps of a surface matrix monitoring method based on hyperspectral technology provided by an embodiment of this application; Figure 5 It is a schematic block diagram of the structure of a surface matrix monitoring device based on hyperspectral technology provided by an embodiment of this application; Figure 6 It is a schematic block diagram of the structure of the terminal module provided by an embodiment of this application.

[0013] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and do not limit this application. Detailed implementation manners

[0014] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are some, but not all, of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the scope of protection of this application.

[0015] The flowcharts shown in the accompanying drawings are merely illustrative, and do not necessarily include all contents and operations / steps, nor do they necessarily need to be executed in the described order. For example, some operations / steps can also be decomposed, combined, or partially merged. Therefore, the actual execution order may be changed according to the actual situation.

[0016] It should be understood that, for the convenience of clearly describing the technical solutions in the embodiments of the present invention, in the embodiments of the present invention, terms such as "first" and "second" are used to distinguish the same items or similar items with basically the same functions and effects. Those skilled in the art can understand that the terms "first" and "second" do not limit the quantity and execution order, and the terms "first" and "second" do not necessarily mean different.

[0017] As the basic substance that nurtures and supports various natural resources such as soil, forests, grasslands, water, and wetlands on the earth's surface, the physical and chemical properties of the surface substrate directly control the spatial distribution pattern of surface agricultural production and vegetation ecology. It is the space where the multi-sphere interactions on the earth are the most frequent, and it is also the material basis for the overall protection, systematic restoration, and comprehensive management of cultivated land and natural ecosystems. Traditional monitoring technologies focus on the point sampling of shallow soil parameters within 0 - 2m, and data is obtained by manually excavating and burying single-point sensors such as temperature, humidity, and salinity. Such methods have significant defects: First, the artificial drilling layout will damage the original structure of the formation, resulting in deviations in monitoring data due to soil disturbance; Second, the layout depth of the sensors is limited by the drilling process and it is difficult to penetrate dense layers or gravel layers to achieve deep substrate monitoring below 2m; Third, single-parameter monitoring cannot reflect the multi-parameter coupling mechanism. Especially under extremely arid conditions, the supporting role of the deep substrate in water transport and nutrient supply cannot be effectively characterized by traditional point data.

[0018] In recent years, research has shown that the surface matrix exhibits significant heterogeneity in the vertical profile. The water-salt migration law, cement chemical composition, and pore structure characteristics of the deep matrix below 2m have a decisive impact on plant root development and the water-holding capacity of the matrix. However, due to technical limitations, existing monitoring equipment can neither achieve non-destructive in-situ burial nor has the ability to establish correlation models between hyperspectral characteristics and multi-physical parameters, resulting in long-term reliance on laboratory core analysis for deep matrix monitoring, with problems such as poor timeliness and low spatial coverage.

[0019] Currently, there is an urgent need for a sensor that can penetrate deep strata, synchronously acquire multi-dimensional parameters, and have intelligent analysis capabilities.

[0020] To solve the above problems, please refer to Figures 1 to 3 , this application provides a surface matrix sensor based on hyperspectral technology, including: a sensor module 20, which is used to extend through the housing into the formation; a hyperspectral module 30, which extends through the housing into the formation; a communication module 10, which is communicatively connected to the sensor module and the hyperspectral module respectively; a terminal module, which is communicatively connected to the communication module. The terminal module stores multiple correlation relationships between hyperspectral curve characteristics and surface matrix physical and chemical property parameters, and each correlation relationship corresponds to a preset regional type; the terminal module acquires the hyperspectral information measured by the hyperspectral module; the terminal module determines the correlation relationship corresponding to the formation according to the regional type corresponding to the formation; the terminal module completes the monitoring of the surface matrix of the formation based on the hyperspectral information, the correlation relationship, and the surface matrix physical and chemical property parameters.

[0021] Specifically, this application provides a surface matrix sensor based on hyperspectral technology, aiming to solve problems existing in traditional surface matrix monitoring technologies such as destructive layout, limited monitoring depth, single parameter acquisition, and insufficient data representation capabilities. The sensor realizes multi-dimensional and high-precision monitoring of the vertical profile of the surface matrix (including below 2m deep) through non-destructive in-situ burial technology combined with hyperspectral analysis and synchronous acquisition of multi-physical parameters, and establishes a correlation model between hyperspectral characteristics and surface matrix physical and chemical property parameters based on intelligent analysis technology, providing real-time dynamic data support for land use and cultivated land protection, ecological restoration, and national scientific greening, etc.

[0022] The sensor module integrates a temperature and humidity sensor, a salinity sensor, a pressure sensor, and a porosity detection unit. By performing multi-parameter collaborative measurements, physical and chemical property data of the surface substrate (such as water content, salt concentration, pore pressure, cement distribution, etc.) can be obtained. A low-friction nano-coating can be applied to the surface of the probe corresponding to the sensor module, which can penetrate through the dense layer or gravel layer through hydraulic drive, and the maximum monitoring depth can reach more than 5m, avoiding the damage to the original formation structure caused by traditional drilling techniques.

[0023] The hyperspectral module includes a hyperspectral imaging unit (covering the 400 - 2500nm band) and a spectral analysis algorithm library. It directly contacts the surface substrate through an optical fiber probe to collect reflected spectral data. By analyzing the hyperspectral characteristics of the surface substrate (such as absorption peaks, reflectance, spectral slope), chemical components of the cement (such as carbonates, silicates), pore structure characteristics, and water migration trajectory information can be extracted. The hyperspectral probe and the sensor module are arranged coaxially to ensure the spatial consistency of the spectral data and physical parameters.

