Resistivity parameter-based soil organic pollution concentration prediction method and system

Abstract

The invention provides a soil organic pollution concentration prediction method and system based on a resistivity parameter, and the method comprises the steps: synchronously obtaining dielectric constant distribution data, conductivity spatial distribution data, a resistivity parameter and a polarizability parameter of a target region in an organic pollution detection scene in a saline-alkali soil environment, and obtaining the dielectric constant distribution data, the conductivity spatial distribution data, the resistivity parameter and the polarizability parameter of the target region; and constructing a bimodal characteristic matrix reflecting the electrical characteristics of the soil. And performing multi-source coupling on the bimodal characteristic matrix, the dielectric constant and the conductivity data to generate a dynamic data set, and analyzing and extracting the polarization attenuation characteristic difference between the organic pollutants and the salt ions by using a frequency domain dielectric spectrum. And finally outputting a spatial concentration distribution prediction result of the organic pollutants in the high salinization environment in combination with a machine learning model. According to the method, the detection accuracy and the spatial positioning precision of the organic pollutants in the high-saline-alkali soil are improved.

Description

Technical Field

[0001] This application relates to the technical fields of resistivity parameters and machine learning, and in particular, to a method and system for predicting soil organic pollution concentration based on resistivity parameters. Background Art

[0002] In a salinized soil environment, the accurate detection of organic pollutants faces severe challenges. Due to the cross-interference between salt ions and organic pollutants in electrochemical characteristics, traditional single-parameter detection methods are difficult to effectively distinguish between the two, resulting in inaccurate pollution assessment.

[0003] Currently, a method for detecting soil pollution based on near-infrared spectroscopy technology has been proposed. This method collects near-infrared spectral data of the soil, extracts the spectral reflectance of the characteristic bands, and combines the support vector machine algorithm to establish a prediction model of spectral characteristics and pollutant concentration, so as to achieve quantitative analysis of pollutants.

[0004] Although this method realizes non-contact detection using spectral technology, it lacks effective means to deal with the spectral interference caused by high salinity in salinized soil, making the characteristic bands of organic pollutants easily masked. At the same time, this method is difficult to distinguish the response differences between salt ions and organic pollutants, and the detection accuracy decreases significantly in areas with a high degree of salinization. Summary of the Invention

[0005] This application provides a method and system for predicting soil organic pollution concentration based on resistivity parameters to solve the problems of low accuracy in detecting organic pollutants and poor spatial positioning accuracy in highly salinized soil in the prior art.

[0006] In a first aspect, this application provides a method for predicting soil organic pollution concentration based on resistivity parameters, including:

[0007] In an organic pollution detection scenario in a salinized soil environment, obtain the dielectric constant distribution data, conductivity spatial distribution data, resistivity parameters, and polarizability parameters of the soil in the target area;

[0008] Generate a bimodal feature matrix based on the resistivity parameters and the polarizability parameters;

[0009] Perform multi-source coupling correlation on the bimodal feature matrix, the dielectric constant distribution data, and the conductivity spatial distribution data to generate a dynamic data set;

[0010] Combine the frequency-domain dielectric spectroscopy analysis method to screen the frequency bands of the dynamic data set to extract the polarization attenuation characteristic differences between organic pollutants and salt ions within a preset frequency band range;

[0011] Based on the polarization attenuation characteristic differences and the bimodal feature matrix, use the trained machine learning model to output the prediction result of the spatial concentration distribution of organic pollutants in the salinized soil environment.

[0012] Optionally, in combination with the frequency domain dielectric spectroscopy analysis method, perform frequency band screening on the dynamic data set to extract the polarization attenuation characteristic differences between organic pollutants and salt ions within a preset frequency band range, including:

[0013] Convert the polarization response signal in the dynamic data set into frequency domain distribution data through the frequency domain dielectric spectroscopy analysis method;

[0014] Based on the polarization characteristic differences between organic pollutants and salt ions in the salinized soil environment, determine the intersection interval between the low frequency band and the middle frequency band within the preset frequency band range;

[0015] Within the intersection interval, according to the frequency domain distribution data, screen out the frequency points where the signal intensity first increases and then drops steeply as the frequency increases to form a set of frequency points;

[0016] Extract the characteristic frequency band corresponding to the organic pollutant and the interference frequency band corresponding to the salt ion from the set of frequency points, and calculate the difference in signal attenuation slope between the characteristic frequency band and the interference frequency band, and use the difference in signal attenuation slope as the polarization attenuation characteristic difference.

[0017] Optionally, the conversion of the polarization response signal in the dynamic data set into frequency domain distribution data through the frequency domain dielectric spectroscopy analysis method includes:

[0018] Extract the polarization response signal from the dynamic data set;

[0019] Intercept the polarization response signal according to a preset segmentation length to generate multiple time domain signal sequences;

[0020] Perform frequency domain conversion processing on each time domain signal sequence to generate a target signal intensity distribution sequence corresponding to each time domain signal sequence;

[0021] Determine the frequency range of the target signal intensity distribution sequence according to the proportional relationship between the conductivity parameter and the polarizability parameter in the conductivity spatial distribution data;

[0022] Perform superposition processing on the target signal intensity distribution sequences corresponding to all time domain signal sequences to generate frequency domain distribution data covering the frequency band range.

[0023] Optionally, the extraction of the characteristic frequency band corresponding to the organic pollutant and the interference frequency band corresponding to the salt ion from the set of frequency points, and the calculation of the difference in signal attenuation slope between the characteristic frequency band and the interference frequency band, include:

[0024] According to the difference in the steep decline timing sequence of the signal intensities in the frequency point set, N consecutive frequency points in the frequency point set whose steep decline timing is in the starting region within the intersection interval are divided into the characteristic frequency band of organic pollutants, and the consecutive frequency points whose steep decline timing is in the end region of the intersection interval are divided into the salt ion interference frequency band, where N is greater than or equal to 3;

[0025] Within the characteristic frequency band of organic pollutants, select the turning frequency point where the signal intensity changes from increasing to steep decline as the first reference point, and generate a first analysis window by extending a preset number of frequency points to both sides centered on the first reference point;

[0026] Within the salt ion interference frequency band, select the frequency points where the steep decline amplitude of the signal intensity is higher than the preset amplitude threshold as the second reference point, and generate a second analysis window by extending the same preset number of frequency points to both sides centered on the second reference point;

[0027] Calculate the first average decline rate of the signal intensity of the frequency points after the turning frequency point in the first analysis window and the second average decline rate of the signal intensity of the frequency points after the turning frequency point in the second analysis window respectively;

[0028] Take the absolute value difference between the first average decline rate and the second average decline rate as the signal attenuation slope difference.

[0029] Optionally, the performing frequency domain conversion processing on each time domain signal sequence to generate a target signal intensity distribution sequence corresponding to each time domain signal sequence includes:

[0030] Perform point-by-point multiplication processing on each time domain signal sequence and a preset window function sequence to obtain a time domain signal subsequence;

[0031] Perform spectrum analysis processing on the time domain signal subsequence to generate an initial signal intensity distribution sequence;

[0032] Perform zeroing processing on the frequency units in the initial signal intensity distribution sequence that are lower than the preset environmental noise threshold to generate a target signal intensity distribution sequence.

[0033] Optionally, the generating a bimodal feature matrix based on the resistivity parameter and the polarizability parameter includes:

[0034] Under the saline-alkali soil environment, perform synchronous alignment processing on the resistivity parameter and the polarizability parameter at the same spatial position to generate a set of parameter pairs;

[0035] Convert the resistivity parameter in each parameter pair in the set of parameter pairs into a first eigenvalue, and convert the polarizability parameter into a second eigenvalue;

[0036] Combine the first eigenvalue and the second eigenvalue at the same spatial position into a two-dimensional feature vector;

[0037] Arrange the two-dimensional feature vectors at all spatial positions to form a bimodal feature matrix.

