Rapid diagnosis method and system for fruit mulberry bacterial wilt based on electrical impedance spectroscopy analysis

Abstract

This invention discloses a rapid diagnostic method and system for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis. The method includes: acquiring raw electrical impedance spectral data from the tested mulberry plants through non-invasive frequency sweep measurement; extracting multidimensional real-time diagnostic feature vectors after denoising and parasitic parameter compensation; performing nonlinear dimensionality reduction on the feature vectors and constructing a classification model; calculating the decision function value and offset percentage to jointly determine and generate real-time disease diagnosis information; if suspected latent infection is identified, reconstructing the spectral data by frequency logarithmic differentiation and relaxation time distribution spectrum inversion; using the expectation-maximization algorithm to separate the bimodal distribution spectrum using a Gaussian mixture model to construct a vascular bundle sap flow continuity index; performing a secondary precise determination based on this index; and, if necessary, acquiring spectral data at noon to calculate the diurnal variation amplitude to complete the final diagnosis. This invention achieves latent period early warning and quantitative grading of embolism severity, improving diagnostic efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of plant disease analysis technology, and in particular to a rapid diagnostic method and system for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis. Background Technology

[0002] Bacterial wilt of mulberry is a devastating soil-borne vascular disease caused by Ralstonia solanacearum. The pathogen invades the root system and spreads upwards along the xylem vessels, multiplying rapidly and secreting extracellular polysaccharides that clog the vessels, ultimately leading to the plant's loss of water transport function and death. This disease spreads rapidly and has a high mortality rate in mulberry-growing areas, becoming a significant factor hindering the healthy development of the mulberry industry. Currently, field diagnosis of bacterial wilt in mulberry mainly relies on symptom observation, namely, judging by observing external symptoms such as browning of the vascular bundles in the stem, bacterial ooze, or acute wilting of the leaves. However, by the time these external symptoms appear, the vascular system of the plant has often already suffered irreversible and widespread blockage, making treatment extremely ineffective, and the diseased plants become a source of secondary infection in the field. While laboratory pathogen isolation and culture identification methods can provide a definitive diagnosis, they require two to three days to obtain results, failing to meet the practical needs for rapid field surveys and immediate treatment.

[0003] In recent years, electrical impedance spectroscopy has shown potential in the non-destructive testing of plant physiological states. By measuring the impedance response of tissues under a broadband alternating electric field, it can sensitively reflect changes in physiological indicators such as cell membrane permeability, intracellular liquid ion concentration, and extracellular liquid conductivity. After bacterial wilt infection of mulberry trees, the blockage of vascular bundles and damage to cell membranes inevitably lead to regular changes in the electrical impedance characteristics of stem tissues, providing a physical basis for disease diagnosis based on electrical impedance spectroscopy analysis. However, existing plant disease diagnosis methods based on electrical impedance spectroscopy mostly rely on fitting equivalent circuit model parameters. The extracted feature parameters lack a direct physical interpretation of the degree of vascular bundle blockage, and it is difficult to effectively distinguish the similar dispersion characteristics exhibited in the impedance spectrum between latent infection and mild drought stress, resulting in insufficient accuracy in early diagnosis. Furthermore, existing methods generally only output binary diagnostic conclusions, failing to quantify and grade the degree of vascular bundle blockage, making it difficult to provide accurate basis for differentiated pesticide application decisions. Therefore, there is an urgent need to develop a rapid diagnostic method for bacterial wilt of mulberry trees that can achieve early warning, quantitative grading, and specific identification of the incubation period. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides a rapid diagnostic method and system for bacterial wilt of mulberry fruit based on electrical impedance spectroscopy analysis. Its main purpose is to improve the efficiency and accuracy of the diagnosis of bacterial wilt of mulberry fruit.

[0005] To achieve the above objectives, the first aspect of this invention provides a rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis, comprising: Non-invasive frequency sweep measurement was performed on the mulberry plant under test, and the response current was collected. The raw data of the impedance spectrum was obtained by fast Fourier transform calculation. After real-time denoising and parasitic parameter compensation, impedance vector feature extraction was performed to obtain multi-dimensional real-time diagnostic feature vector. The multidimensional real-time diagnostic feature vector is nonlinearly reduced in dimensionality to construct a real-time diagnostic classification model. The reduced low-dimensional feature representation is then input into the model. The signed decision function value relative to the classification hyperplane and the percentage of offset from the healthy control baseline are calculated. The disease status of the target mulberry plant is jointly determined to generate real-time disease diagnosis information. If the real-time disease diagnosis information determines that the disease status of the target mulberry plant is suspected latent infection, then the frequency logarithmic derivative of the collected electrical impedance spectrum raw data is taken to obtain enhanced relaxation features and relaxation time distribution spectrum inversion and reconstruction are performed to generate relaxation time distribution spectrum curve. Based on the expectation-maximization algorithm, Gaussian mixture model adaptive peak analysis is performed on the relaxation time distribution spectrum curve to separate the first relaxation peak characterizing the cell membrane polarization process and the second relaxation peak characterizing the vascular bundle obstruction effect. The first relaxation peak and the second relaxation peak are numerically integrated to construct the vascular bundle fluid flow continuity index. The disease status of the target mulberry plant is determined a second time based on the vascular bundle sap flow continuity index. If the determination result is still suspected latent infection, it is suggested to obtain the electrical impedance spectrum data of the target mulberry plant at noon to calculate the diurnal drop amplitude for final diagnosis, and generate a final diagnosis report of mulberry plant disease.

[0006] In this scheme, the non-invasive frequency sweep measurement of the mulberry plant to be tested and the acquisition of response current are used to obtain the raw data of impedance spectrum by fast Fourier transform. After real-time denoising and parasitic parameter compensation, impedance vector feature extraction is performed to obtain a multi-dimensional real-time diagnostic feature vector, which specifically includes: The four-electrode measuring fixture is attached along the stem axis to the internode of the main stem of the mulberry fruit to be tested. A wideband excitation voltage signal containing multiple prime multiples of the fundamental frequency is generated by a direct digital frequency synthesizer. Simultaneously, the response current analog signal flowing through the stem tissue is collected. After differential amplification, anti-aliasing filtering and analog-to-digital conversion, discrete time domain current sequence and discrete time domain voltage sequence are obtained. After applying window functions to the discrete time-domain current sequence and the discrete time-domain voltage sequence, a fast Fourier transform is performed to synchronously convert the two time-domain sequences to the frequency domain. At each discrete frequency point, the complex ratio of the voltage spectrum component to the current spectrum component is calculated to obtain the complex impedance frequency domain distribution data as the uncorrected impedance spectrum raw data. The original impedance spectrum data is subjected to moving median filtering along the frequency axis. The median values ​​of impedance magnitude and phase angle within a preset window are used to replace the original measured values ​​of the center point of the window. While preserving the overall dispersion trend of the spectrum, power frequency interference and random impulse noise are suppressed to obtain the denoised impedance spectrum data. Parasitic parameter compensation is performed on the denoised impedance spectrum data. The parasitic admittance spectrum under no-load condition of the test fixture is obtained in advance through open-circuit calibration, and the parasitic impedance spectrum under short-circuit condition of the fixture is obtained through short-circuit calibration. The series contribution of short-circuit impedance and the parallel contribution of open-circuit admittance are deducted from the measured impedance data in sequence to obtain the calibrated impedance spectrum. On the Nyquist plot of the impedance complex plane of the calibrated impedance spectrum, the mid-frequency capacitive reactance arc vertex is located, and the phase angle extreme value corresponding to the vertex is extracted as the first feature dimension. At the same time, a preset analysis interval is selected on the low-frequency impedance magnitude versus frequency curve, and the least squares method is used to fit the linear regression slope of the logarithm of the impedance magnitude versus the logarithm of the frequency in this interval as the second feature dimension. The characteristic frequency corresponding to the maximum absolute value of the imaginary part is searched in the impedance imaginary part spectrum. The offset relative to the reference characteristic frequency of a healthy plant is calculated as the third feature dimension. The mid-frequency capacitive arc is fitted on the Nyquist plot, and the radius of curvature of the fitted arc is extracted as the fourth feature dimension. The four feature dimensions together constitute a multidimensional real-time diagnostic feature vector.