[0024] The communication module supports a hybrid transmission mode of wired (RS-485, optical fiber) and wireless (LoRa, NB-IoT), adapting to real-time data transmission in different terrain environments. Shielded cables and frequency band adaptive technologies are used to eliminate the influence of formation electromagnetic interference on signals.

[0025] The terminal module incorporates machine learning algorithms (such as random forest, convolutional neural network) to establish an association model between hyperspectral characteristics and physical and chemical property parameters of the surface substrate, supporting parameter inversion and prediction. An association relationship database for different geographical types (such as plain basin areas, mountain and hilly areas, and special geomorphic areas such as loess areas, alpine and high-latitude permafrost areas, karst areas, desert areas, coastal zones, etc.) is pre-stored, and the optimal model is automatically matched through geographical coordinates. For extreme drought or flood scenarios, combining real-time monitoring data with historical spectral curves, the water transport model and nutrient supply prediction results are dynamically corrected.

[0026] In some embodiments, a hydraulic propulsion device is used to vertically press the probes of the sensor module and the hyperspectral module into the surface matrix layer. The probe tip is made of tungsten carbide alloy and penetrates through the dense layer or gravel layer by rotary cutting and vibration-assisted penetration. A spiral flow guide groove is provided on the outer wall of the probe to reduce soil disturbance and ensure the integrity of the original formation structure. The sensor module starts the distributed acquisition mode and measures the temperature, humidity, salinity, and pore pressure layer by layer along the probe axis. After preprocessing by the built-in microprocessor, the data is transmitted to the communication module. The hyperspectral module continuously scans the formation profile at a rate of 100 frames per second, generates high-resolution spectral images, and uploads the original spectral data to the terminal module through optical fibers. The terminal module calls the preset regional type database (for example, the cement in the Loess Plateau is mainly calcium carbonate), and loads the corresponding hyperspectral feature-parameter correlation model. Using principal component analysis (PCA) for dimensionality reduction and partial least squares regression (PLSR) algorithm, the real-time hyperspectral data is coupled and analyzed with the measured parameters of the sensor to invert the water-salt migration rate, cement content, and pore connectivity index of the deep matrix. Based on the time series data, a thermal map of the dynamic distribution of moisture and salinity in the vertical profile is generated to identify potential areas at risk of drought stress or salinization. When an abnormal increase in the pore pressure of the deep matrix is detected, a geological disaster warning signal is triggered and pushed to the user terminal through the communication module.

[0027] Through the design of self-sharpening probes and low-friction coatings, minimize the disturbance during formation penetration, retain the original matrix structure, and ensure data authenticity. Synchronously collect physical parameters (temperature, humidity, salinity) and hyperspectral features (cement composition, pore structure) to reveal the multi-parameter interaction mechanism and improve the prediction accuracy of water and nutrient supply under extreme conditions. Based on the machine learning-based regional association model library, realize the intelligent inversion and dynamic correction of monitoring data to adapt to the different needs of different geological environments. The combination of the distributed sensor array and hyperspectral imaging technology provides continuous monitoring data of the vertical profile, replacing traditional laboratory core analysis, with a timeliness improvement of more than 80%.

[0028] This sensor can be further applied to fields such as agricultural precision irrigation (dynamic regulation of root water absorption depth), mine ecological restoration (assessment of tailings matrix stability), and urban underground space development (real-time monitoring of foundation bearing capacity), with broad technical migration value.

[0029] In some embodiments, the hyperspectral module includes: an optical fiber module that extends through the housing into the formation; a hyperspectral photosensitive chip connected to the optical fiber module; a hyperspectral lens connected to the hyperspectral photosensitive chip, the hyperspectral lens being disposed in the formation; a light source for supplementing light to the formation; wherein the hyperspectral lens measures the hyperspectral light information corresponding to the formation and sends the hyperspectral light information to the hyperspectral photosensitive chip, and the hyperspectral photosensitive chip generates the hyperspectral information.

[0030] The hyperspectral module of the embodiment is composed of the following core components for realizing non-destructive in-situ acquisition of hyperspectral data of the formation.

[0031] The optical fiber module uses a multi-mode quartz fiber bundle (diameter ≤ 2 mm), which is externally wrapped with a compression-resistant ceramic sleeve and extends along the sensor housing to the target depth of the formation (such as below 2 m). A micro prism is integrated at the end of the fiber bundle, and the scattered light of the formation profile is guided to the hyperspectral photosensitive chip through the principle of total internal reflection.

[0032] The hyperspectral photosensitive chip is equipped with a CMOS image sensor and a spectral splitting element (such as a liquid crystal tunable filter LCTF or a grating), and supports high-resolution spectral splitting in the wavelength range of 360 nm to 1100 nm. An analog-to-digital conversion unit is built into the chip to convert the received optical signal into digital spectral curve data.

[0033] The hyperspectral lens uses an ultra-wide-angle fixed-focus lens (focal length 5 mm, F value 1.8), which is embedded in the formation and coaxially aligned with the optical fiber module to ensure that the light collection range covers the vertical section of the formation profile. The lens surface is coated with an anti-fouling coating to prevent soil particles from attaching and affecting the light transmittance.

[0034] The light source integrates a high-brightness LED array (wavelength covering 360 - 1100 nm), emits broadband light through pulse modulation, and supplements light to the formation evenly. The light source module is arranged at a 45° angle with the hyperspectral lens to avoid direct light interfering with the collection of reflected signals.

