A method for detecting water content of hydraulic engineering materials

Abstract

The application provides a kind of water conservancy engineering material moisture content detection method, belongs to water conservancy engineering material detection technical field, comprising: according to water conservancy engineering site, according to relevant standard selects different types of representative sample, based on resistance measurement mode, capacitance measurement mode is sequentially measured to corresponding representative sample multiple times in multiple points, establishes resistance-moisture content relationship function, capacitance-moisture content relationship function;Based on infrared spectrum measurement mode, corresponding representative sample is scanned at different set wavelengths to obtain several infrared spectrograms, and all infrared spectrograms are analyzed, and combined with the pre-established infrared spectrum-moisture content database, construct spectrum-moisture content relationship function;Resistance-moisture content relationship function, capacitance-moisture content relationship function, spectrum-moisture content relationship function are fused and analyzed, and the moisture content of the corresponding representative sample is obtained, to provide reliable data support for water conservancy engineering quality control.

Description

Technical Field

[0001] This invention relates to the field of water conservancy engineering material testing technology, and in particular to a method for testing the moisture content of water conservancy engineering materials. Background Technology

[0002] In water conservancy projects, such as farmland irrigation systems, irrigation and drainage canals, and slope protection projects, the moisture content of materials has a crucial impact on the quality and stability of the project. For example, in the construction of irrigation and drainage canals, if the moisture content of the canal bank materials is too high, it may cause the canal bank to collapse under the action of water flow; in slope protection projects, changes in the moisture content of the slope protection materials will affect their resistance to weathering and erosion. Traditional moisture content testing methods, such as the drying method, although highly accurate, have the problems of long testing time and complicated operation, making it difficult to meet the needs of rapid on-site testing in water conservancy projects; rapid testing methods such as the alcohol combustion method have relatively low detection accuracy and are not suitable for some water conservancy projects with strict requirements for moisture content.

[0003] Therefore, this invention proposes a method for detecting the moisture content of materials used in water conservancy projects. Summary of the Invention

[0004] This invention provides a method for detecting the moisture content of materials used in water conservancy projects, in order to solve the aforementioned technical problems.

[0005] This invention provides a method for detecting the moisture content of materials used in hydraulic engineering, comprising:

[0006] Step 1: Based on the site of the water conservancy project, select representative samples of different types according to relevant standards. The different types include soil, concrete or sand and gravel. The representative samples are samples of raw materials processed from the selected location.

[0007] Step 2: Based on the resistance measurement method and the capacitance measurement method, perform multiple measurements at multiple points on the corresponding representative samples to establish the resistance-moisture content relationship function and the capacitance-moisture content relationship function;

[0008] Step 3: Based on infrared spectroscopy measurement, scan the corresponding representative samples at different set wavelengths to obtain several infrared spectra, analyze all infrared spectra, and construct the spectrum-moisture content relationship function in combination with the pre-established infrared spectrum-moisture content database.

[0009] Step 4: Perform a fusion analysis on the resistance-moisture content relationship function, capacitance-moisture content relationship function, and spectrum-moisture content relationship function to obtain the moisture content of the corresponding representative samples.

[0010] Preferably, before selecting representative samples of different types according to relevant standards based on the site conditions of the water conservancy project, the following steps are taken:

[0011] The project structure is obtained by scanning the water conservancy project site with a drone, and the target material selection area is automatically locked based on the construction type of each sub-project in the project structure.

[0012] Based on the construction batch record log of the target material selection area, the corresponding area is divided, and a significant hierarchical structure of the target material selection area is constructed. The division result includes the range of three-dimensional construction location points, construction edge line, and construction materials of each batch of construction in the corresponding area.

[0013] The hierarchical structure is divided into three dimensions using standard units. The construction vector of each three-dimensional block is determined, and the construction vector is input into the vector analysis model to obtain the moisture content influence coefficient caused by construction differences.

[0014] Based on all moisture content influence coefficients, determine the coefficient variance and coefficient mean, and select the initial sampling quantity that matches the target material region, coefficient variance, and coefficient mean from the variance-mean-region size-quantity comparison table;

[0015] The moisture content influence coefficient is mapped to the target material area and the coefficient position is classified. The highlight coefficient is filtered and the first position of the highlight coefficient in the target material area is locked.

[0016] When the number of the first positions is less than 3, random selection is made in the target material selection area according to the initial sampling quantity;

[0017] When the number of the first position is greater than or equal to 3, the initial sampling quantity is divided into a first quantity based on the first position and a second quantity based on the non-first position. The maximum boundary of the target material selection area is drawn according to the distribution of the first position, and the longest distance and the shortest distance between the two nearest points in the target material selection area are locked. The number of additional non-first positions is determined. At the same time, the initial sampling quantity is divided into a first quantity based on the first position and a second quantity based on the non-first position.

[0018] Random selection is performed on non-first positions according to the sum of the added quantity and the second quantity, and random selection is performed on the first position according to the first quantity;

[0019] Representative samples were obtained from the selected locations according to relevant standards.

[0020] Preferably, determining the number of additional units in non-first positions includes:

[0021]

[0022] in, This represents the rounding function; Indicates the number of additions; Select the area of ​​the inner region below the maximum boundary of the target material region; The basic sample density is determined based on the type of engineering materials; Let be the weighting coefficient, satisfying , Indicates the longest distance; Indicates the shortest distance; This represents the theoretical average distance to the first location in the corresponding type of project; This represents the theoretical standard deviation of the average distance to the first location in the corresponding type of project; Select the area of ​​the region with significant construction differences in the target material area; This represents the total number of items in the first position. This is the initial sample size; This is the discrete point correction coefficient, with a value ranging from 0.1 to 0.3; The standard deviation of the coefficient; The mean of the coefficients; To increase the number.

[0023] Preferably, the initial sampling quantity is divided into a first quantity based on a first position and a second quantity based on non-first positions, including:

[0024] Determine the weight percentage of the first position ,in, Select the total number of all locations within the target material area; Let be the moisture content influence coefficient for the i-th first position; Let be the influence coefficient of moisture content at the j-th position;

[0025] Determine the first quantity ,in, This is a correction factor for the proportion of locations, with a value ranging from 0.2 to 0.5;

[0026] Calculate the second quantity And must meet If the calculated If it is less than the corresponding minimum value, readjust. The calculation method, and The upper limit is set to Then calculate n1 and n2 again.