[0038] Optionally, based on the polarization attenuation feature difference and the bimodal feature matrix, use a pre-trained machine learning model to output the predicted result of the spatial concentration distribution of organic pollutants in the salinized soil environment, including:

[0039] Combine each two-dimensional feature vector in the bimodal feature matrix with the polarization attenuation feature difference at the corresponding spatial position to generate an enhanced feature vector;

[0040] Perform dynamic weighting processing on the enhanced feature vector according to the salt ion concentration gradient generated from the conductivity parameter in the conductivity spatial distribution data;

[0041] Input the weighted enhanced feature vector into the pre-trained machine learning model according to the preset spatial distribution order to generate the initial pollution concentration value at each spatial position;

[0042] Perform dielectric correction processing on the initial pollution concentration value based on the dielectric constant distribution data;

[0043] Associate the corrected pollution concentration value with the corresponding spatial position coordinates and output the predicted result of the spatial concentration distribution.

[0044] In a second aspect, the present application provides a soil organic pollution concentration prediction system based on resistivity parameters, including:

[0045] An acquisition module, configured to acquire the dielectric constant distribution data, conductivity spatial distribution data, resistivity parameters, and polarization rate parameters of the soil in the target area in the organic pollution detection scenario in the salinized soil environment;

[0046] A first generation module, configured to generate a bimodal feature matrix based on the resistivity parameters and the polarization rate parameters;

[0047] A second generation module, configured to perform multi-source coupling association on the bimodal feature matrix, the dielectric constant distribution data, and the conductivity spatial distribution data to generate a dynamic data set;

[0048] A screening module, configured to perform frequency band screening on the dynamic data set through frequency domain dielectric spectroscopy analysis to extract the polarization attenuation feature difference between organic pollutants and salt ions within a preset frequency band range;

[0049] An output module, configured to output a prediction result of the spatial concentration distribution of organic pollutants in a salinized soil environment by using a trained machine learning model based on the polarization attenuation feature difference and the bimodal feature matrix.

[0050] In a third aspect, the present application provides a computing device, including a processor and a memory. A computer program is stored in the memory, and the processor is configured to run the computer program to execute a method for predicting the concentration of soil organic pollution based on resistivity parameters according to any one of the first aspects.

[0051] In a fourth aspect, the present application provides a computer storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, a method for predicting the concentration of soil organic pollution based on resistivity parameters according to any one of the first aspects is implemented.

[0052] In the present application, a method for predicting the concentration of soil organic pollution based on resistivity parameters is provided. The method includes: in an organic pollution detection scenario in a salinized soil environment, obtaining dielectric constant distribution data, conductivity spatial distribution data, resistivity parameters, and polarizability parameters of the soil in a target area; generating a bimodal feature matrix based on the resistivity parameters and the polarizability parameters; performing multi-source coupling association on the bimodal feature matrix, the dielectric constant distribution data, and the conductivity spatial distribution data to generate a dynamic data set; combining a frequency domain dielectric spectroscopy analysis method to perform frequency band screening on the dynamic data set to extract the polarization attenuation feature difference between organic pollutants and salt ions within a preset frequency band range; and outputting a prediction result of the spatial concentration distribution of organic pollutants in a salinized soil environment by using a trained machine learning model based on the polarization attenuation feature difference and the bimodal feature matrix.

[0053] The technical solution provided by the present application has the following beneficial effects:

[0054] The present application establishes a comprehensive soil electrical property feature database to provide multi-dimensional data support for subsequent analysis. Construct a feature expression that can simultaneously reflect the conductivity and polarization characteristics of the soil to enhance the feature representation ability. Realize the collaborative analysis of multi-source electrical parameters to improve the complementarity and integrity of data. Effectively separate the characteristic frequency bands of organic pollutants and salt ions to suppress the interference of the salinized background. Realize the visualization prediction of the spatial distribution of organic pollutants with high precision.

[0055] Furthermore, the present application also converts the polarization response signal into frequency domain distribution data through frequency domain dielectric spectroscopy analysis, determines the intersection interval based on the polarization property difference between organic pollutants and salt ions in a salinized environment, screens a set of frequency points with a specific signal change pattern, and further extracts the characteristic frequency band and the interference frequency band and calculates the signal attenuation slope difference therebetween as the feature difference.

[0056] Moreover, through the frequency-domain feature analysis of the polarization response signal, this technical solution effectively distinguishes the polarization characteristic differences between organic pollutants and salt ions, and improves the recognition accuracy of organic pollutants and the reliability of feature extraction under the background of high salinization by quantitatively comparing the signal attenuation slopes.

[0057] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following briefly introduces the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0059] Figure 1 It is a flowchart of a method for predicting the concentration of soil organic pollution based on resistivity parameters provided by an embodiment of the present application;

[0060] Figure 2 It is a schematic structural diagram of a system for predicting the concentration of soil organic pollution based on resistivity parameters provided by an embodiment of the present application;

[0061] Figure 3 It is a schematic structural diagram of a computing device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] To enable those skilled in the art to better understand the solutions of the present application, the following clearly and completely describes the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application.

[0063] In some processes described in the specification, claims, and above-mentioned drawings of the present application, there are multiple operations that appear in a specific order. However, it should be clearly understood that these operations can be executed not in the order in which they appear in this article or in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish each different operation, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations can be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0064] Researchers have found that in a salinized soil environment, traditional methods for detecting organic pollutants are difficult to effectively distinguish between salt ion interference and organic pollution characteristics, resulting in insufficient detection accuracy. Based on this, a method for predicting the concentration of soil organic pollutants based on resistivity parameters is provided. This method synchronously collects soil dielectric constant, conductivity, resistivity, and polarizability parameters, constructs a bimodal feature matrix and performs multi-source data coupling, combines frequency-domain dielectric spectroscopy analysis to extract the polarization attenuation characteristic differences between organic pollutants and salt ions, and finally uses a machine learning model to achieve accurate spatial distribution prediction of organic pollutants in a highly salinized environment. The technical solution of this application can be applied to scenarios of organic pollution detection and assessment in highly salinized soil environments such as saline-alkali land and coastal tidal flats.

[0065] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present application.

[0066] Figure 1 The flowchart of a method for predicting the concentration of soil organic pollutants based on resistivity parameters provided by an embodiment of the present application is as Figure 1 shown, and the method includes:

[0067] Step 101: In the organic pollution detection scenario in a salinized soil environment, obtain the dielectric constant distribution data, conductivity spatial distribution data, resistivity parameters, and polarizability parameters of the soil in the target area.

[0068] In this step, the dielectric constant distribution data represents the ability of the soil to respond to the electric field at different positions, and is used to characterize the electrical properties of the soil. The conductivity spatial distribution data is used to describe the spatial variation of the soil's conductivity and is used to evaluate the degree of salinization. The resistivity parameter characterizes the ability of the soil to impede the passage of current and is reciprocal to the conductivity. The polarizability parameter represents the degree of polarization phenomenon generated by the soil under the action of an external electric field.

[0069] In the embodiment of the present application, first determine the target detection area, use professional soil electrical property measurement equipment, and according to a preset grid sampling scheme, synchronously measure the four parameters of dielectric constant, conductivity, resistivity, and polarizability at each sampling point. Convert the discrete point measurement data into continuous spatial distribution data through spatial interpolation methods, and finally obtain a complete regional soil electrical property parameter dataset.

[0070] For example, taking the detection of a saline-alkali farmland pollution as an example, the target area is divided into grids of 10 meters × 10 meters. At the center point of each grid, a four-electrode soil resistivity meter is used to measure the resistivity parameter, and at the same time, a frequency-domain dielectric spectrometer is used to measure the dielectric constant and polarization rate parameters, and conductivity data is obtained through an electromagnetic induction instrument. All measurement data is generated into a continuous spatial distribution map through Kriging interpolation method, providing basic data for subsequent analysis.

[0071] Step 102: Generate a bimodal feature matrix based on the resistivity parameter and the polarization rate parameter.

[0072] In this step, the bimodal feature matrix represents a matrix composed of feature data of two modalities, resistivity and polarization rate, and is used to comprehensively characterize the electrical properties of the soil.