[0035] The supplementary light rays are emitted into the formation through the light source, penetrate the soil matrix and stimulate the specific spectral responses of components such as cements and pore water. The hyperspectral lens receives the reflected light from the formation and transmits it to the hyperspectral photosensitive chip through the optical fiber module. The photosensitive chip decomposes the optical signal into continuous spectral band data with a resolution of 1 nm, generates hyperspectral information and uploads it to the terminal module.

[0036] The ultra-fine design of the optical fiber module and the micro lens (diameter ≤ 5 mm) can penetrate the dense layer without drilling, realizing non-destructive spectral acquisition of deep matrix below 2 m. Compared with traditional core sampling, the data timeliness is improved by 90%, and the spatial resolution reaches the millimeter level.

[0037] Covering the visible to near-infrared range through the 360 - 1100nm band, it can simultaneously detect organic matter (absorption peak at 450nm), moisture (absorption peak at 1450nm), and minerals (such as the characteristic spectrum of clay minerals at 2200nm). The 1nm spectral resolution can distinguish the subtle spectral differences between carbonates and silicates in cements, improving the accuracy of component inversion. The combination of pulsed light source and synchronous phase-locked technology effectively suppresses ambient light noise and ensures the signal-to-noise ratio of weak reflection signals in deep layers (SNR≥30dB).

[0038] Exemplarily, it further includes: a convex lens, and the light rays emitted by the light source and irradiating the formation are converged by the convex lens into the hyperspectral lens.

[0039] On the basis of the above embodiments, the optical path efficiency is optimized by adding a convex lens assembly. The convex lens is designed as an aspherical quartz convex lens (diameter 10mm, radius of curvature 15mm), installed between the light source and the formation, and collimates and focuses the supplementary light rays. The lens is coated with an antireflection film (reflectivity ≤ 0.5%) to reduce the light energy loss in the 360 - 1100nm band.

[0040] The light rays of the light source are converged into a parallel beam by the convex lens and perpendicularly incident on the surface of the formation, improving the illumination uniformity. The reflected light from the formation is secondarily focused by the convex lens onto the hyperspectral lens, increasing the effective signal acquisition angle (field of view angle ≥ 60°).

[0041] The scattered light emitted by the light source is collimated by the convex lens to form high-intensity parallel light, which penetrates the deep matrix of the formation. The scattered light reflected from the formation is converged by the convex lens onto the hyperspectral lens, reducing the optical path attenuation and enhancing the signal intensity.

[0042] The convex lens compresses the divergence angle of the light source from 120° to 30°, increasing the light intensity density by 4 times, and increasing the intensity of the reflected signal in deep layers (such as 5m) by more than 50%. The focused optical path reduces the scattering interference of soil particles, and the signal-to-noise ratio (SNR) of the spectral data is increased from 30dB to 40dB. The concentration of light energy extends the effective detection depth, breaking through the penetration limit of traditional hyperspectral technology.

[0043] It should be noted that in some embodiments, the range of the photosensitive wavelengths of the hyperspectral photosensitive chip and the light source is 360nm to 1100nm, and the resolution of the convex lens is 1nm.

[0044] By further defining the spectral parameters, including strictly defining the photosensitive wavelength of the hyperspectral photosensitive chip and the light source to be 360 - 1100 nm, covering the visible light (VIS) and the near-infrared (NIR) I region (700 - 1100 nm). A spectral resolution of 1 nm is achieved through grating spectroscopy, and the number of spectral channels reaches 740 (one channel for every 1 nm within the range of 360 - 1100 nm). The light source adopts a combination of blue LED (peak 450 nm) and near-infrared LED (peak 850 nm) to cover the key absorption bands. The photosensitive chip selects a back-illuminated CMOS, and the efficiency is ≥80% in the wavelength band of 400 - 1000 nm.

[0045] Heavy metal ions (such as Fe 3+ are identified in the 360 - 450 nm wavelength band (absorption peak at 380 nm), chlorophyll and organic matter are detected in the 450 - 700 nm wavelength band, and water content and mineral content are analyzed in the 700 - 1100 nm wavelength band. The 1 nm resolution can distinguish nitrate (absorption peak at 357 nm) and nitrite (absorption peak at 360 nm) in the soil, and the detection limit is as low as 0.1 ppm. Defining the spectral range reduces the complexity of data processing and improves the operation efficiency of the inversion algorithm of the terminal module by 70%.

[0046] It should be noted that in some embodiments, it further includes: a reflector, the reflector is arranged in the water body of the formation, the light of the light source irradiates the water body and then is reflected back to the hyperspectral lens by the reflector, the hyperspectral lens sends the received water body reflected light to the hyperspectral photosensitive chip, and the hyperspectral photosensitive chip generates water quality monitoring information.

[0047] By adding a reflector assembly to expand the water quality monitoring function, the reflector is arranged with a corrosion-resistant titanium alloy reflector (reflectivity ≥95%), embedded in the formation water body (such as groundwater or pore water), and the mirror surface forms a 45° angle with the light source and the hyperspectral lens. Microchannels are etched on the surface of the reflector to guide the water body flow to avoid sediment coverage.

[0048] After the light of the light source irradiates the water body, it is reflected by the reflector to the hyperspectral lens, and the hyperspectral data of the formation matrix and water quality are collected synchronously. The light emitted by the light source penetrates the water body and stimulates the characteristic spectral response of the dissolved substances in the water (such as COD, ammonia nitrogen). The water body reflected light is guided by the reflector to the hyperspectral lens, and the photosensitive chip generates water quality monitoring information (such as turbidity, pollutant concentration).

