Rapid fruit and vegetable pesticide residue screening method based on enzyme inhibition ratio method

Abstract

The invention discloses a fruit and vegetable pesticide residue rapid screening method based on an enzyme inhibition ratio method, and relates to the technical field of pesticide detection.The method comprises the following steps that S100, visible light, near-infrared and hyperspectral data of fruit and vegetable samples are synchronously collected through a multispectral sensing module, and a pH adjusting module is dynamically triggered based on spectral characteristic peak intensity; s200, executing dynamic pH regulation and enzyme reaction monitoring based on the output of the step S100; s300, based on the output of the S200, fusing the multispectral data and the dynamic curve through an intelligent analysis platform, outputting a pH sensitivity response curved surface through a transfer learning module, and dynamically updating proportional-integral control parameters; and S400, establishing a feedback verification closed loop based on a feedback module, automatically generating a unique identification code and triggering a GC-MS laboratory for confirmation when a positive sample of which the inhibition ratio is greater than a preset value is detected, and finally, synchronously and reversely calibrating the spectrum fusion weight in S300 and the pH adjustment parameter in S200 based on the deviation value.

Description

Technical Field

[0001] This invention relates to the field of pesticide detection technology, and in particular to a rapid screening method for pesticide residues in fruits and vegetables based on enzyme inhibition rate method. Background Technology

[0002] With the rapid development of modern agriculture, the extensive use of pesticides has become an important way to improve grain yield and quality. However, the unreasonable use, abuse, or excessive residues of pesticides not only directly threaten human health, but also cause long-term damage to the ecosystem through environmental pollution and food chain accumulation.

[0003] Chinese Patent Application No. 2014108376889 discloses an auxiliary detection device and method for pesticide residues. The auxiliary detection device includes a pesticide residue rapid test card, a power supply, a device body, a temperature control component, a camera component, and a controller. The temperature control component and the camera component are used for constant temperature heating and imaging, respectively, according to the control of the controller. The detection method includes: ① using the auxiliary detection device to detect the cholinesterase inhibition time of surface extracts of various samples with known cholinesterase inhibition rates. The cholinesterase inhibition rate and inhibition time of various sample surface extracts correspond one-to-one to form multiple sets of sample data. ② using the cholinesterase inhibition rate as the variable and the inhibition time as the independent variable in the multiple sets of sample data, a fourth-order Hermite interpolation spline with shape parameters is obtained. ③ The cholinesterase inhibition time of the sample to be tested is obtained using the S1 method, and then substituted into the fourth-order Hermite interpolation spline to obtain the pesticide residue rate in the sample to be tested. This method is simple and has high detection accuracy.

[0004] Similar to the existing technologies mentioned above, the chlorophyll and anthocyanins in dark-colored fruits and vegetables, such as purple cabbage and blueberries, strongly absorb the visible light spectrum, causing the absorbance signal to be masked, significantly reducing detection sensitivity and increasing the false positive rate. At the same time, allium vegetables are rich in sulfides, and the sulfur groups in their chemical structure can directly inhibit cholinesterase activity, producing an inhibitory effect similar to that of pesticides, resulting in false positive results. In addition, traditional methods are carried out under constant pH conditions, which makes it difficult to cope with the acidification effect caused by sulfur-containing matrices. Furthermore, the lack of an effective closed-loop calibration mechanism between rapid screening results and laboratory confirmatory methods makes it difficult to eliminate systematic errors.

[0005] While near-infrared spectroscopy can penetrate pigments on the surface of dark samples, its accuracy in identifying sulfide characteristic peaks is limited, and it is not effectively integrated with enzyme inhibition kinetic data. Local slicing methods for onion and garlic samples can partially reduce sulfide interference, but they cannot eliminate the influence of internal osmotic residues, and the procedure is cumbersome. Existing enzyme reaction monitoring methods mostly rely on static slope calculations, failing to effectively identify abnormal changes in reaction kinetics caused by sulfides.

[0006] Given the aforementioned challenges, achieving accurate and rapid screening of pesticide residues in fruits and vegetables using the enzyme inhibition rate method under high matrix interference conditions has become a core technological hurdle that needs to be overcome for rapid screening of pesticide residues in fruits and vegetables.