[0073] In the embodiment of the present application, first, the resistivity and polarization rate data at the same spatial position are normalized to eliminate the dimension difference, and then the two parameters are converted into characteristic values with physical meanings through a feature fusion algorithm. Finally, the characteristic values of all points are arranged and combined in the order of spatial positions to form a feature matrix containing double information of resistivity and polarization rate.

[0074] For example, the resistivity and polarization rate data of each sampling point in the farmland are respectively standardized, the main characteristic components of the two parameters are extracted by using the principal component analysis method, and the processed characteristic values are arranged in the order of the sampling point coordinates to construct a 100×2 bimodal feature matrix, where each row represents a sampling point, and the two columns respectively correspond to the resistivity and polarization rate characteristic values.

[0075] Step 103: Perform multi-source coupling and association on the bimodal feature matrix, the dielectric constant distribution data, and the conductivity spatial distribution data to generate a dynamic data set.

[0076] In this step, the dynamic data set represents a data set with spatio-temporal characteristics formed after integrating multi-source data.

[0077] In the embodiment of the present application, first, the bimodal feature matrix and the dielectric constant distribution data are spatially registered to ensure the correspondence of data positions, and then the conductivity spatial distribution data is introduced as a weight factor, and the three types of data are associated and integrated through a data fusion algorithm to finally generate a dynamic data set containing complete electrical properties.

[0078] For example, the spatial coordinates of the bimodal feature matrix and the dielectric constant distribution map are matched, the fusion weights of each point are calculated by using the conductivity data, and the three types of data are integrated by using a weighted fusion algorithm to form a dynamic data set containing resistivity, polarization rate, dielectric constant, and conductivity information, providing a data basis for subsequent frequency-domain analysis.

[0079] Step 104: Combine the frequency-domain dielectric spectroscopy analysis method to perform frequency band screening on the dynamic data set, so as to extract the polarization attenuation characteristic differences between organic pollutants and salt ions within a preset frequency band range.

[0080] In this step, the polarization attenuation characteristic difference represents the distinguishing characteristics of the polarization attenuation characteristics between organic pollutants and salt ions.

[0081] In the embodiment of the present application, first perform a frequency-domain transformation on the polarization response signal in the dynamic data set, determine the characteristic analysis frequency band according to the characteristics of saline-alkali soil, then identify the characteristic signal change points within a specific frequency band, respectively extract the characteristic frequency bands of organic pollutants and salt ions, and finally calculate the signal attenuation difference between the two frequency bands as the distinguishing feature.

[0082] For example, for the farmland dynamic data set, select the 50 - 500 Hz frequency band for frequency-domain analysis, identify the signal turning points within this frequency band, divide 80 - 200 Hz into the organic matter characteristic frequency band, and 300 - 450 Hz into the salt ion interference frequency band, and calculate the signal attenuation slope difference between the two frequency bands as the characteristic difference index.

[0083] Step 105: Based on the polarization attenuation characteristic difference and the bimodal feature matrix, use the trained machine learning model to output the prediction result of the spatial concentration distribution of organic pollutants in the saline-alkali soil environment.

[0084] In this step, the trained machine learning model refers to a prediction model that has been pre-trained through a large number of sample data. The spatial concentration distribution prediction result refers to the spatial distribution information of organic pollutants generated by model calculation.

[0085] In the embodiment of the present application, first perform feature-level fusion on the polarization attenuation characteristic difference and the bimodal feature matrix, then input it into the pre-trained machine learning model. The model outputs a prediction value according to the learned relationship between the features and the concentration, and finally generates a concentration distribution map in combination with the spatial position information.

[0086] For example, after fusing the feature differences and bimodal features in the farmland data, input it into the trained random forest model. The model outputs the predicted concentration values of each sampling point, and generates a heat map of the organic pollutant concentration distribution in the entire farmland area through spatial interpolation.

[0087] This method effectively overcomes the interference problem in the detection of organic pollutants in saline-alkali soil through the comprehensive acquisition and analysis of multi-source electrical parameters, realizes the accurate identification and visualization of the spatial distribution of organic pollutants, and provides a reliable technical means for the treatment of saline-alkali land pollution.

[0088] In order to solve the technical problem of accurately distinguishing organic pollutants from salt ion interference in a highly saline and alkaline soil environment, in some embodiments, step 104: combining the frequency domain dielectric spectrum analysis method to perform frequency band screening on the dynamic data set to extract the polarization attenuation characteristic differences between organic pollutants and salt ions within a preset frequency band, including:

[0089] Step 201: converting the polarization response signal in the dynamic data set into frequency domain distribution data by frequency domain dielectric spectrum analysis.

[0090] In step 201, the polarization response signal represents the time-varying electrical response signal generated by the soil under the action of an applied alternating electric field. The frequency domain distribution data represents the frequency-signal intensity distribution spectrum obtained by converting the time domain polarization response signal through a spectrum, which is used to analyze the electrical response characteristics at different frequencies.

[0091] In the embodiment of the present application, the discrete Fourier transform technology is used to convert the collected time domain polarization signal into a frequency domain signal. The specific process is: first, the original time domain signal in the dynamic data set is windowed, and then a fast Fourier transform is performed to finally generate frequency domain distribution data containing amplitude and phase information. The data is presented in the form of a spectrum with frequency as the horizontal axis and signal strength as the vertical axis.

[0092] Step 202: Based on the difference in polarization characteristics between organic pollutants and salt ions in a saline-alkali soil environment, determine the boundary interval between the low frequency band and the middle frequency band within a preset frequency band.

[0093] In step 202, the difference in polarization characteristics indicates the difference between the low-frequency relaxation characteristics of organic matter and the medium-frequency migration characteristics of salt ions. For example, in highly saline and alkaline soils, the polarizability attenuation amplitude of sodium chloride solution at a frequency of 10kHz is 3-5 times higher than that of petroleum hydrocarbon pollutants, and the polarizability peak frequencies of the two are separated by at least one order of magnitude. The boundary interval represents the transition frequency band region where the polarization response characteristics of organic pollutants and salt ions intersect. For example, the polarization response of organic pollutants (such as petroleum hydrocarbons and benzene series) is mainly in the low-frequency band (such as 1Hz-100Hz) because of their longer molecular polarization relaxation time. The polarization response of salt ions (such as Na⁺, Cl⁻) is concentrated in the medium-frequency band (such as 100Hz-10kHz) because of their faster ion migration polarization rate.

[0094] In the embodiment of the present application, based on the statistical analysis of a large amount of experimental data, 50-500 Hz is determined as the characteristic analysis frequency band. Among them, 80-200 Hz is defined as the low frequency band (dominated by organic matter), 200-450 Hz is defined as the mid-frequency band (dominated by salt ions), and the overlapping 150-250 Hz is the key boundary interval. The basis for this division is the essential difference in the frequency response characteristics between the molecular polarization relaxation of organic pollutants and the ion migration of salt ions.

[0095] Step 203: Within the intersection interval, based on the frequency-domain distribution data, filter out the frequency points where the signal strength first increases and then drops steeply as the frequency increases, and form a set of frequency points.

[0096] In step 203, the set of frequency points represents the set of discrete frequency points that satisfy a specific signal change pattern within the intersection interval.

[0097] In the embodiment of the present application, the frequency-domain data within the intersection interval is scanned through an automatic peak detection algorithm to identify the frequency points that meet the following characteristics: 1. There are obvious signal rising and falling edges; 2. The falling slope exceeds a set threshold; 3. The peak amplitude reaches a specific multiple of the background noise. The frequency points that meet all the conditions are arranged in the order of frequency to form a set of characteristic frequency points.

[0098] Step 204: Extract the characteristic frequency band corresponding to the organic pollutant and the interference frequency band corresponding to the salt ions from the set of frequency points, and calculate the difference in the signal attenuation slope between the characteristic frequency band and the interference frequency band, and use the difference in the signal attenuation slope as the polarization attenuation characteristic difference.

