Dam termite dynamic gas sensing and rapid surveying method based on true value calibration

Abstract

The application discloses a kind of dam termite dynamic gas sensing and rapid survey method based on true value calibration, comprising: establishing the correctable reference frame including general multi-gas reference and multi-gas environment correlation model;Obtain the field true value point data in target area, calculate the standard environmental equivalent concentration of field true value point data, and derive the regional correction coefficient accordingly;The general multi-gas reference is corrected using the regional correction coefficient, and the regional multi-gas dynamic reference adapted to the target area is generated;In the target area, perform survey detection, collect survey point data in real time, and convert survey point data based on multi-gas environment correlation model;The converted survey point data is judged based on the regional multi-gas dynamic reference, and the position information of suspected termite nest area is output.The application solves the problem that traditional gas detection lacks depth inversion mechanism and passive scanning efficiency is low, and improves the positioning accuracy and environmental adaptability of dam termite survey.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of dam safety monitoring, and particularly relates to a dam termite dynamic gas sensing and rapid survey method based on true value calibration. BACKGROUND

[0002] As a core water conservancy facility for resisting floods and guaranteeing water resource allocation, the structural integrity of a dam is directly related to the safety of a river basin. Termites use dam soil as a nesting medium and form a network of channels through erosion, which greatly reduces the soil density and easily causes piping, leakage and even dam failure, so early and accurate detection of hidden termite nests in the dam is a key link to ensure the safe operation of water conservancy projects.

[0003] The current mainstream termite detection technology mainly includes traditional patrol relying on manual experience to identify surface features, and geophysical detection technology using high-density electrical method and ground penetrating radar to detect internal cavities. With the development of gas sensing technology, monitoring methods based on characteristic gases such as carbon dioxide and methane produced by termite metabolism have gradually emerged. Currently, fixed sensor networking monitoring or portable equipment is mainly used for single-point sampling, and whether there is termite damage is qualitatively judged by detecting whether the gas concentration in the area exceeds the standard.

[0004] However, the existing gas detection technology has technical problems of source-field mapping loss and passive blind sensing in deep quantification and dynamic adaptability. Specifically, the existing method only uses surface detection concentration as the basis for judgment, ignores the nonlinear attenuation law (Fick's law) of gas transmission in soil pore medium, lacks the physical mechanism of inverting underground termite nest source concentration and depth from the surface concentration field, and cannot accurately locate the depth of the nest; at the same time, the detection path is mainly passive scanning along the preset grid or spiral line, and cannot use real-time concentration gradient vectors for active source tracking similar to biological olfaction; in addition, the rigid universal judgment standard cannot adapt to the differences in soil background noise and environmental background of different dam areas, resulting in high false alarm rate and difficulty in continuous optimization in complex dynamic environment. SUMMARY

[0005] The application aims to provide a dam termite dynamic gas sensing and rapid survey method based on true value calibration to solve the above problems in the prior art.

[0006] Technical scheme, the dam termite dynamic gas sensing and rapid survey method based on true value calibration comprises:

[0007] A revisable benchmark framework including a universal multi-gas benchmark and a multi-gas environment correlation model is established, and the universal multi-gas benchmark includes multi-gas basic parameters in a standard environment;

[0008] Obtaining field true value point data in the target area, calculating the standard environmental equivalent concentration of the field true value point data, and deriving the area correction coefficient based on the same;

[0009] Correcting the general multi-gas reference by using the area correction coefficient to generate an area multi-gas dynamic reference suitable for the target area;

[0010] Performing a general survey detection in the target area, collecting general survey point data in real time, and converting the general survey point data based on a multi-gas environment correlation model;

[0011] Determining the converted general survey point data based on the area multi-gas dynamic reference, and outputting the position information of the suspected termite nest area.

[0012] Beneficial effects, the present application solves the problems of lack of depth inversion mechanism and low passive scanning efficiency of traditional gas detection, and improves the positioning accuracy and environmental adaptability of dam termite general survey. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 A step flowchart of a dam termite dynamic gas sensing and rapid general survey method based on true value calibration provided for the embodiments of the present application.

[0014] Figure 2 A step flowchart of performing a general survey detection in the target area provided for the embodiments of the present application.

[0015] Figure 3 A step flowchart of deriving the area correction coefficient provided for the embodiments of the present application.

[0016] Figure 4 A step flowchart of outputting the position information of the suspected termite nest area provided for the embodiments of the present application.

[0017] Figure 5 An assembly schematic diagram of a termite gas sensing general survey equipment provided for the embodiments of the present application.

[0018] Figure 6 A rapid general survey schematic diagram in the dam field provided for the embodiments of the present application.

[0019] Wherein the reference signs are: 1, Beidou positioning system; 2, gas sensitive sensor array; 3, wireless data transmission terminal; 4, analysis software system. DETAILED DESCRIPTION

[0020] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings in the embodiments of the present application, so that those skilled in the art can better understand the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present application.

[0021] It should be noted that the terms first, second, etc. in the specification of the present application are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms include and have and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to the clearly listed steps or units, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0022] As shown in Figure 1 A dam termite dynamic gas sensing and rapid survey method based on true value calibration includes the following steps:

[0023] A revisable reference framework including a general multi-gas reference and a multi-gas environment correlation model is established, and the general multi-gas reference includes multi-gas basic parameters in a standard environment.

[0024] In other words, a revisable reference framework including a general multi-gas reference and a multi-gas environment correlation model is established, and the general multi-gas reference includes reference concentration values and gas characteristic ratios of a plurality of core target gases in a standard environment.

[0025] In the present embodiment, the revisable reference framework is the basis of the entire system, providing a unified and quantifiable reference system for subsequent detection. Specifically, the revisable reference framework includes at least two core parts: a general multi-gas reference and a multi-gas environment correlation model. The general multi-gas reference is a database constructed based on a large amount of laboratory test data, which includes gas characteristic parameters released by different types of termite nests in a predetermined standard environment. In order to ensure the comparability of the data, the standard environment is preferably set as: temperature T0 is 25 degrees Celsius, relative humidity RH0 is 65%, and wind speed V0 is 0.5 meters per second. Under the standard environment, the system pre-stores reference concentration values C i0 and characteristic ratios R i，j0For example, for a typical black-winged termite, the baseline concentration C10 of carbon dioxide is set to 1200 ppm, and the baseline concentration C20 of methane is set to 4.8 ppm. These values are obtained by collecting N termite live nest samples (e.g., N ≥ 8) and M sets of non-termite damage background samples (e.g., M ≥ 200) and statistically analyzing them, which are statistically representative. In addition, the multi-gas environment correlation model is a mathematical model for describing how environmental factors affect gas sensor readings, establishing a functional relationship between gas concentration and environmental parameters such as temperature, humidity, and wind speed, providing algorithmic support for subsequent data standardization.

[0026] In some optional embodiments, the correctable reference framework also reserves a two-stage dynamic correction interface, which supports updating the reference parameters by the field calibration data and manual verification data obtained in subsequent steps, and realizes the long-term evolution of the system.

[0027] Obtain field true value point data in the target area, calculate the standard environment equivalent concentration of the field true value point data, and derive the regional correction coefficient based on the standard environment equivalent concentration.

[0028] In other words, select field true value points in the target area, collect the core target gas concentration and real-time environmental parameters of the field true value points as field true value point data, calculate the standard environment equivalent concentration of the field true value point data, and derive the regional correction coefficient accordingly.

[0029] Specifically, before conducting a general survey of a specific target dam area, the detection personnel need to select several representative true value points in the area. These true value points can be known termite nest locations or temporary reference points determined by environmental analogy. After collecting the original gas concentration data of the true value points, the multi-gas environment correlation model is used to eliminate the effects of real-time environmental parameters such as current temperature, humidity, and wind speed on the measurement values, and convert the original data into standard environment equivalent concentrations. For example, if the field temperature is 28 degrees Celsius, which is higher than the standard temperature of 25 degrees Celsius, the model will adjust the readings according to the temperature sensitivity coefficient to obtain the theoretical concentration value under the condition of 25 degrees Celsius. On this basis, the average of the calculated true value point standard environment equivalent concentrations is compared with the corresponding parameters in the universal multi-gas reference to derive the regional correction coefficient. The regional correction coefficient reflects the systematic deviation of the target area's specific soil background and vegetation conditions on gas concentration.

[0030] In some optional embodiments, the regional correction coefficient includes not only the correction coefficient K CiAlso included is a correction coefficient KR for the gas characteristic ratio. For example, if the true value point of the carbon dioxide equivalent concentration is 1416 ppm, and the general reference is 1200 ppm, the regional correction coefficient K1 of carbon dioxide is calculated as 1.18.

[0031] The general multi-gas reference is corrected by the regional correction coefficient to generate a regional multi-gas dynamic reference suitable for the target region.

[0032] In this embodiment, the general reference is corrected by the regional correction coefficient, which is actually to translate or scale the general reference standard to the background level of the current target region. Specifically, each parameter in the general multi-gas reference, such as the reference concentration C i0 , the interference threshold C int，k0 , etc., is multiplied by the corresponding regional correction coefficient to obtain a new set of reference parameters, i.e., a regional multi-gas dynamic reference. The new reference parameters are suitable for the soil and climate conditions of the current target region, and can effectively eliminate the risk of false judgment caused by regional differences. For example, by multiplying the coefficient 1.18, the carbon dioxide judgment reference of this region is raised from 1200 ppm to 1416 ppm, avoiding false alarms caused by high organic matter content in the soil of this region. In some optional embodiments, the generated regional multi-gas dynamic reference is temporarily stored in the regional reference branch of the system and only takes effect in the current census task, without affecting the core data of the general reference, ensuring the flexibility and stability of the system.

[0033] The census detection is performed in the target region, real-time acquisition of census point data, and conversion of census point data based on the multi-gas environment correlation model.

[0034] It can also be said that the census detection is performed in the target region, and the core target gas concentration and real-time environmental parameters of each census point are collected by the sensor array as the census point data in real time, and the census point data is converted into the standard environmental equivalent concentration based on the multi-gas environment correlation model.

[0035] Specifically, performing census detection is the process of applying a theoretical model to actual operations. The detection carrier carries a sensor array to move in the target region according to the planned path, and real-time acquisition of gas concentration data, environmental parameters and position coordinates of each census point. The detection carrier can be a robot dog, an unmanned vehicle or a handheld probe. Each set of census point data collected is immediately input into the multi-gas environment correlation model for processing, eliminating the interference of real-time fluctuations of environmental parameters on sensor readings during the census process, and normalizing all measured values to standard environmental conditions, where the real-time fluctuations of environmental parameters can be temperature changes caused by gusts or cloud cover. Make the data collected at different times and different locations physically comparable, laying a data foundation for subsequent accurate judgment.

[0036] In some optional embodiments, in order to improve the data transmission efficiency, the system classifies the collected data by priority, and preferentially transmits the concentration data of the core target gas, so that the key information has real-time performance in the weak network environment in the field. The core target gas includes but is not limited to: carbon dioxide (CO2), hydrogen (H2), hydrogen sulfide (H2S), ammonia (NH3), methane, hydrogen cyanide (HCN), methylbenzene, 2,4-di-tert-butylphenol, 1-octene-3-alcohol, hexanal, carboxylic acid, terpene, branched alkane, etc.

[0037] Based on the regional multi-gas dynamic benchmark, the converted census point data is judged, and the position information of the suspected termite nest area is output.

[0038] That is, the converted census point data is compared and judged with the regional multi-gas dynamic benchmark, and when the census point data meets the preset abnormal judgment condition, the position information of the suspected termite nest area where the corresponding census point is located is output.

[0039] In the present embodiment, the judgment based on the regional multi-gas dynamic benchmark is the core link of identifying termite damage. The standard environmental equivalent concentration obtained by conversion is compared one by one with the regional multi-gas dynamic benchmark. If the data of the census point exceeds the threshold range set by the regional benchmark, for example, the concentration of the core gas is higher than 1.2 times the benchmark value, and the characteristic ratio meets the metabolic law of termites, then the census point is judged as an abnormal point. In order to improve the accuracy of the judgment, the system can combine the spatial distribution of multiple consecutive abnormal points to delineate the suspected termite nest area, and output the center coordinates and suspected level of the area. Optionally, before outputting the position information, a time-space-environment dual-dimension verification is also performed to eliminate false positive results caused by sudden environmental interference, so that the suspected area output finally has high confidence, wherein the sudden environmental interference includes automobile exhaust and pesticide spraying.

