A method and device for rapid non-destructive quality detection of aquatic products

Abstract

The application provides a kind of for aquatic product rapid nondestructive quality detection method and device, it is related to food safety detection technical field, including spectral analysis module, biosensor module, electrochemical detection module, data processing module, data output module and portable equipment.The application adopts the above-mentioned one for aquatic product rapid nondestructive quality detection method and device, and the fusion of spectral analysis, biosensor and electrochemical detection technology is realized Multi-index synchronous detection of aquatic product freshness, drug residue and heavy metal content.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of food safety detection, in particular to a method and device for rapid non-destructive quality detection of aquatic products. BACKGROUND

[0002] Quality detection of aquatic products in the circulation link is an important link to ensure food safety. Traditional detection methods rely on large laboratory instruments (such as high-performance liquid chromatographs, atomic absorption spectrometers, etc.), which have defects such as large equipment size, long detection period (usually several hours to several days), complex operation and high cost, and are difficult to meet the needs of on-site rapid screening.

[0003] In the prior art, a single detection technology portable device (such as a near-infrared spectrometer or an electrochemical sensor) can only achieve partial index detection and cannot comprehensively cover the core parameters (freshness, drug residues and heavy metals) of aquatic product quality. In addition, the multi-module data fusion analysis capability is insufficient, the detection result is easily disturbed by the environment, and there is a lack of automatic calibration mechanism, and long-term use is prone to data drift problems.

[0004] More importantly, the existing multi-detection technology integrated device generally has the problems of simple data fusion algorithm and insufficient feature extraction. Traditional methods mostly use simple data splicing or linear weighted fusion, and fail to fully exploit the internal correlation and complementary information between different detection technology data. In addition, there is a lack of special deep learning model for multi-index detection of aquatic products, resulting in difficulty in meeting the actual application requirements of detection accuracy and stability. Therefore, it is urgent to develop an intelligent fusion algorithm and deep learning model for multi-source heterogeneous detection data, integrate multi-technology, and support a portable device for rapid non-destructive detection, in order to solve the limitations of traditional methods. SUMMARY

[0005] The purpose of the present application is to provide a method and device for rapid non-destructive quality detection of aquatic products, which solves the technical problems of low data fusion efficiency, insufficient feature extraction and poor model generalization ability in traditional multi-technology integrated detection through innovative heterogeneous data preprocessing algorithm, adaptive feature fusion mechanism and special deep residual network model, realizes rapid detection of multiple indexes of aquatic products, and breaks through the bottleneck of large size, long time consumption and low precision of traditional equipment.

[0006] To achieve the above purpose, the present application provides a device for rapid non-destructive quality detection of aquatic products, comprising:

[0007] The spectral analysis module is used for scanning the surface reflectance spectrum of the aquatic product, acquiring spectral data in the 400-2500nm waveband, and transmitting the spectral data to the data processing unit;

[0008] The biosensing module comprises a microfluidic chip, an antibody modified electrode, and an impedance measurement unit, is used for automatically injecting a sample extraction liquid into the microfluidic chip, detecting a rate of change of an electrode interface impedance, generating a drug residue detection curve, and transmitting detected drug residue detection curve data to a data processing unit;

[0009] The electrochemical detection module is used for performing pre-enrichment and recording heavy metal elution peak integral area through a dissolution scan, and transmitting detection data to the data processing unit;

[0010] The data processing module is used for processing data transmitted from the spectral analysis module, the biosensing module, and the electrochemical detection module, and outputting processed data through a data output module;

[0011] The data output module comprises a touch display screen and a wireless transmission unit, the touch display screen is used for displaying detection results and quality scores in real time, and the wireless transmission unit is used for uploading detection data to a block chain storage platform and generating a detection report through an encryption protocol;

[0012] The self-calibration system comprises a standard substance warehouse built-in, contains freshness standard samples, drug residue standard liquids, and heavy metal standard solutions, automatically performs zero-point calibration every day through a mechanical arm, and the calibration data are used for correcting light intensity drift of the spectral analysis module, baseline offset of the biosensing module, and background current of the electrochemical detection module.

[0013] Preferably, the microfluidic chip drives a sample liquid flow through a micropump, triggers an antibody-antigen specific binding reaction, the impedance measurement unit monitors an electrode interface impedance change in real time, generates a drug residue detection signal, and transmits the signal to the data processing unit;

[0014] The microfluidic chip is composed of a PDMS base layer and a glass substrate bonded together, and comprises a serpentine mixing channel, a reaction chamber, and a waste pool.