[0049] By obtaining the formation matrix parameters (water and salt migration) and water quality data (pollutant distribution) simultaneously through a single measurement, the all-element analysis of the ecosystem is realized. Oil pollutants in the water body (C - H bond absorption peak) are detected in the 1100 nm wavelength band, and the detection sensitivity reaches 0.01 mg / L. Reusing the hyperspectral module to achieve water quality monitoring without additional deployment of sensors reduces the equipment cost by 40%.

[0050] In some embodiments, one end of the housing connected to the formation includes a plurality of sensor monitoring holes and a hyperspectral lens hole, and the sensor monitoring holes are arranged around the hyperspectral lens hole; the sensor module includes a plurality of sensor probes, the sensor probes are arranged in the formation through the sensor monitoring holes, and the hyperspectral lens of the hyperspectral module is arranged in the formation through the hyperspectral lens hole.

[0051] In this embodiment, an optimized design is carried out for the sensor housing structure. Through the porous surrounding layout, the collaborative layout of the hyperspectral module and the multi-physical parameter sensors is realized. The bottom of the housing is a conical piercing end (made of titanium alloy), and one hyperspectral lens hole and a plurality of sensor monitoring holes are arranged at the end in contact with the formation. The sensor monitoring holes are distributed around the hyperspectral lens hole in a concentric circle manner, with 6 in number (evenly distributed at an interval of 60°), and the aperture is 5 mm, which is used to arrange the sensor probes. The hyperspectral lens hole is located at the center of the housing, with an aperture of 8 mm, and a waterproof sealing ring is arranged inside to fix the hyperspectral lens.

[0052] The sensor module includes, for example, 5 probes (1 for each of the temperature and humidity, salinity, pressure, pH value, and redox potential sensors), which are respectively inserted into the formation vertically or horizontally through the sensor monitoring holes. The number and type of the sensor probes in the embodiments of the present application are not limited, and can be specifically selected according to actual requirements. The length of the probe is adjustable (1 - 5 m), and a segmented design is adopted. A micro spiral drill bit is arranged at the end of each segment of the probe to assist in penetrating the dense layer.

[0053] The hyperspectral lens is vertically or horizontally embedded into the formation through the central hole, and a transparent sapphire protection window is arranged at the front end of the lens to prevent soil particles from contaminating the optical elements. The field of view angle of the lens is 120°, covering a cylindrical monitoring area centered on the housing with a radius of 15 cm, and spatially coinciding with the monitoring points of the sensor probes.

[0054] After the piercing end of the housing is pressed into the ground surface, the sensor probes and the hyperspectral lens are synchronously inserted into the target depth of the formation (such as 3 m). The sensor probes collect the physical and chemical parameters of the formation, and the hyperspectral lens collects the spectral data at the same spatial position, and uploads them to the terminal module through the communication module. The terminal module fuses the multi-parameter data and the hyperspectral information based on the spatial coordinate matching to construct a three-dimensional matrix model of the vertical or horizontal profile.

[0055] The circumferential arrangement of the sensor probe and the hyperspectral lens ensures that the physical parameter and spectral data acquisition points are located in the same spatial area, eliminating the position deviation of multi-point data in traditional technologies and improving the accuracy of the parameter correlation model by more than 30%. The piercing end of the housing provides guidance and support for the sensor probe. The probe does not require independent drilling, and the resistance is reduced by 60% when penetrating the gravel layer, and the maximum deployment depth can reach 6m. The hyperspectral lens is located at the center of the housing and is protected by the surrounding annular structure of the sensor probe to avoid damage to the optical elements caused by formation extrusion or gravel impact.

[0056] Exemplarily, as Figure 1 shown, the hyperspectral lens and the sensor probe are vertically arranged in the formation; or, as Figure 3 shown, the hyperspectral lens and the sensor probe are horizontally arranged in the formation.

[0057] Based on the above embodiments, by further defining the deployment directions of the hyperspectral lens and the sensor probe, two optional modes are provided: Vertical deployment mode: The hyperspectral lens is vertically inserted into the formation along the central axis of the housing, and the sensor probe is deployed in the same direction to form a linear monitoring array along the depth direction. It is suitable for monitoring the longitudinal distribution laws of water-salt migration and cementitious precipitate.

[0058] Horizontal deployment mode: The hyperspectral lens is horizontally inserted into the formation (parallel to the ground), and the sensor probe is deployed radially horizontally to form a radial monitoring network centered on the housing. It can be installed as Figure 3 described by the surface matrix installation tool 200. It is suitable for evaluating horizontal direction matrix parameter changes such as slope stability and underground cavity expansion.

[0059] In the vertical mode, the probe and the lens are rigidly connected, and the insertion angle (0 - 90°) is adjusted through the slide rail inside the housing.

[0060] In the horizontal mode, the piercing end of the housing is changed to a flat bottom structure, and a lateral piercing edge is provided at the end of the probe to minimize the formation disturbance during horizontal penetration.

[0061] The advantages of the vertical deployment mode lie in the improvement of longitudinal resolution: Spectral and physical parameters are synchronously collected every 10 cm along the depth direction, and the interface position between the cemented layer and the aquifer below 2m can be accurately identified (error ≤ 5 cm). Root development analysis: Combining vertical spectral data with temperature and humidity parameters can invert the water absorption behavior of plant roots in the deep layer and guide precise irrigation.