[0040] In one possible implementation, a correctable benchmark framework is established, including constructing a soil gas transmission attenuation model, specifically:

[0041] A soil gas transmission attenuation model is constructed for each core target gas.

[0042] In the present embodiment, the soil gas transport attenuation model is constructed to quantify the physical law of gas molecules moving in the soil medium. Exemplarily, for the termite-related products or ant trail entrance in the ground, the soil gas transport attenuation model for each core target gas can be constructed based on the Fick diffusion law and the soil porosity characteristics. Specifically, unlike the traditional black box statistical method of directly equating the ground concentration to the hazard level, an analytical model is established based on the Fick first diffusion law to describe the process of gas transport from the underground source (termite nest) to the ground surface, and the proportional relationship between the gas flux and the concentration gradient is determined, and combined with the porosity characteristics of the soil (such as porosity, tortuosity), the attenuation function between the concentration of the gas at the ground surface and the source concentration is derived. Specifically, the soil gas transport attenuation model considers that the transport of the gas is mainly dominated by the molecular diffusion mechanism, and the concentration of the gas at the ground surface is the result of the source concentration after being blocked and attenuated by a certain thickness of the soil layer. Alternatively, for the termite-related products or ant trail entrance at the ground surface, other methods can be used to construct the soil gas transport attenuation model for each core target gas. In some optional embodiments, in order to improve the applicability of the model, for different soil types commonly found in dams, such as clay, loam, and sandy clay, corresponding model parameter libraries are measured and established, so that the model can adapt to the complex internal structure of the dam.

[0043] The soil gas transport attenuation model defines the mapping relationship between the ground detection concentration and the nest source concentration, which is determined by the nest depth, the characteristic diffusion length of the gas, and the soil state influence factor.

[0044] Exemplarily, the soil gas transport attenuation model can be expressed in the following exponential decay form:

[0045] C i，surface = C i，source *exp(-H / (λ i * g i (φ,θ,T soil )));

[0046] where C i，surface represents the detection concentration of the i-th core target gas at the ground surface, which is directly measured by the sensor; C i，source represents the source concentration of the gas in the termite nest, which is usually pre-stored in a general multi-gas reference; H represents the unknown parameter to be inverted, i.e. the nest depth; λ i represents the characteristic diffusion length of the i-th gas, which is a physical constant related to the molecular weight of the gas and the soil type; g i (φ, θ, T soil ) represents the soil state influence factor. The formula reveals that the ground concentration not only depends on the source concentration, but also depends on the depth H and the physical state of the soil through an exponential relationship.

[0047] In some optional embodiments, in order to facilitate actual calculation, the system pre-sets a feature diffusion length table under typical soil environment. For example, in clay environment, the feature diffusion length λ1 of carbon dioxide ranges from 0.15 to 0.25 meters, and the feature diffusion length λ2 of methane ranges from 0.20 to 0.30 meters; in loam environment, the feature diffusion length λ1 of carbon dioxide ranges from 0.30 to 0.45 meters, and the feature diffusion length λ2 of methane ranges from 0.40 to 0.55 meters.

[0048] In another optional embodiment, the process of determining the feature diffusion length λ is as follows: collect a representative soil sample of the target dam, prepare a standard soil column with a diameter of 10 cm and a height of 50 cm; set a gas injection cavity at the bottom of the soil column and open the top; set gas sampling holes every 5 cm along the height of the soil column, and use a micro sensor for real-time monitoring; after the concentration stabilizes, fit the λ value according to the concentration value at each depth and the exponential decay model; repeat the experiment for 3 groups for each soil-gas combination, and take the mean value as the median value of λ, and the standard deviation is used to determine the fluctuation range.

[0049] The soil state influencing factor is calculated based on the soil porosity, volume water content and soil temperature.

[0050] In other words, the soil porosity, volume water content and soil temperature of the target area are collected, and the soil state influencing factor is calculated based on the soil porosity, volume water content and soil temperature, which is used to correct the feature diffusion length.

[0051] In this embodiment, the soil state influencing factor g i is introduced to correct the influence of environmental changes on the gas diffusion path. Soil is not a homogeneous medium, its porosity φ determines the number of gas diffusion channels, the volume water content θ determines the proportion of channels occupied by water (water will block gas diffusion), and the soil temperature T soil affects the rate of thermal motion of gas molecules. Specifically, the soil state influencing factor can be calculated by the following empirical formula:

[0052] g i (φ, θ, T soil ) = ((φ-θ) / (φ0-θ0)) ai *exp(b i *(T soil -T soil，0 ));

[0053] where φ0, θ0, T soil，0 are the soil parameters under standard conditions, a i and b iis an empirical coefficient. The formula of soil state influence factor shows that when the soil water content θ increases, the effective porosity decreases, g i decreases, leading to the intensification of gas attenuation on the ground surface; when the temperature rises, g i increases, and the diffusion effect is enhanced.

[0054] In some alternative embodiments, the system collects φ, θ and T soil data in real time through the on-board soil temperature and humidity sensor, dynamically calculates the current g i value, so that the model can maintain high inversion accuracy under different weather conditions such as after rain or drought.

[0055] Alternatively, the empirical coefficients a i and b i may be determined by multi-factor orthogonal experiment: constructing an orthogonal experiment of porosity φ, water content θ, and soil temperature T soil ; determining the gas diffusion flux under each condition; and using nonlinear regression fitting to obtain the a i , b i values of each gas. The typical value range is: a i ∈ [0.3, 0.8], b i ∈ [0.01, 0.05].

[0056] In further implementations, constructing the soil gas transport attenuation model also includes establishing an ant nest depth inversion mechanism, specifically:

[0057] selecting a first core target gas and a second core target gas with different characteristic diffusion lengths;

[0058] using the surface detection concentration ratio of the first core target gas and the second core target gas, combined with the pre-stored ant nest source concentration ratio, to inversely calculate the ant nest depth.

[0059] In this embodiment, the surface concentration of a single gas is greatly affected by the fluctuation of the source concentration, and direct inversion of the depth has multiple solutions. Therefore, two gases with significantly different physical properties are selected, for example, terpenes with slower diffusion as the first gas and methane with faster diffusion as the second gas, and an equation set is constructed by using their different attenuation characteristics when transported to the ground surface at the same depth H. Since the two gases originate from the same ant nest, their depth H is the same, and the ratio of their source concentrations (such as α-pinene / methane) is relatively stable biologically. Therefore, by comparing their concentration ratio at the surface with the concentration ratio at the source, the depth H can be uniquely determined. In a preferred embodiment, gases with the largest difference in characteristic diffusion length are selected for inversion, for example, volatile organic compounds (such as α-pinene) with larger molecular weight are combined with methane with smaller molecular weight to maximize the sensitivity of the ratio to depth.

[0060] The formula for calculating the nest depth H is:

[0061] H = (ln(R surface ) - ln(R source )) / (1 / λ2 - 1 / λ1);

[0062] wherein R surface is the detected concentration ratio on the ground, R source is the pre-stored nest source concentration ratio, and λ1 and λ2 are the characteristic diffusion lengths of the first and second core target gases, respectively.

[0063] Specifically, R surface represents the concentration ratio (C 1，surface / C 2，surface ) of the two gases detected on the ground; R source represents the pre-stored nest source concentration ratio (C 1，source / C 2，source ) of the two gases, and λ1 and λ2 are the characteristic diffusion lengths of the two gases under the current soil type. By logarithmic operation to linearize the exponential decay relationship, the depth H can be directly calculated. For example, assuming that the detected ratio R surface on the ground is 0.5, and the known source ratio R source is 1.0, i.e., the source concentrations of the two gases are equal, and the λ1 of the first gas is 0.2 meters and the λ2 of the second gas is 0.4 meters. Substituting into the formula: the numerator is ln(0.5) - ln(1.0) ≈ -0.693, and the denominator is 1 / 0.4 - 1 / 0.2 = 2.5 - 5.0 = -2.5. Then the depth H = -0.693 / -2.5 ≈ 0.277 meters. It shows that the nest is located about 0.28 meters underground. Alternatively, in order to further improve the accuracy, the system can use multiple gas pairs, such as gas A / B and gas A / C, to calculate the depth H respectively, and then take the average as the final determination result, in order to reduce the influence of single sensor measurement error. Preferably, it also includes selecting the gas combination with the largest λ difference, so that |1 / λ2 - 1 / λ1| ≥ 2m -1 , as the inversion gas selection constraint condition.

[0064] This embodiment introduces the Fick diffusion law and the gas transmission decay model to invert the gas concentration signal detected on the ground into the nest depth information underground, realizing the expansion of the detection dimension from two dimensions to three dimensions.

[0065] In an exemplary embodiment, after collecting the survey point data in real time, and before converting into the standard environmental equivalent concentration, it further includes a multi-gas cross interference correction step, specifically:

[0066] A cross-response matrix of the sensor array is obtained, which defines the response coefficients of each sensor to non-target gases.

[0067] In the present embodiment, the multi-gas cross-interference correction is the first step to ensure data purity. Since existing electrochemical or metal oxide semiconductor sensors are difficult to achieve 100% specificity, they often produce bypass responses to non-target gases. For example, a carbon dioxide sensor may produce a positive response to ambient humidity, and a hydrogen sulfide sensor may respond to high concentrations of alcohols. To eliminate this effect, the system pre-stores a cross-response matrix M. The cross-response matrix M can be constructed by testing the response amplitude of each sensor to other interfering gases one by one in a laboratory standard gas chamber. The element a i，j in the matrix represents the sensitivity coefficient of the i-th gas sensor to the j-th gas.

[0068] In some optional embodiments, the specific application form of the cross-response matrix can be the inverse operation of a linear equation system. For example, for a simplified system containing carbon dioxide concentration C1 and humidity RH, if the response coefficient of the carbon dioxide sensor to humidity is 0.03, then the measured concentration value C1 raw = C1 true + 0.03 * RH. The system can inversely calculate the true carbon dioxide concentration C1 true by reading the value RH of the humidity sensor.

[0069] Optionally, the sampling parameters of the sensor array are set as follows: the sampling frequency of the gas concentration sensor is set to 2Hz to 10Hz, preferably 5Hz, to balance real-time performance and data stability; the analog-to-digital conversion (ADC) accuracy is set to 12 bits to 16 bits, preferably 14 bits, to ensure the resolution of concentration measurement; the sampling frequency of environmental parameters (temperature, humidity, wind speed) can be set to 0.5Hz to 2Hz, since environmental parameters change relatively slowly. The above sampling parameters can be adjusted according to the specific sensor model and application scenario.

[0070] The cross-response matrix is used to compensate and calculate the collected raw gas concentration data, eliminating the cross-response components of interfering gases to the core target gas sensor.

[0071] Specifically, the compensation calculation can be performed in real time within the microprocessor. The system synchronously acquires the raw voltage or current signals from all sensor channels and converts them into preliminary concentration values. Using the cross-response matrix, interference components are subtracted from the target gas readings through matrix multiplication or iterative elimination. For example, if a high concentration of methanol gas (an interfering gas) is detected in the environment, and the methane sensor is known to have a weak cross-response to terpenes, the system will automatically subtract this false methane reading caused by terpenes. This reduces false positives caused by complex background gases. Preferably, a deep decoupling design is implemented for core target gases such as carbon dioxide, methane, ammonia, hydrogen cyanide, formic acid, and 1-octen-3-ol to ensure the purity of these key indicators.

[0072] The compensated gas concentration data is used as input to perform a step of converting it into the equivalent concentration of a standard environment.

[0073] In this embodiment, although the data after cross-interference correction eliminates interference between components, it is still affected by physical environmental factors such as temperature, humidity, and wind speed. Therefore, this data needs to be passed to the environmental correlation model for normalization processing, serving as a bridge connecting the underlying hardware data and the upper-level decision logic.

[0074] like Figure 3 As shown, according to one aspect of this application, the regional correction coefficient is derived, including:

[0075] Real-time environmental parameters are collected from the field ground truth points. The influence of real-time environmental parameters is eliminated by using a multi-gas environment correlation model. The field ground truth point data is converted into concentration values ​​under standard environment conditions and used as equivalent concentrations under standard environment conditions.