[0007] In this scheme, the nonlinear dimensionality reduction of the multidimensional real-time diagnostic feature vector is performed to construct a real-time diagnostic classification model and input the dimensionality-reduced low-dimensional feature representation. The signed decision function value relative to the classification hyperplane and the percentage offset from the healthy control baseline are calculated, and the disease status of the target mulberry plant is jointly determined to generate real-time disease diagnostic information. Specifically, this includes: Kernel principal component analysis is introduced to perform nonlinear dimensionality reduction on the multidimensional real-time diagnostic feature vector. Gaussian radial basis function is used as kernel function to map the multidimensional real-time diagnostic feature vector to a high-dimensional regenerating kernel Hilbert space. In the high-dimensional space, the kernel matrix elements between any two sample feature vectors are calculated and the kernel matrix is ​​centered. The kernel matrix after centering is subjected to eigenvalue decomposition to solve for eigenvalues ​​and corresponding eigenvectors. The eigenvalues ​​are arranged in descending order, and the top k eigenvectors whose cumulative eigenvalue contribution rate exceeds a preset percentage are selected as the principal component projection directions. The selected eigenvectors form a dimensionality reduction mapping relationship from high-dimensional space to low-dimensional space. Based on the dimensionality reduction mapping relationship, the projection coordinate values ​​of the multidimensional real-time diagnostic feature vector are calculated on the projection direction of each principal component, and the projection coordinate values ​​corresponding to each principal component are used as the low-dimensional feature representation after dimensionality reduction. Using low-dimensional feature representations of mulberry plant samples with known healthy and known diseased states as training data, an instant diagnostic classification model is constructed based on support vector machine using Gaussian radial basis kernel function. The optimal Lagrange multiplier vector that maximizes the classification margin is solved by sequence minimum optimization algorithm to determine the set of support vectors constituting the classification hyperplane, the corresponding multiplier coefficients, and the bias term values. The dimensionality-reduced low-dimensional feature representation of the current mulberry plant to be tested is input into the instant diagnostic classification model. The weighted summation result of the kernel function values ​​between the current plant and each support vector is calculated. The summation result is added to the bias term to obtain the signed decision function value of the current sample to be tested relative to the classification hyperplane. The mean value of the decision function values ​​of all known healthy state samples in the training data is statistically analyzed and used as the baseline position of the healthy control. The offset of the decision function value of the current test sample relative to the baseline position is calculated and normalized to obtain the percentage offset characterizing the degree of deviation from the healthy state. A joint determination is performed based on the sign direction, absolute value, and percentage deviation of the signed decision function value. When the decision function value is positive and the absolute value is greater than the health confidence threshold and the percentage deviation is lower than the upper limit of normal fluctuation, the patient is judged to be healthy. When the decision function value is negative and the absolute value is greater than the disease confidence threshold and the percentage deviation exceeds the serious deviation warning line, the patient is judged to be symptomatic. Other intermediate states are judged as suspected latent infection, and real-time disease diagnosis information is generated.

[0008] In this scheme, if the real-time disease diagnosis information determines that the target mulberry plant is in a suspected latent infection stage, then the collected raw electrical impedance spectrum data is subjected to frequency logarithmic derivative to obtain enhanced relaxation features, and relaxation time distribution spectrum inversion and reconstruction are performed to generate a relaxation time distribution spectrum curve, specifically including: Extract the one-dimensional spectral sequence of the imaginary part of impedance as a function of frequency from the original data of the impedance spectrum. Calculate the difference approximate derivative of the imaginary part of impedance with respect to the logarithm of frequency at each discrete frequency point to generate an enhanced relaxation characteristic curve, thereby sharpening the weak relaxation response signal caused by the ion diffusion blockage effect in the vascular bundle vessels during latent infection. The electrical impedance response of mulberry stem tissue is characterized as a continuous parallel superposition of multiple Debye-type relaxation units with different characteristic time constants in the frequency domain. A first-kind Friedholm integral equation with the relaxation time distribution function as the unknown is established to describe the mapping relationship between the enhanced relaxation characteristic curve and the relaxation time distribution function.

[0009] The electrical impedance response process of mulberry stem tissue is uniformly characterized as a continuous parallel superposition of multiple Debye-type relaxation units with different characteristic time constants in the frequency domain. A first-kind Friedholm integral equation with the relaxation time distribution function as the unknown is established to describe the mapping relationship between the enhanced relaxation characteristic curve and the relaxation time distribution function. The first type of Friedholm integral equation is solved using the Tikhonov regularization method. An objective functional composed of the data fitting residual term and the solution function smoothness penalty term is constructed, and the optimal regularization parameter required for Tikhonov regularization is determined by the L-curve criterion. The target functional is iteratively minimized using the optimal regularization parameter through a numerical solver. In each iteration, the estimated value of the relaxation time distribution function is updated and the decrease of the target functional is calculated. The iteration process is terminated when the change of the solution function between two adjacent iterations is lower than the preset convergence tolerance. The relaxation time distribution spectrum curve is plotted with logarithmic relaxation time as the abscissa and relaxation intensity density as the ordinate. The spectral peaks located in the first time constant interval correspond to the relaxation intensity distribution of the cell membrane interface polarization process, and the spectral peaks located in the second time constant interval correspond to the relaxation intensity distribution of the ion diffusion blockage process in the vascular bundle vessels.

[0010] In this scheme, the Gaussian mixture model adaptive peak analysis is performed on the relaxation time distribution spectrum curve based on the expectation-maximization algorithm to separate the first relaxation peak characterizing the cell membrane polarization process and the second relaxation peak characterizing the vascular bundle obstruction effect. Numerical integration is then performed on the first and second relaxation peaks to construct the vascular bundle fluid flow continuity index, specifically including: The relaxation time distribution spectrum curve is regarded as a one-dimensional probability density distribution function formed by the superposition of two independent relaxation processes. The number of components of the Gaussian mixture model is set to two. Two initial mean positions are selected on the time constant axis as the initial centers of the Gaussian components. Each Gaussian component is assigned an equal initial mixture weight and an initial variance to form an initial model parameter set. The Gaussian mixture model parameters are iteratively optimized using the expectation-maximization algorithm. In the expectation step, the posterior probability of the intensity value of the spectral curve at each time constant sampling point belonging to the first Gaussian component and the second Gaussian component is calculated based on the current model parameters. In the maximization step, the posterior probability is used as a weight to re-estimate the mixture weight, mean position and variance parameters of each Gaussian component. After each iteration, the log-likelihood function value between the current Gaussian mixture model and the relaxation time distribution spectrum curve is calculated. The iteration process is terminated when the relative change of the log-likelihood function value between two adjacent iterations is lower than the preset convergence criterion threshold, thus obtaining the optimal Gaussian mixture model description of the relaxation time distribution spectrum curve. Based on the optimal Gaussian mixture model, the first relaxation peak and the second relaxation peak are obtained, and numerical integration is performed within their respective effective time constant intervals to obtain the first and second spectral peak integral areas. The first spectral peak integral area represents the integrity of the cell membrane system, and the second spectral peak integral area represents the degree of vascular bundle embolization. A stem diameter correction coefficient is introduced to perform geometric scaling correction on the integrated area of ​​the second spectral peak. The stem diameter correction coefficient is determined by the perimeter of the cross-section of the stem segment to be measured. The integrated area of ​​the first spectral peak is used as the numerator, and the sum of the integrated area of ​​the first spectral peak and the integrated area of ​​the second spectral peak weighted by the stem diameter correction coefficient is used as the denominator. The vascular bundle fluid flow continuity index is constructed by ratio calculation.

[0011] In this scheme, the disease status of the target mulberry plant is accurately determined a second time based on the vascular bundle sap flow continuity index. If the determination result is still suspected latent infection, it prompts the acquisition of electrical impedance spectrum data of the target mulberry plant at noon to calculate the diurnal drop amplitude for final diagnosis, generating a final diagnosis report of mulberry plant disease, specifically including: The vascular bundle fluid flow continuity index is compared with the preset latent period critical threshold. If it is lower than the latent period critical threshold, the target mulberry plant is directly determined to be in the latent period infection state, and a disease diagnosis report containing the latent period diagnosis conclusion and the quantitative value of the degree of vascular bundle embolization is generated. If it is higher than the reference lower limit threshold for healthy plants, it is determined to be in a healthy state and a health conclusion report is generated. If the vascular bundle sap flow continuity index is within a preset fuzzy range between the latent period critical threshold and the healthy plant reference lower limit threshold, a secondary measurement prompt signal is generated, prompting the operator to perform a second electrical impedance spectrum acquisition operation on the same measurement site of the same target mulberry plant during the midday transpiration peak period. The vascular bundle fluid flow continuity index at noon was obtained by using the raw data of the second electrical impedance spectrum collected at noon. The difference between the vascular bundle fluid flow continuity index obtained from the first measurement in the morning and the vascular bundle fluid flow continuity index obtained from the second measurement at noon was calculated to obtain the diurnal drop amplitude. The day-night drop amplitude is compared with a preset difference threshold. If the day-night drop amplitude is greater than or equal to the difference threshold, it indicates that the vascular bundle sap flow continuity exhibits the high stress-sensitive collapse characteristics unique to latent infection, and the target mulberry plant is determined to be in a latent infection state. If the diurnal drop is less than the dynamic difference threshold, it indicates that the low vascular bundle sap flow continuity index is more likely due to mild water deficit or environmental stress. The target mulberry plant is determined to be in a state of doubt and needs to be observed. A final diagnostic report containing suggestions for short-term retesting and monitoring is generated, the stratified diagnostic process is completed, and the final diagnostic report of the mulberry plant disease is output.