[0027] Preferably, the establishment of the resistance-moisture content relationship function and the capacitance-moisture content relationship function includes:

[0028] Representative samples are laid out evenly, and measurement points are set up according to the three-dimensional grid method. The grid spacing is determined according to the sample size.

[0029] During resistance measurement, the four-electrode method was used to perform N1 repeated tests at each point, and the resistance value R(i1,j1,k1) was recorded for each measurement. The electrode insertion depth was 1 / 3 to 2 / 3 of the sample thickness, and each repeated test was performed according to the following procedure. The measurement depth is incremented, where h is the sample thickness, i1 is the point number, j1 is the material type identifier, and k1 is the number of repeated resistance measurements.

[0030] During capacitance measurement, the sample is fitted to the surface and N2 repeated measurements are performed at each point. The capacitance value C(i1,j1,k2) of each measurement is recorded, where k2 is the number of repeated capacitance measurements.

[0031] The average resistance and average capacitance at each point are calculated. A piecewise fitting method is used to construct the resistance-moisture content relationship function, and a cubic spline interpolation method is used to construct the capacitance-moisture content relationship function.

[0032] Preferably, the spectral-moisture content relationship function is constructed, including:

[0033] The infrared spectral wavelength range is divided into a basic scanning segment and a fine scanning segment. Based on the basic scanning segment, M1 global scans are performed on the uniformly tiled sample to obtain the overall spectral characteristics. Based on the fine scanning segment, M2 local scans are performed on the uniformly tiled sample to obtain the local spectral characteristics.

[0034] The local spectral features and the partial spectral features that are consistent with the local scanning area are extracted from the overall spectral features for feature comparison. Based on the feature comparison results, the remaining areas of the overall spectral features excluding the local scanning area are subjected to fine feature transformation.

[0035] The two spectral features of the local scanning area and the remaining area are fused to obtain the position feature vector of the uniformly tiled sample;

[0036] Based on a pre-established infrared spectrum-moisture content database, a coefficient vector matching the location feature vector is retrieved, and this coefficient vector is input into a function analysis model to output a spectrum-moisture content relationship function. ,in, Output parameters for the model. Let V be the mapping function for the location feature vector V.

[0037] Preferably, fine feature transformation is performed on the remaining regions of the overall spectral features excluding the local scanning regions according to the feature comparison results, including:

[0038] Calculate the feature matching degree between the local spectral features and the partial spectral features. Where E is the feature dimension. are the u-th eigenvalues ​​of the overall spectral characteristics and the local spectral characteristics, respectively;

[0039] If Pt is greater than the preset matching degree, the remaining region features excluding the local scanning region in the overall spectral features are directly used;

[0040] Otherwise, the features of the remaining region are refined according to the feature comparison results, using the following transformation formula. ,in, Let u be the eigenvalue of the remaining region in dimension u. Let be the deviation value between the u-th residual feature and the corresponding local feature. The mean of the deviation values ​​across all dimensions; The standard deviation of the deviation value; Let be the importance weight of the u-th dimension feature.

[0041] Preferably, the resistance-moisture content relationship function, capacitance-moisture content relationship function, and spectrum-moisture content relationship function are fused and analyzed to obtain the moisture content of the corresponding representative samples, including:

[0042] Determine the error variances of the resistance-moisture content relationship function, the capacitance-moisture content relationship function, and the spectrum-moisture content relationship function, and calculate the confidence weight of each function;

[0043] When the material homogeneity coefficient U is less than 1, increase the spectral function weighting correction amount. At the same time, the weight of the resistance function is reduced by the same amount;

[0044] When the material homogeneity coefficient U is greater than or equal to 1, increase the capacitance function weight correction amount. At the same time, the weights of the spectral functions are reduced by the same amount, and the sum of the weights of the adjusted functions is 1.

[0045] Based on the adjusted weights of each function and the aforementioned resistance-moisture content relationship function, capacitance-moisture content relationship function, and spectrum-moisture content relationship function, the moisture content of the representative sample is obtained.

[0046] Compared with the prior art, the beneficial effects of this application are as follows:

[0047] By scientifically selecting and processing samples, and combining three complementary detection methods—resistance, capacitance, and infrared spectroscopy—the moisture content is obtained through fusion analysis. This overcomes the time-consuming nature of traditional drying methods and solves the problem of insufficient accuracy of single rapid detection methods. It can accurately and efficiently detect the moisture content of soil, concrete, and gravel in scenarios such as farmland irrigation, irrigation and drainage canals, and slope protection, providing reliable data support for the quality control of water conservancy projects.

[0048] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0049] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0050] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0051] Figure 1 This is a flowchart of a method for detecting the moisture content of water conservancy engineering materials according to an embodiment of the present invention. Detailed Implementation

[0052] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0053] This invention provides a method for detecting the moisture content of materials used in hydraulic engineering, such as... Figure 1 As shown, it includes:

[0054] Step 1: Based on the site of the water conservancy project, select representative samples of different types according to relevant standards. The different types include soil, concrete or sand and gravel. The representative samples are samples of raw materials processed from the selected location.

[0055] Step 2: Based on the resistance measurement method and the capacitance measurement method, perform multiple measurements at multiple points on the corresponding representative samples to establish the resistance-moisture content relationship function and the capacitance-moisture content relationship function;

[0056] Step 3: Based on infrared spectroscopy measurement, scan the corresponding representative samples at different set wavelengths to obtain several infrared spectra, analyze all infrared spectra, and construct the spectrum-moisture content relationship function in combination with the pre-established infrared spectrum-moisture content database.

[0057] Step 4: Perform a fusion analysis on the resistance-moisture content relationship function, capacitance-moisture content relationship function, and spectrum-moisture content relationship function to obtain the moisture content of the corresponding representative samples.

[0058] In this embodiment, the water conservancy project site refers to the actual construction or operation site of water conservancy projects such as farmland irrigation systems, irrigation and drainage canals, and slope protection. The relevant standards refer to the national standards, industry specifications, or local regulations for sampling materials in water conservancy projects, which are used to standardize the representativeness of samples and the sampling process. For example, the frequency requirements for soil sampling in the "Code for Quality Inspection and Evaluation of Water Conservancy Project Construction" (SL176-2007) and the size standards for concrete core sample sampling in the "Code for Acceptance of Construction Quality of Concrete Structures" (GB50204-2015).