[0076] In this embodiment, the multi-gas environment correlation model C i = C i0 * f i (T) * f i (RH) * f i (V) was used to eliminate environmental differences, where C i Let C be the equivalent concentration of the i-th gas in the field environment converted to the standard environment. i0 Let f be the baseline concentration of the i-th gas under standard conditions. Specifically, for the temperature influence factor f... i (T), the model preferably uses a linear or polynomial fitting formula, such as f i (T) = 1 + k Ti * (T -T0), where T is the current temperature, T0 is the standard temperature of 25 degrees Celsius, and k Ti Let k be the temperature sensitivity coefficient for the i-th gas, for example, for carbon dioxide, k Tican be set to 0.025. For humidity influence factor f i (RH), considering the dilution or adsorption of water vapor on the sensor, the formula can be f i (RH) = 1 + k RHi *(RH - RH0), where RH is the current relative humidity, RH0 is the standard humidity 65%, k RHi is the humidity sensitivity coefficient of the i-th gas, which can be -0.015. For wind speed influence factor f i (V), when the wind speed is greater than the standard wind speed V0 (such as 0.5 meters per second), the gas concentration will be diluted, and the model compensates by using the correction formula f i (V) = 1 - k Vi * (V - V0), where V is the current wind speed, k Vi is the wind speed sensitivity coefficient of the i-th gas. Regardless of whether the field is high temperature and dry or low temperature and humid, all data is converted to a unified standard environment.

[0077] In some optional embodiments, the sensitivity coefficients k Ti , k RHi , etc. are not fixed but can be dynamically adjusted through a closed-loop feedback mechanism to adapt to sensor aging or special environmental characteristics.

[0078] Calculate the average of the standard environment equivalent concentrations of all field true value points.

[0079] Specifically, in order to obtain a stable regional reference, the data of multiple true value points need to be statistically averaged. The system automatically eliminates invalid data collected at moments of dramatic fluctuations in environmental parameters, such as invalid data collected when the wind speed instantaneously exceeds 3 meters per second, and only retains data collected during stable periods for average calculation.

[0080] Calculate the ratio of the average to the corresponding parameter in the universal multi-gas reference to obtain the concentration correction coefficient for each core target gas and the ratio correction coefficient for the gas characteristic ratio, and the concentration correction coefficient and the ratio correction coefficient are collectively referred to as the regional correction coefficient.

[0081] In this embodiment, the regional correction coefficient K is a key parameter that connects universal knowledge and local environment, where the universal knowledge can be the corresponding parameter in the universal multi-gas reference. For the i-th core target gas, the concentration correction coefficient K Ci =C i，cal0，avg / C i0 ; where C i，cal0，avgThe equivalent concentration average of the gas measured by the true value point in the target area. For example, if the true value point in the field measures the average equivalent concentration of methane as 5.76 ppm, and the concentration C20 of methane in the general reference is 4.8 ppm, then K Ci = 1.2. This means that the methane background value of the area is 20% higher than the standard model, and in the subsequent census, the system will adjust the judgment threshold accordingly to prevent false positives due to high background values. Similarly, for the characteristic ratio, the corresponding correction coefficient KR is also calculated. Preferably, if the calculated correction coefficient K deviates too much from the normal range, for example, <0.5 or >2.0, the system will trigger an abnormal alarm to prompt the operator to check whether the true value point selection is appropriate or the sensor is malfunctioning, rather than blindly applying the coefficient.

[0082] The embodiment provides a high-precision data preprocessing scheme, so that the data input into the subsequent judgment model can truly reflect the metabolic level of termites.

[0083] In an embodiment of the present application, before performing the census detection, a grid path planning step based on a risk grading model is further included, specifically:

[0084] Obtain the historical termite damage record factor, vegetation coverage factor, slope factor, and multi-gas background concentration factor of the target area.

[0085] In other words, the historical termite damage record factor of the target area is obtained from the dam management database, the vegetation coverage factor and the slope factor are obtained from remote sensing data, and the multi-gas background concentration factor is obtained from the regional multi-gas dynamic reference.

[0086] In the embodiment, the first step of path planning is to perform digital risk assessment on the target dam. The system retrieves the historical maintenance records of the dam from the water conservancy project management database, extracts the historical termite damage record factor H, and has a record of 1 and no record of 0. The vegetation coverage factor V (such as coverage rate greater than 60% for 1) and the slope factor S (such as slope greater than 30 degrees for 1) are calculated using unmanned aerial vehicle or satellite remote sensing data. At the same time, the multi-gas background concentration factor G is determined using regional reference data. The historical termite damage record factor, vegetation coverage factor, slope factor, and multi-gas background concentration factor constitute the basic dimensions for assessing the risk of termite in the dam.

[0087] Based on the weighted calculation formula R = w H * H + w V * V + w S * S + w G * G to evaluate the risk level value R of the area.

[0088] Specifically, the risk is quantified by a linear weighting model. The setting of each weight reflects the influence degree of different factors on termite breeding. Preferably, the historical termite damage record weight w H is set to 0.3 because termites have the habit of original nest recurrence or dispersal by flying; the vegetation coverage weight w V is set to 0.2 because dense vegetation provides termites with food sources and hiding places; the slope weight w S is set to 0.2, which involves easily overlooked areas such as water-facing slopes; the background concentration weight w G is set to 0.3. The calculated R value is between 0 and 1.

[0089] According to the risk level value, the target area is divided into high-risk areas, ordinary areas and transition areas, and different survey grid densities are allocated to different areas; the grid length of the high-risk area is less than that of the ordinary area.

[0090] In this embodiment, the dam is divided into different areas based on the R value to balance the survey efficiency and accuracy. For the high-risk area with a risk level value R≥0.7, a high-density grid is generated, and the grid length L high is set to 1.0 meter to ensure a thorough investigation of high-risk areas; for the ordinary area with R<0.5, the grid length L low is set to 2.0 meters to improve the speed of advance; the transition area between the two uses a 1.5-meter grid. Through the differential network deployment strategy, the detection benefit is maximized within the limited endurance time.

[0091] As shown in FIG. 8A, Figure 2 in one possible embodiment, a survey detection is performed in the target area, and an adaptive tracking strategy based on gas concentration gradient is adopted, which specifically includes:

[0092] Real-time concentration data of the preset core target gas is collected in each detection direction around the current detection point.

[0093] In this embodiment, when the detection carrier discovers a gas anomaly at a certain grid point, it immediately switches to a microscopic tracking mode. The carrier rotates in place or uses a multi-directional sensor array to uniformly select n directions, for example, 8 directions, with an interval of 45 degrees, on a circumference centered at the current position, and collect the core target gas concentration C i (k, j) in each direction. The olfactory search mechanism of insect antennae is simulated.

[0094] Based on the real-time concentration data, the gas concentration gradient vector of each detection direction is calculated.

[0095] Specifically, the system calculates the concentration difference between the current point and each detection direction point to obtain the gradient vector. Exemplarily, for the jth direction, the gradient modulus |Grad j| = (C i (k, j) - C i (current)) / dL, wherein C i (current) is the gas concentration of the current detection point; dL is the sampling distance, such as 0.5 meters; the direction of the gradient vector is the unit vector of the detection direction. The gradient vector reflects the rate of change of the gas concentration in the direction.

[0096] The gas concentration gradient vectors of all detection directions are integrated to determine the optimal tracking direction, and the detection carrier is controlled to move to the next detection point along the optimal tracking direction.

[0097] That is, the next moving direction of the carrier is not simply directed to the point with the highest concentration, but is determined by all positive gradient directions. The weighted vector sum method is preferably used to calculate the optimal direction.

[0098] In an exemplary implementation, determining the optimal tracking direction comprises:

[0099] Based on the gas concentration gradient vectors, a weighted sum algorithm is used in combination with an indicator function to calculate the combined gradient direction of the predetermined N core target gases as the optimal tracking direction, wherein only the contribution of the positive gradient direction is counted; N is a natural number.

[0100] In this embodiment, the calculation formula of the optimal tracking direction θ* is specifically: for all core target gases i and all detection directions j, the weighted sum ∑(w Ci * Grad ij * I(Grad ij > 0)) is calculated; wherein the indicator function I ensures that only the direction with increasing concentration participates in the voting, w Ci is the weight of different gases, and Grad ij is the gas concentration gradient vector of the i-th core target gas in the j-th detection direction. The final combined vector direction is the next moving direction of the carrier. The single gas plume breakage or local turbulence is effectively overcome, and the source is pointed to by using the collaborative information of multiple gases.

[0101] The steps of collecting, calculating and moving are iteratively performed until the gas source is locked. The gas source is used as the positioning basis of the suspected termite nest area.

[0102] In this embodiment, the carrier repeatedly performs the above tracking process to gradually approach the odor source.

[0103] Exemplarily, the locking gas source comprises: when the concentration change of the continuous detection point meets the preset concentration peak condition and the gas concentration gradient vector in each direction meets the preset gradient convergence condition, it is determined that the current position is the gas channel outlet. The concentration peak condition is that the current detection point concentration is higher than the previous detection point concentration by a preset multiple, and the gradient convergence condition is that the gradient modulus in each direction is less than a preset gradient threshold.

[0104] In the embodiment, the termination condition of the locking is a mathematical local maximum point. Specifically, the concentration peak condition requires that the current point concentration C k is significantly higher than the previous time C k-1 and the surrounding detection points; and the gradient convergence condition requires that the gradient values in all directions are less than a preset small gradient convergence threshold ε, that is, the concentration does not have a significant increase around the current point. The position meeting the condition is marked as the ant nest outlet (ventilation hole or branch hole). Optionally, the gradient convergence threshold ε is set according to the measurement accuracy of the gas concentration and the environmental noise level: ε = δ Ci / ΔL × k safe ; where δ Ci is the sensor resolution of the ith gas, ΔL is the detection step length, and k safe is a safety factor, which can be 2-3. Taking carbon dioxide as an example: δ C1 = 10 ppm, ΔL = 0.4 m, and k safe = 2.5, then ε = 62.5 ppm / m.

[0105] After locking the outlets of the gas channels, the spatial distribution characteristics of the outlets are calculated, and if the outlets of the gas channels present central radiation, the outlets of the gas channels presenting central radiation are associated as multiple outlets of the same ant nest system.

[0106] In other words, after locking the multiple gas channel outlets, the spatial distribution characteristics of the gas channel outlets are calculated, and if the multiple gas channel outlets present central radiation or linear arrangement characteristics, they are associated as multiple outlets of the same ant nest system.

[0107] Specifically, in view of the fact that a mature ant nest usually has multiple ventilation holes, the system performs topological analysis on the multiple locked outlets within a certain range. The geometric center and distribution radius of these points are calculated, and if they present a radial distribution with the main nest as the center, the system merges these points as one ant nest event, thereby avoiding repeated counting and assisting in judging the approximate position and depth of the main nest. Alternatively, linear arrangement characteristics within a preset range can also be used to define linear arrangement, which is regarded as an ant nest event.

[0108] The embodiment combines macro-level hierarchical distribution and micro-level olfactory tracking, which not only ensures the coverage of the general survey, but also improves the accurate positioning ability of the hidden ant nest.

[0109] As shown in Figure 4 , according to one aspect of the present application, the position information of the suspected termite nest area is output, including:

[0110] In combination with the core target gas concentration, gas characteristic ratio in the converted survey point data, and the preset interference gas concentration, the multi-gas principal component value is calculated. The interference gas includes environmental background gas irrelevant to termite metabolism but affecting the sensor reading.

[0111] In the present embodiment, the weighted summation algorithm is preferably adopted to calculate the multi-gas principal component value P Ci as the core quantitative index for determining the termite damage degree. Exemplarily, the multi-gas principal component value integrates information in three dimensions, and the specific formula is:

[0112] P Ci =∑ i (w Ci * (C i / C i0 )) + ∑ j (w Rj * (R j / R j0 )) + ∑ k (w int，k * (1 - C int，k / C int，k0 ));

[0113] Among them, the first term is the concentration ratio term, reflecting the metabolic intensity, and the weight w Ci is higher, such as 0.5; the second term is the characteristic ratio term, reflecting the gas fingerprint characteristics, and the weight w Rj is moderate, such as 0.35; the third term is the interference suppression term, and the weight w int，k is used to reduce the P Ci score when interference gas is detected. R j is the jth gas characteristic ratio, R j0 is the baseline value of the jth characteristic ratio in the standard environment, C int，k is the concentration of the kth interference gas at the current survey point, and C int，k0 is the baseline concentration of the kth interference gas in the standard environment, i.e. the interference threshold. All terms are dimensionless, ensuring the unity of physical meaning. The weight coefficients are preferably determined by combining the analytic hierarchy process (AHP) and entropy weight method with historical data statistics to ensure the scientificity of the index.