[0015] The antibody modified electrode adopts a nanogold / graphene composite substrate, and a chloramphenicol monoclonal antibody is fixed on the surface through an EDC / NHS coupling method;

[0016] The impedance measurement unit is integrated with an electrochemical workstation.

[0017] Preferably, the electrochemical detection module adopts a three-electrode system, including a working electrode, a counter electrode, and a reference electrode, the surface of the working electrode is modified with a heavy metal ion selective membrane, and a heavy metal ion dissolution peak current is detected through a square wave anodic stripping voltammetry.

[0018] Preferably, the data processing module and the data output module comprise the following steps in data processing:

[0019] Baseline calibration stage: PBS buffer is injected through a micropump, an initial impedance spectrum is recorded, and baseline equivalent circuit parameters are established;

[0020] Dynamic detection phase: The sample liquid flows over the surface of the antibody-modified electrode, and the binding of antigen and antibody causes changes in the double-layer capacitance. The system continuously collects impedance amplitude and phase data for multiple cycles.

[0021] Signal processing stage: The Savitzky-Golay filtering algorithm is used to eliminate high-frequency noise, and the real part change of impedance at characteristic frequency points is extracted by the sliding window method to generate a three-dimensional impedance fingerprint spectrum.

[0022] Specificity verification mechanism: A control electrode modified with non-specific antibodies is set up, and matrix interference is eliminated by the difference in impedance changes between the main detection site and the control site;

[0023] Multi-source data reception stage: The data processing unit synchronously receives raw data from the three detection modules through the first communication interface, the second communication interface and the third communication interface, and performs timestamp alignment to ensure data consistency;

[0024] The second communication interface uses the RS485 protocol to transmit structured data packets containing timestamps, impedance amplitude spectra, and rate of change of characteristic frequency points at 115200bps.

[0025] The microfluidic chip is equipped with an automatic cleaning process. After the test is completed, it is sequentially injected with 0.1M NaOH solution, deionized water and nitrogen gas for three rinsing cycles, with a residual rate of less than 0.5%.

[0026] Preferably, the working electrode of the three-electrode system is a glassy carbon electrode with a diameter of 3 mm, and the surface is sequentially modified with a nano-platinum / graphene composite layer and a heavy metal ion selective film.

[0027] The selective membrane is composed of chitosan-thioglycolic acid copolymer and Pb 2+ / Cd 2+ The ion-imprinted polymers are compounded at a mass ratio of 1:3.

[0028] Preferably, the data processing module includes a data fusion submodule and a machine learning analysis submodule. The data fusion submodule normalizes the spectral data, impedance change data, and dissolution peak current data and generates feature vectors. The machine learning analysis submodule classifies and performs regression analysis on the feature vectors using a pre-trained deep neural network model and outputs a comprehensive evaluation report that includes freshness level, drug residue type, and heavy metal exceedance index.

[0029] A rapid and non-destructive quality testing method for aquatic products includes the following steps:

[0030] S1. Place the aquatic product to be tested in the testing chamber and perform a full-band spectral scan through the spectral analysis module to obtain surface reflectance distribution data;

[0031] S2, the biosensor module automatically injects the sample extraction liquid into the microfluidic chip, monitors the impedance change rate of the antibody modified electrode, and generates a drug residue detection curve;

[0032] S3, the electrochemical detection module performs three times of square wave voltammetry scanning in the potential range of-1.2V to +0.5V, and records the heavy metal elution peak integral area;

[0033] S4, the data processing module extracts and fuses the spectral reflectance data, impedance change rate data and elution peak area data, inputs a deep neural network model for multi-index joint analysis;

[0034] S5, a detection report containing freshness score, drug residue over-limit warning and heavy metal content threshold is generated according to the output result of the data processing module, and is displayed and stored through the data output module.