[0059] In this embodiment, representative samples of different types refer to samples that can reflect the overall characteristics of materials in the target area, and the types cover commonly used materials in water conservancy projects, such as:

[0060] Soil samples: taken from the topsoil layer at a depth of 0-30cm on the banks of irrigation and drainage canals, and must contain common clay and loam components in the area;

[0061] Concrete sample: A core sample, 50 mm in diameter and 50 mm in height, drilled from the precast slope protection slab, representing the characteristics of this batch of concrete;

[0062] Sand and gravel samples: graded sand and gravel taken from the bedding layer of farmland irrigation channels, with particle size distribution meeting design requirements (e.g., 60% of particles are 5-20mm in diameter).

[0063] In this embodiment, selecting the raw material samples after processing at the selected location refers to processing the original materials obtained on-site to remove interfering factors and standardize their morphology to ensure consistent testing. For example:

[0064] Soil: After removing weeds and stones, crush the soil to a particle size of ≤2mm using a soil crusher, sieve it, and mix it evenly;

[0065] Concrete: Use a cutting machine to shape the core sample into a regular cylinder, and sand the surface until smooth;

[0066] Sand and gravel: Remove oversized particles (such as stones >20mm) by sieving, mix and take a uniform sample.

[0067] In this embodiment, the resistance measurement method reflects the moisture content by measuring the resistance value of the material (the higher the moisture content, the lower the resistance is usually), and the four-electrode method is commonly used to reduce the influence of contact resistance.

[0068] Capacitance measurement method: Taking advantage of the fact that the dielectric constant of water in the material (about 80) is much higher than that of dry materials (such as soil, about 3-5), the change in capacitance value reflects the moisture content.

[0069] Multiple measurements at multiple points involve setting up multiple detection points on the sample and repeatedly measuring them to reduce errors caused by material inhomogeneity.

[0070] The resistance-moisture content relationship function and the capacitance-moisture content relationship function are mathematical expressions fitted from a large amount of experimental data, used to convert resistance / capacitance values ​​into moisture content.

[0071] Infrared spectroscopy is a method that utilizes the strong absorption characteristics of water molecules at specific infrared wavelengths to analyze water content through spectral features.

[0072] The different wavelength settings are scanning bands divided according to the absorption characteristics of water molecules, taking into account both global characteristics and detailed local analysis.

[0073] An infrared spectrum is a curve with wavelength (or wavenumber) on the horizontal axis and absorbance on the vertical axis, reflecting the absorption intensity of a material for infrared light of different wavelengths.

[0074] A pre-established infrared spectroscopy-moisture content database stores the spectral characteristics of different materials and different moisture contents, which are used to match unknown samples. Specifically, materials such as soil from farmland irrigation areas, concrete from irrigation and drainage ditches, and sand and gravel for slope protection are collected, and the moisture content is adjusted in a 5% gradient (5%~30%). The spectra are measured and correlated with the measured values ​​by the drying method to form a database (containing 1000+ data entries).

[0075] The spectrum-moisture content relationship function is a function that converts the spectral characteristics of a sample into a moisture content function through database matching and modeling.

[0076] Fusion analysis combines the results of three methods: resistance, capacitance, and spectroscopy. By using weighted averages or algorithmic optimization, it reduces the limitations of a single method and improves accuracy. The moisture content of the representative sample is the final output sample moisture content, expressed as a percentage by mass (%), reflecting the actual moisture content of the material.

[0077] The beneficial effects of the above technical solution are: by scientifically selecting and processing samples, and combining three complementary detection methods—resistance, capacitance, and infrared spectroscopy—the moisture content is obtained through fusion analysis. This overcomes the time-consuming defects of traditional drying methods and solves the problem of insufficient accuracy of single rapid detection methods. It can accurately and efficiently detect the moisture content of soil, concrete, and sand in scenarios such as farmland irrigation, irrigation and drainage canals, and slope protection, providing reliable data support for the quality control of water conservancy projects.

[0078] This invention provides a method for detecting the moisture content of materials used in water conservancy projects. Before selecting representative samples of different types according to relevant standards based on the site conditions of the water conservancy project, the method includes:

[0079] The project structure is obtained by scanning the water conservancy project site with a drone, and the target material selection area is automatically locked based on the construction type of each sub-project in the project structure.

[0080] Based on the construction batch record log of the target material selection area, the corresponding area is divided, and a significant hierarchical structure of the target material selection area is constructed. The division result includes the range of three-dimensional construction location points, construction edge line, and construction materials of each batch of construction in the corresponding area.

[0081] The hierarchical structure is divided into three dimensions using standard units. The construction vector of each three-dimensional block is determined, and the construction vector is input into the vector analysis model to obtain the moisture content influence coefficient caused by construction differences.

[0082] Based on all moisture content influence coefficients, determine the coefficient variance and coefficient mean, and select the initial sampling quantity that matches the target material region, coefficient variance, and coefficient mean from the variance-mean-region size-quantity comparison table;

[0083] The moisture content influence coefficient is mapped to the target material area and the coefficient position is classified. The highlight coefficient is filtered and the first position of the highlight coefficient in the target material area is locked.

[0084] When the number of the first positions is less than 3, random selection is made in the target material selection area according to the initial sampling quantity;

[0085] When the number of the first position is greater than or equal to 3, the initial sampling quantity is divided into a first quantity based on the first position and a second quantity based on the non-first position. The maximum boundary of the target material selection area is drawn according to the distribution of the first position, and the longest distance and the shortest distance between the two nearest points in the target material selection area are locked. The number of additional non-first positions is determined. At the same time, the initial sampling quantity is divided into a first quantity based on the first position and a second quantity based on the non-first position.

[0086] Random selection is performed on non-first positions according to the sum of the added quantity and the second quantity, and random selection is performed on the first position according to the first quantity;

[0087] Representative samples were obtained from the selected locations according to relevant standards.

[0088] In this embodiment, if there are only one or two first locations, the area coverage is too narrow (may only represent local points) and cannot reflect the overall risk of the target area. Three locations can initially form a triangular coverage, covering different sub-regions of the target area (such as the upper, middle, and lower parts of the canal embankment), ensuring the spatial representativeness of abnormal risks. Moreover, three samples are the critical value for balancing detection costs and acceptable error: in engineering, the detection cost (time and manpower) of three samples is relatively controllable; from the perspective of error theory, although the mean standard error of three samples is higher than that of a large sample, it can meet the needs of preliminary judgment of abnormal areas and avoid misjudgment due to too few samples (such as one sample being completely abnormal, directly determining the result).