[0114] In some optional embodiments, the weight coefficients w Ci , w Rj , w intkThe determination of the suspected interval [P, P] is based on statistical analysis of historical verification data. Specifically, historical survey data that have completed manual verification are collected and divided into a sample set confirmed to have termite damage and a sample set confirmed to have no termite damage, and the P values of the two sample sets are calculated respectively. The lower limit P of the suspected interval is set as the 5th percentile of the P values of the sample set confirmed to have termite damage, to ensure that more than 95% of the true positive samples are captured; the upper limit P of the suspected interval is set as the 95th percentile of the P values of the sample set confirmed to have no termite damage, to exclude obvious false positives. For example, based on statistical analysis of 200 groups of historical data in a certain area, the 5th percentile of the P values of the sample set confirmed to have termite damage is 0.71, and the 95th percentile of the P values of the sample set confirmed to have no termite damage is 0.95.

[0115] Performing double-track determination logic: if the core target gas concentration and gas characteristic ratio of the converted survey point data both meet the threshold range set by the regional multi-gas dynamic benchmark, and the multi-gas principal component value is within the preset suspected interval, it is determined that the region where the current survey point is located is a suspected termite nest region.

[0116] Specifically, a double-track determination mechanism is adopted. Condition one requires that the single indicator is passing, i.e. the concentration of all core gases must be within a reasonable range, e.g. greater than 0.8 times the benchmark, and the key characteristic ratio must meet the termite metabolic law; condition two requires that the comprehensive indicator is excellent, i.e. the P Ci value must fall within the suspected interval [P Ci_low , P Ci_high ], e.g. 0.71 to 0.95. Only when both conditions are met will it be marked as a suspected zone. This avoids false positives caused by accidental fluctuations in a single gas, and also prevents missed detection caused by a low overall concentration of multiple gases coincidentally.

[0117] Optionally, the determination of the suspected interval [P Ci_low , P Ci_high ] is based on statistical analysis of historical verification data. Specifically, historical survey data that have completed manual verification are collected and divided into a sample set confirmed to have termite damage and a sample set confirmed to have no termite damage, and the P Ci values of the two sample sets are calculated respectively. The lower limit P Ci_low of the suspected interval is set as the 5th percentile of the P Ci values of the sample set confirmed to have termite damage, to ensure that more than 95% of the true positive samples are captured; the upper limit P Ci_high of the suspected interval is set as the 95th percentile of the P Ci values of the sample set confirmed to have no termite damage, to exclude obvious false positives. For example, based on statistical analysis of 200 groups of historical data in a certain area, the 5th percentile of the P Ci values of the sample set confirmed to have termite damage is 0.71, and the 95th percentile of the P CiThe 95th percentile is 0.95, and thus the suspected interval of the region is determined as [0.71, 0.95]. The interval can be dynamically adjusted through a closed-loop feedback mechanism as the verification data accumulates.

[0118] In one possible embodiment, in order to further clean the data, a firewall is added before the output result, that is, before outputting the position information of the suspected termite nest region, a space-time-environment double-dimension verification step is further included, which is specifically:

[0119] Environment verification: It is judged whether the real-time environment parameter when collecting the survey point data is within the effective adaptive range of the multi-gas dynamic benchmark of the region. The effective adaptive range is determined based on the fluctuation amplitude of the environment parameter at the calibration time.

[0120] In the embodiment, it is checked whether the environment parameter when collecting the data is out of range. For example, if the temperature at the calibration time is 25 degrees, and the effective adaptive range of the model is defined as plus or minus 10 degrees, if a sudden high temperature such as 40 degrees is encountered when the survey is conducted, the data will be marked as untrusted, and the data will be supplemented after the environment recovers.

[0121] Space-time verification: When the detection carrier performs secondary detection around the suspected termite nest region, the time sequence fluctuation stability and spatial diffusion trend of the core target gas concentration in the converted survey point data are verified.

[0122] Specifically, for the suspected region determined initially, the carrier will perform secondary scanning. In the time dimension, data of a continuous time period is collected, and the fluctuation coefficient σ is calculated. If the fluctuation is severe (such as σ greater than 10%), it means that it is an exogenous gas such as vehicle exhaust passing by, rather than a source of continuous release underground, and it is determined as a false positive. In the spatial dimension, it is verified whether the concentration distribution conforms to the point source diffusion model, that is, the center concentration is high, and gradually decreases to the surrounding. If the distribution is chaotic or shows a uniform distribution like a large area of fertilization, it is also excluded.

[0123] Only when the environment verification and the space-time verification are passed, the operation of outputting the position information of the suspected termite nest region is performed.

[0124] In the embodiment, only the region that passes through the above-mentioned layer-by-layer screening will be finally confirmed as the suspected termite nest region and output to the artificial review team. Each coordinate point output has a digging value, and the labor cost of invalid excavation is reduced.

[0125] In one embodiment of the present application, after outputting the position information of the suspected termite nest region, a closed-loop feedback and model evolution are further included, which is specifically:

[0126] Obtaining a manual verification truth value for the suspected termite nest area, the manual verification truth value including nest existence results and field truth gas concentration.

[0127] In other words, obtaining a manual verification truth value after excavation verification or field sampling of the suspected termite nest area.

[0128] In the present embodiment, when the system outputs suspected area coordinates, a manual review team intervenes. In order to obtain high-quality truth value data, the excavation verification process follows strict operation specifications. Specifically, before excavation, a ground sampling method is used to collect the field truth gas concentration C field of the point, which represents the true gas leakage level before the nest structure is damaged. Excavation is performed, and the excavation results are recorded, including whether live termites or fungus gardens are found (nest existence results), the actual depth H nest of the nest, the size S nest of the nest, and the main nest position. These physical data constitute the gold standard for evaluating the detection accuracy of the system. Optionally, the calibrated portable probe can also be inserted into the soil surface layer to collect the ground gas concentration C surface and the soil layer gas concentration C depth (d) at different depths (such as 10 cm, 20 cm, and 30 cm). The ground gas concentration is used to calibrate the regional reference, and the soil layer gas concentration at different depths is used to correct the data of the termite nest depth inversion.

[0129] Calculating the comprehensive matching degree between the standard environmental equivalent concentration during detection and the standard environmental equivalent value of the field truth gas concentration.

[0130] Specifically, in order to fairly evaluate the detection bias, the historical data during detection and the current verification data are unified to the same standard environment for comparison. The system retrieves the standard environmental equivalent concentration C corr during detection. At the same time, using the real-time recorded temperature, humidity, and wind speed parameters at the excavation site, the field truth gas concentration C field is converted to the standard environmental equivalent value C field0 . The calculation formula of the comprehensive matching degree γ match is preferably:

[0131] γ match = 1 - (abs(C corr - C field0 ) / C field0 ).

[0132] Where abs() represents the absolute value. The comprehensive matching degree quantifies the degree of agreement between the detection value and the truth value, and the closer the numerical value is to 1, the more accurate the detection is.

[0133] In some alternative embodiments, the comprehensive matching degree not only considers the concentration matching of a single gas, but also the weighted average of all core target gases. For example, γ match_com = ∑(w Ci * γ match_i ).

[0134] If the comprehensive matching degree is lower than the preset iteration threshold, an error characteristic is outputted, and error attribution analysis is performed according to the error characteristic to generate updated parameters for the universal multi-gas reference or the multi-gas environment correlation model.

[0135] In the present embodiment, the system presets an iteration threshold η err , for example, 70%. When the comprehensive matching degree is lower than the value, it means that the system has a significant deviation, triggering the automatic correction process. At this time, the system starts the error attribution analysis module, and the core task of the module is to diagnose the disease, that is, to judge whether the error is from the inaccurate reference, or the incorrect environment model, or the too strict threshold.

[0136] In a preferred implementation, generating updated parameters for the universal multi-gas reference or the multi-gas environment correlation model includes:

[0137] An error attribution matrix including the reference adaptation deviation dimension, the environment model deviation dimension, and the threshold setting deviation dimension is constructed, and a mapping relationship between the error characteristic and each dimension of the error attribution matrix is established.

[0138] In the present embodiment, in order to realize intelligent diagnosis, the system internally builds an error attribution matrix in the form of an expert rule base. The matrix defines the root causes corresponding to different deviation phenomena. Specifically: the typical characteristic of the reference adaptation deviation dimension is that the true positive is overall low or high in detection value. For example, the core gas measured concentration deviates from the target dam exclusive reference by more than 15%, but the characteristic ratio is still within the normal fluctuation range, and there is no interference gas alarm. Usually it indicates that the soil background value of this area is quite different from the universal reference, and the reference concentration parameter C i0 needs to be adjusted. The typical characteristic of the environment model deviation dimension is that the environment parameter is stable but the matching degree is low. For example, when the temperature or humidity is in the extreme range, such as temperature greater than 30 degrees Celsius or humidity greater than 70%, the core gas concentration deviates from the reference by more than 10%, and the characteristic ratio appears abnormal fluctuation. Usually it indicates that the sensitivity coefficient k T or k RH of the environment correlation model under extreme conditions is not accurate and cannot completely eliminate the environmental impact. The typical characteristic of the threshold setting deviation dimension is that the false positive or false negative is concentrated in the threshold boundary. For example, the P Ci indicator fluctuates near the critical value of the suspected area for a long time, causing the system to be not sensitive enough to weak signals or too sensitive to noise. It indicates that the determination threshold P Ci_low or PCi_high Fine-tuning is needed.

[0139] Based on the mapping relationship, weight coefficients are assigned to each dimension of the error attribution matrix, and correction factors of each dimension are calculated according to the comprehensive matching degree.

[0140] In other words, based on the mapping relationship, the deviation features between the artificial verification true value and the standard environmental equivalent concentration at the time of detection are extracted, weight coefficients are assigned to each dimension of the error attribution matrix, and correction factors of each dimension are calculated according to the comprehensive matching degree.

[0141] Specifically, considering that actual errors are often the result of coupling of multiple factors, the system adopts a weighted distribution strategy. Based on the statistical law of historical big data, the weight coefficients W k of each dimension are preset base . For example, the benchmark adaptation deviation weight W mod is set to 0.4, the environmental model deviation weight W thres is set to 0.3, the threshold setting deviation weight W trans is set to 0.2, and the interface transmission deviation weight W k is set to 0.1. The correction factor F k is calculated based on the matching degree gap, for example, F err = 1 - (η match_com -γ err ) / η corr , which means the lower the matching degree, the greater the correction strength.

[0142] The updated parameters are calculated using the weighted correction formula, which is:

[0143] P orig ×∑(W k ×F k );

[0144] where P corr is the updated parameter, P orig is the original parameter before correction, W k is the weight coefficient of the kth dimension, F k is the correction factor of the kth dimension, and ∑ represents weighted summation of all relevant dimensions.

[0145] In this embodiment, the system calculates the new parameter value P corr using the above formula. By weighted accumulation of all relevant dimensions through the summation symbol, the system can smoothly update the parameters and avoid parameter shock caused by single accidental error.

[0146] The general multi-gas reference or multi-gas environment correlation model is corrected by using the updated parameters, and the corrected reference or model is applied to subsequent general survey detection.

[0147] Specifically, the calculated updated parameters are written into the non-volatile memory of the system. If the updated parameters are general reference parameters, such as the standard concentration of a certain gas, the update will be synchronized to all devices of the same type, achieving one correction and whole network evolution; if the updated parameters are region-specific parameters, they will only take effect in the database of the dam project. In subsequent general survey tasks, the system will directly call the corrected reference or model to obtain higher accuracy in the next round of detection.

[0148] The embodiment solves the problem of precision decline of the detection equipment due to environmental drift or sensor aging in long-term use.

[0149] According to one aspect of the present application, in order to achieve full-terrain coverage, a multi-carrier cooperative operation mode can be used. Specifically, for areas with high vegetation coverage (e.g., greater than 60%) or rugged terrain, a quadruped robot such as a robot dog is used as a detection carrier. The robot dog has excellent obstacle avoidance capability, and its sensor array is installed on the belly or back of the body, with a ground clearance of about 0.3 meters, which can penetrate into the grass to perform ground detection. For gentle and large dam slopes, a tracked unmanned vehicle is used as a carrier. The unmanned vehicle has strong endurance, and the single operation coverage area can reach more than 2000 square meters, which is suitable for rapid scanning. For dead corners such as dam shoulders, stone cracks, and drainage ditches that are difficult for machines to reach, a handheld probe rod is provided. The probe rod is made of lightweight alloy material, weighs less than 1.8 kilograms, and the length can be adjusted (1.0 meters to 2.0 meters), which is operated by a person to supplement the detection.