[0035] Therefore, the application adopts the above-mentioned method and device for rapid non-destructive quality detection of aquatic products, and has the following technical effects:

[0036] (1) Intelligent fusion of multi-source heterogeneous data: for the first time, an adaptive fusion algorithm for spectral-biosensor-electrochemical heterogeneous data is proposed, which realizes the deep fusion of data with different physical detection principles through feature space mapping and weight dynamic adjustment;

[0037] (2) Special deep learning model: innovative design of branch-fusion-residual three-level network architecture, compared with traditional machine learning method, the detection accuracy is improved by 15-20%, and the false alarm rate is reduced by more than 60%;

[0038] (3) Adaptive model optimization: through online learning and transfer learning mechanism, the model can automatically adjust parameters according to different types of aquatic products and seasonal changes, and has strong generalization ability. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is a modular structure schematic diagram of the portable device of the application;

[0040] Figure 2 is a flow chart of the detection method;

[0041] Figure 3 is a serpentine channel design diagram of the biosensor module microfluidic chip;

[0042] Figure 4 is a three-electrode system layout diagram of the electrochemical detection module;

[0043] Figure 5 is a deep learning model architecture diagram of the data processing unit. DETAILED DESCRIPTION

[0044] The technical solutions of the present application are further described below with reference to the drawings and examples.

[0045] Unless otherwise defined, technical terms or scientific terms used in the present application shall have the usual meaning understood by a person having ordinary skill in the art to which the present application pertains.

[0046] Example 1

[0047] As shown in Figure 1 , a device for rapid non-destructive quality detection of aquatic products comprises

[0048] The spectral analysis module is used for scanning the reflectance spectrum of the surface of the aquatic product, acquiring spectral data in the 400-2500 nm waveband, and transmitting the spectral data to the data processing unit.

[0049] The spectral analysis module is located on the upper part of the portable device and comprises a high-precision LED light source array, a multi-spectral sensor, and a reflective probe. The LED light source array is composed of 32 groups of adjustable wavelength LED units, covers the 400-2500 nm spectral range, and is arranged in a ring at an incident angle of 45°±2° on the periphery of the reflective probe. The multi-spectral sensor adopts a 2048-pixel InGaAs line array detector and is equipped with a grating spectrometer system, with a spectral resolution of 1.5 nm. The reflective probe adopts a combination structure of a sapphire lens and a seven-core optical fiber bundle, is connected to the main body of the spectrometer through an SMA905 interface, and transmits spectral data to the data processing module through a USB3.0 interface.

[0050] As shown in Figure 3 , the biosensor module comprises a microfluidic chip, an antibody-modified electrode, and an impedance measurement unit, is used for automatically injecting a sample extraction solution into the microfluidic chip, detecting the rate of change of the electrode interface impedance, generating a drug residue detection curve, and transmitting the detected drug residue detection curve data to the data processing unit.

[0051] The biosensor module is located in the middle of the device and is composed of a microfluidic chip, an antibody-modified electrode, and an impedance measurement unit. The microfluidic chip is composed of a PDMS base layer and a glass substrate bonded together, contains a sample inlet, a serpentine mixing channel, a reaction chamber, an antibody-modified electrode array, a waste pool, and a micropump system, and has a channel width of 200 μm and a depth of 50 μm. The antibody-modified electrode adopts a nanogold / graphene composite substrate, and the surface is fixed with a chloramphenicol monoclonal antibody through an EDC / NHS coupling method. The impedance measurement unit is integrated with an electrochemical workstation, and transmits the impedance change data to the data processing unit through an RS485 protocol.

[0052] The basic structure of the microfluidic chip includes: a PDMS base layer with a thickness of 2 mm, which is prepared by soft lithography technology to form a microchannel structure; and a glass substrate with a thickness of 1 mm, which is firmly bonded to the PDMS base layer after plasma activation treatment. The combination of the two forms a closed microfluidic channel system, ensuring the directional flow of liquid in the chip and avoiding cross contamination.

[0053] The sample inlet is located at the upper left corner of the chip and has a diameter of 1 mm. It is designed in a funnel structure to facilitate accurate injection of the sample extraction liquid. The inner wall of the sample inlet is treated with hydrophobicity to prevent liquid residue and ensure the accuracy of quantitative sample injection. The sample inlet is connected to a micro valve to control the accurate injection time and flow rate of the sample liquid.

[0054] The serpentine mixing channel extends from the sample inlet and has an "S" shape. The channel has a width of 200 μm, a depth of 50 μm, and a total length of 35 mm. A micro mixing unit is arranged every 1 mm in the channel, which is arranged in an interlaced diamond-shaped micro column structure to enhance fluid disturbance and promote the thorough mixing of the sample and the buffer. The channel wall is treated with hydrophilicity to reduce surface tension and ensure smooth liquid flow.

[0055] The reaction chamber (303) is located at the end of the serpentine channel and has a diameter of 5 mm and a depth of 100 μm, with a volume of about 2 μL. The bottom of the chamber is a glass substrate and the top is a transparent PDMS layer, which facilitates optical observation. The inner wall of the chamber is treated with BSA (bovine serum albumin) to reduce non-specific adsorption and improve detection sensitivity.