[0089] Preferably, determining the number of additional units in non-first positions includes:

[0090]

[0091] in, This represents the rounding function; Indicates the number of additions; Select the area of ​​the inner region below the maximum boundary of the target material region; The basic sample density is determined based on the type of engineering materials; Let be the weighting coefficient, satisfying , Indicates the longest distance; Indicates the shortest distance; This represents the theoretical average distance to the first location in the corresponding type of project; This represents the theoretical standard deviation of the average distance to the first location in the corresponding type of project; Select the area of ​​the region with significant construction differences in the target material area; This represents the total number of items in the first position. This is the initial sample size; This is the discrete point correction coefficient, with a value ranging from 0.1 to 0.3; The standard deviation of the coefficient; The mean of the coefficients; To increase the number.

[0092] In this embodiment, Based on the material uniformity grade, it is divided into 0.1, 0.2, and 0.3. , The value is 0.3, which enhances the sampling compensation for highly discrete regions.

[0093] In this embodiment, the area of ​​the inner region is the area of ​​the enclosed region of the maximum boundary.

[0094] In this embodiment, The weights are preset according to the material type. For example, if the soil has high dispersion, then k1 is 0.5; if the concrete has good uniformity, then k1 is 0.3.

[0095] In water conservancy projects, the moisture content of materials is affected by complex factors such as construction differences (e.g., compaction degree, paving thickness) and regional distribution (e.g., the dispersion of the first location). Traditional uniform random sampling cannot specifically cover areas with significant construction differences and is prone to missing points with abnormal moisture content. Therefore, it is necessary to determine the additional sampling quantity.

[0096] Preferably, the initial sampling quantity is divided into a first quantity based on a first position and a second quantity based on non-first positions, including:

[0097] Determine the weight percentage of the first position ,in, Select the total number of all locations within the target material area; Let be the moisture content influence coefficient for the i-th first position; Let be the influence coefficient of moisture content at the j-th position;

[0098] Determine the first quantity ,in, This is a correction factor for the proportion of locations, with a value ranging from 0.2 to 0.5;

[0099] Calculate the second quantity And must meet If the calculated If it is less than the corresponding minimum value, readjust. The calculation method, and The upper limit is set to Then calculate n1 and n2 again.

[0100] In this embodiment, the drone is an unmanned aerial vehicle equipped with high-definition scanning equipment (such as lidar and visible light camera) to quickly acquire three-dimensional data of the engineering site. The DJI Matrice 300RTK drone is used, equipped with a lidar module, to collect data from the engineering site in all directions and generate three-dimensional point clouds or models.

[0101] In this embodiment, the engineering structure is a three-dimensional model of a water conservancy project reconstructed from scanned data, which includes the spatial shape and positional relationship of each component. For example, in the engineering structure model of an irrigation and drainage canal, the canal bottom (2m wide), canal slope (slope 1:1.5), and embankment top (3m wide) can be clearly distinguished. Sub-projects are relatively independent components in a water conservancy project, divided according to function or construction stage. For example, the sub-projects of a slope protection project include a foundation cushion (sand and gravel), retaining wall (concrete), and impermeable layer (geomembrane).

[0102] In this embodiment, the construction type is the construction process or material category of the sub-project, used to distinguish areas with different testing needs, such as concrete pouring (e.g., canal embankment retaining wall), soil backfilling (e.g., embankment backfill layer), and gravel laying (e.g., canal bedding layer).

[0103] The target material selection area is the area where the material whose moisture content needs to be tested is located, based on the construction type. For example, for soil backfill construction, the target area is the 0-50cm soil layer on the banks of irrigation and drainage canals.

[0104] In this embodiment, the construction batch record log is a document (paper or electronic) that records the construction time, material source and process parameters of each area of ​​the project. It is used to trace construction differences. For example, the log records that the backfilling of the embankment of the irrigation and drainage canal from K0+100 to K0+200 was carried out by taking soil from material yard A and compaction degree of 93%.

[0105] The hierarchical structure is structured data formed by dividing the target area into hierarchical levels according to construction batches. It clearly presents the spatial and material characteristics of each batch. For example, the target area of ​​irrigation and drainage canals is divided into 3 batch levels: batch 1 (K0+000-K0+100, loam), batch 2 (K0+100-K0+200, sandy loam), and batch 3 (K0+200-K0+300, loam).

[0106] The three-dimensional location range of the construction site is the three-dimensional coordinate boundary of the construction area for each batch, used for precise positioning. For example, the three-dimensional location range of batch 2 is X: 1000-1200m, Y: 500-550m, Z: 20-25m (altitude).

[0107] The construction edge line is a two-dimensional boundary line of the batch area, which is formed by connecting plane coordinate points. For example, the edge line of batch 2 is formed by connecting coordinates (1000,500), (1200,500), (1200,550), and (1000,550) to form a rectangular boundary.

[0108] Construction materials refer to the specific types of materials used in each batch of construction. For example, batch 1 uses loam (25% clay content), batch 2 uses sandy loam (60% sand content), and batch 3 uses C25 concrete (42.5R cement grade).

[0109] The three-dimensional division of standard units divides the batch area into cubic units of fixed size, which facilitates the unified analysis of construction differences. For example, using the standard unit of 1m×1m×0.5m (length×width×height), batch 2 (200m length×50m width×5m height) is divided into 200×50×10=100000 three-dimensional blocks.

[0110] A unit 3D block is the smallest cubic unit after standard unit division, representing a local construction area.

[0111] A construction vector is a set of multidimensional parameters that describe the construction characteristics of a unit three-dimensional block and reflect factors that may affect the moisture content. For example, a construction vector is [construction time (2024.03.10), material moisture content (18%), compaction degree (92%), paving thickness (0.3m)], which is quantified as a numerical vector [3.10,18,92,0.3].

[0112] Vector analysis models are models trained through machine learning (such as random forests and neural networks) to predict the moisture content influence coefficient from construction vectors. The model is trained with historical data (10,000 sets of construction vectors and corresponding measured moisture content differences). After inputting a new construction vector, it outputs the moisture content influence coefficient of the three-dimensional block due to construction differences.

[0113] In this embodiment, the moisture content influence coefficient is a numerical value that quantifies the degree of influence of construction differences on the moisture content of materials. The value ranges from 1 to 2. For example, if a three-dimensional block has a low compaction degree (90%), the influence coefficient is 1.2 (mean 0.15), indicating that its moisture content may be significantly higher than that of the normal area.

[0114] The coefficient variance is an index of the dispersion of all moisture content influence coefficients, reflecting the uniformity of construction differences (the larger the variance, the more uneven the differences).