[0150] In some optional embodiments, the detection carrier is not limited to the three specific forms described above. Other carriers with mobility and sensor mounting capability can be selected according to the actual dam environment and operation requirements, such as wheeled robots, tracked robots, multi-rotor drones (used for preliminary aerial reconnaissance to guide ground carriers), underwater robots (used for water area detection), etc. As long as the carrier can carry the gas sensor array, environmental sensors and positioning modules required by the present application, and can control the path according to the gradient tracking strategy of the present application, it can be applied to the technical solution of the present application.

[0151] According to another aspect of the present application, after collecting the general survey point data in real time, it is uploaded to the processing center. The process of uploading to the processing center follows a multi-priority transmission strategy, specifically:

[0152] The core target gas data for secondary detection is marked as the first priority.

[0153] In the present embodiment, the data transmission system adopts a priority-based queue scheduling mechanism. When the probe carrier enters the suspected area for secondary detection, the collected core target gas concentration data is tagged with P1. Such data is directly related to the confirmation of termite infestation, and has extremely high real-time requirements. Therefore, regardless of the current network conditions, P1 data is always placed at the forefront of the sending queue.

[0154] Grid point data for general surveys are marked as second priority.

[0155] Specifically, a large amount of continuous monitoring data generated during the regular grid survey process is marked with a P2 tag. Such data is mainly used to draw the overall gas distribution map, and has slightly lower real-time requirements than P1 data. When the network bandwidth is normal, P2 data is alternately sent with P1 data.

[0156] Auxiliary target gas data and historical supplementary data are marked as third priority.

[0157] In the present embodiment, auxiliary target gases (non-core discriminant gases), environmental parameter background data, and historical data that have not been successfully sent due to network interruption are marked with a P3 tag. Such data is mainly used for later scientific research or full-quantity data archiving, and is not sensitive to real-time requirements.

[0158] When the data transmission bandwidth is limited, the bandwidth is preferentially allocated to transmit first priority data, and the third priority data is locally cached.

[0159] Specifically, considering that dams are usually located in the wild, 4G / 5G signals may not be stable, such as signal strength below -85 dBm. When the system detects that the uplink bandwidth is limited, the transmission controller will immediately start the traffic shaping strategy: suspend sending P3 data, limit the sending frequency of P2 data, and concentrate all available bandwidth on P1 data to ensure that core alarm information can be uploaded to the cloud server within 1 second. At the same time, the blocked P3 data will be written to the local mass storage of the carrier, and when the signal is restored or the task is completed, the system will automatically perform breakpoint resume transmission to ensure the integrity of the data.

[0160] In an optional embodiment, a calibration strategy in the scenario of lack of historical data in the target area is also included, specifically:

[0161] For new areas without termite infestation records, an environmental analogy calibration method is used to select true value points.

[0162] In this embodiment, when the detection device first enters a new embankment area that has never been surveyed and lacks historical ant damage records, the system cannot directly apply the true value screening strategy for ant damage areas. At this time, the environmental analogy calibration mode is started to find a substitute. The cloud database is traversed to retrieve adjacent areas that are highly similar to the current new area in environmental characteristics, and the adjacent areas have completed the survey and have true value data.

[0163] Specifically, the calculation of similarity is based on three core dimensions: soil type, climate parameters, and vegetation coverage. The system calculates the soil type similarity between the new area and the candidate area, requiring it to be greater than or equal to a preset threshold, such as 0.75, calculated using the weighted Gower similarity model; calculates the climate parameter difference, including annual average temperature, annual relative humidity, monthly average precipitation frequency, etc., requiring each difference to be ≤15%; and calculates the vegetation coverage difference, requiring it to be ≤20%. When a substitute area that meets all the above conditions is found, the system directly calls the true value point data of the substitute area as the temporary anchor point of the new area for initial calibration. At the same time, the system can preset several artificial verification points in the new area based on the spatial distribution pattern of the substitute area in an equidistant or regular grid manner for subsequent true value supplement. This effectively reduces the cold start error of the new area survey.

[0164] A correctable multi-gas reference framework containing a universal reference, an environmental model, and a two-level interface is constructed.

[0165] In this embodiment, in order to realize the safe isolation and efficient synchronization of the universal reference and the regional reference, the system deploys a two-level dynamic correction interface at the software bottom. This interface uses JSON standard data format and supports 4G or LoRa wireless transmission protocol, ensuring that the data transmission delay is less than 1 second.

[0166] Specifically, the interface is divided into two levels of permissions: the first interface (regional level): used to receive field true value calibration data. This interface only has write permission for the regional reference branch area, allowing updates to regional correction coefficients, regional exclusive thresholds, and other localized parameters, but strictly prohibits modification of the universal reference core data. This ensures that the particularity of a single project does not contaminate the global standard. The second interface (system level): used to receive error correction parameters. This interface has the highest permission and can write to both the universal reference core area and the regional reference branch area. This interface is usually activated only after the closed-loop feedback verification is passed, i.e., the simulation matching degree meets the standard, and is used to update the basic parameters involving all devices, such as the standard characteristic diffusion length of a certain gas. In addition, the interface also has complete data verification and permission control functions to prevent illegal data injection.

[0167] Verification area selection priority sorting and error data screening priority sorting are performed.

[0168] In this embodiment, in order to maximize the effect of closed-loop iteration under limited human resources, the system defines a strict job priority strategy. In the manual verification stage, according to the detection results, the suspected area is divided into different verification priorities: the first priority: the time-space-environment double-dimensional verification passes, and the P Ci value of the high-confidence suspected area is extremely high. Such areas are excavated in priority to confirm true positives. The second priority: the point marked as a dynamic monitoring area, that is, the environment is stable but the time-space fluctuation is slightly over-standard. Such areas are verified by encrypting a 1-meter by 1-meter grid, mainly for capturing false positive samples. The third priority: the point marked as a to-be-reviewed area (part of the indicators meet the standard). Such areas are verified as appropriate in combination with historical distribution. In the error attribution stage, the system does not correct all deviations, but rather sorts them by impact: the highest priority: false negative (missed detection) error. That is, the manual finds the anthill but the system does not alarm, which is the most serious failure mode and must be attributed and corrected in priority. The second highest priority: false positive (misjudgment) error. That is, the system alarms but manual does not find, which affects the operation efficiency. The lowest priority: multi-gas data fluctuation anomaly. That is, although the judgment result is correct, the numerical stability is poor, which is handled as an optimization item. Through the hierarchical strategy, the system can ensure that each iteration prioritizes solving the most deadly detection defects. This embodiment solves the problems of difficulty in obtaining true value points in new area general survey and confusion in multi-level benchmark data management.

[0169] In one embodiment of the present application, the dam termite dynamic gas sensing and rapid survey method based on true value calibration can also be used for:

[0170] S1, construct a correctable multi-gas benchmark framework containing universal benchmark, environment model and double-level interface.

[0171] In this embodiment, a multi-gas basic framework and interface support are provided for dynamic correction, a multi-gas benchmark system that can be calibrated on site and iteratively corrected for errors is established, avoiding benchmark rigidity and gas type adaptation limitations. Specifically, a unified standard environment condition is defined, denoted as T0 (temperature), RH0 (relative humidity), V0 (wind speed), S0 (soil parameters), and each parameter is set to allow a fluctuation range ΔT, ΔRH, ΔV, ΔS, ensuring the comparability of laboratory multi-gas benchmark data. The standard environment is determined based on the active habits of termites and the best response conditions of the sensor, providing an anchor point for subsequent environment difference correction. Collect gas samples from N or more live termite nests in the dam, denoted as sample set G, and collect background gas samples from M or more non-termite-damaged dams, denoted as sample set B; the sample set should cover all target gas types that may be produced by termite metabolism, usually 5-20 types, denoted as the total number of target gases m, and the number of a single target gas is i, i=1, 2,..., m. The core parameters in the sample set are determined: the concentration of a single target gas C i (i=1, 2,..., m), and the multi-group characteristic ratio Ri，j = C i / C j (i≠j, i, j∈[1, m]), based on correlation analysis to filter p groups of ratios sensitive to termite damage, p≤m(m-1) / 2, interference gas concentration C int，k (k=1, 2,..., q, q is the number of interference gas species). Based on statistical analysis to determine the basic parameters of the standard environment: single target gas reference concentration C i0 , characteristic ratio reference value R i，j0 , interference gas threshold C int，k0 , establish a structured multi-gas basic parameter reference table, with a gas species expansion field and parameter update field reserved in the table, supporting subsequent correction iteration.

[0172] To eliminate the influence of environmental parameter changes on the concentration of each target gas, collect multi-gas characteristic data under 5 or more gradients of environmental parameters (T a , RH a , V a , a=1, 2,..., 5), based on physical and chemical mechanism and experimental data fitting, establish an environmental correlation model for single target gas (multi-gas is adapted separately, as different gases have different sensitivity to the environment): C i = C i0 ×f i (T)×f i (RH)×f i (V). The model reserves an interface for adjusting the sensitivity coefficient of each gas, supporting updates through field data and error attribution results to improve environmental adaptability of different regions and different gases. To simplify the multi-gas joint determination process, based on the filtered core target gas and characteristic ratios, construct a comprehensive quantitative index P Ci , the formula is as follows: P Ci = ∑ i n (w Ci *(C i / C i0 )) + ∑ i，j=1 p (w Ri，j * (R i，j / R i，j0 )) + ∑ k=1 q (w int，k * (1 - C int，k / C int，k0)) ; wherein n is the number of screened core target gas species, n≤m, the core gas is screened by principal component analysis with variance contribution rate≥85%; p is the number of groups of screened core target gas characteristic ratios included in the model; q is the number of types of interference gases included in the model for interference correction; w Ci is the weight coefficient of the ith core target gas; w Ri，j is the weight coefficient of the ith, j group of characteristic ratios; w int，k is the weight coefficient of the kth interference gas; all weight coefficients satisfy∑ i=1 n w Ci +∑ i，j=1 p w Ri，j +∑ k=1 q w int，k =1, based on the contribution of each parameter to the determination of termite damage, the contribution is determined by statistical analysis; C i , R i，j , C int，k are measured basic parameters; C i0 , R i，j0 , C int，k0 are the reference parameters in the standard environment. Determine the threshold of the suspected area in the standard environment: P Ci_low0 ≤P Ci <P Ci_high0 , the threshold is linked with the multi-gas basic parameter threshold, and supports dynamic updating.

[0173] Design a two-stage dynamic correction interface: the first interface is used to receive field true value calibration data, including the area correction coefficient of each core target gas, the optimized environmental impact factor of each gas area, etc., and is only authorized to write in the area reference branch area, without changing the general reference core; the second interface is used to receive error correction parameters, including multi-gas general reference parameter adjustment value, each gas environment correlation model sensitive coefficient update value, P Ci weight and threshold correction value, authorized to write in the general reference core area and the area reference branch area. The interface supports gas type expansion, adopts JSON standard format, supports 4G / LoRa transmission protocol, delay≤1s, has the function of authority control and data verification, and ensures the safety and accuracy of multi-gas data transmission.

[0174] S2, deploy a multi-target gas anti-interference sensor array to realize accurate collection of multi-gas and environmental parameters.

[0175] The embodiment provides precise data collection support for dynamic correction of multiple gases, so that the collected multiple gas parameters and environmental parameters are accurate and reliable, and meet the data quality requirements of multiple gas reference calibration and correction. Specifically, for n kinds of core target gases, specific response sensors are selected, the detection range of the i-th core target gas sensor is [C imin , C imax ], the response time is ≤t ri , the resolution is ≤δ Ci , and the anti-other target gas cross-response rate is ≤γ i ; wherein t ri is the maximum response time of the i-th core target gas sensor, and γ i is the maximum cross-response rate of the i-th core target gas sensor. For m-n kinds of auxiliary target gases, multi-gas integrated sensors are selected, the detection range covers the concentration interval of each auxiliary gas, and the cross-response rate is ≤γ aux , wherein γ aux is the maximum cross-response rate of the integrated sensor of the auxiliary target gas. For q kinds of interference gases, specific response sensors are selected, the detection range of the k-th interference gas sensor is [C int，kmin , C int，kmax ], the alarm threshold is set to C int，k0 , and when the alarm is triggered, the interference filtering process of the corresponding target gas is automatically started. Temperature, humidity, wind speed and slope sensors are selected synchronously, and the detection accuracies are ±ΔT s , ±ΔRH s , ±ΔV s , and ±ΔS s , respectively, which meet the parameter input requirements of the multi-gas environment correlation model.