[0007] Therefore, it is necessary to invent a rapid screening method for pesticide residues in fruits and vegetables based on enzyme inhibition rate to solve the above problems. Summary of the Invention

[0008] The purpose of this invention is to provide a rapid screening method for pesticide residues in fruits and vegetables based on enzyme inhibition rate method, so as to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a rapid screening method for pesticide residues in fruits and vegetables based on enzyme inhibition rate method, comprising the following steps: S100: Simultaneously acquires visible light, near-infrared and hyperspectral data of fruit and vegetable samples through a multispectral sensing module, and dynamically triggers the pH adjustment module based on the intensity of spectral characteristic peaks. S200: Based on the output of step S100, perform dynamic pH adjustment and enzyme reaction monitoring. When the characteristic peak of sulfide is detected, inject buffer solution through the dynamic fluid control module to adjust the pH value of the solution and collect the enzyme reaction kinetic curve in real time. After noise reduction by the filtering algorithm with a preset window width, calculate the slope change rate in the initial and middle stages of the reaction. S300, based on the output of S200, integrates multispectral data and dynamic curves through an intelligent analysis platform, dynamically weights multispectral data using a cross-modal attention mechanism, adaptively increases the near-infrared weight of dark samples according to the absorbance deviation of sulfide characteristic peaks, outputs pH sensitivity response surface through transfer learning module, and dynamically updates proportional-integral control parameters. S400 establishes a feedback verification closed loop based on the feedback module. When a positive sample with an inhibition rate greater than the preset value is detected, laboratory confirmation is automatically triggered and residual data is generated. The residual is decomposed into four hierarchical sub-vectors through affine transformation, and the near-infrared wavelength of the data acquisition layer, the PID parameters of the decision execution layer, the fusion weight of the intelligent analysis layer, and the coordinate transformation matrix of the feedback layer are calibrated in reverse respectively. Finally, the hardware control and software model are coordinated and iterated through the synchronous distribution of cross-level parameters.

[0010] Preferably, in step S100, when collecting dark-colored samples from fruit and vegetable samples, the short-wave near-infrared band is used to penetrate surface pigments and identify internal pesticide characteristic peaks. When collecting onion and garlic samples from fruit and vegetable samples, the internal tissue is extracted by local slicing and the penetration residue is detected by near-infrared spectroscopy, while simultaneously suppressing sulfide interference.

[0011] Preferably, step S100 includes: S110. Perform local sectioning on onion and garlic samples and generate pesticide penetration heat maps by near-infrared spectroscopy scanning. S120: Identify high-residue areas based on permeation thermal map, simultaneously detect sulfide characteristic peaks, and activate anti-interference mode when absorbance exceeds preset value; S130: Enable 850nm near-infrared penetrating scanning for dark samples, while simultaneously suppressing sulfide interference.

[0012] Preferably, in step S200, the calculation of the slope change rate includes: fitting the initial average slope in the initial stage of the reaction; fitting the average slope in the middle stage of the reaction, and judging the interference based on the relative change rate of the slopes in the two stages. When the change rate exceeds a preset value, it indicates that there is sulfide interference at this time.

[0013] Preferably, the weight optimization in step S300 includes: when the absorbance of the sulfide characteristic peak exceeds a preset threshold, the near-infrared weight is increased to reduce the false positive rate of dark samples, and the weight adjustment amount is exponentially positively correlated with the absorbance deviation.

[0014] Preferably, the proportional-integral parameter update in S400 includes: the proportional coefficient being dynamically adjusted based on the gradient direction and amplitude of the pH sensitivity response surface.

[0015] Preferably, step S400 includes the following steps: S410. When the system detects that the inhibition rate exceeds the preset threshold, it automatically marks the sample as a positive sample and generates a unique identification code containing timestamp and geographic information. Then, it initiates the laboratory confirmation process, uploads the sample-related data to the cloud platform through an encrypted channel, and sends a confirmation analysis request to the cooperating testing institutions to ensure that the sample traceability information is complete and traceable. S420. The laboratory uses chromatography-mass spectrometry to accurately analyze positive samples and obtain qualitative and quantitative results of pesticide residues. The system uses a spatial transformation algorithm to align the rapid detection data with the laboratory confirmation results and calculates the difference between the two as the system deviation. This deviation will serve as the benchmark for subsequent parameter optimization. S430: Based on the absorbance deviation data of sulfide characteristic peaks, the weighting coefficient of the near-infrared spectrum is dynamically adjusted according to a preset exponential function relationship. When significant interference is detected in dark samples, the system automatically increases the weighting ratio of the near-infrared band, thereby effectively suppressing false positive signals caused by matrix interference. S440. By analyzing the response law of pH value to detection sensitivity, a corresponding response surface model is established. Based on the gradient change characteristics of the model, the parameter settings of the proportional-integral control algorithm are automatically adjusted. S450. By analyzing the rate of change of the slope of the kinetic curve in the early and middle stages of the enzyme-catalyzed reaction, an interference substance identification model is established, and the threshold is dynamically calibrated periodically based on newly added confirmatory data to maintain the accuracy of the judgment.