[0056] The antibody-modified electrode array (304) is fixed at the bottom of the reaction chamber and is composed of three groups of parallel gold electrodes. Each group of electrodes includes a working electrode (diameter 500 μm), a reference electrode, and an auxiliary electrode. The working electrode uses a nano-gold / graphene composite substrate, and the surface is fixed with chloramphenicol monoclonal antibody by EDC / NHS coupling method, with an antibody loading density of 1.2 x 10 4 molecules / μm 2 . Among them, the first group of electrodes in the three groups is the main detection site, the second group is the non-specific antibody-modified control electrode, and the third group is the standby electrode.

[0057] The working process of the microfluidic chip is as follows:

[0058] Sample preparation stage: 50 μL of sample liquid is extracted from aquatic products and injected into the sample inlet after simple centrifugal treatment;

[0059] Sample delivery stage: the micro-pump system drives the sample liquid to enter the serpentine mixing channel from the sample inlet (301) at a constant flow rate of 0.1 μL / s;

[0060] Mixed reaction stage: the sample liquid is mixed with pre-set PBS buffer in the serpentine channel, and the disturbance is enhanced by the micro-mixing unit in the channel to ensure uniform distribution of the reactants;

[0061] Antigen-antibody binding stage: the mixed liquid enters the reaction chamber and contacts with the antibody-modified electrode array. The drug residues (antigens) in the sample specifically bind to the monoclonal antibodies on the electrode surface, resulting in a change in the impedance of the electrode interface;

[0062] Signal detection stage: the impedance measurement unit continuously collects 20 cycles of impedance amplitude and phase data at a sampling frequency of 10 Hz, while monitoring the impedance change difference between the main detection site and the control site to eliminate matrix interference;

[0063] Data processing stage: the Savitzky-Golay filtering algorithm is used to eliminate high-frequency noise, and the sliding window method is used to extract the impedance real part change of the characteristic frequency points (1 kHz, 10 kHz, 100 kHz) to generate a three-dimensional impedance fingerprint spectrum;

[0064] Waste liquid discharge stage: after the detection is completed, the reaction liquid flows into the waste liquid pool (305) through the drainage channel, completing a detection cycle.

[0065] After the detection is completed, the microfluidic chip automatically performs a cleaning process: first, 0.1M NaOH solution is injected to flush the channel and reaction chamber. The strong alkali environment can effectively separate the antigen-antibody complex; then deionized water is injected for neutralization and dilution to remove residual alkali; finally, nitrogen gas is introduced to dry the channel and reaction chamber to prevent microbial growth and maintain the long-term stability of the chip.

[0066] The microfluidic chip design has the following advantages: first, the serpentine channel structure prolongs the residence time of the liquid in the chip, ensuring sufficient mixing of the sample and reagents; second, the three parallel electrode design allows the main detection and control to be performed simultaneously, effectively eliminating the influence of environmental interference and non-specific adsorption; finally, the integrated microfluidic structure realizes the integration of sample injection, mixing reaction, signal detection, and waste collection, significantly improving the detection efficiency and accuracy. The microfluidic chip is particularly suitable for rapid screening of trace drug residues in aquatic products, with detection sensitivity and specificity meeting national standard requirements.

[0067] As shown in Figure 4 , the electrochemical detection module is used to perform pre-concentration and dissolution scanning to record the integral area of heavy metal dissolution peaks, and transmit the detection data to the data processing unit;

[0068] The electrochemical detection module is located at the lower part of the device and adopts a three-electrode system, which is composed of a working electrode, a counter electrode, a reference electrode, an electrode fixing seat, a detection cavity, and an electrochemical control unit, forming a complete heavy metal detection system.

[0069] The working electrode is located at the center of the three-electrode system, using a glassy carbon electrode with a diameter of 3 mm, and the surface is sequentially modified with a 50 nm thick nano platinum / graphene composite layer and a 100 μm thick heavy metal ion selective membrane. The nano platinum / graphene composite layer is prepared by electrochemical deposition method to enhance the electron transfer efficiency of the electrode; the heavy metal ion selective membrane is composed of chitosan-mercaptoacetic acid copolymer and Pb 2+ / Cd 2+ ion imprinted polymer is compounded in a mass ratio of 1:3 to achieve selective recognition and enrichment of specific heavy metal ions. The surface area of the working electrode is 7.07 mm 2 , effectively increasing the deposition area of heavy metal ions.