[0062] The advantages of the horizontal layout mode are as follows: Lateral stability assessment: The shear strength of the formation and the pore pressure gradient can be calculated from the horizontal cross-section monitoring data to warn of landslide risks (sensitivity increased by 40%). Underground space modeling: Combining horizontal spectral data with porosity parameters to generate a 3D underground structure model for the stability analysis of mine roadways or subway tunnels. Flexible deployment ability: Quickly switch the layout direction according to the monitoring target (vertical profile or horizontal cross-section) without replacing hardware equipment, significantly enhancing field adaptability.

[0063] In some embodiments, the association relationship includes hyperspectral inversion parameters, and the terminal module inverses the hyperspectral information according to the hyperspectral inversion parameters to obtain the physical and chemical property parameters of the surface matrix corresponding to the hyperspectral information.

[0064] The core of the embodiment is to perform high-precision mapping between hyperspectral information and the physical and chemical property parameters of the surface matrix through the collaborative inversion of machine learning and physical models.

[0065] Perform spatial-spectral joint denoising on the original hyperspectral data by using a convolutional neural network (CNN). Specifically, construct a three-dimensional convolutional kernel (size 5×5×5, covering the spatial neighborhood and adjacent bands), extract the noise pattern through multiple layers of convolution and generate a residual map, and remove the noise from the original data (signal-to-noise ratio increased by ≥15 dB).

[0066] Introduce a channel attention module (SE Block) to calculate the contribution weights of each band to the target parameters (such as moisture content, cement concentration), screen out the characteristic bands (such as the moisture-sensitive band at 1450 nm, the clay mineral characteristic band at 2200 nm), and remove redundant bands (data volume reduced by 50%).

[0067] The construction of the surface matrix inversion model includes: Hybrid architecture design: including Random Forest (RF): used to process discrete parameters (such as pore type classification, cement type identification), and improve the classification robustness through the voting mechanism of multiple decision trees. Deep Belief Network (DBN): stacked by multiple Restricted Boltzmann Machines (RBMs), used for non-linear regression of continuous parameters (such as water-salt concentration, porosity), and solve the small-sample overfitting problem through unsupervised pre-training. Model training: Based on a regional geological sample library (containing 100,000 pairs of hyperspectral-laboratory measured parameters), use an adaptive learning rate optimizer (such as AdamW) for end-to-end training, and the loss function combines the mean square error (MSE) and KL divergence to balance the regression and classification tasks.

[0068] Embedding of physical inversion equation: Introduce a physical driving module after the output layer, which contains a non-linear compensation equation fitted from regional geological data. For example, the form of the moisture inversion equation is: θ = α ×R 1450 + β × ( R 1450 / R 950) 2 + γ × ln ( T ) ; where R 1450, R 950 are the reflectivities corresponding to bands 1450 and 950, T is the measured temperature value, and α, β, γ are region - adaptive weight coefficients. Based on the measured data, the weight coefficients are iteratively optimized through the genetic algorithm (GA) to adapt to the differences in substrate characteristics in different regions. For example, in the Loess Plateau region, since the cement is rich in calcium carbonate, β β weight needs to be increased to compensate for the interference of mineral absorption. Monte Carlo simulation: Gaussian noise (with a standard deviation equal to the reciprocal of the signal - to - noise ratio after pre - processing) is added to the input of the inversion model, and 1000 random samplings are carried out. The standard deviation of the output parameters is statistically calculated as the confidence index. For example, the inversion result of the moisture content can be expressed as 25.3% ± 1.2% (confidence level 95%).

[0069] Machine learning and physical models work together to solve the over - fitting problem of traditional pure data - driven models. The inversion errors of key parameters such as moisture and salinity are ≤ 5% (compared with laboratory measurements), and the accuracy is improved by 40% compared with single models. The non - linear compensation term effectively eliminates the systematic deviation caused by differences in regional mineral components. For example, the interference of clay minerals on the moisture characteristic band is reduced by 90%. The genetic algorithm calibrates the weight coefficients in real - time, enabling the same sensor to adapt to multiple geological environments (such as alluvial plains and karst landforms) without retraining the model, and the deployment efficiency is increased by 70%. Measured parameters (such as temperature T) are embedded in the physical equation to achieve multi - modal information fusion of hyperspectral data and sensor modules, and the spatial consistency of the inversion results is increased by 50%. Monte Carlo simulation provides the parameter uncertainty interval, guiding users to distinguish reliable data from noise interference. Especially in extreme environments (such as enhanced substrate heterogeneity after heavy rain), confidence assessment can avoid misjudgment.

[0070] Exemplarily, the inversion of the hyperspectral information according to the hyperspectral inversion parameters to obtain the physicochemical property parameters of the surface substrate corresponding to the hyperspectral information includes: parsing the preprocessed hyperspectral information according to a pre-trained surface substrate inversion model, the surface substrate inversion model being constructed based on a machine learning algorithm, the input layer receiving the preprocessed spectral feature vector, and the output layer generating a set of inversion parameters; the preprocessing includes performing spectral noise reduction on the original hyperspectral data using a convolutional neural network and extracting the characteristic bands associated with the physicochemical property parameters of the surface substrate through an attention mechanism; inputting the set of inversion parameters into a correlation relationship optimization module, matching the corresponding physical inversion equation based on the regional type, and using a genetic algorithm to dynamically calibrate the weight coefficients in the inversion equation; constructing a region-adaptive inversion matrix according to the calibrated weight coefficients, and mapping the hyperspectral information to the target surface substrate physicochemical property parameter space through matrix operations to generate the physicochemical property parameters of the surface substrate including multiple parameter confidence levels; wherein, the machine learning algorithm adopts a hybrid architecture combining random forest and deep belief network, the physical inversion equation includes a non-linear compensation term trained by a regional geological sample library, and the parameter confidence level is quantified by Monte Carlo simulation for uncertainty.