[0016] S460: All optimized parameters will be sent to each testing terminal through a secure channel to upgrade the overall system performance. At the same time, key indicators will be monitored, and a hardware calibration program will be automatically started when abnormal deviations are detected to ensure that the testing system continues to maintain optimal condition.

[0017] Preferably, the multispectral sensing module is used to output characteristic peaks of sulfides, pesticides and chlorophyll in a timely manner, the dynamic fluid control module can adaptively adjust the buffer injection rate based on the weight adjustment amount, and the intelligent analysis platform has a built-in filtering algorithm, pH sensitivity response surface generator and slope analysis unit. The slope analysis unit can extract the slope parameters of the initial and middle stages of the reaction by dividing the time window.

[0018] Preferably, the feedback module includes a weight synchronization module, which can uniformly express the near-infrared weight values ​​as an absorbance deviation function in the data fusion and feedback process; the feedback module also includes a verification database, which stores the correlation matrix between laboratory confirmation data and slope change rate, for residual reverse calibration.

[0019] Preferably, the feedback module has a built-in real-time feedback channel. When the false positive rate of dark-colored samples exceeds a preset value, it automatically triggers the coordinated update of weights and proportional-integral coefficients. The output of the intelligent analysis platform is directly connected to the microfluidic parameter update interface to form a closed-loop control.

[0020] The technical effects and advantages of this invention are as follows: This invention utilizes a multispectral sensing module to penetrate pigments on the surface of dark-colored samples and identify sulfide characteristic peaks. A dynamic pH adjustment module, based on a PID algorithm, maintains the reaction system in real time, neutralizing sulfide interference. An intelligent analysis platform dynamically weights near-infrared data through a cross-modal attention mechanism, combining transfer learning to optimize control parameters. Finally, GC-MS confirmation data is used to back-calibrate system parameters, forming a closed-loop optimization that overcomes the interference problems of traditional enzyme inhibition methods in dark-colored fruits and vegetables, as well as onion and garlic-based vegetables. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the method of the present invention.

[0022] Figure 2 This is a system framework diagram of the method of the present invention.

[0023] Figure 3 This is the logic control diagram of the method of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] First Embodiment Because the pigments such as chlorophyll and anthocyanins in dark-colored fruits and vegetables (such as purple cabbage and blueberries) strongly absorb the visible light spectrum, the absorbance signal is masked, significantly reducing detection sensitivity and increasing the false positive rate. At the same time, allium vegetables (such as onions and garlic) are rich in sulfides (such as allicin), and the sulfur group (-SH) in their chemical structure can directly inhibit cholinesterase activity, producing an inhibitory effect similar to that of pesticides, resulting in false positive results. Furthermore, traditional methods are carried out under constant pH conditions, which makes it difficult to cope with the acidification effect caused by sulfur-containing matrices. Moreover, the lack of an effective closed-loop calibration mechanism between rapid screening results and laboratory confirmatory methods makes it difficult to eliminate systematic errors.

[0026] like Figures 1-3 As shown, based on the above problems, this embodiment provides a rapid screening method for pesticide residues in fruits and vegetables based on enzyme inhibition rate. This method integrates three modules—multispectral sensing, intelligent analysis, and dynamic pH adjustment—to construct a four-level nested technology matrix, achieving accurate screening of dark-colored samples and interfering matrices such as onions and garlic. The method includes the following steps: S100 synchronously collects visible light, near-infrared and hyperspectral data of fruit and vegetable samples through a multispectral sensing module, and dynamically triggers the pH adjustment module based on the intensity of spectral characteristic peaks.

[0027] In this embodiment, when collecting dark-colored samples from fruits and vegetables, the short-wave near-infrared band is used to penetrate surface pigments and identify internal pesticide characteristic peaks. When collecting onion and garlic samples from fruits and vegetables, the internal tissue is extracted by local slicing, and the penetration residue is detected by near-infrared spectroscopy, while simultaneously suppressing sulfide interference.