[0070] The counter electrode is a spiral platinum wire electrode with a diameter of 0.5 mm and a length of 10 mm, which is spirally wound to increase the surface area and reduce the polarization effect. The counter electrode is located on one side of the working electrode, at an angle of 120° with the working electrode, and the fixed distance is 2 mm±0.1 mm. The spiral design ensures sufficient surface area to support the current flow of the working electrode and prevents current limitation during the reaction process.

[0071] The reference electrode is an Ag / AgCl microelectrode containing 3M KCl electrolyte, with an electrode tip diameter of 0.8 mm. The reference electrode is located on the other side of the working electrode, also at an angle of 120° with the working electrode, and the distance is 2 mm±0.1 mm, forming a ring-shaped distribution structure of the three electrodes. The reference electrode provides a stable reference potential to ensure the accuracy and repeatability of the measured potential.

[0072] The electrode fixing seat is made of polytetrafluoroethylene material, which is corrosion resistant and has high chemical inertness. The fixing seat is designed with precise electrode insertion holes to ensure the accurate fixing of the relative positions of the three electrodes and prevent signal fluctuations caused by position deviation during detection. The bottom of the fixing seat is provided with a sealing ring, which is tightly connected with the detection cavity to prevent liquid leakage.

[0073] The detection cavity is a cylindrical structure with an inner diameter of 12 mm and a height of 10 mm, with a volume of about 1.1 mL, made of corrosion-resistant PEEK material. The inner wall of the cavity is treated specially to prevent heavy metal ion adsorption. The bottom of the cavity is provided with a sample injection port and a liquid discharge port, and the liquid flow is controlled by a micro electromagnetic valve to realize automatic sampling and liquid discharge.

[0074] The electrochemical control unit includes a constant potential instrument (precision ±0.1 mV) and a micro-current detection unit (range 0.1 nA-10 mA), which is responsible for controlling the electrode potential and measuring the current response. The control unit is connected with the three-electrode system through flexible flat cable to realize data acquisition and potential control.

[0075] The working process of the electrochemical detection module is as follows:

[0076] Preparation stage: Before detection, the system automatically takes the heavy metal standard solution from the standard substance warehouse for calibration, and establishes the standard curve of the dissolution peak current and the heavy metal concentration;

[0077] Sample preparation stage: Extract the test liquid from the aquatic product sample, and inject it into the detection cavity after simple filtration, and the liquid completely covers the three-electrode system;

[0078] Pre-enrichment stage: The electrochemical control unit applies a deposition potential of-1.2V on the working electrode for 300s, so that the target heavy metal ions such as Pb 2+ , Cd 2+ in the sample solution are reduced and deposited on the electrode surface. The specific binding sites on the selective membrane help to enrich the target ions and improve the detection sensitivity;

[0079] Equilibrium and standing stage: The system pauses for 30s to make the electrode interface reach a stable state and reduce the background current interference;

[0080] Dissolution scanning stage: The system performs dissolution scanning in the potential range of-1.2V to +0.5V with the parameters of square wave frequency 25Hz, potential increment 4mV, and amplitude 50mV. As the potential gradually increases, the heavy metals enriched on the surface of the working electrode are oxidized and dissolved in turn, generating characteristic dissolution peaks;

[0081] Data acquisition stage: The system continuously performs three repeated scans, records the average value of the dissolution peak current, and monitors the potential position of each characteristic peak (Pb 2+ :-0.56V±0.02V, Cd 2+ :-0.76V±0.02V);

[0082] Signal processing stage: Moving average filtering algorithm is used to eliminate the double-layer charging current, and wavelet denoising technology (db4 wavelet basis, decomposition level 5) is used to remove environmental electromagnetic interference. The Gaussian fitting algorithm is used to separate the overlapping dissolution peaks, and the peak current integral area of each heavy metal ion is accurately calculated;

[0083] Concentration calculation stage: The system converts the peak current integral area into heavy metal concentration value according to the standard curve established in advance, and automatically compares it with the threshold value of GB 2762-2017 National Food Safety Standard to generate an over-limit warning information;

[0084] Verification and correction stage: The system automatically compares the dissolution peak current difference between the main working electrode and the control electrode without ion imprinted polymer modification, corrects the matrix effect, and calculates the true heavy metal content.