[0099] In step 204, the characteristic frequency band represents the continuous frequency interval where the polarization response of the organic pollutant dominates. The interference frequency band represents the continuous frequency interval where the polarization response of the salt ions dominates. The difference in the signal attenuation slope represents the difference in the signal drop rate between the two frequency bands.

[0100] In the embodiment of the present application, first, cluster analysis is performed based on the time-series characteristics of the set of frequency points, and the frequency points with a steep drop occurring earlier are classified into the characteristic frequency band (organic matter), and those occurring later are classified into the interference frequency band (salt ions). Then, the average attenuation slopes of the two frequency bands are calculated respectively, and finally, the arithmetic difference between the two slopes is used as the polarization attenuation characteristic difference value.

[0101] The following is a specific example:

[0102] Taking the detection of oil pollution in a saline-alkali farmland as an example, based on the completion of 10-meter by 10-meter grid sampling and the acquisition of resistivity, polarizability, dielectric constant, and conductivity data, frequency band screening is performed on the dynamic data set: First, the polarization response signals of each sampling point are converted into frequency domain data through fast Fourier transform. According to the historical test data of this area, the characteristic frequency band of organic pollutants (petroleum hydrocarbons) is determined to be 80 - 200 Hz (low frequency band), and the salt ion interference frequency band is 300 - 450 Hz (medium frequency band). The 150 - 250 Hz between the two frequency bands is selected as the boundary interval; within this interval, the frequency domain data is analyzed, and 18 characteristic frequency points with signal intensity first rising to a peak and then rapidly decreasing (the decrease slope exceeds 5 decibels per second) are screened out. Among them, 9 frequency points in the range of 160 - 190 Hz (the average decrease slope is 1.2 units per minute) are determined as the petroleum characteristic frequency band, and 9 frequency points in the range of 210 - 240 Hz (the average decrease slope is 2.8 units per minute) are determined as the salt interference frequency band; the polarization attenuation characteristic difference value is obtained by calculating the difference between the average slopes of the two frequency bands (2.8 - 1.2 = 1.6). The calculation process of the characteristic difference value is as follows: First, calculate the arithmetic average of the instantaneous slopes of each frequency point, and then take the difference between the average values of the two frequency bands to ensure that the quantization result can accurately reflect the polarization response difference between organic matter and salt ions.

[0103] In the embodiment of the present application, this technical solution effectively separates the electrical response characteristics of organic pollutants and salt ions through precise frequency band division and feature extraction, provides a reliable feature basis for accurately predicting the pollutant concentration subsequently, and improves the accuracy of organic pollution detection in a high-salt environment.

[0104] In order to further improve the accuracy and anti-interference ability of frequency domain analysis, in some embodiments, step 201: The conversion of the polarization response signal in the dynamic data set into frequency domain distribution data by means of frequency domain dielectric spectroscopy analysis includes:

[0105] Step 301: Extract the polarization response signal from the dynamic data set.

[0106] In step 301, the polarization response signal represents the electrical response data that changes with time generated by the soil under the action of an alternating electric field, including amplitude and phase information.

[0107] In the embodiment of the present application, the polarization response signal data is located and read from the storage structure of the dynamic data set. This signal is waveform data with time as the abscissa and voltage value as the ordinate, and the sampling frequency needs to meet the requirements of the Nyquist sampling theorem to ensure signal integrity.

[0108] Step 302: Intercept the polarization response signal according to a preset segment length to generate multiple time domain signal sequences.

[0109] In step 302, the preset segmentation length represents the time window length set according to the signal characteristics and analysis requirements. The time-domain signal sequence represents discrete signal segments segmented by a fixed time duration.

[0110] In the embodiment of the present application, the continuous polarization signal is segmented by using an overlapping frame division method. The length of each segment is set to an integer multiple of the fundamental wave period of the signal (usually taking 5 - 10 periods), and a partial overlapping area is reserved between adjacent segments to reduce the spectral leakage caused by the truncation effect.

[0111] Step 303: Perform a frequency-domain conversion process on each time-domain signal sequence to generate a target signal intensity distribution sequence corresponding to each time-domain signal sequence.

[0112] In step 303, the frequency-domain conversion process represents a digital signal processing method for converting a time-domain signal into a frequency-domain representation. The target signal intensity distribution sequence represents the frequency-energy distribution data obtained after the frequency-domain conversion.

[0113] In the embodiment of the present application, each time-domain signal is first windowed (using a Hanning window function), then the spectrum is calculated by the fast Fourier transform algorithm, and finally the modulus value squared is taken to obtain the signal intensity spectrum, forming a target signal intensity distribution sequence corresponding to each segment.

[0114] Step 304: Determine the frequency range of the target signal intensity distribution sequence according to the proportional relationship between the conductivity parameter and the polarizability parameter in the conductivity spatial distribution data.

[0115] In step 304, the conductivity-polarizability ratio represents a characteristic parameter reflecting the degree of soil salinization. The proportional relationship is established through statistical analysis of historical sample data. The specific process is as follows: A large number of sample points are collected in a typical salinized soil area, the conductivity values and polarizability values of each point are measured synchronously, and the ratio of the two types of parameters is calculated after normalization processing; through cluster analysis, it is found that this ratio is correlated with the characteristic frequency band of organic pollutants. When the ratio is greater than 1, it indicates that the interference of salt ions is strong, and the middle and high frequency bands (200 - 500 Hz) need to be focused on for analysis. When the ratio is less than 0.5, it indicates that the characteristics of organic matter are obvious, and the low frequency band (50 - 200 Hz) should be mainly investigated. When it is between 0.5 - 1, full-band analysis is adopted; finally, these empirical relationships are compiled into a query comparison table, and the optimal analysis frequency band range is determined by looking up the table according to the real-time measured conductivity and polarizability values during actual detection. The frequency range represents the effective analysis frequency band of the signal intensity distribution sequence. The frequency range represents the effective analysis frequency band of the signal intensity distribution sequence.

[0116] In the embodiments of the present application, according to the conductivity - polarizability ratio and the corresponding relationship table of characteristic frequency bands established based on measured data, the analysis frequency range of each segment of the signal is dynamically determined. When the ratio is relatively high, the focus is on medium - high frequency analysis; when the ratio is relatively low, the low - frequency range is extended.

[0117] Step 305: Perform superposition processing on the target signal intensity distribution sequences corresponding to all time - domain signal sequences to generate frequency - domain distribution data covering the frequency band range.

[0118] In step 305, the superposition processing represents a method of fusing the frequency - domain results of multiple segments of signals.

[0119] In the embodiments of the present application, incoherent superposition is performed on each segment of the target signal intensity distribution sequence, that is, the maximum intensity value at the same frequency point is taken as the final result, retaining the effective information of all frequency bands to generate complete frequency - domain distribution data.

[0120] The following is a specific example:

[0121] Taking the detection of petroleum pollution in a salinized farmland as an example, on the basis of having completed grid sampling and established a dynamic data set, signal frequency - domain conversion is performed on sampling point No. 3 (conductivity value is 1200 μS / cm, polarizability value is 0.15): First, extract the polarization response signal of this point from the dynamic data set (sampling rate 1000 Hz, duration 2 seconds), frame it every 200 milliseconds (including 10 complete cycles), with an overlap of 100 milliseconds between adjacent frames, and a total of 19 time - domain signal sequences are obtained; after applying a Hamming window to each segment of the signal, perform a 1024 - point Fourier transform to obtain the signal intensity spectrum in the range of 0 - 500 Hz; according to the ratio of the conductivity to the polarizability of this point, which is 8 (1200 / 150 = 8, where the polarizability value of 0.15 is converted to 150 for unified dimension), look up the pre - established ratio - frequency band comparison table to determine the analysis frequency band as 100 - 400 Hz; superimpose the maximum values of each frequency point of the intensity spectra of the 19 segments of signals in the range of 100 - 400 Hz, and finally generate the frequency - domain distribution data of this point. There is an obvious peak at 180 Hz (petroleum characteristic peak) and a secondary peak at 350 Hz (salt ion interference peak), providing accurate frequency - domain basic data for subsequent characteristic difference analysis. The frame length of 200 milliseconds is determined according to 10 times the period of the signal fundamental frequency of 50 Hz (period 20 milliseconds), and the frequency band range of 100 - 400 Hz is obtained according to the rule that the ratio of 8 falls into the preset "5 - 10" interval corresponding to "focus on medium - high frequency analysis".