[0176] At the hardware layer, the sensor array is packaged in a dustproof and waterproof shell with IP level IPXY, and a built-in metal oxide filter film (filter particle size ≤ the maximum filter particle size d filter of the built-in metal oxide filter film of the sensor shell) is provided to reduce the pollution of soil particles and water vapor to the sensor probe; an independent temperature compensation module is added for each core target gas to correct the influence of environmental temperature on the sensor response value in real time. At the algorithm layer, the sliding average filter + multi-gas adaptive threshold denoising algorithm is integrated, the mean value of n filter sampling point data of each target gas is taken, and the abnormal value deviating from the mean value ±η filter is removed; a multi-gas cross-interference correction algorithm is added, based on the cross-response matrix of each gas sensor, the measured concentration is compensated for interference, and the anti-interference ability of multi-gas joint detection is improved. Set the sampling frequency f sample , and synchronously collect m kinds of target gas concentrations (C1, C2,..., C m ) + q kinds of interference gas concentrations (Cint，1 ,..., C int，q ) + environmental parameters (T, RH, V, S slope ) + positioning data (Beidou / GPS, accuracy ± δ pos ), single group data volume ≤ D single . Upload priority: secondary detection of core target gas data is marked as the highest priority P1, occupying transmission bandwidth; ordinary census of core target gas data is marked as medium priority P2 according to grid batch; auxiliary target gas and historical supplementary data are marked as low priority P3. The local cache capacity of the signal weak area is ≥ D cache , and after the signal is restored, it is supplemented according to the priority order to ensure that there is no loss of multi-gas data. Wherein S slope is the real-time slope value collected by the slope sensor, δ pos is the error accuracy of Beidou / GPS positioning data, D single is the maximum data volume of a single group of collected data, and D cache is the minimum capacity of the local cache of the sensor in the signal weak area. The collected data is packaged according to the requirements of the dynamic correction interface, including gas type number, parameter value, collection time, positioning coordinates, environmental label and other core fields, to ensure that it can be directly used for multi-gas calibration and joint analysis.

[0177] S3, an intelligent detection and closed-loop path planning system adapted to dam terrain.

[0178] Specifically, the robot dog carrier adopts a multi-wheel shock-absorbing chassis, the body height ≤ the maximum height limit value H dog of the body, which is suitable for dense areas with vegetation coverage ≤ the upper limit value ρ veg1 of the vegetation coverage that the robot dog can adapt to work; the multi-gas sensor array is installed at a height H sen1 , the endurance time ≥ the minimum endurance time T bat1 of the robot dog, the single coverage area ≥ the minimum detection area A dog that the robot dog can cover in a single full-power operation, and supports autonomous obstacle avoidance and path correction. The unmanned vehicle carrier adopts a tracked chassis, the ground contact area ≥ the minimum ground contact area S track of the tracked chassis of the unmanned vehicle, the slope adaptation range ≤ the upper limit value α car of the terrain slope that the unmanned vehicle can adapt to work, the soft soil anti-sinking rate ≥ the minimum anti-sinking rate β car of the unmanned vehicle in soft soil environment; the sensor support height can be adjusted in the range [H sen2min , H sen2max ], the endurance time ≥ the minimum endurance time T bat2 of the unmanned vehicle, and the single coverage area ≥ the minimum detection area A carSupport remote control and automatic cruise. The handheld probe carrier is made of lightweight alloy material, and the telescopic rod length range [L polemin , L polemax ] is IP pole waterproof level, and the weight is ≤ the maximum overall weight limit value W pole of the handheld probe. It is suitable for machine / vehicle blind area (dam shoulder gap, water surface edge, etc.), supports single-person continuous operation ≥ the minimum time length T pole supported by the handheld probe for single-person continuous holding operation, and integrates the core target gas rapid detection module.

[0179] Based on the dam environment characteristics, ant damage distribution law and multi-gas background concentration, a risk assessment model is constructed to divide different risk level areas, match different grid densities, and improve the efficiency of multi-gas true value point collection and general survey. The ant damage record area model is: R = w H *H + w V *V + w S *S + w G *G; The new area model without ant damage record is: R = w T *T + w V *V + w S *S + w G' *G'; wherein T is the soil type factor, which is valued according to the suitability of ant damage, and the range is [0, 1]; G' is the multi-gas anomaly factor; w T , w G' are weight coefficients. The grid density is divided as follows: the grid length L high in the high-risk area (R≥risk value upper limit critical value R high ), the grid length L low in the ordinary area (R low ), and the grid length L mid in the transition area (R low ≤R<R high ), to ensure fine detection in the high-risk area and efficient coverage in the ordinary area. The double-standard conditions triggering the secondary detection closed-loop mechanism are: multi-gas basic parameters meet C ilow0 ≤C i <C ihigh0 (i=1, 2,..., n), the selected characteristic ratio meets R i，j ≥R i，j0 (i, j ∈ [1, p]), and all interference gas concentrations meet C int，k ≤C int，k0 (k=1, 2,..., q); P Ci meets P Ci_low0 ≤PCi<P Ci_high0 ; wherein C ilow0C is the lower limit threshold for the concentration of the i-th core target gas. ihigh0 Let be the upper limit threshold for the concentration of the i-th core target gas. Let the center of the suspected area be a circle with radius R. det The spiral path, with an adjacent loop spacing L det , each interval D det Data was collected once, and the sensor height remained at H. sen The fluctuation range of multi-gas data exceeds the maximum permissible fluctuation range threshold η for multi-gas data acquisition. fluc In such cases, priority is given to correcting the environmental parameters using a single gas and performing multi-gas spatiotemporal interpolation. If the fluctuation still exceeds the standard after correction, a second sampling is performed during a period of environmental stability. If the second sampling still results in abnormalities, the sampling point ΔL is finely adjusted. adj Subsequent retesting ensures the reliability of multi-gas data.

[0180] S4. Generate a regional multi-gas dynamic benchmark through on-site true value calibration to complete the first-level correction.

[0181] In this embodiment, a region-specific dynamic multi-gas benchmark is generated by calibrating a general benchmark using on-site multi-gas true data. This forms the first closed loop of dynamic correction and solves the problem of adapting a general multi-gas benchmark to the specificity of the region. Specifically, in areas with existing ant infestations, n... cal1 For each live nesting site verified by manual excavation within one year, the following conditions must be met: the main soil type, vegetation cover, and slope type within the covered area; the distance between sites must be greater than or equal to the minimum distance requirement D. cal There are no external sources of interference, such as pesticides or industrial waste gas. In new areas free from termite infestations, an environmental analogy calibration method is used to screen areas with soil types ≥ the minimum soil type similarity γ of the new area. sim Climate parameter differences ≤ Maximum permissible difference value Δ cli The difference in vegetation coverage is less than or equal to the maximum allowable difference in vegetation coverage Δ. veg Multiple gas data points from neighboring regions are collected simultaneously, along with new data from n regions. cal2 One high-risk candidate point (R ≥ risk threshold R) cand Multi-gas data are used as temporary anchor points. Each true value point has n values ​​set within a 1m × 1m area. sub There are n sampling sub-points, and the sampling is repeated n times. rep Simultaneously record the concentrations of m target gases, q interfering gases, and environmental parameters; remove abnormal environmental data (wind speed > wind speed threshold V). abn Sudden humidity change > Maximum permissible value of sudden humidity change ΔRH abn The minimum number of times n to obtain valid sampled data. val The mean of ≥2 valid data points is used as the multi-gas representative value for the true value point.

[0182] To eliminate the difference between the field environment and the standard environment, the actual environmental concentration C ical of the i-th target gas collected by the true value point is corrected to the standard environmental equivalent concentration C ical0 by the corresponding single-gas environment correlation model: C ical =C i (T cal )÷f i (RH cal )÷f i (V cal ); wherein C ical0 is the true value point standard environmental equivalent concentration of the i-th target gas; C ical is the true value point actual environmental collection concentration of the i-th target gas; f i (T cal ), f i (RH cal ), f i (V cal ) are the environmental impact factors of the i-th gas calculated based on the true value point field environment parameters. Similarly, the standard environmental equivalent concentrations C 1cal0 , C 2cal0 ,..., C ncal0 of all core target gases and the standard environmental equivalent values R i，jcal0 =C ical0 / C jcal0 of the characteristic ratios are calculated. The single-gas concentration correction coefficient is calculated based on the ratio of the standard environmental equivalent concentration to the universal reference, and the formula is as follows: K Ci = C i，cal0，avg / C i0 ; wherein K Ci is the regional concentration correction coefficient of the i-th core target gas; C i，cal0，avg is the average of the standard environmental equivalent concentrations of the i-th target gas of all true value points; C i0 is the universal reference concentration of the i-th target gas. The characteristic ratio correction coefficient is: K Ri，j =R i，jcal0，avg / R i，j0 ; wherein R i，jcal0，avg =C i，cal0，avg / C j，cal0，avg is the average of the standard environmental equivalent values of the i, j group of characteristic ratios of all true value points; R i，j0 is the universal reference value of the i, j group of characteristic ratios. The multi-gas comprehensive correction coefficient is calculated based on the correction coefficients of all core target gases and characteristic ratios, which is used for P Ci area threshold calibration, and the formula is as follows: K com =(∑ i= 1 n w Ci ×KCi +∑ i，j=1 p w Ri，j ×K Ri，j ) / (∑ i=1 n w Ci +∑ i，j=1 p w Ri，j Based on the derivation of the single-gas correction coefficient, the regional core target gas benchmark is generated, as shown in the following formula: C ireg =C i0 ×K Ci ; where C ireg Let R be the region-specific baseline concentration of the i-th core target gas. The region characteristic ratio baseline is: R i，jreg =R i，j0 ×K Ri，j ;where R i，jreg This is the region-specific baseline value for the feature ratios of the i-th and j-th groups. Region P Ci The threshold is: P Ci_lowreg =P Ci_low0 ×K com ;P Ci_highreg =P Ci_high0 ×K com ;where P Ci_lowreg P Ci_highreg For regional exclusive P Ci Suspected area threshold. Based on the true value point field environmental parameters and multi-gas concentration data pairs, the environmental impact factor sensitivity coefficient (k) of each core target gas is refitted. Tireg k RHireg k Vireg This generates a region-specific multi-gas environment correlation model, improving the accuracy of local environment correction. Where k... Tireg k is a region-specific temperature sensitivity coefficient. RHireg k is a region-specific humidity sensitivity coefficient. Vireg This is a region-specific wind speed sensitivity coefficient.

[0183] Through the first-level dynamic correction interface, the regional core target gas benchmark, regional characteristic ratio benchmark, and P are integrated. Ci The regional threshold and optimized regional multi-gas environment correlation model are written into the regional benchmark branch area of ​​the benchmark database, isolated from the general benchmark core area to ensure the stability of the general benchmark. Cross-validation is performed: multi-gas data are collected at the same ground truth point using a robot dog and an unmanned vehicle. The deviation from the regional dynamic benchmark is ≤ the cross-validation deviation threshold η. cross ; Perform analog verification in the new region: the deviation between the temporary anchor point multi-gas data and the analog true point multi-gas data ≤ analog verification deviation threshold η anaIf the verification fails, re-screen the true value points or optimize the multi-gas environment correlation model until the deviation requirement is met.

[0184] S5, dam scene multi-module closed-loop coupling analysis and positioning.

[0185] In this embodiment, the multi-gas dynamic reference of the calling area is called, combined with real-time environment correction, to complete the accurate positioning of the suspected area, which is the core application link of dynamic correction results, and ensures the comparability and accuracy of multi-gas survey data. Specifically, the multi-gas basic parameters synchronously collected by the sensor array are received, and the multi-gas P Ci value is calculated based on the formula, and the measured basic parameters and the regional multi-gas basic parameter reference are substituted into the formula. Ci value. The measured concentration of the i-th core target gas is corrected to the standard environment equivalent concentration by calling the regional exclusive multi-gas environment correlation model, and the formula is as follows: C icorr =C imeas ÷f i (T meas )÷f i (RH meas )÷f i (V meas ); wherein C icorr is the corrected standard environment equivalent concentration of the i-th core target gas; C imeas is the measured concentration of the i-th core target gas; f i (T meas ), f i (RH meas ), and f i (V meas ) are regional environment influence factors calculated based on the measured environment parameters. Substitute the corrected basic parameters of all core target gases into the formula to recalculate the P Ci value and obtain the corrected P Ci_corr . Dual-track determination is adopted to ensure the accuracy of the determination and avoid misjudgment of a single gas indicator: multi-gas basic parameter determination: the corrected standard environment equivalent concentration of all core target gases meets the regional dynamic reference characteristic label: the lower limit of the regional core gas concentration reference C ilowreg ≤C icorr < the upper limit of the regional core gas concentration reference C ihighreg (i=1, 2,..., n), the selected characteristic ratio meets R i，jcorr ≥R i，jreg (i, j ∈ [1, p]), and the concentration of all interference gases meets C int，k ≤C int，k0 (k=1, 2,..., q); PCi determination: P Ci_lowreg ≤P Ci_corr <P Ci_highreg . wherein Ri，jcorr is the corrected equivalent value of the i, j group characteristic ratio; P Ci_corr is the regional exclusive suspected interval of the corrected comprehensive index. Final judgment: both meet, preliminary confirmation as suspected area; either core target gas or characteristic ratio does not meet, exclude suspected area; only part of the core gas meets the standard but P Ci meets the standard, marked as a pending review area, and subsequent focus.