[0028] It should be noted that when simultaneously collecting visible and near-infrared light from fruit and vegetable samples using a multispectral sensing module, the near-infrared light bands of 850nm and 1050nm can be selected. When collecting data from dark-colored samples, since organophosphorus compounds are at 412nm and carbamates at 270nm, the 850nm near-infrared band is used to penetrate surface pigments and identify internal pesticide characteristic peaks. When collecting features from onion and garlic samples, the internal tissue is extracted using a local slicing method, and the penetration residue is detected by 1050nm near-infrared spectroscopy, while simultaneously suppressing sulfide interference.

[0029] S200: Based on the output of step S100, perform dynamic pH adjustment and enzyme reaction monitoring. When the characteristic peak of sulfide is detected, inject buffer solution through the dynamic fluid control module to adjust the pH value of the solution and collect the enzyme reaction kinetic curve in real time. After noise reduction by the filtering algorithm with a preset window width, calculate the slope change rate in the initial and middle stages of the reaction.

[0030] It should be noted that when the characteristic peak of sulfide is identified as 230nm±5nm and the absorbance is >0.25, the pH of the system is adjusted to 7.8±0.1 by injecting buffer through the PID control formula (V=0.5·|ΔpH|+0.2·∫|ΔpH|dt), with an accuracy of ±0.2. When acquiring the enzyme reaction kinetic curve in real time, the sampling frequency is controlled at 10Hz, and the noise is removed by Savitzky-Golay filtering with a window width of 15 points and a polynomial order of 3. The slope change rate Δk is also calculated.

[0031] The S300, based on the output of the S200, integrates multispectral data and dynamic curves through an intelligent analysis platform. It adopts a cross-modal attention mechanism to dynamically weight multispectral data. The near-infrared weight of dark samples is adaptively increased according to the absorbance deviation of sulfide characteristic peaks. The pH sensitivity response surface is output through the transfer learning module, and the proportional-integral control parameters are dynamically updated.

[0032] It should be noted that the cross-modal attention mechanism is used to weight the data for analysis, and the weight optimization formula is as follows: in, 0.1 represents the absorbance of the characteristic peak of the sulfide; 0.1 is the amplitude control coefficient. The response function is nonlinear. A saturated response curve is constructed through an exponential decay structure to achieve weight adjustment in a smooth transition manner. The steepness factor of 0.3 controls the growth rate of the function curve. The larger the value, the more sensitive the system is to small absorbance deviations. The absorbance deviation of the characteristic peaks of sulfides is due to... The measured absorbance deviation at 230nm wavelength directly reflects the sulfide concentration; 0.25 is the threshold for sulfide interference, set based on noise distribution and signal separation requirements. When the absorbance deviation of the sulfide characteristic peak... When the value is zero, the function output is zero and no weight adjustment is triggered. However, as the deviation increases, the function value increases non-linearly and gradually approaches saturation.

[0033] It should be noted that the transfer learning module outputs the pH sensitivity response surface and dynamically updates the PID parameters. For example, the update logic is that K_p changes from 0.3 to 0.52 based on the surface gradient change rate being >5% / pH.

[0034] In this embodiment, the weight optimization in step S300 includes: when the absorbance of the sulfide characteristic peak exceeds a preset threshold, the near-infrared weight is increased to reduce the false positive rate of dark samples, and the weight adjustment amount is exponentially positively correlated with the absorbance deviation.

[0035] It should be noted that the S300 weight optimization formula In the middle, when When the absorbance is greater than 0.25, the absorbance of the sulfide characteristic peak increases, thereby reducing the false positive rate of dark-colored samples.

[0036] S400 establishes a feedback verification closed loop based on the feedback module. When a positive sample with an inhibition rate greater than the preset value is detected, laboratory confirmation is automatically triggered and residual data is generated. The residual is decomposed into four hierarchical sub-vectors through affine transformation, and the near-infrared wavelength of the data acquisition layer, the PID parameters of the decision execution layer, the fusion weight of the intelligent analysis layer, and the coordinate transformation matrix of the feedback layer are calibrated in reverse respectively. Finally, the hardware control and software model are coordinated and iterated through the synchronous distribution of cross-level parameters.

[0037] It should be noted that the positive samples triggered GC-MS confirmation, and the residual data inversely optimized the weight parameters. and PID coefficients.

[0038] In this embodiment, the proportional-integral parameter update in S400 includes: dynamic adjustment of the proportional coefficient based on the gradient direction and amplitude of the pH sensitivity response surface.