[0122] In the embodiments of the present application, through scientific and reasonable signal segmentation and frequency - domain conversion methods, combined with dynamic adjustment of the analysis frequency band according to soil electrical parameters, the spectrum features with physical significance are effectively extracted, laying a solid foundation for accurately distinguishing organic pollutants and salt ion interference subsequently, and improving the reliability of the detection results.

[0123] In order to further improve the accuracy of identifying the characteristic frequency bands of organic pollutants and salt ions, in some embodiments, step 204: extracting the characteristic frequency bands corresponding to the organic pollutants and the interference frequency bands corresponding to the salt ions from the frequency point set, and calculating the signal attenuation slope difference between the characteristic frequency bands and the interference frequency bands, includes:

[0124] Step 401: According to the difference in the steep drop time sequence of the signal intensity in the frequency point set, divide the continuous N frequency points with the steep drop time sequence in the starting region within the intersection interval in the frequency point set into the characteristic frequency bands of organic pollutants, and divide the continuous frequency points with the steep drop time sequence in the end region of the intersection interval into the interference frequency bands of salt ions, where N is greater than or equal to 3.

[0125] In step 401, the difference in the steep drop time sequence represents the relative order of appearance of the falling edge of the signal intensity on the time axis. The intersection interval is only divided into two regions, the starting region (corresponding to the characteristic frequency bands of organic matter) and the end region (corresponding to the interference frequency bands of salt ions). The relationship and difference between the intersection interval and the starting region: The starting region is the first half (low-frequency side) of the intersection interval, used to capture the polarization response of organic matter. The intersection interval is the overall range (such as 50Hz–200Hz), and the starting region is its subset (such as 50Hz–120Hz). The relationship and difference between the intersection interval and the end region: The end region is the second half (high-frequency side) of the intersection interval, used to capture the polarization response of salt ions. The end region (such as 120Hz–200Hz) does not overlap with the starting region, and the two together constitute the complete intersection interval.

[0126] In the embodiments of the present application, first, all the frequency points within the intersection interval are sorted in ascending order of frequency, the starting time of the signal intensity drop of each frequency point is calculated, the continuous 5 frequency points starting to drop within the first one-third frequency band are classified as the characteristic frequency bands of organic pollutants, and the continuous 5 frequency points starting to drop within the last one-third frequency band are classified as the interference frequency bands of salt ions, ensuring that the two types of characteristic frequency bands are completely separated in terms of frequency distribution.

[0127] Step 402: Within the characteristic frequency bands of the organic pollutants, select the turning frequency point where the signal intensity changes from increasing to steeply decreasing as the first reference point, and generate a first analysis window by extending a preset number of frequency points to both sides centered on the first reference point.

[0128] In step 402, the turning frequency point represents the critical frequency point where the change trend of the signal intensity changes from rising to falling. The first analysis window represents the local frequency band analysis range centered on the turning point.

[0129] In the embodiments of the present application, the rate of change of the signal strength at each frequency point within the characteristic frequency band is calculated by the difference method. When the rate of change turns from positive to negative, it is determined as the turning frequency point. Taking 2 adjacent frequency points on both the high-frequency and low-frequency sides centered on this point, a first analysis window containing 5 frequency points is formed.

[0130] Step 403: Within the salt ion interference frequency band, select the frequency points whose signal strength steep drop amplitude is higher than the preset amplitude threshold as the second reference points, and generate a second analysis window by extending the same preset number of frequency points to both sides centered on the second reference points.

[0131] In step 403, the preset amplitude threshold represents the minimum signal strength change amount for determining an effective steep drop. The second analysis window represents the interference frequency band analysis range symmetric to the first window.

[0132] In the embodiments of the present application, within the interference frequency band, calculate the signal strength drop amplitude at each frequency point, select the frequency points whose amplitude exceeds 3 times the background noise level as the second reference points, and also extend 2 frequency points to both sides to form the second analysis window, maintaining the same analysis scale as the first window.

[0133] Step 404: Calculate the first average drop rate of the signal strength of the frequency points after the turning frequency point in the first analysis window and the second average drop rate of the signal strength of the frequency points after the turning frequency point in the second analysis window respectively.

[0134] In step 404, the average drop rate represents the average drop speed of the signal strength with respect to the frequency change within the analysis window.

[0135] In the embodiments of the present application, calculate the signal strength change in the frequency band after the turning point for the two analysis windows respectively, fit the linear part of the intensity-frequency curve using the least squares method, and take the absolute value of the slope of the fitted straight line as the average drop rate of the corresponding frequency band.

[0136] Step 405: Take the absolute value difference between the first average drop rate and the second average drop rate as the signal attenuation slope difference.

[0137] In step 405, the absolute value difference represents the absolute value of the difference between the two drop rate values.

[0138] In the embodiments of the present application, subtract the organic matter drop rate obtained from the first analysis window from the salt ion drop rate of the second analysis window and then take the absolute value as the final characteristic difference index.

[0139] The following is a specific example:

[0140] Taking the detection of oil pollution in a saline-alkali farmland as an example, 18 characteristic frequency points (the signal intensity first increases and then decreases, and the decreasing slope exceeds 5 decibels per second) screened at the 7th sampling point in the range of 150 - 250 Hz are processed as follows: First, the first 9 frequency points (in the range of 160 - 190 Hz) are divided into the oil characteristic frequency band according to the order of frequency point appearance, and the last 9 frequency points (in the range of 210 - 240 Hz) are divided into the salt interference frequency band; in the oil characteristic frequency band, the turning point where the signal intensity reaches the peak at 175 Hz and then starts to decline is selected as the first reference point, and 2 frequency points are extended to both sides to form the first analysis window of 170 - 180 Hz; in the salt interference frequency band, the frequency point where the decline amplitude reaches the preset threshold (more than 3 times the background noise) at 230 Hz is selected as the second reference point, and 2 frequency points are also extended to both sides to form the second analysis window of 225 - 235 Hz; the signal intensity decline rates of the frequency bands after the turning points in the two windows are calculated respectively. The decline rates of the 3 frequency points (176 Hz, 178 Hz, 180 Hz) in the first analysis window are 1.1, 1.2, and 1.3 units per minute respectively, and the arithmetic mean is taken to obtain the first average decline rate of 1.2 units. The decline rates of the 3 frequency points (231 Hz, 233 Hz, 235 Hz) in the second analysis window are 2.7, 2.8, and 2.9 units respectively, and the average value is taken to obtain the second average decline rate of 2.8 units; finally, the absolute value of the difference between the two rates is taken to obtain the signal attenuation slope difference of 1.6 (2.8 - 1.2 = 1.6). In the calculation process of this value, the instantaneous slope of each frequency point is obtained by dividing the signal intensity difference between adjacent frequency points by the frequency difference. For example, the slope of the 176 Hz frequency point = (signal intensity at 176 Hz - signal intensity at 174 Hz) / (176 Hz - 174 Hz).

[0141] In the embodiment of the present application, this technical solution effectively captures the essential differences in the polarization attenuation characteristics between organic pollutants and salt ions by accurately dividing the characteristic frequency bands and scientifically calculating the attenuation differences, provides a stable and reliable characteristic index for subsequent pollution identification, and improves the accuracy of organic pollution detection in high saline-alkali environments.

[0142] In order to further improve the accuracy and anti-interference ability of frequency domain conversion, in some embodiments, step 303: The frequency domain conversion process for each time domain signal sequence to generate a target signal intensity distribution sequence corresponding to each time domain signal sequence includes:

[0143] Step 501: Multiply each time domain signal sequence point by point with a preset window function sequence to obtain a time domain signal subsequence.