[0186] Temporal-spatial and environmental two-dimensional verification: environmental verification (preliminary threshold): the measured environmental parameters need to meet the environmental adaptation range in the regional characteristic label: T regmin ≤T meas ≤T regmax , RH regmin ≤RH meas ≤RH regmax , V meas ≤V regmax ; if not passed, trigger supplementary sampling or site adjustment, and do not enter the temporal-spatial verification; wherein T regmin and T regmax , RH regmin and RH regmax , V regmax are the environmental characteristic labels of the dam area, representing the optimal environmental interval for stable dispersion of ant pest gases in this area; exceeding this interval will cause abnormal gas diffusion / dissolution, and the collected data has no judgment value and must be supplemented. T meas is the measured temperature, RH regmin is the measured humidity, and V meas is the measured wind speed. Temporal-spatial verification (post-confirmation): time dimension: take n temp groups of continuous data in the stable environment period of the second detection, calculate the fluctuation coefficient σ icorr =(C Ci_corr -C ti ) / C imax , σ imin =(P iavg -P t ) / P Ci_max of each core target gas C Ci_min and P Ci_avg , requiring all σ ti ≤η tfluc and σ t ≤η tfluc ; wherein C iavg is the average corrected concentration of the i-th gas, P Ci_max is the maximum multi-gas principal component value, P Ci_min is the minimum multi-gas principal component value, P Ci_avg is the average multi-gas principal component value, C imax is the maximum corrected concentration of the i-th gas, Cimin is the minimum correction concentration of the i-th gas, σ ti is the corrected concentration of the i-th core target gas C icorr is the time fluctuation coefficient of the i-th core target gas, σ t is the time fluctuation coefficient of the corrected comprehensive index P Ci_corr is the time fluctuation coefficient of the corrected comprehensive index P tfluc is the maximum allowed threshold of the time fluctuation coefficient. Spatial dimension: C 1corr ,..., C ncorr , P Ci_corr trend, which needs to comply with the multi-gas diffusion law, such as the concentration of the core target gas gradually decreasing from the center to the outside. If the verification is passed, it is confirmed as the final suspected area; if the verification is not passed, the multi-gas deviation reason is analyzed in combination with the environmental parameters, and it is marked as a dynamic monitoring area and included in the subsequent review.

[0187] S6, manual excavation verification and feedback of multi-gas true value.

[0188] The multi-gas physical true value is obtained by manual excavation, the deviation of the detection data from the regional multi-gas dynamic benchmark is quantified, and the core basis for error correction is provided. Specifically, the verification area selection priority is as follows: first priority: high-priority suspected area that passes the time-space-environmental dual-dimension verification and meets the double-track judgment, positioned at the highest point of the secondary detection characteristic value (P Ci_corr ) with a positioning deviation ≤ the maximum allowed threshold of the positioning deviation δ loc ; second priority: dynamic monitoring area, environmental stability but time-space fluctuation exceeds the standard, verification points are selected by 1m×1m grid encryption; third priority: area to be reviewed, part of the core gas meets the standard but P Ci does not meet the standard, verification points are selected in combination with the historical distribution of regional termite pests; points with clear interference reasons such as sustained gust and external gas are temporarily not verified and included in the subsequent review during the stable environmental period. The excavation range is: high-priority area excavation size L exc1 ×W exc1 ×H exc1 , dynamic monitoring area excavation size L exc2 ×W exc2 ×H exc2 , which adapts to the common burial depth range of termite nests. Whether live termites / nests are found, nest depth H nest , nest size S nest are recorded; at the same time, the calibrated portable multi-gas sensor is used to collect the on-site core target gas concentration C ifield (i=1,2,...,n); the on-site environmental parameters T field , RH field , V field are recorded. Based on the on-site true value data, the matching degree of the detection data of each core target gas and the regional dynamic benchmark is calculated, and the formula is as follows: γmatch，i = (1 - (|C icorr ifield0 |) / C ifield0 ) × 100%; wherein γ match，i is the concentration matching degree of the i-th core target gas; C ifield0 = C ifield ÷ f i (T field ) ÷ f i (RH field ) ÷ f i (V field ) is the standard ambient equivalent concentration of the i-th core target gas field true value; C icorr is the standard ambient equivalent concentration of the i-th core target gas detection after correction. Comprehensive matching degree: γ match，com =∑ i=1 n w Ci ×γ match，i ; if γ match，com ≥η match and a living nest is found, it is a true positive; if γ match，com <η match and no nest is found, it is a false positive; no suspected area is determined but a nest is found, which is a false negative. wherein η match is the qualified judgment threshold of multi-gas comprehensive matching degree.

[0189] Record the verification data in a unified format to form a structured multi-gas data set, which includes the following core fields: verification ID, detection point coordinate, each core target gas detection corrected parameter (C 1corr ,..., C ncorr ), each core target gas field true value parameter (C 1field0 ,..., C nfield0 ), single gas matching degree γ match，i , comprehensive matching degree γ match，com , excavation result, preliminary analysis of deviation reason; through a dedicated interface, it is synchronized to the error attribution module, and data transmission adopts an encryption protocol to ensure the reliability of multi-gas true value data.

[0190] S7, error data attribution and model correction, complete secondary dynamic correction.

[0191] In this embodiment, based on multi-gas true value feedback, the error root related to dynamic correction is located, and multi-gas reference and model are optimized, which is the second heavy closed loop of dynamic correction, and solves the problem of long-term iterative optimization of multi-gas reference and model. Specifically, if any of the following conditions is met, it is determined as error data: comprehensive matching degree γ match，com <comprehensive matching degree error judgment threshold η errmatch ​; the deviation of any core target gas detection corrected data from the regional dynamic baseline is greater than the deviation error threshold η of the corrected data from the regional dynamic baseline errdev ; multiple re-sampling still cannot meet the verification standard. According to the influence degree on the accuracy of the census, the order is: false negative (missed detection) > false positive (misjudgment) > multiple gas data fluctuation anomaly, ensuring that high-impact errors are corrected first. Establish an error attribution matrix to focus on locating the core reasons related to multiple gas baseline, model, and threshold, avoiding interference from irrelevant factors: if the attribution dimension is the multiple gas baseline fitting deviation, the typical data characteristics are true positive but the detection does not meet the standard, and the deviation of the regional multiple gas baseline from the equivalent concentration of the true value on site is greater than the deviation threshold η of the regional multiple gas baseline from the equivalent concentration of the true value on site basedev , then the correction object is the general multiple gas baseline parameter and the regional multiple gas baseline parameter; if the attribution dimension is the multiple gas environment model deviation, the typical data characteristics are environmental stability, i.e. fluctuation ≤ environmental stability judgment threshold η envstab , but the multiple gas correction matching degree is less than the multiple gas environment model matching degree threshold η modmatch , then the correction object is the general multiple gas environment correlation model and the regional multiple gas environment correlation model; if the attribution dimension is the multiple gas threshold setting deviation, the typical data characteristics are false positive / false negative concentrated at the threshold boundary (|P Ci_corr - Ci_lowreg / highreg | < threshold boundary judgment interval Δ thres ), then the correction object is the standard environment P Ci threshold and the regional P Ci threshold; if the attribution dimension is the multiple gas interface transmission deviation, the typical data characteristics are multiple gas parameter distortion during data synchronization, and the deviation from the local cache data of the sensor is greater than the interface transmission deviation threshold η transdev , then the correction object is the dynamic correction interface data verification rule.

[0192] Based on historical error rate statistics, set weight coefficients for each attribution dimension to ensure that the correction priority matches the error impact degree: weight setting: W = [w base , w mod , w thres , w trans ] satisfies w base +w mod +w thres +w trans =1, where the multiple gas baseline fitting deviation weight w base > the multiple gas environment model deviation weight w mod > the multiple gas threshold setting deviation weight w thres > the multiple gas interface transmission deviation weight w trans . Correction formula: P corr= P orig ×∑ k=14 (W k ×F k ); where P corr These are the corrected parameter values, including multi-gas reference parameters, model sensitivity coefficients, thresholds, etc.; P orig The original parameter values ​​before correction; W k F represents the weighting coefficient. k F is a correction factor for each attribution dimension, derived based on the overall fit and bias value. k =1-(η err -γ match，com ) / η err η err To allow the maximum deviation, for different correction targets, substitute the corresponding parameters and correction factors to calculate the specific correction value: Multi-gas reference parameter correction: C i0corr =C i0 ×(W base ×F base +W mod ×F mod Multi-gas environment model correction: k Ticorr =k Ti ×(W mod ×F mod Multi-gas threshold correction: P Ci_low0corr =P Ci_low0 ×(W thres ×F thres ) Calculate the multi-gas simulation overall matching degree γ corresponding to the corrected parameters. match，sim , requires γ match，sim ≥η corr If the conditions are not met, the weights and correction factors will be readjusted; where F base F is a correction factor for multi-gas reference adaptation bias. mod F is a correction factor for the bias of multi-gas environment models. thres A correction factor, η, is set for the deviation of the multi-gas threshold. corr To correct the overall matching threshold of the simulation, the following content is updated synchronously through the secondary dynamic correction interface: multi-gas benchmark parameters in the general benchmark core area, sensitivity coefficients of each gas environment correlation model, and P. Ci Weights and thresholds; regional multi-gas dynamic benchmarks and regional multi-gas environmental models for regional benchmark branch areas. The updated multi-gas benchmarks and models are automatically synchronized to the coupling analysis module of subsequent surveys, realizing a closed-loop iteration of correction → application → re-verification → re-correction, ensuring continuous improvement in multi-gas detection accuracy.

[0193] like Figure 5As shown, according to one aspect of the present application, four core functional modules are arranged: Beidou positioning system, gas sensor array, wireless data transmission terminal and analysis software system, and each module constitutes a whole framework for collaborative operation through the intermediate connecting structure. The Beidou positioning system is used to obtain the geographic position information of the equipment in real time, providing data support for the spatial positioning of termite monitoring; the gas sensor array as the core sensing component, when there is termite-related characteristic gas in the environment, gas molecules will be adsorbed by the sensitive material on the surface of the gas sensitive element, causing specific changes in the physical properties of the sensitive material, which can be used to collect characteristic gas signals related to termite activity in the environment; the wireless data transmission terminal is responsible for wireless transmission of the gas data collected by the gas sensor array and the position data obtained by the Beidou positioning system; the analysis software system adopts a neural network structure including an input layer, a hidden layer and an output layer to analyze and process the transmitted gas data, realizing the identification and judgment of termite activity. Each module works collaboratively through signal interaction to complete the sensing, transmission and analysis functions of termite monitoring survey. Figure 6 As shown, before the survey personnel carry out the survey, the dam slope is divided into several independent detection units; the survey equipment host of the present application is deployed in the working area at the top of the dam, and the collection end of the gas sensing survey equipment is attached to the slope detection unit area, and the equipment relies on the pipeline system to move along the reciprocating flow path shown by the arrow in the figure to cyclically and fully cover the gas in each unit; in this process, the equipment synchronously completes real-time analysis of gas signals and identification of termite activity, and quickly locates the suspected termite activity area in combination with the spatial partition information of the dam, finally realizing efficient survey operation in the dam scene.