[0012] A second aspect of the present invention provides a rapid diagnostic system for bacterial wilt of mulberry fruit based on electrical impedance spectroscopy analysis. The system includes: a memory, a processor, and a communication interface. The memory contains a program for a rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectroscopy analysis. When the program for a rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectroscopy analysis is executed by the processor, it implements the steps of the rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectroscopy analysis as described in any of the above claims. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments or examples of the present invention, the drawings used in the embodiments or examples will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained according to these drawings without creative effort.

[0014] Figure 1 This is a flowchart of the first method for a rapid diagnosis of bacterial wilt of mulberry fruit based on electrical impedance spectrum analysis, provided in an embodiment of the present invention. Figure 2 This is a flowchart of the second method for a rapid diagnosis of bacterial wilt of mulberry fruit based on electrical impedance spectrum analysis, provided in an embodiment of the present invention. Figure 3 A block diagram of a rapid diagnostic system for bacterial wilt of mulberry fruit based on electrical impedance spectrum analysis is provided in an embodiment of the present invention. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0015] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0016] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0017] Figure 1 This is a flowchart of the first method for a rapid diagnosis of bacterial wilt of mulberry fruit based on electrical impedance spectrum analysis, provided in an embodiment of the present invention. like Figure 1 As shown, the present invention provides a first method flowchart for a rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectroscopy analysis, comprising: S102, non-invasive frequency sweep measurement is performed on the mulberry plant to be tested and the response current is collected. The raw data of the impedance spectrum is obtained by fast Fourier transform calculation. After real-time denoising and parasitic parameter compensation, impedance vector feature extraction is performed to obtain multi-dimensional real-time diagnostic feature vector. S104, perform nonlinear dimensionality reduction on the multidimensional real-time diagnostic feature vector, construct a real-time diagnostic classification model and input the dimensionality-reduced low-dimensional feature representation, calculate the signed decision function value relative to the classification hyperplane and the percentage of offset from the healthy control baseline, and jointly determine the disease status of the target mulberry plant to generate real-time disease diagnosis information. S106, if the real-time disease diagnosis information determines that the disease status of the target mulberry plant is suspected latent infection, then the frequency logarithmic derivative of the collected electrical impedance spectrum raw data is performed to obtain enhanced relaxation features and relaxation time distribution spectrum inversion reconstruction is performed to generate relaxation time distribution spectrum curve. S108, Based on the expectation-maximization algorithm, Gaussian mixture model adaptive peak analysis is performed on the relaxation time distribution spectrum curve to separate the first relaxation peak characterizing the cell membrane polarization process and the second relaxation peak characterizing the vascular bundle obstruction effect. The first relaxation peak and the second relaxation peak are numerically integrated to construct the vascular bundle fluid flow continuity index. S110, based on the vascular bundle sap flow continuity index, a second precise determination of the disease status of the target mulberry plant is made. If the determination result is still suspected latent infection, it is suggested to obtain the electrical impedance spectrum data of the target mulberry plant at noon to calculate the diurnal drop amplitude for final diagnosis, and generate a final diagnosis report of mulberry plant disease.

[0018] Furthermore, in a preferred embodiment of the present invention, the non-invasive frequency sweep measurement of the mulberry plant to be tested and the acquisition of response current are used to obtain the raw data of impedance spectrum by fast Fourier transform calculation. After real-time denoising and parasitic parameter compensation, impedance vector feature extraction is performed to obtain a multi-dimensional real-time diagnostic feature vector, specifically including: The four-electrode measuring fixture is attached along the stem axis to the internode of the main stem of the mulberry fruit to be tested. A wideband excitation voltage signal containing multiple prime multiples of the fundamental frequency is generated by a direct digital frequency synthesizer. Simultaneously, the response current analog signal flowing through the stem tissue is collected. After differential amplification, anti-aliasing filtering and analog-to-digital conversion, discrete time domain current sequence and discrete time domain voltage sequence are obtained. After applying window functions to the discrete time-domain current sequence and the discrete time-domain voltage sequence, a fast Fourier transform is performed to synchronously convert the two time-domain sequences to the frequency domain. At each discrete frequency point, the complex ratio of the voltage spectrum component to the current spectrum component is calculated to obtain the complex impedance frequency domain distribution data as the uncorrected impedance spectrum raw data. The original impedance spectrum data is subjected to moving median filtering along the frequency axis. The median values ​​of impedance magnitude and phase angle within a preset window are used to replace the original measured values ​​of the center point of the window. While preserving the overall dispersion trend of the spectrum, power frequency interference and random impulse noise are suppressed to obtain the denoised impedance spectrum data. Parasitic parameter compensation is performed on the denoised impedance spectrum data. The parasitic admittance spectrum under no-load condition of the test fixture is obtained in advance through open-circuit calibration, and the parasitic impedance spectrum under short-circuit condition of the fixture is obtained through short-circuit calibration. The series contribution of short-circuit impedance and the parallel contribution of open-circuit admittance are deducted from the measured impedance data in sequence to obtain the calibrated impedance spectrum. On the Nyquist plot of the impedance complex plane of the calibrated impedance spectrum, the mid-frequency capacitive reactance arc vertex is located, and the phase angle extreme value corresponding to the vertex is extracted as the first feature dimension. At the same time, a preset analysis interval is selected on the low-frequency impedance magnitude versus frequency curve, and the least squares method is used to fit the linear regression slope of the logarithm of the impedance magnitude versus the logarithm of the frequency in this interval as the second feature dimension. The characteristic frequency corresponding to the maximum absolute value of the imaginary part is searched in the impedance imaginary part spectrum. The offset relative to the reference characteristic frequency of a healthy plant is calculated as the third feature dimension. The mid-frequency capacitive arc is fitted on the Nyquist plot, and the radius of curvature of the fitted arc is extracted as the fourth feature dimension. The four feature dimensions together constitute a multidimensional real-time diagnostic feature vector.

[0019] It should be noted that, firstly, the four-electrode measuring fixture is attached along the stem axis to a specific internode position on the epidermis of the main stem of the mulberry tree to be tested. The four electrodes are arranged with two outer electrodes injecting excitation current and two inner electrodes detecting response voltage to eliminate the interference of contact impedance changes between the electrodes and the epidermis on measurement accuracy. A broadband excitation voltage signal containing multiple prime multiples of the fundamental frequency is generated by a direct digital frequency synthesizer. After this synthesized signal is applied to the two outer electrodes, the analog signal of the response current flowing through the stem tissue is simultaneously acquired. After being suppressed by a differential amplifier circuit to suppress common-mode noise and filtered out high-frequency spurious components by an anti-aliasing low-pass filter, it is discretized by a high-speed analog-to-digital converter at a sampling rate no less than five times that of the highest frequency component of the excitation signal, finally obtaining two sets of synchronously sampled data: a discrete-time domain current sequence and a discrete-time domain voltage sequence. Hanning window functions are applied to the above two sets of discrete-time domain sequences to suppress spectral leakage effects, and then a fast Fourier transform is performed to synchronously convert the time-domain data to the frequency domain. The complex ratio of the voltage spectrum component to the current spectrum component is calculated at each discrete frequency point, thereby obtaining the complex impedance frequency domain distribution data, which is the original data of the uncorrected impedance spectrum.

[0020] To eliminate power frequency interference and occasional random pulse noise during measurement, a moving median filter is applied along the frequency axis to the original impedance spectrum. The median values ​​of the impedance magnitude and phase angle within a preset window width are used to replace the original measured values ​​at the center point of the window. This processing method can effectively suppress narrowband spike noise while preserving the overall dispersion trend of the spectrum. After denoising, parasitic parameter compensation is further performed on the spectrum. The parasitic admittance spectrum of the test fixture under no-load conditions is obtained through open-circuit calibration measurement, and the parasitic impedance spectrum of the fixture under short-circuit conditions is obtained through short-circuit calibration measurement. The series contribution of the short-circuit impedance and the parallel contribution of the open-circuit admittance are successively subtracted from the impedance data obtained by actual measurement, thereby shifting the measurement reference plane from the instrument port to the actual contact interface between the electrode and the stem epidermis, and obtaining the calibrated impedance spectrum.