[0144] In step 501, the window function sequence represents a weighted coefficient sequence used to suppress the signal truncation effect. The time domain signal subsequence represents a finite-length signal segment after windowing processing.

[0145] In the embodiment of the present application, a Hanning window function is used to window each segment of the time-domain signal. The length of the window function is strictly consistent with the length of the signal segment. The signal is smoothly attenuated at both ends by point-by-point multiplication, effectively reducing spectral leakage. The window function coefficients are generated according to a cosine-squared distribution that is centrosymmetric, ensuring that the main lobe is concentrated and the side lobes decay rapidly.

[0146] Step 502: Perform spectral analysis processing on the time-domain signal subsequence to generate an initial signal intensity distribution sequence.

[0147] In step 502, spectral analysis processing refers to a digital signal processing method that converts a time-domain signal into a frequency-domain representation. The initial signal intensity distribution sequence represents the unoptimized original spectral analysis result.

[0148] In the embodiment of the present application, a discrete Fourier transform is performed on the windowed signal subsequence. The number of transform points is taken as the smallest power of 2 greater than or equal to the signal length. After calculating the complex spectrum, the modulus value is squared to generate an initial signal intensity distribution sequence containing amplitude information. The frequency resolution is jointly determined by the sampling rate and the number of transform points.

[0149] Step 503: Set to zero the frequency units in the initial signal intensity distribution sequence that are lower than a preset environmental noise threshold to generate a target signal intensity distribution sequence.

[0150] In step 503, the environmental noise threshold represents a signal intensity threshold value set according to the background noise level.

[0151] In the embodiment of the present application, first, the average intensity of the lowest frequency band in the initial spectrum is calculated as the noise benchmark. A specific multiple of this benchmark value is set as the noise threshold. The intensity values of all frequency units are judged against the threshold, and the units below the threshold are set to zero, retaining the effective signal components, and finally generating a denoised target signal intensity distribution sequence.

[0152] The following is a specific example:

[0153] Taking the 5th segment of the time-domain signal (200 milliseconds in duration, 200 sampling points) at the 3rd sampling point of a certain salinized farmland as an example for frequency-domain conversion processing: First, multiply the original signal point by point using a 200-point Hamming window function (the window function coefficients are calculated and generated according to the formula 0.54 - 0.46 * cos(2πn / 199)) to obtain the windowed time-domain signal subsequence; after padding this subsequence with zeros to 1024 points, perform a fast Fourier transform to obtain the initial signal intensity spectrum of 512 frequency points in the range of 0 - 500 Hz, where the intensity value of each frequency point is the square of the modulus value of the corresponding frequency component; select the average intensity of 0.02 in the frequency band of 120 - 150 Hz (this frequency band theoretically only contains environmental noise) as the benchmark, set the preset environmental noise threshold to 2.5 times the benchmark value, that is, 0.05, and set to zero all the frequency points in the initial spectrum with intensities lower than 0.05 (such as 0.03 at 80 Hz, 0.04 at 420 Hz, etc.). In the finally generated target signal intensity distribution sequence, the petroleum characteristic peak at 180 Hz is retained intact (intensity 0.78), the salt ion interference peak at 350 Hz is partially retained (intensity 0.12), and the remaining noise frequency points are effectively filtered out. The calculation formula of the Hamming window coefficient can ensure smooth transitions at both ends of the signal, and the noise threshold of 0.05 is a reasonable threshold value determined through statistical analysis of typical noise frequency bands, which can effectively distinguish useful signals from background noise.

[0154] In the embodiment of the present application, this technical solution improves the accuracy and signal-to-noise ratio of frequency-domain analysis through scientific windowing processing and noise suppression means, provides high-quality spectral data for subsequent feature extraction, and effectively guarantees the accuracy of organic pollutant detection.

[0155] In order to further improve the characterization ability of soil electrical properties, in some embodiments, step 102: generating a bimodal feature matrix based on the resistivity parameter and the polarizability parameter includes:

[0156] Step 601: In the salinized soil environment, perform synchronous alignment processing on the resistivity parameter and the polarizability parameter at the same spatial position to generate a set of parameter pairs.

[0157] In step 601, the spatial position means dividing the target soil area into equally spaced square / hexagonal grids, and the center point of each grid is used as a spatial position (such as a 1m × 1m grid). The synchronous alignment processing represents a data matching process that ensures strict correspondence of different parameters in the spatial position. The set of parameter pairs represents a paired data set composed of resistivity and polarizability data at the same position point.

[0158] In the embodiments of the present application, first, a unified spatial coordinate indexing system is established. Through the geographic information system, the resistivity and polarizability measurement data of each sampling point are position-matched, and the data points with position offsets exceeding the allowable error are eliminated. Finally, a set of resistivity-polarizability parameter pairs corresponding strictly to each sampling point is formed.

[0159] Step 602: Convert the resistivity parameter in each parameter pair in the parameter pair set into a first eigenvalue, and convert the polarizability parameter into a second eigenvalue.

[0160] In step 602, the first eigenvalue is a standardized value used to reflect the soil resistance characteristics. The second eigenvalue is a normalized value characterizing the soil polarization characteristics.

[0161] In the embodiments of the present application, take the logarithm of the resistivity parameter with base 10 and perform linear normalization processing, and convert it to the 0-1 interval as the first eigenvalue; perform angular normalization processing on the polarizability parameter, and divide its phase angle by 90 degrees as the second eigenvalue to ensure the comparability of the two types of parameters.

[0162] Step 603: Combine the first eigenvalue and the second eigenvalue at the same spatial position into a two-dimensional feature vector.

[0163] In step 603, the two-dimensional feature vector is a two-element data unit that simultaneously contains resistivity and polarizability characteristics.

[0164] In the embodiments of the present application, combine the first eigenvalue and the second eigenvalue of the same sampling point in a fixed order (resistivity first, polarizability second) to form a two-dimensional vector representing the comprehensive electrical characteristics of the point, and the vector elements retain the original physical meaning.

[0165] Step 604: Arrange the two-dimensional feature vectors at all spatial positions to form a bimodal feature matrix.

[0166] In the embodiments of the present application, according to the spatial distribution order of the sampling points (such as from left to right, from top to bottom), arrange the two-dimensional feature vectors of each point as the row vectors of the matrix in turn to construct an m×2 bimodal feature matrix, where m is the number of sampling points, and the two columns correspond to the resistivity and polarizability characteristics respectively.

[0167] The following is a specific example:

[0168] Taking the pollution detection of a saline-alkali farmland (200 m × 150 m) as an example, based on the completed 10 m × 10 m grid sampling (a total of 315 sampling points), the collected resistivity and polarizability data are processed as follows: First, through the built-in GPS module of the measuring device, ensure that the resistivity (e.g., the measured value at point 15 is 92 Ω·m) and polarizability (e.g., the measured value at point 15 is 0.22) data of each sampling point are strictly spatially aligned; after taking the common logarithm of the resistivity value of each point (lg92 ≈ 1.96), perform linear normalization according to the regional resistivity range (50 - 300 Ω·m), and the calculation formula is (1.96 - lg50) / (lg300 - lg50) ≈ 0.72; the polarizability value is directly divided by the regional maximum value of 0.3 to get 0.73, and thus generate the two-dimensional feature vector [0.72, 0.73] of point 15; according to the order of the farmland sampling point numbers (starting from the northwest corner, arranged from west to east and from north to south), take the feature vectors of all 315 points as row vectors in turn, and construct a bimodal feature matrix of 315 rows × 2 columns. The data in the first column (resistivity feature) is obtained through the above logarithmic normalization process, and the data in the second column (polarizability feature) is obtained through linear normalization. After this matrix is fused with the dielectric constant and conductivity data, it provides a standardized feature input for subsequent frequency-domain analysis and machine learning prediction. The normalization process can eliminate the dimensional differences between different parameters and ensure that each feature has the same importance in model training.