[0115] The mean coefficient is the average value of all moisture content influence coefficients, reflecting the average impact of overall construction differences on moisture content.

[0116] In this embodiment, the variance-mean-region size-number comparison table is a pre-established table of correlation coefficient variance, mean, target region area and initial sampling number, used to quickly determine the sampling quantity. For example, a variance of 0.02, a mean of 0.15 and a region area of ​​1000 square meters correspond to an initial sampling quantity of 20; a variance of 0.1, a mean of 0.3 and a region area of ​​500 square meters correspond to 30.

[0117] In this embodiment, the location mapping associates each moisture content influence coefficient with the three-dimensional coordinates of the target area to form a coefficient-location correspondence. Specifically, the influence coefficient is marked on the corresponding three-dimensional block position of the three-dimensional model through the GIS system to generate a visual heat map (red indicates high coefficient areas).

[0118] The coefficient location classification group is based on the magnitude of the influence coefficient to group the locations of three-dimensional blocks. For example, in the target area, 30 three-dimensional blocks are classified into the high coefficient group (>0.3), and the rest are classified into the medium and low coefficient group.

[0119] The salience coefficient is an influence coefficient that is significantly higher than the mean coefficient (e.g., greater than the mean + 2 times the standard deviation), reflecting areas with extremely large differences in construction. For example, if the mean is 0.15 and the standard deviation is 0.05, the threshold for the salience coefficient is 0.15 + 2 × 0.05 = 0.25. All coefficients greater than 0.25 (e.g., 0.3, 0.4) are salience coefficients.

[0120] The first position is the three-dimensional block position corresponding to the salience coefficient, which is the key detection area where the moisture content may be abnormal. The first quantity is the number of samples selected from the first position, reflecting the detection intensity of the key area. The second quantity is the number of samples selected from non-first positions (ordinary areas) to ensure overall representativeness.

[0121] The maximum boundary is the smallest closed region that includes all first positions. It is used to define the range of additional samples that are not first positions. For example, if 10 first positions are distributed in X: 1000-1200m and Y: 500-550m, the maximum boundary is this rectangular region. The longest distance and the shortest distance are the extreme values ​​of the distance between the two nearest first positions within the maximum boundary. The number of additional samples is the number of samples added to make up for the lack of representativeness of non-first positions.

[0122] In this embodiment,

[0123] The beneficial effects of the above technical solution are: by integrating drone scanning with construction data, the key detection area (first location) can be accurately located and the number of samples can be scientifically allocated, taking into account both areas with significant differences in construction and the representativeness of ordinary areas. This solves the problem that key areas may be missed by traditional random sampling, provides a high-quality sample basis for subsequent moisture content testing, and ultimately improves the accuracy and efficiency of moisture content testing of water conservancy engineering materials.

[0124] This invention provides a method for detecting the moisture content of materials used in hydraulic engineering, establishing resistance-moisture content relationship functions and capacitance-moisture content relationship functions, including:

[0125] Representative samples are laid out evenly, and measurement points are set up according to the three-dimensional grid method. The grid spacing is determined according to the sample size.

[0126] During resistance measurement, the four-electrode method was used to perform N1 repeated tests at each point, and the resistance value R(i1,j1,k1) was recorded for each measurement. The electrode insertion depth was 1 / 3 to 2 / 3 of the sample thickness, and each repeated test was performed according to the following procedure. The measurement depth is incremented, where h is the sample thickness, i1 is the point number, j1 is the material type identifier, and k1 is the number of repeated resistance measurements.

[0127] During capacitance measurement, the sample is fitted to the surface and N2 repeated measurements are performed at each point. The capacitance value C(i1,j1,k2) of each measurement is recorded, where k2 is the number of repeated capacitance measurements.

[0128] The average resistance and average capacitance at each point are calculated. A piecewise fitting method is used to construct the resistance-moisture content relationship function, and a cubic spline interpolation method is used to construct the capacitance-moisture content relationship function.

[0129] In this embodiment, uniform spreading involves evenly distributing the sample material into a thin layer of a certain thickness and area to ensure uniform material distribution and representativeness of subsequent measurement points. Specifically, for soil samples, the sample can be poured into a square mold with sides of 50cm and a thickness of 5cm, and gently smoothed with a scraper. For concrete samples, the sample is cut into regular small pieces and then spread into a uniform layer on a flat surface. For example, soil samples can be spread into a 5cm thick, 50cm x 50cm thin layer.

[0130] In this embodiment, the grid spacing is the distance between two adjacent measurement points in the length, width, and height directions. The determination is based on the following criteria: larger sample sizes allow for larger spacing; poor sample uniformity requires smaller spacing. For example, for concrete samples with good uniformity, the grid spacing can be set to 20mm; for soil samples with poor uniformity, the spacing can be set to 10mm.

[0131] In this embodiment, the index of the point i1 in the 1st row, 1st column and 1st layer of the 3D mesh is 1, and the material type identifier is 1 for soil and 2 for concrete.

[0132] In this embodiment, a flexible capacitive sensor is used for samples with irregular surfaces (such as rough concrete blocks); a planar capacitive sensor is used for flat soil surfaces. For example, when measuring concrete samples, an arc-shaped flexible capacitive sensor is fitted to the curved surface, and a 1kHz alternating voltage is applied. Measurements are taken five times at each point, and the capacitance value (unit: pF) is recorded. In the sand and gravel sample, the capacitance values ​​measured at a certain point were 80pF, 82pF, 79pF, 81pF, and 78pF, respectively, and the average value of 80pF was taken.

[0133] In this embodiment, a four-electrode resistance meter is used. Four metal electrodes are inserted into the soil sample (or close to the surface of the concrete sample) at a spacing of 10 mm. The resistance value (unit: Ω) is recorded at each point. For example, in the soil sample, the resistance values ​​of a certain point measured three times are 500Ω, 510Ω and 490Ω respectively. The average value of 500Ω is taken for subsequent analysis.

[0134] In this embodiment, points are set up using a three-dimensional grid method (e.g., for a soil sample of 10cm×10cm×5cm, 3×3×2 points are set up). Each point is measured 3 to 5 times. For example, on a concrete core sample (50mm in diameter and 50mm in height), one point is set up every 10mm along the height direction. The resistance and capacitance are measured 3 times and 5 times at each point.