[0021] The vertex of the mid-frequency capacitive arc is located on the Nyquist plot of the impedance complex plane of the calibrated impedance spectrum. The frequency point where the phase angle reaches a local maximum is identified at this vertex, and this phase angle extreme value is extracted as the first feature dimension. Simultaneously, a preset analysis interval is selected on the double logarithmic curve of the impedance magnitude versus frequency in the low-frequency band, for example, the low-frequency range of 100 Hz to 500 Hz. The least squares method is used to fit the linear regression slope of the logarithm of the impedance magnitude versus the logarithm of the frequency within this interval, and the slope exponent is used as the second feature dimension. The characteristic frequency corresponding to the maximum absolute value of the imaginary part is searched in the impedance imaginary part spectrum, and the offset of this characteristic frequency relative to the pre-calibrated reference characteristic frequency of the healthy plant is calculated as the third feature dimension. Finally, a circular arc is fitted to the mid-frequency capacitive arc on the Nyquist plot, and the radius of curvature of the fitted arc is extracted as the fourth feature dimension. These four feature dimensions characterize the feature information of the impedance spectrum from different frequency bands and morphological perspectives, collectively forming a multi-dimensional real-time diagnostic feature vector.

[0022] Furthermore, in a preferred embodiment of the present invention, the step of performing nonlinear dimensionality reduction on the multidimensional real-time diagnostic feature vector, constructing a real-time diagnostic classification model and inputting the dimensionality-reduced low-dimensional feature representation, calculating the signed decision function value relative to the classification hyperplane and the percentage offset from the healthy control baseline, and jointly determining the disease status of the target mulberry plant to generate real-time disease diagnosis information specifically includes: Kernel principal component analysis is introduced to perform nonlinear dimensionality reduction on the multidimensional real-time diagnostic feature vector. Gaussian radial basis function is used as kernel function to map the multidimensional real-time diagnostic feature vector to a high-dimensional regenerating kernel Hilbert space. In the high-dimensional space, the kernel matrix elements between any two sample feature vectors are calculated and the kernel matrix is ​​centered. The kernel matrix after centering is subjected to eigenvalue decomposition to solve for eigenvalues ​​and corresponding eigenvectors. The eigenvalues ​​are arranged in descending order, and the top k eigenvectors whose cumulative eigenvalue contribution rate exceeds a preset percentage are selected as the principal component projection directions. The selected eigenvectors form a dimensionality reduction mapping relationship from high-dimensional space to low-dimensional space. Based on the dimensionality reduction mapping relationship, the projection coordinate values ​​of the multidimensional real-time diagnostic feature vector are calculated on the projection direction of each principal component, and the projection coordinate values ​​corresponding to each principal component are used as the low-dimensional feature representation after dimensionality reduction. Using low-dimensional feature representations of mulberry plant samples with known healthy and known diseased states as training data, an instant diagnostic classification model is constructed based on support vector machine using Gaussian radial basis kernel function. The optimal Lagrange multiplier vector that maximizes the classification margin is solved by sequence minimum optimization algorithm to determine the set of support vectors constituting the classification hyperplane, the corresponding multiplier coefficients, and the bias term values. The dimensionality-reduced low-dimensional feature representation of the current mulberry plant to be tested is input into the instant diagnostic classification model. The weighted summation result of the kernel function values ​​between the current plant and each support vector is calculated. The summation result is added to the bias term to obtain the signed decision function value of the current sample to be tested relative to the classification hyperplane. The mean value of the decision function values ​​of all known healthy state samples in the training data is statistically analyzed and used as the baseline position of the healthy control. The offset of the decision function value of the current test sample relative to the baseline position is calculated and normalized to obtain the percentage offset characterizing the degree of deviation from the healthy state. A joint determination is performed based on the sign direction, absolute value, and percentage deviation of the signed decision function value. When the decision function value is positive and the absolute value is greater than the health confidence threshold and the percentage deviation is lower than the upper limit of normal fluctuation, the patient is judged to be healthy. When the decision function value is negative and the absolute value is greater than the disease confidence threshold and the percentage deviation exceeds the serious deviation warning line, the patient is judged to be symptomatic. Other intermediate states are judged as suspected latent infection, and real-time disease diagnosis information is generated.

[0023] It should be noted that after obtaining the multidimensional real-time diagnostic feature vector composed of four feature dimensions, kernel principal component analysis is introduced to perform nonlinear dimensionality reduction on this feature vector. Since the four feature dimensions originate from different frequency bands and morphological representations of the electrical impedance spectrum, there is information redundancy between the dimensions due to nonlinear coupling of the plant electrophysiological system. Directly using these dimensions for classification would increase the complexity of the decision boundary. Therefore, a Gaussian radial basis function is used as the kernel function to implicitly map the original four-dimensional feature vector to a higher-dimensional regenerated kernel Hilbert space. In this higher-dimensional space, the kernel matrix elements between any two sample feature vectors are calculated. Specifically, the squared Euclidean distance between the feature vectors of two samples is calculated and mapped to the kernel function value in negative exponential form. Subsequently, the kernel matrix is ​​centered to eliminate the influence of mean bias in the feature space. The centered kernel matrix is ​​subjected to eigenvalue decomposition to solve for eigenvalues ​​and their corresponding eigenvectors. The eigenvalues ​​are arranged in descending order, and the proportion of each eigenvalue in the sum of all eigenvalues ​​is calculated as the contribution rate. The top k eigenvectors with a cumulative contribution rate exceeding a preset percentage, such as 95%, are selected as the principal component projection directions. These selected eigenvectors form a dimensionality reduction mapping relationship from the high-dimensional space to the low-dimensional space. Based on this dimensionality reduction mapping relationship, the original multidimensional real-time diagnostic feature vectors are projected one by one onto each principal component projection direction to obtain a dimensionality-reduced low-dimensional feature representation composed of k projected coordinate values.

[0024] Using the low-dimensional feature representations of known healthy mulberry plant samples and known bacterial wilt disease-inducing mulberry plant samples collected and labeled in advance as training data, an instant diagnostic classification model based on support vector machine was constructed using Gaussian radial basis kernel function. Healthy samples were labeled as positive class labels and diseased samples were labeled as negative class labels. The optimal Lagrange multiplier vector that maximizes the classification margin was solved by the sequence minimum optimization algorithm. During the solution process, only samples with Lagrange multipliers greater than zero were retained as support vectors. Thus, the set of support vectors constituting the classification hyperplane, the multiplier coefficients corresponding to each support vector, and the value of the bias term were determined.

[0025] The dimensionality-reduced low-dimensional feature representation of the current mulberry plant to be tested is input into the trained instant diagnostic classification model. The weighted sum of the multiplier coefficients of the kernel function values ​​between the low-dimensional feature representation and each support vector is calculated. The summation result is added to the bias term to obtain the signed decision function value of the current test sample relative to the classification hyperplane. At the same time, the decision function values ​​of all known healthy samples in the training data are statistically analyzed, and their mean is calculated as the baseline position for healthy controls. The deviation of the decision function value of the current test sample from this baseline position is calculated, and the deviation is normalized by dividing it by the standard deviation of the decision function values ​​of healthy samples to obtain the percentage deviation. Finally, a joint determination is performed based on the sign direction, absolute value, and offset percentage of the signed decision function value. A healthy state is determined when the decision function value is positive, the absolute value is greater than a preset health confidence threshold, and the offset percentage is lower than the upper limit of normal fluctuation. A symptomatic disease state is determined when the decision function value is negative, the absolute value is greater than a preset disease confidence threshold, and the offset percentage exceeds the severe deviation warning line. Other intermediate states are determined as suspected latent infection states. Based on this, real-time disease diagnosis information containing the above determination results is generated. This real-time disease diagnosis information also serves as the basis for determining whether to trigger the subsequent precise verification mode.

[0026] Furthermore, in a preferred embodiment of the present invention, if the real-time disease diagnosis information determines that the disease status of the target mulberry plant is suspected latent infection, then the frequency logarithmic derivative of the collected electrical impedance spectrum raw data is performed to obtain enhanced relaxation features, and the relaxation time distribution spectrum is inverted and reconstructed to generate a relaxation time distribution spectrum curve, specifically including: Extract the one-dimensional spectral sequence of the imaginary part of impedance as a function of frequency from the original data of the impedance spectrum. Calculate the difference approximate derivative of the imaginary part of impedance with respect to the logarithm of frequency at each discrete frequency point to generate an enhanced relaxation characteristic curve, thereby sharpening the weak relaxation response signal caused by the ion diffusion blockage effect in the vascular bundle vessels during latent infection. The electrical impedance response of mulberry stem tissue is characterized as a continuous parallel superposition of multiple Debye-type relaxation units with different characteristic time constants in the frequency domain. A first-kind Friedholm integral equation with the relaxation time distribution function as the unknown is established to describe the mapping relationship between the enhanced relaxation characteristic curve and the relaxation time distribution function.