[0071] Based on the embodiment, the hybrid model architecture and the dynamic optimization process are refined. By using the classification probability output by RF (such as the probability of 80% for the pore type of "connected type") as one of the input features of DBN, the physical constraints for continuous parameter inversion are enhanced. First, fix the RF model parameters and train DBN to fit the continuous parameters; then jointly fine-tune RF and DBN to minimize the total loss function. Extract key variables from the regional geological sample library and automatically generate the equation form through symbolic regression. For example, the salinity inversion equation may include polynomial combinations of reflectance ratios such as R1650 / R550 and temperature and pressure. Non-linear compensation terms (such as exponential functions, piecewise functions) are embedded in the equation to correct the deviation of the machine learning model in the extreme value interval.

[0072] The genetic algorithm optimization process includes: population initialization: the initial value ranges of the weight coefficients α, β, γ are set according to the regional type (such as α ∈ [0.1, 0.3] in desert areas). The fitness function is defined as the weighted sum of the correlation coefficient (R²) and the error (RMSE) between the inversion parameters and the measured values. The simulated binary crossover (SBX) and non-uniform mutation operators are used to ensure the convergence of the weight coefficients within a reasonable range.

[0073] Add noise (standard deviation = 1 / SNR) to the preprocessed hyperspectral data to generate a perturbed data set. Statistically analyze the distribution of the inversion results of all perturbed samples, and calculate the 95% confidence interval and the coefficient of variation (CV).

[0074] The cascaded training of RF and DBN takes into account both classification and regression tasks. The accuracy of pore type identification is ≥92%, and the salinity inversion error is ≤3%. The physical equation and the data-driven model complement each other, and high inversion accuracy (error ≤8%) can still be maintained in areas with scarce samples (such as permafrost zones). Symbolic regression automatically mines physical laws from data, reduces the dependence on manual experience, and the interpretability of the equation is increased by 60%. The genetic algorithm completes the optimization of weight coefficients within 10 seconds, supporting the sensor to quickly adapt to new regions in mobile monitoring scenarios (such as on-vehicle landslide inspections). The parameter confidence level is directly related to the warning threshold. For example, when the lower limit of the confidence interval of the surface matrix shear strength inversion value is lower than the safety threshold, a red warning is immediately triggered.

[0075] In some embodiments, the sensor probe is made of an alloy material, and the alloy material includes at least any one of stainless steel and zinc. The use of the alloy material enhances the durability of the sensor probe and extends its service life. At the same time, it improves the corrosion resistance of the probe in harsh environments and is suitable for various environmental monitoring. The sensor probe adopts a zinc-nickel alloy coating (with a thickness of 15 - 20μm) plated on the surface of a SUS316L stainless steel substrate, and the zinc content is controlled at 12 - 15wt% to optimize the corrosion resistance. The tip of the probe is designed as a triangular pyramid nanostructure (cone angle 60°, surface roughness Ra ≤0.8μm), and the surface activity is enhanced through an electro-chemical polishing process. The conical geometric design reduces the insertion resistance of the probe and can still ensure the measurement accuracy in compacted soil; the sacrificial anode effect of zinc extends the service life of the probe to more than 5 years.

[0076] In some embodiments, as Figure 3 shown, it further includes: a gas sensor 40. The hyperspectral sensor and the gas sensor are connected to the terminal module and are used to send the measured hyperspectral information and gas information to the terminal module. The terminal module analyzes the measured parameters, hyperspectral information, and gas information according to a preset analysis algorithm to generate the physical and chemical property parameters of the surface matrix. These two sensors are connected to the terminal module and are used to send the measured hyperspectral information and gas information to the terminal module. The terminal module analyzes the measured parameters, hyperspectral information, and gas information according to a preset analysis algorithm to generate the physical and chemical property parameters of the surface matrix. The addition of the hyperspectral sensor and the gas sensor enables the system to monitor more physical and chemical property parameters of the surface matrix, such as soil organic matter value, etc. It increases the richness of data and provides more data support for environmental monitoring and geological disaster warning.

[0077] A hyperspectral imaging module (wavelength range 400 - 2500 nm, spectral resolution 3 nm) and a MEMS gas sensor array (including a PID photoionization detector and an NDIR carbon dioxide sensor) are respectively integrated at both ends of the probe array. The hyperspectral data is preprocessed by SG smoothing - first derivative and then input into a random forest model for organic matter content inversion.

[0078] Exemplarily, as Figure 3 shown, the hyperspectral sensor and the gas sensor are respectively located at both ends of a plurality of the sensor probes.

[0079] The hyperspectral sensor and the gas sensor are distributed at both ends of a plurality of sensor probes to achieve more comprehensive monitoring. The distribution of the sensors optimizes the monitoring layout and improves the monitoring efficiency. The sensors distributed at both ends can cover a wider monitoring area and improve the comprehensiveness of the monitoring.

[0080] The hyperspectral module and the gas sensor are respectively installed at the north and south poles of the probe array, and the distance is maintained above 50 mm. The hyperspectral probe faces outward at an inclination angle of 45°, and the gas collection port is designed as a cyclone separation structure to avoid soil particle pollution.

[0081] In some embodiments, as Figure 3 shown, a plurality of the sensor probes are arranged in a straight line on the terminal module.