[0135] In this embodiment, the piecewise fitting method is to divide the sample into different intervals according to the range of sample moisture content. Within each interval, a suitable function (such as a linear function or a quadratic function) is used to fit the relationship between the mean resistance and the moisture content. First, the mean resistance of samples with different known moisture contents is measured by the drying method. For example, the soil sample is divided into three intervals with moisture content of 5% to 15%, 15% to 25%, and 25% to 35%. Multiple sets of data (mean resistance and moisture content) are collected in each interval, and the least squares method is used to fit the function.

[0136] In this embodiment, cubic spline interpolation is a method for smoothing interpolation between known data points to construct a continuous relationship function between the mean capacitance and the moisture content. Multiple sets of data points (mean capacitance, moisture content) are obtained through the drying method, such as (80pF, 5%), (90pF, 10%), and (100pF, 15%) for concrete samples. Using the cubic spline interpolation algorithm, a smooth curve is constructed between these points to obtain the capacitance-moisture content relationship function. The corresponding moisture content can be calculated based on any mean capacitance.

[0137] The beneficial effects of the above technical solution are as follows: by uniformly laying out the sample and using a three-dimensional grid, combined with multiple depth-increment resistance measurements using the four-electrode method and capacitance measurements adapted to the sample, multiple sets of data can be accurately obtained. Then, through mean calculation and scientific fitting / interpolation, a relationship function is constructed. This not only considers the characteristics of different positions and depths of the sample, but also reduces errors through multiple measurements and function construction. It can accurately and comprehensively establish the relationship between resistance, capacitance and moisture content, providing reliable and precise methodological support for moisture content detection of hydraulic engineering materials, and improving the accuracy and reliability of detection.

[0138] This invention provides a method for detecting the moisture content of materials used in hydraulic engineering, comprising constructing a spectrum-moisture content relationship function, including:

[0139] The infrared spectral wavelength range is divided into a basic scanning segment and a fine scanning segment. Based on the basic scanning segment, M1 global scans are performed on the uniformly tiled sample to obtain the overall spectral characteristics. Based on the fine scanning segment, M2 local scans are performed on the uniformly tiled sample to obtain the local spectral characteristics.

[0140] The local spectral features and the partial spectral features that are consistent with the local scanning area are extracted from the overall spectral features for feature comparison. Based on the feature comparison results, the remaining areas of the overall spectral features excluding the local scanning area are subjected to fine feature transformation.

[0141] The two spectral features of the local scanning area and the remaining area are fused to obtain the position feature vector of the uniformly tiled sample;

[0142] Based on a pre-established infrared spectrum-moisture content database, a coefficient vector matching the location feature vector is retrieved, and this coefficient vector is input into a function analysis model to output a spectrum-moisture content relationship function. ,in, Output parameters for the model. Let V be the mapping function for the location feature vector V.

[0143] Preferably, fine feature transformation is performed on the remaining regions of the overall spectral features excluding the local scanning regions according to the feature comparison results, including:

[0144] Calculate the feature matching degree between the local spectral features and the partial spectral features. Where E is the feature dimension. are the u-th eigenvalues ​​of the overall spectral characteristics and the local spectral characteristics, respectively;

[0145] If Pt is greater than the preset matching degree, the remaining region features excluding the local scanning region in the overall spectral features are directly used;

[0146] Otherwise, the features of the remaining region are refined according to the feature comparison results, using the following transformation formula. ,in, Let u be the eigenvalue of the remaining region in dimension u. Let be the deviation value between the u-th residual feature and the corresponding local feature. The mean of the deviation values ​​across all dimensions; The standard deviation of the deviation value; Let be the importance weight of the u-th dimension feature.

[0147] In this embodiment, the infrared spectral wavelength range is typically between 780 nm and 1000 nm. Different wavelengths within a given range exhibit varying degrees of sensitivity to moisture in materials (for example, the area around 1450nm is particularly sensitive to moisture vibrations).

[0148] Basic scanning range: Select a wavelength range that covers a wide range of moisture characteristics (e.g., 1000-2500nm) to quickly capture the overall moisture characteristics of the sample. Use an infrared spectrometer, set the scanning wavelength to 1000-2500nm, and the step size to 10nm. Scan the flat soil sample (spread into a 5cm×5cm thin layer) three times (M1=3) to obtain the overall spectral curve (like an electrocardiogram). Figure 1 The same fluctuation curve, with different wavelengths corresponding to different absorbance.

[0149] The fine scanning segment selects the narrow band with the most obvious moisture characteristics (such as 1400-1500nm, which is the strong absorption peak region of water). The wavelength is set to 1400-1500nm, the step size is 1nm, and the same soil sample is scanned 5 times (M2=5) to obtain the local spectrum (the curve will have a clear dip at 1450nm, which represents water absorption).

[0150] Full-area scanning: The basic segment scans the entire sample surface (5cm×5cm full scan) to capture overall features;

[0151] Local scanning: Fine-scan only scan small areas in the sample where the moisture content may be abnormal (e.g., 2cm×2cm), or fix the central area of ​​the sample to capture local details.

[0152] Local spectral features are spectral data obtained from fine scanning segments. They are a set of wavelength-absorbance values ​​(e.g., absorbance of 0.8 at 1400nm and absorbance of 1.2 at 1450nm), representing the moisture characteristics of a local area.

[0153] The overall spectral characteristics are the full sample spectral data (absorbance sequence of 1000-2500 nm) obtained from the basic scan segment.

[0154] In this embodiment, the feature matching degree Pt is an index that quantifies the spectral similarity between corresponding regions in the local and global samples. The closer the value is to 1, the more similar the samples are. For example, the absorbance of the local and global soil samples at 5 wavelengths is shown in Table 1:

[0155] Table 1 Absorbance at wavelength points

[0156] Wavelength (nm) Overall absorbance Local absorbance min max 1400 0.7 0.8 0.7 0.8 1420 0.6 0.7 0.6 0.7 1440 0.5 0.6 0.5 0.6 1460 0.4 0.5 0.4 0.5 1480 0.3 0.4 0.3 0.4

[0157] when A preset matching degree (e.g., 0.8) is used to correct the remaining portion of the overall spectrum except for local regions, and to address the reliability of spectra in local anomalous regions.

[0158] It is the absorbance of the u-th wavelength in the overall spectrum, excluding local regions (for example, if the overall absorbance at 1400nm is 0.7, and 0.8 is used in a local area, then the absorbance in the remaining region is 0.7).

[0159] It is the deviation between residual features and local features (e.g., residual 0.7, local 0.8, =0.7−0.8=−0.1).