[0027] The electrical impedance response process of mulberry stem tissue is uniformly characterized as a continuous parallel superposition of multiple Debye-type relaxation units with different characteristic time constants in the frequency domain. A first-kind Friedholm integral equation with the relaxation time distribution function as the unknown is established to describe the mapping relationship between the enhanced relaxation characteristic curve and the relaxation time distribution function. The first type of Friedholm integral equation is solved using the Tikhonov regularization method. An objective functional composed of the data fitting residual term and the solution function smoothness penalty term is constructed, and the optimal regularization parameter required for Tikhonov regularization is determined by the L-curve criterion. The target functional is iteratively minimized using the optimal regularization parameter through a numerical solver. In each iteration, the estimated value of the relaxation time distribution function is updated and the decrease of the target functional is calculated. The iteration process is terminated when the change of the solution function between two adjacent iterations is lower than the preset convergence tolerance. The relaxation time distribution spectrum curve is plotted with logarithmic relaxation time as the abscissa and relaxation intensity density as the ordinate. The spectral peaks located in the first time constant interval correspond to the relaxation intensity distribution of the cell membrane interface polarization process, and the spectral peaks located in the second time constant interval correspond to the relaxation intensity distribution of the ion diffusion blockage process in the vascular bundle vessels.

[0028] It should be noted that when the real-time disease diagnosis information determines that the target mulberry plant is in a suspected latent infection stage, a deep analysis process is immediately performed on the collected raw electrical impedance spectrum data. First, a one-dimensional spectral sequence of the imaginary part of the impedance as a function of frequency is extracted from the raw electrical impedance spectrum data. This imaginary part spectrum directly corresponds to the energy dissipation characteristics of each relaxation process in the stem tissue in the frequency domain.

[0029] To enhance the identification of implicit relaxation peaks in the imaginary part spectrum and suppress the interference of broadband measurement noise on subsequent inversion operations, the approximate derivative of the imaginary part of the impedance with respect to the logarithm of the frequency is calculated at each discrete frequency point. Specifically, the difference between the imaginary part value at the current frequency point and the imaginary part value at the previous frequency point is divided by the difference between the logarithms of the two frequency points, thereby generating an enhanced relaxation characteristic curve showing the change of the imaginary part derivative with frequency. In this curve, the relaxation information that originally appeared as a gentle slope or a slight bulge in the original imaginary part spectrum is significantly sharpened into a clearly identifiable peak response. In particular, the weak relaxation signal caused by the ion diffusion blockage effect in the vascular bundle vessels during latent infection is effectively amplified.

[0030] Subsequently, the electrical impedance response process of the mulberry stem tissue was uniformly characterized as a continuous parallel superposition of multiple Debye-type relaxation units with different characteristic time constants in the frequency domain. Each Debye-type relaxation unit is described by a series combination of a resistor and a capacitor, and its characteristic time constant is equal to the product of the resistance and capacitance values. Based on the above characterization, a first-kind Friedholm integral equation was established with the relaxation time distribution function as the unknown. This equation describes the mapping relationship between the enhanced relaxation characteristic curve and the relaxation time distribution function in the frequency and time domains. Since solving the first-kind Friedholm integral equation is a typical ill-posed problem, even small noise in the measurement data can cause severe oscillations in the solution function. Therefore, the Tikhonov regularization method was used to transform the original ill-posed problem into a well-posed optimization problem.

[0031] A target functional is constructed by combining a data fitting residual term and a smoothness penalty term based on the second derivative norm of the solution function. The data fitting residual term measures the deviation between the calculated frequency domain response corresponding to the preset relaxation time distribution function and the measured value of the enhanced relaxation characteristic curve. The smoothness penalty term constrains the second derivative norm of the solution function to suppress spurious relaxation peaks caused by measurement noise amplification. The magnitude of the regularization parameter determines the trade-off between fitting accuracy and solution stability. The optimal regularization parameter is adaptively determined using the L-curve criterion. A curve showing the relationship between the norm of the data fitting residual term and the norm of the smoothness penalty term is plotted in a logarithmic coordinate system. This curve exhibits a distinct L-shaped bend. The regularization parameter corresponding to the point of maximum curvature on the curve is taken as the optimal value. Using this optimal regularization parameter, an iterative minimization operation is performed on the target functional through a numerical solver. In each iteration, the estimated value of the relaxation time distribution function is updated, and the decrease in the target functional is calculated. The iteration process terminates when the change in the solution function obtained from two consecutive iterations is lower than the preset convergence tolerance. After iteration terminates, a relaxation time distribution spectrum curve is output, with logarithmic relaxation time as the abscissa and relaxation intensity density as the ordinate. In this distribution spectrum curve, the peaks located in the shorter time constant range correspond to the relaxation intensity distribution of the cell membrane interface polarization process, while the peaks located in the longer time constant range correspond to the relaxation intensity distribution of the ion diffusion blockage process caused by Ralstonia solanacearum secretion in the vascular bundle vessels. The relative intensity change of the two peaks directly reflects the physical development degree of vascular bundle embolism caused by latent infection.

[0032] Furthermore, in a preferred embodiment of the present invention, the method of performing Gaussian mixture model adaptive peak analysis on the relaxation time distribution spectrum curve based on the expectation-maximization algorithm to separate and obtain the first relaxation peak characterizing the cell membrane polarization process and the second relaxation peak characterizing the vascular bundle obstruction effect, and numerically integrating the first relaxation peak and the second relaxation peak to construct the vascular bundle fluid flow continuity index, specifically includes: The relaxation time distribution spectrum curve is regarded as a one-dimensional probability density distribution function formed by the superposition of two independent relaxation processes. The number of components of the Gaussian mixture model is set to two. Two initial mean positions are selected on the time constant axis as the initial centers of the Gaussian components. Each Gaussian component is assigned an equal initial mixture weight and an initial variance to form an initial model parameter set. The Gaussian mixture model parameters are iteratively optimized using the expectation-maximization algorithm. In the expectation step, the posterior probability of the intensity value of the spectral curve at each time constant sampling point belonging to the first Gaussian component and the second Gaussian component is calculated based on the current model parameters. In the maximization step, the posterior probability is used as a weight to re-estimate the mixture weight, mean position and variance parameters of each Gaussian component. After each iteration, the log-likelihood function value between the current Gaussian mixture model and the relaxation time distribution spectrum curve is calculated. The iteration process is terminated when the relative change of the log-likelihood function value between two adjacent iterations is lower than the preset convergence criterion threshold, thus obtaining the optimal Gaussian mixture model description of the relaxation time distribution spectrum curve. Based on the optimal Gaussian mixture model, the first relaxation peak and the second relaxation peak are obtained, and numerical integration is performed within their respective effective time constant intervals to obtain the first and second spectral peak integral areas. The first spectral peak integral area represents the integrity of the cell membrane system, and the second spectral peak integral area represents the degree of vascular bundle embolization. A stem diameter correction coefficient is introduced to perform geometric scaling correction on the integrated area of ​​the second spectral peak. The stem diameter correction coefficient is determined by the perimeter of the cross-section of the stem segment to be measured. The integrated area of ​​the first spectral peak is used as the numerator, and the sum of the integrated area of ​​the first spectral peak and the integrated area of ​​the second spectral peak weighted by the stem diameter correction coefficient is used as the denominator. The vascular bundle fluid flow continuity index is constructed by ratio calculation.

[0033] It should be noted that after obtaining the relaxation time distribution spectrum curve, it is necessary to separate the two independent relaxation peaks corresponding to the cell membrane polarization process and the vascular bundle ion diffusion blockage process, respectively, and quantify their respective contributions. First, the relaxation time distribution spectrum curve is considered as a one-dimensional probability density distribution function composed of the superposition of two independent relaxation processes. The Gaussian mixture model is set to have two components, corresponding to the two types of relaxation processes with different physical origins. Two initial mean positions are selected on the time constant axis as the initial centers of the Gaussian components. For example, the first initial mean is set in the region with a shorter time constant, and the second initial mean is set in the region with a longer time constant. Each Gaussian component is assigned an equal initial mixing weight of half and a suitable initial variance, thus forming the initial model parameter set.