[0039] It should be noted that the formula for adjusting the proportional coefficient Kp is: in, This is the updated value of the scaling factor; 0.2 is the current scaling factor; 0.2 is the learning rate factor, defined as the step size constraint for gradient updates, to prevent... Mutations trigger system oscillations; The partial derivative of the response surface function S with respect to pH represents the magnitude of the effect of a unit pH change on the system's detection sensitivity S; S is a multidimensional surface function generated by the transfer learning module.

[0040] In this embodiment, the multispectral sensing module is used to output characteristic peaks of sulfides, pesticides and chlorophyll in a timely manner, the dynamic fluid control module can adaptively adjust the buffer injection rate based on the weight adjustment amount, and the intelligent analysis platform has built-in filtering algorithm, pH sensitivity response surface generator and slope analysis unit. The slope analysis unit can extract the slope parameters of the initial and middle stages of the reaction by dividing the time window.

[0041] It should be noted that the resolution of the multispectral sensing module can be set to 10nm, and it can output the required characteristic peak data in real time. The dynamic fluid control module can then... Adjust the injection rate of the buffer solution to control the response delay within 10ms. The intelligent analysis platform can incorporate the Savitzky-Golay filtering algorithm and pH-sensitivity response surface generator.

[0042] In this embodiment, the feedback module includes a weight synchronization module, which can uniformly express the near-infrared weight values ​​as an absorbance deviation function in the data fusion and feedback process; the feedback module also includes a verification database, which stores the correlation matrix between laboratory confirmation data and slope change rate, for residual reverse calibration.

[0043] In this embodiment, the feedback module has a built-in real-time feedback channel. When the false positive rate of dark-colored samples exceeds the preset value, it automatically triggers the coordinated update of weights and proportional-integral coefficients. The output of the intelligent analysis platform is directly connected to the microfluidic parameter update interface to form a closed-loop control.

[0044] It should be noted that the real-time feedback channel can be automatically triggered when the false positive rate of dark-colored samples exceeds the preset value. Update synchronously with the PID coefficients.

[0045] In use, the multispectral sensing module first uses the 850nm near-infrared band to penetrate the surface pigment layer and directly identify the internal pesticide characteristic peaks for dark-colored samples such as purple cabbage and blueberries. For onion, garlic and other garlic-like samples, the module detects penetration residues by combining local slicing with 1050nm spectroscopy and simultaneously monitors the sulfide characteristic peak at 230nm, thus effectively avoiding the absorption of visible light by chlorophyll and the false positive interference caused by sulfides.

[0046] When the system detects that the characteristic peak of sulfide exceeds the preset value, the pH adjustment module is immediately activated and a buffer solution controlled by PID is injected to adjust the pH of the sample solution to the required value, thereby neutralizing the acidification effect caused by sulfide. At the same time, the enzyme reaction kinetic curve is acquired at a frequency of 10Hz, and after being filtered and denoised by Savitzky-Golay filtering, the rate of change of the reaction slope is analyzed, thereby accurately quantifying the degree of enzyme activity inhibition and reducing the misclassification rate of onion and garlic samples.

[0047] During the data integration phase, the intelligent analysis platform uses a cross-modal attention mechanism to fuse and analyze multispectral data and kinetic curves. It can adaptively increase the weight of near-infrared data based on the sulfide absorbance deviation, enhance the extraction capability of useful signals in dark samples, and dynamically adjust PID parameters by combining the pH sensitivity response surface generated by transfer learning to achieve adaptive matching for different matrix characteristics, thereby further reducing the false positive rate of dark samples.

[0048] When a positive sample with an inhibition rate exceeding a preset value is detected, the GC-MS laboratory confirmation process is automatically triggered. The rapid test data is aligned with the confirmation results through spatial coordinate transformation to generate a system bias. Based on this bias, the system reverse-calibrates the spectral fusion weights and pH adjustment parameters, forming a self-learning ability for continuous optimization, thereby reducing errors.

[0049] Second Embodiment Based on the above methods, technicians found that while near-infrared spectroscopy can penetrate pigments on the surface of dark samples, its accuracy in identifying sulfide characteristic peaks is limited, and it is not effectively integrated with enzyme inhibition kinetic data. Although local slicing of onion and garlic samples can partially reduce sulfide interference, it cannot eliminate the influence of internal permeation residues, and the operation process is cumbersome. Existing enzyme reaction monitoring methods mostly rely on static slope calculations, which fail to effectively identify abnormal changes in reaction kinetics caused by sulfides.

[0050] In this embodiment, step S100 includes: S110. Perform local sectioning on onion and garlic samples, and generate pesticide penetration heat maps by scanning near-infrared spectroscopy in the 1050nm band.