[0160] It is the average value of all wavelength deviations (e.g., 5 wavelength deviations -0.1, -0.1, -0.1, -0.1, -0.1, da = −0.1).

[0161] It represents the importance of the u-th dimension feature; for example, 1450 nm is a strong absorption peak for water. It is 0.8; 1000nm is not important. It is 0.2.

[0162] The location feature vector is a set of numbers concatenated from the local and the transformed residual spectral features. It represents the spatial location and spectral features of the sample. For example, the local features are selected from the absorbance of 5 wavelengths (1400-1500nm) (0.8, 1.2, 1.0, 0.9, 0.7), and the residual features after transformation are selected from the absorbance of 5 wavelengths (1000-1200nm) (0.3, 0.4, 0.5, 0.6, 0.7), which are then fused into a vector (0.8, 1.2, 1.0, 0.9, 0.7, 0.3, 0.4, 0.5, 0.6, 0.7).

[0163] The infrared spectroscopy-moisture content database is a database that pre-stores a large amount of spectral vector-moisture content data (for example, the spectral vector of a certain soil sample corresponds to a moisture content of 18%).

[0164] The coefficient vector is a set of parameters in the database that is most similar to the location feature vector. For example, when matching a sample with a moisture content of 18%, the coefficient vector is (0.5, 0.3, 0.2).

[0165] Function analysis models are models trained using machine learning (such as neural networks), taking a coefficient vector as input and outputting a spectrum-water content relationship.

[0166] In this embodiment, water conservancy engineering materials (such as soil, concrete, and sand) have differences in construction batches and material inhomogeneity (such as soil stratification and uneven concrete vibration). Directly mixing the overall and local spectra can lead to misjudgment of moisture content. By matching degree judgment and dynamic correction, the spectral feature deviation is accurately identified and compensated, making the subsequent moisture content calculation more reliable.

[0167] The beneficial effects of the above technical solution are as follows: by scanning the basic and fine dual spectral bands, the overall picture is captured first and then the details are examined. Feature matching is used to quantify spectral differences, and mismatched areas are accurately corrected (considering deviation, importance, etc.). Finally, the results are fused into a location feature vector associated with the moisture content. This solves the problem that local anomalies in traditional spectral detection are easily masked by global averaging, making the spectrum-moisture content relationship more accurate. It is especially suitable for scenarios with uneven materials in water conservancy projects (such as different local compaction degrees of soil in canal embankments), providing a more reliable function model for subsequent moisture content detection.

[0168] This invention provides a method for detecting the moisture content of materials used in hydraulic engineering. The method involves fusing and analyzing the resistance-moisture content relationship function, capacitance-moisture content relationship function, and spectrum-moisture content relationship function to obtain the moisture content of representative samples. The method includes:

[0169] Determine the error variances of the resistance-moisture content relationship function, the capacitance-moisture content relationship function, and the spectrum-moisture content relationship function, and calculate the confidence weight of each function;

[0170] When the material homogeneity coefficient U is less than 1, increase the spectral function weighting correction amount. At the same time, the weight of the resistance function is reduced by the same amount;

[0171] When the material homogeneity coefficient U is greater than or equal to 1, increase the capacitance function weight correction amount. At the same time, the weights of the spectral functions are reduced by the same amount, and the sum of the weights of the adjusted functions is 1.

[0172] Based on the adjusted weights of each function and the aforementioned resistance-moisture content relationship function, capacitance-moisture content relationship function, and spectrum-moisture content relationship function, the moisture content of the representative sample is obtained.

[0173] In this embodiment, the error variance is an indicator that measures the degree of deviation between the function's predicted value and the actual value. It is used to calculate the predicted moisture content of multiple samples with known moisture content using a function, and then compare it with the actual moisture content. The mean of the squared difference is then calculated.

[0174] The confidence weight is calculated based on the error variance. The smaller the error variance, the more reliable the function, and the larger the confidence weight. First, the derivative of the error variance of the three functions is calculated, and then normalized.

[0175] In this embodiment, the material uniformity coefficient U is an indicator for measuring the uniformity of materials in hydraulic engineering; the smaller the value, the less uniform the material.

[0176] In this embodiment, the final moisture content is calculated by weighted averaging of the three functions based on the adjusted weights. For example, the adjusted resistance weight is 0.27, and the moisture content of the resistance measurement is 10%; the adjusted capacitance weight is 0.14, and the moisture content of the capacitance measurement is 12%; the adjusted spectral weight is 0.59, and the moisture content of the spectral measurement is 11%. The weighted calculation result is 10.87%.

[0177] The beneficial effects of the above technical solution are as follows: For moisture content detection of materials in hydraulic engineering, considering the error characteristics of different detection functions (resistance, capacitance, and spectrum), the initial confidence weight is determined by calculating the error variance. Then, the weight is dynamically adjusted based on the material homogeneity coefficient, allowing the function more suitable for the current material characteristics to play a greater role. Finally, the moisture content is obtained by weighted fusion. This approach combines the advantages of different detection methods and adapts to differences in material homogeneity, effectively improving the accuracy and reliability of moisture content detection.