[0034] Subsequently, the Gaussian mixture model parameters are iteratively optimized using the expectation-maximization algorithm. In the expectation step, based on the current model parameters, the posterior probability of the intensity value of the spectral curve at each time constant sampling point belonging to the first Gaussian component and the second Gaussian component, respectively, is calculated. This posterior probability characterizes the likelihood that the sampling point belongs to a specific relaxation process component under the current model parameters. In the maximization step, the above posterior probabilities are used as weights to re-estimate the mixing weight, mean position, and variance parameters of each Gaussian component, so that the log-likelihood function of the observed data monotonically increases under the current posterior probability distribution. After each iteration, the log-likelihood function value between the current Gaussian mixture model and the relaxation time distribution spectral curve is calculated and compared with the log-likelihood function value of the previous iteration. The iteration process terminates when the relative change in the log-likelihood function value between two adjacent iterations is lower than a preset convergence criterion threshold. The Gaussian mixture model parameters obtained at this time are the optimal statistical description of the relaxation time distribution spectral curve.

[0035] Based on this optimal Gaussian mixture model, the Gaussian component with a smaller mean corresponds to the first relaxation peak, which characterizes the polarization-charge-discharge relaxation process at the interface of active cell membranes in the stem tissue. The Gaussian component with a larger mean corresponds to the second relaxation peak, which characterizes the ion diffusion-blockage relaxation process within the vascular bundle vessels. The first relaxation peak is integrated over an area twice the standard deviation centered on its mean position. Within this interval, the relaxation intensity density curve described by the first Gaussian component is numerically integrated to obtain the integrated area of ​​the first peak. The second relaxation peak is numerically integrated in the same way to obtain its integrated area. The integrated area of ​​the first peak is proportional to the integrity of the cell membrane system, and the integrated area of ​​the second peak is proportional to the degree of embolization of the vascular bundle vessels. Because the stem diameters of different mulberry plants vary, the integrated area of ​​the second spectral peak is affected by the scaling effect of the stem segment's geometric dimensions. Therefore, a stem diameter correction coefficient is introduced to correct it. This coefficient is determined by the perimeter of the cross-section of the stem segment according to a pre-defined empirical relationship. The corrected second spectral peak area is obtained by multiplying the integrated area of ​​the second spectral peak by the correction coefficient. Finally, the integrated area of ​​the first spectral peak is used as the numerator, and the sum of the integrated area of ​​the first spectral peak and the corrected second spectral peak area is used as the denominator. A ratio calculation is then performed to construct the vascular bundle flow continuity index (VCI). This index is normalized to a range of zero to one. A value closer to one indicates better flow continuity and less embolism within the vascular bundle vessels, while a value closer to zero indicates more severe embolism. Figure 2 This is a flowchart of the second method for a rapid diagnosis of bacterial wilt of mulberry fruit based on electrical impedance spectrum analysis, provided in an embodiment of the present invention. like Figure 2 As shown, the present invention provides a second method flowchart for a rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectroscopy analysis, including: S202, compare the vascular bundle fluid flow continuity index with the preset latent period critical threshold. If it is lower than the latent period critical threshold, the target mulberry plant is directly determined to be in the latent period infection state, and a disease diagnosis report containing the latent period diagnosis conclusion and the quantitative value of the degree of vascular bundle embolization is generated. If it is higher than the reference lower limit threshold for healthy plants, it is determined to be in a healthy state and a health conclusion report is generated. S204, if the vascular bundle sap flow continuity index is within a preset fuzzy range between the latent period critical threshold and the healthy plant reference lower limit threshold, a secondary measurement prompt signal is generated to prompt the operator to perform a second electrical impedance spectrum acquisition operation on the same measurement site of the same target mulberry plant during the noon transpiration peak period of the day. S206. The vascular bundle fluid flow continuity index at noon is obtained by using the raw data of the second electrical impedance spectrum collected at noon. The difference between the vascular bundle fluid flow continuity index obtained from the first measurement in the morning and the vascular bundle fluid flow continuity index obtained from the second measurement at noon is calculated to obtain the diurnal drop amplitude. S208, compare the day and night drop amplitude with the preset difference threshold. If the day and night drop amplitude is greater than or equal to the difference threshold, it indicates that the vascular bundle sap flow continuity exhibits the high stress-sensitive collapse characteristics unique to latent infection, and the target mulberry plant is determined to be in a latent infection state. S210, if the diurnal drop is less than the dynamic difference threshold, it indicates that the low vascular bundle sap flow continuity index is more likely to be due to mild water deficit or environmental stress. The target mulberry plant is determined to be in a state of doubt and needs to be observed. A final diagnostic report containing suggestions for short-term retesting and monitoring is generated, the stratified diagnostic process is completed, and the final diagnostic report of the mulberry plant disease is output.

[0036] It should be noted that after obtaining the vascular bundle sap flow continuity index (VCI), the index is first compared with a preset latent period critical threshold, which can be set to 0.65, and the lower reference threshold for healthy plants can be set to 0.85. If the vascular bundle sap flow continuity index is below 0.65, it indicates that a significant ion diffusion blockage effect has formed in the vascular bundle vessels of the stem. Based on this, the target mulberry plant is directly determined to be in a latent infection state, and a disease diagnosis report containing the latent period diagnosis conclusion and the specific value of the vascular bundle sap flow continuity index is generated. If the vascular bundle sap flow continuity index is above 0.85, it indicates that the continuity of sap flow in the vessels is no different from that of a healthy plant, and it is judged to be in a healthy state. If the vascular bundle sap flow continuity index is between 0.65 and 0.85, it indicates that a single measurement in the early morning cannot distinguish between latent infection and mild drought stress, suggesting that a second electrical impedance spectrum acquisition operation be performed during the peak transpiration period at noon on the same day. At noon, transpiration stress reaches its maximum value for the day. In plants infected during the latent period, the vascular bundles that were initially weakly attached to Ralstonia solanacearum will undergo cavitation and blockage under strong negative pressure, leading to a significant drop in the vascular bundle sap flow continuity index. In contrast, drought-stressed plants have intact vascular structures, and their diurnal drop is smaller. The difference between the morning and noon indices is calculated to determine the diurnal drop. If this difference is greater than or equal to 0.15, it is considered a latent infection; if it is less than 0.15, it is considered questionable and requires further observation, with a recommendation for retesting within 2 days. This diurnal dual-phase verification mechanism fully utilizes the unique pathological characteristics of latent Ralstonia solanacearum infection in terms of vascular bundle mechanical vulnerability. It can effectively eliminate the interference of simple environmental stress on impedance spectra, raising the diagnostic specificity of the latent period to a higher level, avoiding unnecessary pesticide application due to drought misdiagnosis, and simultaneously providing a valuable time window for early disease intervention.

[0037] Furthermore, the rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis provided by this invention also includes the following steps: Read the vascular bundle fluid flow continuity index value and disease stage determination results recorded in the final diagnosis report. If the determination result is a healthy state, output a recommendation that no medication is needed. If the determination result is an obvious disease stage, output a decision to remove diseased plants and disinfect the soil. If the determination result is an incubation period infection state, proceed to the medication plan matching stage. When the determination result is latent infection status, the vascular bundle fluid flow continuity index is compared with the preset first, second and third numerical intervals. The three numerical intervals correspond to the continuous grading of the degree of vascular bundle embolization from mild to severe. Each numerical interval is preset with a corresponding differentiated drug administration strategy type. If the vascular bundle sap flow continuity index falls within the first value range, a mild intervention application plan is generated, recommending the use of biological control agents for root soil irrigation. The output includes application operation parameters such as agent type, dilution ratio, ring trench application method, and application frequency. If the vascular bundle fluid flow continuity index falls within the second value range, a moderate intervention drug administration plan is generated. It is recommended to use the stem injection method for precise drug administration. The injection depth, injection angle and injection point are calculated based on the diameter of the stem segment to be tested, and the output includes the drug type and injection volume parameters, including kasugamycin or kasugamycin aqueous solution. If the vascular bundle sap flow continuity index falls within the third value range, an enhanced intervention application plan is generated. It is recommended to adopt a combined application strategy of stem injection and foliar spraying. The stem injection operation outputs the increased drug concentration and seals the injection hole with wax for protection. The foliar spraying outputs the type of amino oligosaccharide or salicylic acid inducer and provides application guidance for spraying in the evening. The application parameters are dynamically fine-tuned based on the phenological stage of the mulberry plant under test and the ambient temperature conditions. If the plant is in the fruit enlargement or harvesting stage, the chemical agent is replaced with a microbial agent or plant-derived pesticide. If the ambient temperature is higher than the preset high temperature threshold, the injection time is adjusted and the agent concentration is reduced. Finally, a differentiated application decision report is generated.

[0038] It should be noted that after obtaining the final diagnosis report of the target mulberry plant and the corresponding vascular bundle sap flow continuity index, a differentiated treatment decision-making process is further initiated. Based on the quantified value of the degree of embolism, a precisely matched treatment plan is generated, thus transforming the traditional extensive, uniform treatment into precise treatment based on the level of vascular bundle embolism. First, the vascular bundle sap flow continuity index value and disease stage determination result recorded in the final diagnosis report are read. If the determination result is a healthy state, a recommendation not to treat is output. If the determination result is a symptomatic disease stage, a decision to remove the diseased plant and disinfect the soil is output. If the determination result is a latent infection stage, the treatment plan matching stage begins.