[0051] S120: Based on the penetration thermal map, high residual areas are identified, sulfide characteristic peaks are detected simultaneously, and the anti-interference mode is activated when the absorbance exceeds the preset value.

[0052] It should be noted that since sulfides such as allicin and diallyl disulfide have characteristic absorption peaks in the short-wave ultraviolet region, with their maximum absorption wavelength located in the range of 225–235 nm, garlic extract exhibits a significant absorption peak at 230 nm. Other interfering substances, such as phenols, have weak absorption in this band. Furthermore, hyperspectral databases such as NIST Chemistry WebBook show that the characteristic absorption peaks of organosulfur compounds are concentrated around 230 nm. Therefore, the simultaneous detection of sulfide characteristic peaks is set at 230 nm ± 5 nm. Setting the preset absorbance value to 0.25, i.e., activating the anti-interference mode when the absorbance is greater than 0.25, ensures the elimination of matrix background interference and the capture of true sulfide signals.

[0053] S130: Enable 850nm near-infrared penetrating scanning for dark samples, while simultaneously suppressing sulfide interference.

[0054] In this embodiment, the calculation of the slope change rate in step S200 includes: fitting the initial average slope in the early stage of the reaction; fitting the average slope in the middle stage of the reaction, and judging the interference based on the relative change rate of the slopes in the two stages. When the change rate exceeds the preset value, it indicates that there is sulfide interference at this time.

[0055] It should be noted that, for ease of understanding, the initial stage of the reaction, t1-t2, is defined as 0-30 seconds, and its initial fitting slope is k1; the middle stage of the reaction, t3-t4, is defined as 30-60 seconds, and its middle fitting slope is k2. The formula for calculating the slope is: In this embodiment, step S400 includes the following steps: S410. When the system detects that the inhibition rate exceeds the preset threshold, it automatically marks the sample as a positive sample and generates a unique identification code containing timestamp and geographic information. Then, it initiates the laboratory confirmation process, uploads the sample-related data to the cloud platform through an encrypted channel, and sends a confirmation analysis request to the cooperating testing institutions to ensure that the sample traceability information is complete and traceable.

[0056] S420. The laboratory uses chromatography-mass spectrometry (GC-MS) to accurately analyze positive samples, obtaining qualitative and quantitative results of pesticide residues. The system uses a spatial transformation algorithm to align the rapid detection data with the laboratory confirmatory results, calculating the difference between the two as the system bias. This bias will serve as the benchmark for subsequent parameter optimization.

[0057] It should be noted that coordinate alignment uses affine transformation alignment: Where Y is the GC-MS confirmatory data; X is the rapid detection data, namely the suppression rate and characteristic peak absorbance; A is the rotation scaling matrix fitted by the least squares method; and b is the translation vector.

[0058] S430: Based on the absorbance deviation data of sulfide characteristic peaks, the weighting coefficient of the near-infrared spectrum is dynamically adjusted according to a preset exponential function relationship. When significant interference is detected in dark samples, the system automatically increases the weighting ratio of the near-infrared band, thereby effectively suppressing false positive signals caused by matrix interference.

[0059] S440. By analyzing the response law of pH value to detection sensitivity, a corresponding response surface model is established. Based on the gradient change characteristics of the model, the parameter settings of the proportional-integral control algorithm are automatically adjusted.

[0060] S450. By analyzing the rate of change of the slope of the kinetic curve in the early and middle stages of the enzyme-catalyzed reaction, an interference substance identification model is established, and the threshold is dynamically calibrated periodically based on newly added confirmatory data to maintain the accuracy of the judgment.

[0061] S460: All optimized parameters will be sent to each testing terminal through a secure channel to upgrade the overall system performance. At the same time, key indicators will be monitored, and a hardware calibration program will be automatically started when abnormal deviations are detected to ensure that the testing system continues to maintain optimal condition.

[0062] In use, firstly, to address the issue of strong absorption of visible light by pigments such as chlorophyll and anthocyanins in dark samples, the 850nm near-infrared band is used to penetrate the surface pigments and identify the characteristic peaks of pesticides inside. For the interference of sulfides on enzyme activity in onion and garlic samples, a local slicing method combined with 1050nm near-infrared spectral scanning is used to generate a pesticide penetration heat map, and the characteristic peaks of sulfides are monitored at 230nm±5nm. When the absorbance exceeds the set value, the anti-interference mode is automatically activated.