[0178] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for detecting the water content of hydraulic engineering materials, characterized by, include: Step 1: Based on the site of the water conservancy project, select representative samples of different types according to relevant standards. The different types include soil, concrete or sand and gravel. The representative samples are the samples after processing the raw materials at the selected location. Step 2: Based on the resistance measurement method and the capacitance measurement method, perform multiple measurements at multiple points on the corresponding representative samples to establish the resistance-moisture content relationship function and the capacitance-moisture content relationship function; Step 3: Based on infrared spectroscopy measurement, scan the corresponding representative samples at different set wavelengths to obtain several infrared spectra, analyze all infrared spectra, and construct the spectrum-moisture content relationship function in combination with the pre-established infrared spectrum-moisture content database. Step 4: Perform a fusion analysis on the resistance-moisture content relationship function, capacitance-moisture content relationship function, and spectrum-moisture content relationship function to obtain the moisture content of the corresponding representative samples; Establish the resistance-moisture content relationship function and the capacitance-moisture content relationship function, including: Representative samples are laid out evenly, and measurement points are set up according to the three-dimensional grid method. The grid spacing is determined according to the sample size. During resistance measurement, the four-electrode method was used to perform N1 repeated tests at each point, and the resistance value R(i1,j1,k1) was recorded for each measurement. The electrode insertion depth was 1 / 3 to 2 / 3 of the sample thickness, and each repeated test was performed according to the following procedure. The measurement depth is incremented, where h is the sample thickness, i1 is the point number, j1 is the material type identifier, and k1 is the number of repeated resistance measurements. During capacitance measurement, the sample is fitted to the surface and N2 repeated measurements are performed at each point. The capacitance value C(i1,j1,k2) of each measurement is recorded, where k2 is the number of repeated capacitance measurements. Calculate the mean resistance and mean capacitance at each point, and construct the resistance-moisture content relationship function using a piecewise fitting method and the capacitance-moisture content relationship function using a cubic spline interpolation method. Constructing the spectrum-moisture content relationship function includes: The infrared spectral wavelength range is divided into a basic scanning segment and a fine scanning segment. Based on the basic scanning segment, M1 global scans are performed on the uniformly tiled sample to obtain the overall spectral characteristics. Based on the fine scanning segment, M2 local scans are performed on the uniformly tiled sample to obtain the local spectral characteristics. The local spectral features and the partial spectral features that are consistent with the local scanning area are compared. Based on the feature comparison results, the remaining areas of the overall spectral features excluding the local scanning area are subjected to fine feature transformation. The two spectral features of the local scanning area and the remaining area are fused to obtain the position feature vector of the uniformly tiled sample; Based on a pre-established infrared spectrum-moisture content database, a coefficient vector matching the location feature vector is retrieved, and this coefficient vector is input into a function analysis model to output a spectrum-moisture content relationship function. ,in, Output parameters for the model. Let V be the mapping function for the location feature vector V.

2. The method for detecting the water content of hydraulic engineering materials according to claim 1, characterized in that, Based on the site conditions of water conservancy projects, before selecting representative samples of different types according to relevant standards, the following should be included: The project structure is obtained by scanning the water conservancy project site with a drone, and the target material selection area is automatically locked based on the construction type of each sub-project in the project structure. Based on the construction batch record log of the target material selection area, the corresponding area is divided, and a significant hierarchical structure of the target material selection area is constructed. The division result includes the range of three-dimensional construction location points, construction edge line, and construction materials of each batch of construction in the corresponding area. The hierarchical structure is divided into three-dimensional units of standard units. The construction vector of each three-dimensional unit is determined, and the construction vector is input into the vector analysis model to obtain the moisture content influence coefficient caused by construction differences. Based on all moisture content influence coefficients, determine the coefficient variance and coefficient mean, and select the initial sampling quantity that matches the target material region, coefficient variance, and coefficient mean from the variance-mean-region size-quantity comparison table; The moisture content influence coefficient is mapped to the target material area and the coefficient position is classified. The highlight coefficient is filtered and the first position of the highlight coefficient in the target material area is locked. When the number of the first positions is less than 3, random selection is made in the target material selection area according to the initial sampling quantity; When the number of the first position is greater than or equal to 3, the initial sampling quantity is divided into a first quantity based on the first position and a second quantity based on the non-first position. The maximum boundary of the target material selection area is drawn according to the distribution of the first position, and the longest distance and the shortest distance between the two nearest points in the target material selection area are locked. The number of additional non-first positions is determined. At the same time, the initial sampling quantity is divided into a first quantity based on the first position and a second quantity based on the non-first position. Random selection is performed on non-first positions according to the sum of the added quantity and the second quantity, and random selection is performed on the first position according to the first quantity; Representative samples were obtained from the selected locations according to relevant standards.

3. The method for detecting the water content of hydraulic engineering materials according to claim 2, characterized in that, Determine the number of additional units not in the first position, including: ； in, This represents the rounding function; Indicates the number of additions; Select the area of ​​the inner region below the maximum boundary of the target material region; The basic sample density is determined based on the type of engineering materials; For the weighting coefficients, satisfying , Indicates the longest distance; Indicates the shortest distance; This represents the theoretical average distance to the first location in the corresponding type of project; This represents the theoretical standard deviation of the average distance to the first location in the corresponding type of project; Select the area of ​​the region with significant differences in construction between the target material areas; This represents the total number of items in the first position. This is the initial sample size; This is the discrete point correction coefficient, with a value ranging from 0.1 to 0.3; The standard deviation of the coefficient; The mean of the coefficients; To increase the number.

4. The method of claim 3, wherein The initial sample size is divided into a first sample size based on the first position and a second sample size based on non-first positions, including: Determine the weight percentage of the first position ,in, Select the total number of all locations within the target material area; Let be the moisture content influence coefficient for the i-th first position; Let be the influence coefficient of moisture content at the j-th position; determining the first quantity wherein, is a position quantity proportion correction coefficient, and the value range of the position quantity proportion correction coefficient is 0.2 to 0.5; Calculate the second quantity And must meet If the calculated If it is less than the corresponding minimum value, readjust. The calculation method, and The upper limit is set to Then recalculate n1 and n2.

5. The method of claim 1, wherein Based on the feature comparison results, fine feature transformation is performed on the remaining regions of the overall spectral features, excluding the local scanning regions, including: calculating a feature matching degree of the local spectral feature and the partial spectral feature wherein E is a feature dimension, respectively, are the u-th feature values of the overall spectral feature and the local spectral feature. If Pt is greater than the preset matching degree, the remaining region features excluding the local scanning region in the overall spectral features are directly used; Otherwise, the features of the remaining region are refined according to the feature comparison results, using the following transformation formula. ,in, Let u be the eigenvalue of the remaining region in dimension u. Let be the deviation value between the u-th residual feature and the corresponding local feature. The mean of the deviation values ​​across all dimensions; The standard deviation of the deviation value; Let be the importance weight of the u-th dimension feature.

6. The method of claim 1, wherein The resistance-moisture content, capacitance-moisture content, and spectral-moisture content functions are fused and analyzed to obtain the moisture content of representative samples, including: Determine the error variances of the resistance-moisture content relationship function, the capacitance-moisture content relationship function, and the spectrum-moisture content relationship function, and calculate the confidence weight of each function; When the material homogeneity coefficient U is less than 1, increase the spectral function weighting correction amount. At the same time, the weight of the resistance function is reduced by the same amount; When the material homogeneity coefficient U is greater than or equal to 1, increase the capacitance function weight correction amount. At the same time, the weights of the spectral functions are reduced by the same amount, and the sum of the weights of the adjusted functions is 1. Based on the adjusted weights of each function and the aforementioned resistance-moisture content relationship function, capacitance-moisture content relationship function, and spectrum-moisture content relationship function, the moisture content of the representative sample is obtained.

Images