[0194] In one detailed embodiment, a certain dam on the main stream of the Yangtze River is 3.2 kilometers long, the dam body is clay, and the vegetation coverage rate is about 55%. Historical records show that termite activity signs were found in this dam section in 2019. The survey time is in the middle of June, the temperature is 28°C, the relative humidity is 72%, and the wind speed is 0.8 meters per second. The detection team carries the robot dog carrier into the target dam section. The system loads the general multi-gas benchmark, in which the carbon dioxide benchmark concentration C10 is 1200 ppm, the methane benchmark concentration C20 is 4.8 ppm, and the standard environment is set to temperature 25°C, humidity 65%, and wind speed 0.5 meters per second. Since there is a historical termite damage record in this dam section, the team selects four positions near the discovery point in 2019 as true value points for calibration sampling. After the collected raw data is converted by the environmental correlation model, the standard environment equivalent concentration mean value is obtained: carbon dioxide is 1380 ppm, and methane is 5.28 ppm. According to this, the regional correction coefficients are calculated: carbon dioxide correction coefficient K C1 = 1380 / 1200 = 1.15, and methane correction coefficient K C2= 5.28 / 4.8 = 1.10. The system generates a regional multi-gas dynamic baseline, with the carbon dioxide decision threshold raised to 1380 ppm and the methane decision threshold raised to 5.28 ppm. The historical termite damage record factor H = 1, the vegetation coverage factor V = 0.55 (after normalization, about 0.92), the slope factor S = 0.3 (the average slope is about 18 degrees), and the multi-gas background concentration factor G = 0.85 are obtained for this section of the embankment. Using the formula R = 0.3 x 1 + 0.2 x 0.92 + 0.2 x 0.3 + 0.3 x 0.85 = 0.799, the risk level value R = 0.799 > 0.7, and the overall section of the embankment is classified as a high-risk area, and the general survey grid length is set to 1.0 meters.

[0195] The robotic dog starts scanning at a 1.0-meter grid spacing. When it moves to the coordinate point (152.3, 45.7), the sensor detects a carbon dioxide concentration of 1650 ppm and a methane concentration of 6.2 ppm, and after environmental normalization conversion, the standard environmental equivalent concentrations are 1485 ppm and 5.58 ppm, respectively, both of which exceed the regional dynamic baseline, triggering the gradient tracking mode. The system samples in 8 directions (45 degrees apart) around this point, and calculates that the composite gradient in the northeast direction (45 degrees) is the largest. The robotic dog moves 0.5 meters in the northeast direction, and the concentration continues to rise to 1720 ppm of carbon dioxide. After 6 rounds of iterative tracking, a concentration peak (carbon dioxide 1890 ppm, methane 7.1 ppm) is detected at the coordinate point (153.8, 47.2), and the gradients in all directions around it are less than the threshold value, i.e. 0.05 ppm / m, so the system locks this position as the gas channel exit. The multi-gas principal component value P Ci at this point is calculated using the formula ∑ i (w Ci * (C i / C i0 )) + ∑ j (w Rj * (R j / R j0 )) + ∑ k (w int，k * (1 - C int，k / C int，k0 )) = 0.3 x 0.85 + 0.2 x 0.92 + 0.2 x 0.3 + 0.3 x 0.85 = 0.799. Ci= 0.83, which is in the suspicious interval [0.71, 0.95]. Meanwhile, both the core gas concentration and the characteristic ratio meet the threshold requirements, and the dual-track determination passes. The space-time-environmental dual-dimension verification is performed. Environmental verification: the current temperature of 28°C is within the effective range (25 ± 10°C), and the humidity of 72% is within the effective range (65 ± 15%), so the environmental verification passes. Space-time verification: the robot dog performs secondary detection around the point, and the concentration fluctuation coefficient σ = 3.2% < 10% within 30 seconds, and the spatial distribution conforms to the point source diffusion model, i.e., the center concentration is high and decreases outward, so the space-time verification passes. The suspected termite nest area is output, with coordinates (153.8, 47.2) and high confidence.

[0196] The soil gas transport attenuation model is called for depth inversion. Carbon dioxide (λ1=0.20 m) and methane (λ2=0.35 m) are selected as the dual-gas pair. The surface detection concentration ratio R surface = 1890 / 7.1 ≈ 266, and the pre-stored ant nest source concentration ratio R source = 250. Substituting into the formula H = (ln(R surface ) - ln(R source )) / (1 / λ2 - 1 / λ1) ≈ -0.028. Since the calculation result is negative, it indicates that the current soil state influence factor needs to be corrected. The system retrieves real-time soil parameters: porosity φ = 0.42, water content θ = 0.28, soil temperature T soil = 24°C, and after recalculating the soil state influence factor, the corrected depth estimate H ≈ 0.45 meters. In the output results, it is marked that the suspected nest depth is about 0.45 meters, and excavation verification is recommended. The verification team excavates at coordinates (153.8, 47.2). Before excavation, the calibrated probe is used to collect the field true value gas concentration. When excavated to a depth of 0.52 meters, live termite colonies and fungus garden structures are found, confirming the existence of the nest. The calculation of the comprehensive matching degree: the matching degree γ match = 92% between the standard environmental equivalent concentration during detection and the field true value standard environmental equivalent value is greater than the iteration threshold of 70%, and no error attribution correction is triggered. The depth inversion error = (0.52-0.45) / 0.52 = 13.5% is within the acceptable range. The system records this verification data for subsequent statistical analysis and model optimization.

[0197] Overall, the dam termite dynamic gas sensing and rapid census method based on true value calibration includes: constructing a correctable framework containing a universal reference and an environment-related model; using field true value points to derive regional correction coefficients to generate a regional dynamic reference that adapts to the current environment; in the census stage, an adaptive tracking strategy based on gas concentration gradient is used to actively lock the abnormal source; a soil gas transport attenuation model is constructed based on Fick's diffusion law, the depth of the nest is physically inverted, and a double-track judgment is made in combination with the principal component values of multiple gases; through artificial verification data feedback, the reference and model are iteratively corrected in a closed loop.

[0198] The application constructs a soil gas transport attenuation model based on Fick's diffusion law, uses the physical differences in the characteristic diffusion lengths of different gas molecules in soil media to establish a quantitative mapping of surface detection concentration to underground nest source concentration, and realizes accurate calculation of the depth of hidden nests through a double-gas ratio inversion algorithm. An adaptive tracking strategy based on gas concentration gradient is used to replace the traditional fixed grid scanning, and the multidirectional concentration gradient vector calculated in real time is used to simulate the biological olfactory mechanism to actively plan the path and lock the gas channel outlet, thereby improving the census efficiency and positioning accuracy. A two-stage dynamic correction closed-loop mechanism is implemented to generate a regional exclusive dynamic reference through field true value calibration, thereby solving the cold start problem of water and soil. Through error attribution feedback of artificial verification data, long-term evolution of model parameters is realized, thereby solving the problem of poor environmental adaptability.

[0199] The above describes the preferred embodiments of the application, but the application is not limited to the specific details in the above embodiments. Within the technical concept of the application, various equivalent transformations of the technical solutions of the application can be made, and these equivalent transformations all belong to the protection scope of the application.

Claims

1. A method for embankment termite dynamic gas sensing and rapid surveying based on true value calibration, characterized in that, The method comprises the following steps: Establishing a correctable reference framework including a universal multi-gas reference and a multi-gas environment correlation model, wherein the universal multi-gas reference comprises multi-gas basic parameters in a standard environment; Obtaining field true value point data in a target area, calculating the standard environment equivalent concentration of the field true value point data, and deriving a regional correction coefficient based on the same; Correcting the universal multi-gas reference by using the regional correction coefficient to generate a regional multi-gas dynamic reference suitable for the target area; Performing a census detection in the target area, collecting census point data in real time, and converting the census point data based on the multi-gas environment correlation model; Judging the converted census point data based on the regional multi-gas dynamic reference, and outputting the position information of suspected termite nest areas; Establishing a correctable reference framework, including constructing a soil gas transmission attenuation model, specifically: Constructing a soil gas transmission attenuation model for each core target gas; The soil gas transmission attenuation model defines the mapping relationship between the ground detection concentration and the nest source concentration, which is determined by the nest depth, the characteristic diffusion length of the gas, and the soil state influence factor; The soil state influence factor is calculated based on the soil porosity, volume water content, and soil temperature; Performing a census detection in the target area, and adopting an adaptive tracking strategy based on the gas concentration gradient, specifically including: Collecting real-time concentration data of the core target gas in each detection direction around the current detection point; Calculating the gas concentration gradient vector of each detection direction based on the real-time concentration data; Integrating the gas concentration gradient vectors of each detection direction to determine the optimal tracking direction, and controlling the detection carrier to move to the next detection point along the optimal tracking direction; Iteratively performing the collection, calculation, and movement steps until the gas source is locked; Determining the optimal tracking direction and locking the gas source, including: Based on the gas concentration gradient vector, using a weighted summation algorithm combined with an indicator function to calculate the synthetic gradient direction of the predetermined N core target gases as the optimal tracking direction, wherein only the contribution of the positive gradient direction is counted; N is a natural number; When the concentration change of the consecutive detection points meets the preset concentration peak condition and the gas concentration gradient vectors of each direction meet the preset gradient convergence condition, it is determined that the current position is the outlet of the gas channel; After locking the outlets of each gas channel, the spatial distribution characteristics of each outlet are calculated, and if each gas channel outlet presents a central radiation, it is associated as a multi-outlet of the same termite nest system; Deriving the regional correction coefficient, including: Collecting real-time environmental parameters of the field true value points, eliminating the influence of the real-time environmental parameters by using the multi-gas environment correlation model, converting the field true value point data into concentration values in the standard environment as the standard environment equivalent concentration; Calculating the mean value of the standard environment equivalent concentrations of all field true value points; Calculating the ratio of the mean value to the corresponding parameter in the universal multi-gas reference to obtain the concentration correction coefficient for each core target gas and the ratio correction coefficient for the gas characteristic ratio, and collectively referring to the concentration correction coefficient and the ratio correction coefficient as the regional correction coefficient.

2. The method of claim 1, wherein, The soil gas transmission attenuation model also includes establishing a nest depth inversion mechanism, specifically: Selecting a first core target gas and a second core target gas with different characteristic diffusion lengths; Inversion calculation of the termite nest depth by using the surface detection concentration ratio of the first core target gas and the second core target gas, combined with the pre-stored termite nest source concentration ratio; The calculation formula of the termite nest depth H is: H = (ln(R surface ) - ln(R source )) / (1 / λ2 - 1 / λ1); where R surface is the surface detected concentration ratio, R source is the pre-stored nest source concentration ratio, and λ1 and λ2 are the characteristic diffusion lengths of the first and second core target gases, respectively.

3. The method of claim 1, wherein, Output the position information of the suspected termite nest area, including: Combined with the core target gas concentration, gas characteristic ratio in the converted general survey point data and the preset interference gas concentration, the multi-gas principal component value is calculated; If the core target gas concentration and the gas characteristic ratio of the converted general survey point data both meet the threshold range of the regional multi-gas dynamic benchmark, and the multi-gas principal component value is in the preset suspected interval, it is determined that the region where the current general survey point is located is a suspected termite nest area.

4. The method of claim 1, wherein, Before outputting the position information of the suspected termite nest area, it also includes time-space-environmental two-dimensional verification, specifically: Environmental verification: judge whether the real-time environmental parameters when collecting the general survey point data are within the effective adaptation range of the regional multi-gas dynamic benchmark; Time-space verification: when the detection carrier performs secondary detection around the suspected termite nest area, verify the time sequence fluctuation stability and spatial diffusion trend of the core target gas concentration in the converted general survey point data; Only when the environmental verification and the time-space verification both pass, the operation of outputting the position information of the suspected termite nest area is performed.

5. The method of claim 1, wherein, After outputting the position information of the suspected termite nest area, it also includes closed-loop feedback and model evolution, specifically: Obtain the artificial verification true value for the suspected termite nest area, which contains the nest existence result and the field true value gas concentration; Calculate the comprehensive matching degree between the standard environmental equivalent concentration at the time of detection and the standard environmental equivalent value of the field true value gas concentration; If the comprehensive matching degree is lower than the preset iteration threshold, output the deviation feature, and perform error attribution analysis accordingly to generate update parameters for the general multi-gas benchmark or the multi-gas environmental correlation model; Use the update parameters to modify the general multi-gas benchmark or the multi-gas environmental correlation model, and apply the modified benchmark or model to subsequent general survey detection.

6. The method of claim 5, wherein, Generating update parameters includes: Constructing an error attribution matrix including benchmark adaptation bias dimensions, environmental model bias dimensions and threshold setting bias dimensions, and establishing a mapping relationship between the deviation feature and each dimension of the error attribution matrix; Based on the mapping relationship, assign weight coefficients to each dimension of the error attribution matrix, and calculate the correction factor of each dimension according to the comprehensive matching degree; Use the weighted correction formula to calculate the update parameters, and the weighted correction formula is: P corr = P orig ×∑(W k ×F k ); where P corr is the updated parameter, P orig is the original parameter before correction, W k is the weight coefficient of the kth dimension, F k is the correction factor of the kth dimension, and ∑ represents the weighted sum over all relevant dimensions.

Images