[0039] When the infection is determined to be in the latent stage, the vascular bundle sap flow continuity index is compared with three preset numerical ranges. The first range can be set to 0.60 to 0.65, corresponding to a mild stage of vascular bundle embolization, generating a mild intervention plan. It is recommended to use biological control agents such as Bacillus subtilis for root soil irrigation treatment, utilizing the root system to actively absorb antagonistic bacteria into the vascular bundle system to inhibit further colonization of Ralstonia solanacearum. The second range can be set to 0.45 to 0.60, corresponding to a moderate stage of embolization, generating a moderate intervention plan. It is recommended to use stem injection to directly inject kasugamycin or streptomycin aqueous solution into the xylem vessels. The injection depth is automatically calculated based on the stem diameter and controlled to one-third of the stem diameter, and the injection angle is controlled to be 30 degrees with the stem axis to ensure that the drug solution is evenly diffused throughout the entire plant with the transpiration flow. The third numerical range can be set below 0.45, corresponding to a more severe stage of embolism. This generates a strengthened intervention plan, recommending a combined application of stem injection and foliar spraying. For stem injection, the pesticide concentration should be appropriately increased, and the injection holes should be sealed with wax to prevent air from entering the vascular bundles and inducing secondary cavitation embolism. For foliar spraying, an amino oligosaccharide inducer should be applied to induce systemic acquired resistance in the plant. This differentiated application decision mechanism directly maps the continuous quantitative index of vascular bundle sap flow to a tiered application strategy, achieving a quantitative correlation between diagnostic results and treatment measures. This avoids the problems of over- or under-application caused by traditional binary diagnosis, improving the success rate of latent period intervention while reducing pesticide dosage.

[0040] Figure 3 A rapid diagnostic system 3 for mulberry bacterial wilt based on electrical impedance tomography analysis is provided in one embodiment of the present invention. The system includes: a memory 301, a processor 302, and a communication interface 303. The memory 301 contains a program for a rapid diagnostic method for mulberry bacterial wilt based on electrical impedance tomography analysis. When the program for a rapid diagnostic method for mulberry bacterial wilt based on electrical impedance tomography analysis is executed by the processor 302, it implements the steps of the rapid diagnostic method for mulberry bacterial wilt based on electrical impedance tomography analysis as described above.

[0041] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis, characterized in that, include: Non-invasive frequency sweep measurement was performed on the mulberry plant under test, and the response current was collected. The raw data of the impedance spectrum was obtained by fast Fourier transform calculation. After real-time denoising and parasitic parameter compensation, impedance vector feature extraction was performed to obtain multi-dimensional real-time diagnostic feature vector. The multidimensional real-time diagnostic feature vector is nonlinearly reduced in dimensionality to construct a real-time diagnostic classification model. The reduced low-dimensional feature representation is then input into the model. The signed decision function value relative to the classification hyperplane and the percentage of offset from the healthy control baseline are calculated. The disease status of the target mulberry plant is jointly determined to generate real-time disease diagnosis information. If the real-time disease diagnosis information determines that the disease status of the target mulberry plant is suspected latent infection, then the frequency logarithmic derivative of the collected electrical impedance spectrum raw data is taken to obtain enhanced relaxation features and relaxation time distribution spectrum inversion and reconstruction are performed to generate relaxation time distribution spectrum curve. Based on the expectation-maximization algorithm, Gaussian mixture model adaptive peak analysis is performed on the relaxation time distribution spectrum curve to separate the first relaxation peak characterizing the cell membrane polarization process and the second relaxation peak characterizing the vascular bundle obstruction effect. The first relaxation peak and the second relaxation peak are numerically integrated to construct the vascular bundle fluid flow continuity index. The disease status of the target mulberry plant is determined a second time based on the vascular bundle sap flow continuity index. If the determination result is still suspected latent infection, it is suggested to obtain the electrical impedance spectrum data of the target mulberry plant at noon to calculate the diurnal drop amplitude for final diagnosis, and generate a final diagnosis report of mulberry plant disease.

2. The rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis according to claim 1, characterized in that, The process involves non-invasive frequency sweep measurement of the mulberry plant under test, acquisition of response current, and fast Fourier transform to obtain raw impedance spectrum data. After real-time denoising and parasitic parameter compensation, impedance vector feature extraction is performed to obtain a multi-dimensional real-time diagnostic feature vector, specifically including: The four-electrode measuring fixture is attached along the stem axis to the internode of the main stem of the mulberry fruit to be tested. A wideband excitation voltage signal containing multiple prime multiples of the fundamental frequency is generated by a direct digital frequency synthesizer. Simultaneously, the response current analog signal flowing through the stem tissue is collected. After differential amplification, anti-aliasing filtering and analog-to-digital conversion, discrete time domain current sequence and discrete time domain voltage sequence are obtained. After applying window functions to the discrete time-domain current sequence and the discrete time-domain voltage sequence, a fast Fourier transform is performed to synchronously convert the two time-domain sequences to the frequency domain. At each discrete frequency point, the complex ratio of the voltage spectrum component to the current spectrum component is calculated to obtain the complex impedance frequency domain distribution data as the uncorrected impedance spectrum raw data. The original impedance spectrum data is subjected to moving median filtering along the frequency axis. The median values ​​of impedance magnitude and phase angle within a preset window are used to replace the original measured values ​​of the center point of the window. While preserving the overall dispersion trend of the spectrum, power frequency interference and random impulse noise are suppressed to obtain the denoised impedance spectrum data. Parasitic parameter compensation is performed on the denoised impedance spectrum data. The parasitic admittance spectrum under no-load condition of the test fixture is obtained in advance through open-circuit calibration, and the parasitic impedance spectrum under short-circuit condition of the fixture is obtained through short-circuit calibration. The series contribution of short-circuit impedance and the parallel contribution of open-circuit admittance are deducted from the measured impedance data in sequence to obtain the calibrated impedance spectrum. On the Nyquist plot of the impedance complex plane of the calibrated impedance spectrum, the mid-frequency capacitive reactance arc vertex is located, and the phase angle extreme value corresponding to the vertex is extracted as the first feature dimension. At the same time, a preset analysis interval is selected on the low-frequency impedance magnitude versus frequency curve, and the least squares method is used to fit the linear regression slope of the logarithm of the impedance magnitude versus the logarithm of the frequency in this interval as the second feature dimension. The characteristic frequency corresponding to the maximum absolute value of the imaginary part is searched in the impedance imaginary part spectrum. The offset relative to the reference characteristic frequency of a healthy plant is calculated as the third feature dimension. The mid-frequency capacitive arc is fitted on the Nyquist plot, and the radius of curvature of the fitted arc is extracted as the fourth feature dimension. The four feature dimensions together constitute a multidimensional real-time diagnostic feature vector.

3. The rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis according to claim 1, characterized in that, The process involves nonlinearly reducing the dimensionality of the multidimensional real-time diagnostic feature vector, constructing a real-time diagnostic classification model, inputting the reduced low-dimensional feature representation, calculating the signed decision function value relative to the classification hyperplane and the percentage offset from the healthy control baseline, and jointly determining the disease status of the target mulberry plant to generate real-time disease diagnostic information. Specifically, this includes: Kernel principal component analysis is introduced to perform nonlinear dimensionality reduction on the multidimensional real-time diagnostic feature vector. Gaussian radial basis function is used as kernel function to map the multidimensional real-time diagnostic feature vector to a high-dimensional regenerating kernel Hilbert space. In the high-dimensional space, the kernel matrix elements between any two sample feature vectors are calculated and the kernel matrix is ​​centered. The kernel matrix after centering is subjected to eigenvalue decomposition to solve for eigenvalues ​​and corresponding eigenvectors. The eigenvalues ​​are arranged in descending order, and the top k eigenvectors whose cumulative eigenvalue contribution rate exceeds a preset percentage are selected as the principal component projection directions. The selected eigenvectors form a dimensionality reduction mapping relationship from high-dimensional space to low-dimensional space. Based on the dimensionality reduction mapping relationship, the projection coordinate values ​​of the multidimensional real-time diagnostic feature vector are calculated on the projection direction of each principal component, and the projection coordinate values ​​corresponding to each principal component are used as the low-dimensional feature representation after dimensionality reduction. Using low-dimensional feature representations of mulberry plant samples with known healthy and known diseased states as training data, an instant diagnostic classification model is constructed based on support vector machine using Gaussian radial basis kernel function. The optimal Lagrange multiplier vector that maximizes the classification margin is solved by sequence minimum optimization algorithm to determine the set of support vectors constituting the classification hyperplane, the corresponding multiplier coefficients, and the bias term values. The dimensionality-reduced low-dimensional feature representation of the current mulberry plant to be tested is input into the instant diagnostic classification model. The weighted summation result of the kernel function values ​​between the current plant and each support vector is calculated. The summation result is added to the bias term to obtain the signed decision function value of the current sample to be tested relative to the classification hyperplane. The mean value of the decision function values ​​of all known healthy state samples in the training data is statistically analyzed and used as the baseline position of the healthy control. The offset of the decision function value of the current test sample relative to the baseline position is calculated and normalized to obtain the percentage offset characterizing the degree of deviation from the healthy state. A joint determination is performed based on the sign direction, absolute value, and percentage deviation of the signed decision function value. When the decision function value is positive and the absolute value is greater than the health confidence threshold and the percentage deviation is lower than the upper limit of normal fluctuation, the patient is judged to be healthy. When the decision function value is negative and the absolute value is greater than the disease confidence threshold and the percentage deviation exceeds the serious deviation warning line, the patient is judged to be symptomatic. Other intermediate states are judged as suspected latent infection, and real-time disease diagnosis information is generated.