[0063] Secondly, the dynamic pH adjustment module dynamically adjusts the pH value of the reaction system based on the real-time collected sulfide characteristic peak signal through a PID-controlled buffer injection mechanism. At the same time, it collects the enzyme reaction kinetic curve at a frequency of 10Hz, and after filtering and noise reduction, it analyzes the slope change rate in the initial and middle stages of the reaction. When the relative change of the slope exceeds the set threshold, it is determined to be sulfide interference and triggers the correction mechanism.

[0064] Furthermore, the intelligent analysis platform employs a cross-modal attention mechanism to fuse multispectral data with enzyme kinetic parameters. It adaptively adjusts the fusion weights of near-infrared data through weight optimization. When the absorbance of the sulfide characteristic peak exceeds the set value, the near-infrared weight is increased, effectively suppressing false positive signals from dark samples. The platform also integrates a transfer learning module, outputting the pH sensitivity response surface to the inhibition rate, thereby dynamically updating the PID control parameters.

[0065] Finally, when a positive sample with an inhibition rate exceeding the set value is detected, a unique identification code is automatically generated and GC-MS confirmatory analysis is triggered. The rapid detection data is spatially aligned with the laboratory results through an affine transformation algorithm, the system deviation is calculated, and the near-infrared weights and PID parameters are optimized accordingly. All optimized parameters are sent to each detection terminal through a secure channel to achieve continuous calibration and performance upgrade of system parameters.

[0066] It should be noted that this invention uses a multispectral sensing module to penetrate the pigments on the surface of dark samples and identify sulfide characteristic peaks; then, a dynamic pH adjustment module maintains the reaction system in real time based on a PID algorithm to neutralize sulfide interference; the intelligent analysis platform dynamically weights near-infrared data through a cross-modal attention mechanism and optimizes control parameters by combining transfer learning; finally, the system parameters are back-calibrated using GC-MS confirmation data to form a closed-loop optimization, overcoming the interference problem of traditional enzyme inhibition methods in dark fruits and vegetables and onion and garlic vegetables.

[0067] Finally, it should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A rapid screening method for pesticide residues in fruits and vegetables based on enzyme inhibition rate, characterized in that, This method integrates three modules—multispectral sensing, intelligent analysis, and dynamic pH adjustment—to construct a four-level nested technology matrix, enabling accurate screening of dark-colored samples and interfering matrices such as onions and garlic. The steps include: S100: Simultaneously acquires visible light, near-infrared and hyperspectral data of fruit and vegetable samples through a multispectral sensing module, and dynamically triggers the pH adjustment module based on the intensity of spectral characteristic peaks. S200: Based on the output of step S100, perform dynamic pH adjustment and enzyme reaction monitoring. When the characteristic peak of sulfide is detected, inject buffer solution through the dynamic fluid control module to adjust the pH value of the solution and collect the enzyme reaction kinetic curve in real time. After noise reduction by the filtering algorithm with a preset window width, calculate the slope change rate in the initial and middle stages of the reaction. S300, based on the output of S200, integrates multispectral data and dynamic curves through an intelligent analysis platform, dynamically weights multispectral data using a cross-modal attention mechanism, adaptively increases the near-infrared weight of dark samples according to the absorbance deviation of sulfide characteristic peaks, outputs pH sensitivity response surface through transfer learning module, and dynamically updates proportional-integral control parameters. S400 establishes a feedback verification closed loop based on the feedback module. When a positive sample with an inhibition rate greater than the preset value is detected, laboratory confirmation is automatically triggered and residual data is generated. The residual is decomposed into four hierarchical sub-vectors through affine transformation, and the near-infrared wavelength of the data acquisition layer, the PID parameters of the decision execution layer, the fusion weight of the intelligent analysis layer, and the coordinate transformation matrix of the feedback layer are calibrated in reverse respectively. Finally, the hardware control and software model are coordinated and iterated through the synchronous distribution of cross-level parameters.

2. The rapid screening method for pesticide residues in fruits and vegetables according to claim 1, characterized in that, In step S100, when collecting dark-colored samples from fruits and vegetables, the short-wave near-infrared band is used to penetrate surface pigments and identify internal pesticide characteristic peaks. When collecting onion and garlic samples from fruits and vegetables, the internal tissue is extracted by local slicing and the penetration residue is detected by near-infrared spectroscopy, while simultaneously suppressing sulfide interference.