[0082] The sensor probes arranged in a straight line make the monitoring more orderly, facilitating management and maintenance. It optimizes the space utilization and reduces the floor area of the device. The sensor probes arranged in a straight line are arranged linearly along the axis, and the spacing follows the principle such as λ / 4 (λ is the working wavelength of each sensor), and the minimum spacing is 15 mm. The probe base adopts a polyetheretherketone (PEEK) insulating frame, and an EMI shielding layer (copper mesh coverage rate ≥ 95%) is arranged on the surface of the frame. The linear layout controls the diameter of the device within the range applicable to the standard drilling hole diameter; the wavelength - adapted spacing eliminates the electromagnetic coupling interference between sensors and improves the signal - to - noise ratio; the shielding structure attenuates the external 50 Hz power frequency interference to - 60 dB.

[0083] Please refer to Figure 4 , Figure 4 which is a schematic flowchart of a surface matrix monitoring method based on hyperspectral technology provided by an embodiment of the present application. The surface matrix monitoring method based on hyperspectral technology can be implemented by the terminal module of the surface matrix sensor provided by any embodiment of the present application, and the terminal module can be deployed on a single server or a server cluster. It can also be deployed on a handheld terminal, a laptop computer, a wearable device or a robot, etc.

[0084] As Figure 4As shown, the provided method for monitoring surface matrix based on hyperspectral technology includes steps S101 to S104. Details are as follows: Step S101. Obtain the correlation relationships corresponding to the hyperspectral curve features and the physical and chemical property parameters of the surface matrix, where each of the correlation relationships corresponds to a preset regional type; Step S102. Obtain the hyperspectral information measured by the hyperspectral module; Step S103. Determine the correlation relationship corresponding to the formation according to the regional type corresponding to the formation; Step S104. Complete the monitoring of the surface matrix of the formation based on the hyperspectral information, the correlation relationship, and the physical and chemical property parameters of the surface matrix.

[0085] It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the above-described method for monitoring the surface matrix based on hyperspectral technology and the specific working processes of each step can refer to the corresponding processes in the embodiments of the hyperspectral technology-based surface matrix sensor described above, and will not be elaborated here.

[0086] Please refer to Figure 5 as shown Figure 5 is a schematic structural diagram of a surface matrix monitoring device 300 based on hyperspectral technology provided by an embodiment of the present application. The surface matrix monitoring device 300 based on hyperspectral technology is used to execute the steps of the method for monitoring the surface matrix based on hyperspectral technology shown in the above embodiments. The surface matrix monitoring device 300 based on hyperspectral technology can be a single server or a server cluster, or the surface matrix monitoring device 300 based on hyperspectral technology can be a terminal, and the terminal can be a handheld terminal, a laptop computer, a wearable device, or a robot, etc.

[0087] As Figure 5 shown, the surface matrix monitoring device 300 based on hyperspectral technology includes: A parameter acquisition unit 301, configured to obtain the correlation relationships corresponding to the hyperspectral curve features and the physical and chemical property parameters of the surface matrix, where each of the correlation relationships corresponds to a preset regional type; An information acquisition unit 302, configured to obtain the hyperspectral information measured by the hyperspectral module; A relationship acquisition unit 303, configured to determine the correlation relationship corresponding to the formation according to the regional type corresponding to the formation; A monitoring completion unit 304, configured to complete the monitoring of the surface matrix of the formation based on the hyperspectral information, the correlation relationship, and the physical and chemical property parameters of the surface matrix.

[0088] It should be noted that those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described surface substrate monitoring device based on hyperspectral technology and each module can refer to the corresponding processes in the embodiments of the surface substrate sensor based on hyperspectral technology described in the above embodiments, and will not be repeated here.

[0089] The above-described surface substrate monitoring method based on hyperspectral technology can be implemented in the form of a computer program, and this computer program can run on a device as shown in Figure 5 the figure.

[0090] Please refer to Figure 6 , Figure 6 which is a schematic block diagram of the structure of the terminal module provided in the embodiment of the present application. The terminal module includes a processor, a memory, and a network interface connected through a device bus. Among them, the memory can include a storage medium and an internal memory.

[0091] The storage medium can store an operating device and a computer program. This computer program includes program instructions, and when the program instructions are executed, the processor can be made to execute any surface substrate monitoring method based on hyperspectral technology.

[0092] The processor is used to provide computing and control capabilities to support the operation of the entire terminal module.

[0093] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When this computer program is executed by the processor, the processor can be made to execute any surface substrate monitoring method based on hyperspectral technology.

[0094] This network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art can understand that Figure 6 the structure shown in is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the terminal to which the solution of the present application is applied. The specific terminal module may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0095] It should be understood that the processor can be a Central Processing Unit (CPU), and the processor can also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0096] Among them, in one embodiment, the processor is used to run a computer program stored in a memory to implement the following steps: Obtain the correlation relationships corresponding to the hyperspectral curve features and the physical and chemical property parameters of the surface matrix, and each of the correlation relationships corresponds to a preset regional type; Obtain the hyperspectral information measured by the hyperspectral module; Determine the correlation relationship corresponding to the formation according to the regional type corresponding to the formation; Complete the monitoring of the surface matrix of the formation according to the hyperspectral information, the correlation relationship, and the physical and chemical property parameters of the surface matrix.

[0097] It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described steps can refer to the corresponding processes in the embodiments of the surface matrix sensor based on hyperspectral technology described in the above embodiments, and will not be elaborated here.