4. The rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis according to claim 1, characterized in that, If the real-time disease diagnosis information determines that the target mulberry plant is in a suspected latent infection stage, then the collected raw electrical impedance spectrum data is subjected to frequency logarithmic derivative to obtain enhanced relaxation features, and relaxation time distribution spectrum inversion and reconstruction are performed to generate a relaxation time distribution spectrum curve, specifically including: Extract the one-dimensional spectral sequence of the imaginary part of impedance as a function of frequency from the original data of the impedance spectrum. Calculate the difference approximate derivative of the imaginary part of impedance with respect to the logarithm of frequency at each discrete frequency point to generate an enhanced relaxation characteristic curve, thereby sharpening the weak relaxation response signal caused by the ion diffusion blockage effect in the vascular bundle vessels during latent infection. The electrical impedance response of the mulberry stem tissue is characterized as a continuous parallel superposition of multiple Debye-type relaxation units with different characteristic time constants in the frequency domain. A first-kind Friedholm integral equation with the relaxation time distribution function as the unknown is established to describe the mapping relationship between the enhanced relaxation characteristic curve and the relaxation time distribution function. The electrical impedance response process of mulberry stem tissue is uniformly characterized as a continuous parallel superposition of multiple Debye-type relaxation units with different characteristic time constants in the frequency domain. A first-kind Friedholm integral equation with the relaxation time distribution function as the unknown is established to describe the mapping relationship between the enhanced relaxation characteristic curve and the relaxation time distribution function. The first type of Friedholm integral equation is solved using the Tikhonov regularization method. An objective functional composed of the data fitting residual term and the solution function smoothness penalty term is constructed, and the optimal regularization parameter required for Tikhonov regularization is determined by the L-curve criterion. The target functional is iteratively minimized using the optimal regularization parameter through a numerical solver. In each iteration, the estimated value of the relaxation time distribution function is updated and the decrease of the target functional is calculated. The iteration process is terminated when the change of the solution function between two adjacent iterations is lower than the preset convergence tolerance. The relaxation time distribution spectrum curve is plotted with logarithmic relaxation time as the abscissa and relaxation intensity density as the ordinate. The spectral peaks located in the first time constant interval correspond to the relaxation intensity distribution of the cell membrane interface polarization process, and the spectral peaks located in the second time constant interval correspond to the relaxation intensity distribution of the ion diffusion blockage process in the vascular bundle vessels.

5. The rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis according to claim 1, characterized in that, The method involves applying a Gaussian mixture model adaptive peak analysis to the relaxation time distribution spectrum curve using the expectation-maximization algorithm, separating the first relaxation peak characterizing cell membrane polarization and the second relaxation peak characterizing the vascular bundle obstruction effect. Numerical integration is then performed on the first and second relaxation peaks to construct a vascular bundle fluid flow continuity index. Specifically, this includes: The relaxation time distribution spectrum curve is regarded as a one-dimensional probability density distribution function formed by the superposition of two independent relaxation processes. The number of components of the Gaussian mixture model is set to two. Two initial mean positions are selected on the time constant axis as the initial centers of the Gaussian components. Each Gaussian component is assigned an equal initial mixture weight and an initial variance to form an initial model parameter set. The Gaussian mixture model parameters are iteratively optimized using the expectation-maximization algorithm. In the expectation step, the posterior probability of the intensity value of the spectral curve at each time constant sampling point belonging to the first Gaussian component and the second Gaussian component is calculated based on the current model parameters. In the maximization step, the posterior probability is used as a weight to re-estimate the mixture weight, mean position and variance parameters of each Gaussian component. After each iteration, the log-likelihood function value between the current Gaussian mixture model and the relaxation time distribution spectrum curve is calculated. The iteration process is terminated when the relative change of the log-likelihood function value between two adjacent iterations is lower than the preset convergence criterion threshold, thus obtaining the optimal Gaussian mixture model description of the relaxation time distribution spectrum curve. Based on the optimal Gaussian mixture model, the first relaxation peak and the second relaxation peak are obtained, and numerical integration is performed within their respective effective time constant intervals to obtain the first and second spectral peak integral areas. The first spectral peak integral area represents the integrity of the cell membrane system, and the second spectral peak integral area represents the degree of vascular bundle embolization. A stem diameter correction coefficient is introduced to perform geometric scaling correction on the integrated area of ​​the second spectral peak. The stem diameter correction coefficient is determined by the perimeter of the cross-section of the stem segment to be measured. The integrated area of ​​the first spectral peak is used as the numerator, and the sum of the integrated area of ​​the first spectral peak and the integrated area of ​​the second spectral peak weighted by the stem diameter correction coefficient is used as the denominator. The vascular bundle fluid flow continuity index is constructed by ratio calculation.

6. The rapid diagnostic method for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis according to claim 1, characterized in that, The process involves a second, precise assessment of the disease status of the target mulberry plant based on the vascular bundle sap flow continuity index. If the assessment result still indicates a suspected latent infection, it suggests obtaining electrical impedance data of the target mulberry plant at noon to calculate the diurnal variation for a final diagnosis, generating a final diagnosis report of the mulberry plant disease, which specifically includes: The vascular bundle fluid flow continuity index is compared with the preset latent period critical threshold. If it is lower than the latent period critical threshold, the target mulberry plant is directly determined to be in the latent period infection state, and a disease diagnosis report containing the latent period diagnosis conclusion and the quantitative value of the degree of vascular bundle embolization is generated. If it is higher than the reference lower limit threshold for healthy plants, it is determined to be in a healthy state and a health conclusion report is generated. If the vascular bundle sap flow continuity index is within a preset fuzzy range between the latent period critical threshold and the healthy plant reference lower limit threshold, a secondary measurement prompt signal is generated, prompting the operator to perform a second electrical impedance spectrum acquisition operation on the same measurement site of the same target mulberry plant during the midday transpiration peak period. The vascular bundle fluid flow continuity index at noon was obtained by using the raw data of the second electrical impedance spectrum collected at noon. The difference between the vascular bundle fluid flow continuity index obtained from the first measurement in the morning and the vascular bundle fluid flow continuity index obtained from the second measurement at noon was calculated to obtain the diurnal drop amplitude. The day-night drop amplitude is compared with a preset difference threshold. If the day-night drop amplitude is greater than or equal to the difference threshold, it indicates that the vascular bundle sap flow continuity exhibits the high stress-sensitive collapse characteristics unique to latent infection, and the target mulberry plant is determined to be in a latent infection state. If the diurnal drop is less than the dynamic difference threshold, it indicates that the low vascular bundle sap flow continuity index is more likely due to mild water deficit or environmental stress. The target mulberry plant is determined to be in a state of doubt and needs to be observed. A final diagnostic report containing suggestions for short-term retesting and monitoring is generated, the stratified diagnostic process is completed, and the final diagnostic report of the mulberry plant disease is output.

7. A rapid diagnostic system for bacterial wilt of mulberry fruit based on electrical impedance spectral analysis, characterized in that, The system includes: a memory, a processor, and a communication interface. The memory contains a program for a rapid diagnosis method of bacterial wilt of mulberry fruit based on electrical impedance spectroscopy analysis. When the processor executes the program for a rapid diagnosis method of bacterial wilt of mulberry fruit based on electrical impedance spectroscopy analysis, it implements the steps of the rapid diagnosis method of bacterial wilt of mulberry fruit based on electrical impedance spectroscopy analysis as described in any one of claims 1-6.

Images