[0169] In the embodiment of the present application, through scientific data alignment and feature transformation methods, this technical solution constructs a comprehensive feature matrix that can simultaneously reflect the conductive and polarizing characteristics of the soil, provides a multi-dimensional electrical feature basis for the identification of organic pollutants, and improves the feature representation ability of the detection model.

[0170] In order to further improve the accuracy of predicting the concentration of organic pollutants, in some embodiments, step 105: Based on the polarization attenuation feature difference and the bimodal feature matrix, use the trained machine learning model to output the predicted result of the spatial concentration distribution of organic pollutants in the saline-alkali soil environment, including:

[0171] Step 701: Combine each two-dimensional feature vector in the bimodal feature matrix with the polarization attenuation feature difference at the corresponding spatial position to generate an enhanced feature vector.

[0172] In step 701, the enhanced feature vector refers to a multi-dimensional feature representation that fuses electrical features and spectral features.

[0173] In the embodiment of the present application, the two-dimensional electrical feature vector (resistivity + polarizability) of each sampling point is spliced with the polarization attenuation feature difference value to form a three-dimensional enhanced feature vector, ensuring that it contains both time-domain and frequency-domain feature information.

[0174] Step 702: Perform dynamic weighting on the enhanced feature vector according to the salt ion concentration gradient generated from the conductivity parameters in the conductivity spatial distribution data.

[0175] In step 702, the generation process of the salt ion concentration gradient is as follows: First, perform Gaussian filtering and smoothing on the conductivity spatial distribution data to eliminate measurement noise; then, based on the linear relationship model between soil conductivity and salt ion concentration (salt ion concentration = 0.8 × conductivity value + 20, where the conductivity unit is μS / cm and the concentration unit is mg / kg), convert the conductivity parameters at each point into salt ion concentration values; then calculate the spatial change rate of the concentration values, and obtain the gradient components in the east-west direction and the north-south direction respectively; finally, perform vector synthesis on the gradient components in the two directions to obtain the gradient value representing the intensity of the spatial change of the salt ion concentration, where the gradient direction points to the direction with the fastest increasing concentration, and the gradient magnitude reflects the severity of the concentration change. This gradient value is used for weight calculation in dynamic weighting to ensure the contribution degree of enhanced spectral features in the area where the salt ion concentration changes. Dynamic weighting represents a method of adjusting the feature importance according to the salt ion interference intensity.

[0176] In the embodiment of the present application, the salt ion concentration gradient is calculated through conductivity parameters, and the polarization attenuation feature difference components in the enhanced feature vector are weighted. The higher the salt ion concentration, the greater the weight of this feature (weight coefficient = 1 + 0.5 × normalized salt ion concentration value).

[0177] Step 703: Input the weighted enhanced feature vector into the trained machine learning model according to the preset spatial distribution order to generate the initial pollution concentration value at each spatial position.

[0178] In step 703, the initial pollution concentration value represents the original prediction result directly output by the machine learning model.

[0179] In the embodiment of the present application, the weighted feature vector is input into the pre-trained gradient boosting tree model according to the spatial order of the sampling points, and the model outputs the pollutant concentration prediction values at each point based on the learned feature-concentration mapping relationship.

[0180] Step 704: Perform dielectric correction processing on the initial pollution concentration value based on the dielectric constant distribution data.

[0181] In step 704, dielectric correction processing represents a post-processing method of adjusting the concentration prediction value according to dielectric characteristics.

[0182] In the embodiment of the present application, a calibration coefficient table between the dielectric constant and the pollutant concentration is established to compensate the initial prediction value (correction coefficient = 1 + 0.3 × (dielectric constant - regional average value) / standard deviation) to improve the prediction accuracy in the high-dielectric region.

[0183] Step 705: Associate the corrected pollution concentration values with the corresponding spatial position coordinates, and output the predicted results of the spatial concentration distribution.

[0184] In the embodiment of the present application, the corrected concentration values are associated with the sampling point coordinates, and a concentration distribution grid map with consistent resolution is generated through spatial interpolation, and different colors are used to represent the pollution degree levels.

[0185] The following is a specific example:

[0186] Taking the pollution detection of a certain salinized farmland (100 meters × 100 meters) as an example, based on the completion of data collection and processing of 100 sampling points: First, combine the bimodal feature vectors of each sampling point (such as the 23rd point [0.68, 0.71]) with the corresponding polarization attenuation feature difference value (1.5) to form a three-dimensional enhanced feature vector [0.68, 0.71, 1.5]; According to the conductivity value of 1500 μS / cm at this point, calculate the salt ion concentration of 1220 mg / kg through the formula "salt ion concentration = 0.8 × conductivity + 20", and then calculate the gradient magnitude in combination with the concentration values of the surrounding 8 points (using the central difference method, the east-west direction gradient = (1260 - 1180) / 20 = 4, the north-south direction gradient = (1240 - 1200) / 20 = 2, and the combined gradient value is √(4² + 2²) ≈ 4.47), determine the dynamic weight coefficient as 1 + 0.2 × 4.47 ≈ 1.89, and obtain [0.68, 0.71, 2.84] after weighting the third-dimensional feature; Input this vector into the pre-trained gradient boosting decision tree model, and output the initial predicted concentration value of 18.6 mg / kg; Calculate the correction coefficient 1 + 0.25 × (9.2 - 8.0) / 1.5 = 1.2 according to the dielectric constant of 9.2 at this point (regional mean 8.0, standard deviation 1.5), and obtain the corrected concentration of 22.3 mg / kg; Finally, import the corrected concentration values and coordinate information of all 100 sampling points into the geographic information system, and generate a spatial concentration distribution map with a resolution of 10 meters through inverse distance weighted interpolation, where the gradient calculation uses a 20-meter adjacent sampling point spacing as the differential step size, and 0.25 in the dielectric correction coefficient is an empirical adjustment parameter to ensure reasonable correction of the pollutant concentration in the high-dielectric region.

[0187] In the embodiment of the present application, this technical solution improves the accuracy and spatial resolution of organic pollution prediction in high-saline-alkali environments through innovative processing such as feature fusion, dynamic weighting, and dielectric correction, providing a reliable basis for precise treatment. The whole set of methods fully considers the particularity of salinized soil and has important practical value.

[0188] Figure 2 It is a schematic structural diagram of a soil organic pollution concentration prediction system based on resistivity parameters provided for the embodiment of the present application, asFigure 2 As shown, the system includes:

[0189] An acquisition module 21, configured to acquire the dielectric constant distribution data, conductivity spatial distribution data, resistivity parameters, and polarizability parameters of the soil in the target area in the organic pollution detection scenario under the saline-alkali soil environment.

[0190] A first generation module 22, configured to generate a bimodal feature matrix based on the resistivity parameters and the polarizability parameters.

[0191] A second generation module 23, configured to perform multi-source coupling association on the bimodal feature matrix, the dielectric constant distribution data, and the conductivity spatial distribution data to generate a dynamic data set.

[0192] A screening module 24, configured to perform frequency band screening on the dynamic data set through frequency domain dielectric spectroscopy analysis to extract the polarization attenuation feature differences between organic pollutants and salt ions within a preset frequency band range.

[0193] An output module 25, configured to output the predicted result of the spatial concentration distribution of organic pollutants in the saline-alkali soil environment based on the polarization attenuation feature differences and the bimodal feature matrix by using a trained machine learning model.

[0194] Figure 2 The described soil organic pollution concentration prediction system based on resistivity parameters can execute Figure 1 The described soil organic pollution concentration prediction method based on resistivity parameters in the embodiments shown. Its implementation principle and technical effects will not be elaborated further. For the soil organic pollution concentration prediction system based on resistivity parameters in the above embodiments, the specific manners in which each module and unit perform operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0195] In a possible design, Figure 2 The soil organic pollution concentration prediction system based on resistivity parameters in the embodiments shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0196] The storage component 31 stores one or more computer instructions, where the one or more computer instructions are called and executed by the processing component 32.