3. The rapid screening method for pesticide residues in fruits and vegetables according to claim 1, characterized in that, Step S100 includes: S110. Perform local sectioning on onion and garlic samples and generate pesticide penetration heat maps by near-infrared spectroscopy scanning. S120: Identify high-residue areas based on permeation thermal map, simultaneously detect sulfide characteristic peaks, and activate anti-interference mode when absorbance exceeds preset value; S130: Enable 850nm near-infrared penetrating scanning for dark samples, while simultaneously suppressing sulfide interference.

4. The rapid screening method for pesticide residues in fruits and vegetables according to claim 1, characterized in that, In step S200, the calculation of the slope change rate includes: fitting the initial average slope in the early stage of the reaction; fitting the average slope in the middle stage of the reaction, and judging the interference based on the relative change rate of the slopes in the two stages. When the change rate exceeds the preset value, it indicates that there is sulfide interference at this time.

5. The rapid screening method for pesticide residues in fruits and vegetables according to claim 1, characterized in that, The weight optimization in step S300 includes: when the absorbance of the sulfide characteristic peak exceeds a preset threshold, the near-infrared weight is increased to reduce the false positive rate of dark samples, and the weight adjustment amount is exponentially positively correlated with the absorbance deviation.

6. The rapid screening method for pesticide residues in fruits and vegetables according to claim 1, characterized in that, The proportional-integral parameter update of the S400 includes: the proportional coefficient is dynamically adjusted based on the gradient direction and amplitude of the pH sensitivity response surface.

7. The rapid screening method for pesticide residues in fruits and vegetables according to claim 1, characterized in that, Step S400 includes the following steps: S410. When the system detects that the inhibition rate exceeds the preset threshold, it automatically marks the sample as a positive sample and generates a unique identification code containing timestamp and geographic information. Then, it initiates the laboratory confirmation process, uploads the sample-related data to the cloud platform through an encrypted channel, and sends a confirmation analysis request to the cooperating testing institutions to ensure that the sample traceability information is complete and traceable. S420. The laboratory uses chromatography-mass spectrometry to accurately analyze positive samples and obtain qualitative and quantitative results of pesticide residues. The system uses a spatial transformation algorithm to align the rapid detection data with the laboratory confirmation results and calculates the difference between the two as the system deviation. This deviation will serve as the benchmark for subsequent parameter optimization. S430: Based on the absorbance deviation data of sulfide characteristic peaks, the weighting coefficient of the near-infrared spectrum is dynamically adjusted according to a preset exponential function relationship. When significant interference is detected in dark samples, the system automatically increases the weighting ratio of the near-infrared band, thereby effectively suppressing false positive signals caused by matrix interference. S440. By analyzing the response law of pH value to detection sensitivity, a corresponding response surface model is established. Based on the gradient change characteristics of the model, the parameter settings of the proportional-integral control algorithm are automatically adjusted. S450. By analyzing the rate of change of the slope of the kinetic curve in the early and middle stages of the enzyme-catalyzed reaction, an interference substance identification model is established, and the threshold is dynamically calibrated periodically based on newly added confirmatory data to maintain the accuracy of the judgment. S460: All optimized parameters will be sent to each testing terminal through a secure channel to upgrade the overall system performance. At the same time, key indicators will be monitored, and a hardware calibration program will be automatically started when abnormal deviations are detected to ensure that the testing system continues to maintain optimal condition.

8. The rapid screening method for pesticide residues in fruits and vegetables according to claim 1, characterized in that, The multispectral sensing module is used to output characteristic peaks of sulfides, pesticides and chlorophyll in a timely manner. The dynamic fluid control module can adaptively adjust the buffer injection rate based on the weight adjustment amount. The intelligent analysis platform has built-in filtering algorithm, pH sensitivity response surface generator and slope analysis unit. The slope analysis unit can extract slope parameters in the early and middle stages of the reaction by dividing the time window.

9. The rapid screening method for pesticide residues in fruits and vegetables according to claim 1, characterized in that, The feedback module includes a weight synchronization module, which can uniformly express the near-infrared weight values ​​as an absorbance deviation function in the data fusion and feedback process; the feedback module also includes a verification database, which stores the correlation matrix between laboratory confirmation data and slope change rate, and is used for residual reverse calibration.

10. The rapid screening method for pesticide residues in fruits and vegetables according to claim 1, characterized in that, The feedback module has a built-in real-time feedback channel. When the false positive rate of dark samples exceeds the preset value, it automatically triggers the coordinated update of weights and proportional-integral coefficients. The output of the intelligent analysis platform is directly connected to the microfluidic parameter update interface to form a closed-loop control.

Images