[0098] The present application also provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor is caused to implement the steps of the surface matrix monitoring method based on hyperspectral technology as described above.

[0099] Among them, the computer-readable storage medium can be an internal storage unit of the terminal module described in the foregoing embodiments, such as the hard disk or memory of the terminal module. The computer-readable storage medium can also be an external storage device of the terminal module, such as a plug-in hard disk equipped on the terminal module, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc.

[0100] As described above, it is only the specific implementation manner of the present application. However, the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present application.

Claims

1. A surface matrix sensor based on hyperspectral technology, characterized in that, Comprising: A sensor module, which is used to extend into the formation through the housing; A hyperspectral module, which extends into the formation through the housing; A communication module, which is communicatively connected to the sensor module and the hyperspectral module respectively; A terminal module, which is communicatively connected to the communication module. The terminal module stores a plurality of correlation relationships between hyperspectral curve features and surface matrix physical and chemical property parameters, and each correlation relationship corresponds to a preset regional type; the terminal module acquires the hyperspectral information measured by the hyperspectral module; the terminal module determines the correlation relationship corresponding to the formation according to the regional type corresponding to the formation; the terminal module completes the monitoring of the surface matrix of the formation according to the hyperspectral information, the correlation relationship, and the surface matrix physical and chemical property parameters.

2. The surface matrix sensor according to claim 1, characterized in that, The hyperspectral module includes: An optical fiber module, which extends into the formation through the housing; A hyperspectral photosensitive chip, which is connected to the optical fiber module; A hyperspectral lens, which is connected to the hyperspectral photosensitive chip and is disposed in the formation; A light source, which is used to supplement light to the formation; Wherein, the hyperspectral lens measures the hyperspectral light information corresponding to the formation and sends the hyperspectral light information to the hyperspectral photosensitive chip, and the hyperspectral photosensitive chip generates the hyperspectral information.

3. The surface matrix sensor according to claim 2, characterized in that, Further comprising: A convex lens, and the light from the light source irradiating the formation is converged into the hyperspectral lens by the convex lens.

4. The surface matrix sensor according to claim 3, characterized in that, The photosensitive wavelength range corresponding to the hyperspectral photosensitive chip and the light source is 360nm to 1100nm, and the resolution of the convex lens is 1nm.

5. The surface matrix sensor according to claim 4, characterized in that, Further comprising: A reflector, which is disposed in the water body of the formation. The light from the light source irradiates the water body and is reflected back to the hyperspectral lens by the reflector. The hyperspectral lens sends the received water body reflected light to the hyperspectral photosensitive chip, and the hyperspectral photosensitive chip generates water quality monitoring information.

6. The surface matrix sensor according to claim 1, characterized in that, One end of the housing connected to the formation includes a plurality of sensor monitoring holes and a hyperspectral lens hole, and the sensor monitoring holes are arranged around the hyperspectral lens hole; The sensor module includes a plurality of sensor probes, the sensor probes are disposed in the formation through the sensor monitoring holes, and the hyperspectral lens of the hyperspectral module is disposed in the formation through the hyperspectral lens hole.

7. The surface matrix sensor according to claim 6, characterized in that, The hyperspectral lens and the sensor probe are vertically disposed in the formation; or, the hyperspectral lens and the sensor probe are horizontally disposed in the formation.

8. The surface matrix sensor according to claim 1, characterized in that, The correlation relationship includes hyperspectral inversion parameters, and the terminal module inverses the hyperspectral information according to the hyperspectral inversion parameters to obtain the surface matrix physical and chemical property parameters corresponding to the hyperspectral information.

9. The surface matrix sensor according to claim 8, characterized in that, The inversing the hyperspectral information according to the hyperspectral inversion parameters to obtain the surface matrix physical and chemical property parameters corresponding to the hyperspectral information includes: Analyze the preprocessed hyperspectral information according to a pre-trained surface substrate inversion model. The surface substrate inversion model is constructed based on a machine learning algorithm. The input layer receives the preprocessed spectral feature vectors, and the output layer generates a set of inversion parameters. The preprocessing includes performing spectral noise reduction on the original hyperspectral data using a convolutional neural network and extracting feature bands associated with the physical and chemical property parameters of the surface substrate through an attention mechanism. Input the set of inversion parameters into a correlation relationship optimization module, match the corresponding physical inversion equation based on the regional type, and use a genetic algorithm to dynamically calibrate the weight coefficients in the inversion equation. Construct a region-adaptive inversion matrix according to the calibrated weight coefficients, map the hyperspectral information to the target physical and chemical property parameter space of the surface substrate through matrix operations, and generate the physical and chemical property parameters of the surface substrate including multiple parameter confidence levels. Among them, the machine learning algorithm uses a hybrid architecture combining random forest and deep belief network. The physical inversion equation includes a non-linear compensation term trained by a regional geological sample library, and the parameter confidence level is quantified for uncertainty through Monte Carlo simulation.

10. A surface matrix monitoring method based on hyperspectral technology, characterized in that, Applied to the surface substrate sensor based on hyperspectral technology according to any one of claims 1-9. The method includes: Obtain the correlation relationships corresponding to the hyperspectral curve features and the physical and chemical property parameters of the surface substrate, and each correlation relationship corresponds to a preset regional type. Obtain the hyperspectral information measured by the hyperspectral module. Determine the correlation relationship corresponding to the formation according to the regional type corresponding to the formation. Complete the monitoring of the surface substrate of the formation according to the hyperspectral information, the correlation relationship, and the physical and chemical property parameters of the surface substrate.

Images