[0197] The processing component 32 is the above Figure 1 The soil organic pollution concentration prediction method based on resistivity parameters in the embodiments.

[0198] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above methods. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above methods.

[0199] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disks or optical discs.

[0200] Of course, the computing device may necessarily further include other components, such as input / output interfaces, display components, communication components, etc.

[0201] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above peripheral interface module may be an output device, an input device, etc.

[0202] The communication component is configured to facilitate wired or wireless communication between the computing device and other devices, etc.

[0203] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above processing component, storage component, etc. may be basic server resources leased or purchased from a cloud computing platform.

[0204] The embodiments of the present application also provide a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above-mentioned Figure 1 method for predicting soil organic pollution concentration based on resistivity parameters in the embodiments shown.

[0205] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0206] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0207] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting the concentration of soil organic pollution based on resistivity parameters, characterized in that, Including: In the scenario of organic pollution detection in saline-alkali soil environment, obtain the dielectric constant distribution data, conductivity spatial distribution data, resistivity parameters and polarizability parameters of the soil in the target area; Generate a bimodal feature matrix based on the resistivity parameters and the polarizability parameters; Perform multi-source coupling correlation on the bimodal feature matrix, the dielectric constant distribution data and the conductivity spatial distribution data to generate a dynamic data set; Combined with the frequency-domain dielectric spectroscopy analysis method, perform frequency band screening on the dynamic data set to extract the polarization attenuation characteristic differences between organic pollutants and salt ions within a preset frequency band range; Based on the polarization attenuation characteristic differences and the bimodal feature matrix, use the trained machine learning model to output the prediction result of the spatial concentration distribution of organic pollutants in the saline-alkali soil environment.

2. The method according to claim 1, wherein The combining the frequency-domain dielectric spectroscopy analysis method to perform frequency band screening on the dynamic data set to extract the polarization attenuation characteristic differences between organic pollutants and salt ions within a preset frequency band range includes: Convert the polarization response signal in the dynamic data set into frequency-domain distribution data through the frequency-domain dielectric spectroscopy analysis method; Based on the polarization characteristic differences between organic pollutants and salt ions in the saline-alkali soil environment, determine the junction interval between the low-frequency band and the middle-frequency band within the preset frequency band range; Within the junction interval, according to the frequency-domain distribution data, screen out the frequency points where the signal intensity first increases and then steeply decreases as the frequency increases to form a frequency point set; Extract the characteristic frequency band corresponding to the organic pollutant and the interference frequency band corresponding to the salt ion from the frequency point set, and calculate the signal attenuation slope difference between the characteristic frequency band and the interference frequency band, and use the signal attenuation slope difference as the polarization attenuation characteristic difference.

3. The method according to claim 2, wherein The converting the polarization response signal in the dynamic data set into frequency-domain distribution data through the frequency-domain dielectric spectroscopy analysis method includes: Extract the polarization response signal from the dynamic data set; Intercept the polarization response signal according to a preset segmentation length to generate multiple time-domain signal sequences; Perform frequency-domain conversion processing on each time-domain signal sequence to generate a target signal intensity distribution sequence corresponding to each time-domain signal sequence; Determine the frequency range of the target signal intensity distribution sequence according to the proportional relationship between the conductivity parameter in the conductivity spatial distribution data and the polarizability parameter; Perform superposition processing on the target signal intensity distribution sequences corresponding to all time-domain signal sequences to generate frequency-domain distribution data covering the frequency band range.

4. The method according to claim 2, characterized in that, The extracting the characteristic frequency band corresponding to the organic pollutant and the interference frequency band corresponding to the salt ion from the frequency point set, and calculating the signal attenuation slope difference between the characteristic frequency band and the interference frequency band includes: According to the steep drop time sequence difference of the signal intensity in the frequency point set, divide the continuous N frequency points with the steep drop time sequence located in the starting area within the junction interval in the frequency point set into the organic pollutant characteristic frequency band, and divide the continuous frequency points with the steep drop time sequence located in the end area of the junction interval into the salt ion interference frequency band, where N is greater than or equal to 3; Within the characteristic frequency band of the organic pollutants, select the turning frequency point where the signal intensity changes from increasing to steeply decreasing as the first reference point, and generate a first analysis window by extending a preset number of frequency points to both sides centered on the first reference point; Within the salt ion interference frequency band, select the frequency points where the steep drop amplitude of the signal intensity is higher than the preset amplitude threshold as the second reference point, and generate a second analysis window by extending the same preset number of frequency points to both sides centered on the second reference point; Calculate the first average decline rate of the signal intensity of the frequency points after the turning frequency point in the first analysis window and the second average decline rate of the signal intensity of the frequency points after the turning frequency point in the second analysis window respectively; Take the absolute value difference between the first average decline rate and the second average decline rate as the signal attenuation slope difference.

5. The method according to claim 3, characterized in that, The frequency domain conversion process for each time domain signal sequence to generate the target signal intensity distribution sequence corresponding to each time domain signal sequence includes: Perform point-by-point multiplication processing on each time domain signal sequence and a preset window function sequence to obtain a time domain signal subsequence; Perform spectral analysis processing on the time domain signal subsequence to generate an initial signal intensity distribution sequence; Set to zero the frequency units in the initial signal intensity distribution sequence that are lower than the preset environmental noise threshold to generate the target signal intensity distribution sequence.

6. The method according to claim 1, characterized in that, The generation of the dual-mode feature matrix based on the resistivity parameter and the polarizability parameter includes: Under the saline-alkali soil environment, perform synchronous alignment processing on the resistivity parameter and the polarizability parameter at the same spatial position to generate a set of parameter pairs; Convert the resistivity parameter in each parameter pair in the set of parameter pairs into a first eigenvalue, and convert the polarizability parameter into a second eigenvalue; Combine the first eigenvalue and the second eigenvalue at the same spatial position into a two-dimensional feature vector; Arrange the two-dimensional feature vectors at all spatial positions to form a dual-mode feature matrix.

7. The method according to claim 1, characterized in that The output of the spatial concentration distribution prediction result of the organic pollutants in the saline-alkali soil environment based on the polarization attenuation feature difference and the dual-mode feature matrix using a trained machine learning model includes: Combine each two-dimensional feature vector in the dual-mode feature matrix with the polarization attenuation feature difference at the corresponding spatial position to generate an enhanced feature vector; Perform dynamic weighting processing on the enhanced feature vector according to the salt ion concentration gradient generated from the conductivity parameter in the conductivity spatial distribution data; Input the weighted enhanced feature vectors into the trained machine learning model in the preset spatial distribution order to generate an initial pollution concentration value for each spatial position; Perform dielectric correction processing on the initial pollution concentration value based on the dielectric constant distribution data; Associate the corrected pollution concentration value with the corresponding spatial position coordinates and output the spatial concentration distribution prediction result.

8. A soil organic pollution concentration prediction system based on resistivity parameters, characterized in that, Including: An acquisition module for acquiring the dielectric constant distribution data, conductivity spatial distribution data, resistivity parameter, and polarizability parameter of the soil in the target area in the organic pollution detection scenario under the saline-alkali soil environment; A first generation module for generating a bimodal feature matrix based on the resistivity parameter and the polarizability parameter; A second generation module for performing multi-source coupling correlation on the bimodal feature matrix, the dielectric constant distribution data, and the conductivity spatial distribution data to generate a dynamic data set; A screening module for screening frequency bands of the dynamic data set through frequency domain dielectric spectroscopy analysis to extract the polarization attenuation characteristic differences between organic pollutants and salt ions within a preset frequency band range; An output module for outputting a predicted result of the spatial concentration distribution of organic pollutants in a salinized soil environment based on the polarization attenuation characteristic differences and the bimodal feature matrix by using a trained machine learning model.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for predicting the concentration of soil organic pollution based on resistivity parameters as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by a computer, it implements a method for predicting the concentration of soil organic pollution based on resistivity parameters as described in any one of claims 1 to 7.

Images