[0085] After the detection is completed, the system automatically performs the electrode cleaning process:

[0086] Firstly, 0.1M HNO3 solution was injected, and electrochemical cleaning was performed at-0.2V for 30s to oxidize and dissolve the residual metal on the electrode surface;

[0087] Subsequently, deionized water was injected to flush the electrode surface and neutralize the acidic environment;

[0088] Finally, anhydrous ethanol was injected for final cleaning, and dry nitrogen was introduced to dry the electrode surface, so that the residual metal was less than 0.1ng / cm 2 , ensuring the accuracy of the next detection.

[0089] The electrochemical detection module has the following technical advantages:

[0090] Firstly, the three-electrode ring distribution design optimizes the electric field distribution and reduces the mutual interference between electrodes; secondly, the chitosan-mercaptoacetic acid / ion imprinted polymer composite film improves the selective recognition ability of heavy metal ions and reduces matrix interference; finally, the design of multiple repeated scanning combined with the comparison of the reference electrode significantly improves the accuracy and reliability of the detection. The linear detection range of the system for Pb 2+ is 0.5-200μg / kg, and the detection limit is 0.2μg / kg (S / N=3); the linear detection range for Cd 2+ is 0.3-150μg / kg, and the detection limit is 0.1μg / kg (S / N=3), which fully meets the requirements of GB 2762-2017 National Food Safety Standard, and provides reliable technical support for on-site rapid screening of heavy metal content in aquatic products.

[0091] As shown in Figure 5 , the data processing module is used to process the data transmitted from the spectral analysis module, the biosensor module and the electrochemical detection module, and output the processed data through the data output module;

[0092] The data processing module is located in the central control area of the device, including a data fusion sub-module and a machine learning analysis sub-module. The data fusion sub-module is composed of a multi-channel data acquisition card, a preprocessing unit and a feature fusion processor, which is responsible for preprocessing and feature extraction of the data transmitted by the three detection modules; the machine learning analysis sub-module is equipped with an ARM Cortex-A72 quad-core processor and an NPU acceleration unit, which runs a pre-trained deep residual network model and realizes multi-index joint analysis through 4 residual blocks.

[0093] The core innovation of the data processing module lies in the intelligent fusion algorithm of multi-source heterogeneous data. Firstly, for the 2048 wavelength points of spectral data, the improved Savitzky-Golay filter (window width adaptive adjustment of 5-25 points) is used to eliminate high-frequency noise, combined with SNV standardization to eliminate scattering effects, and then the first 10 principal components (cumulative variance contribution rate > 98%) are extracted by PCA. Secondly, for the impedance data of the biosensor, the FFT transform is used to extract the impedance real and imaginary part changes of 1 kHz, 10 kHz, and 100 kHz characteristic frequencies, and the sliding window method (window size 20 points) is used to extract the impedance change trend characteristics. Thirdly, for the electrochemical data, the db4 wavelet basis 5-layer decomposition denoising is used, combined with the Gaussian fitting algorithm to separate the overlapping stripping peaks, and four types of features are extracted, including peak position, peak height, peak area, and peak symmetry.

[0094] The deep learning model architecture of the data processing module consists of an input layer, a data preprocessing layer, a feature fusion layer, a deep residual network layer, a multi-task output layer, and a model optimization unit, forming a complete intelligent analysis system.

[0095] The input layer includes three parallel data interfaces, which receive reflectance data (2048-point spectral data in the range of 400-2500 nm) from the spectral analysis module, impedance change data (including timestamp, impedance amplitude spectrum, and characteristic frequency point change rate) from the biosensor module, and stripping peak current data (including heavy metal type, peak potential value, and peak current integral area) from the electrochemical detection module. The input layer is equipped with a data buffer, which can temporarily store 100 groups of historical detection data for model adaptive optimization.

[0096] The data preprocessing layer performs specific preprocessing operations on different types of input data. For spectral data, Savitzky-Golay filter algorithm (window width 15 points, polynomial order 3) is applied for smoothing and denoising, and standard normal transformation (SNV) is used to eliminate scattering effects. For impedance data, FFT algorithm is used for frequency domain transformation to extract impedance real and imaginary part changes at 1 kHz, 10 kHz, and 100 kHz characteristic frequency points. For electrochemical data, wavelet denoising technology (db4 wavelet basis, decomposition level 5) is used to eliminate background noise, and Gaussian fitting algorithm is used to accurately extract peak position and peak area. The preprocessing layer outputs three types of standardized feature data.

[0097] The three-level architecture of branch-fusion-residual is adopted: the first level is a dedicated feature extraction branch, and dedicated feature extraction networks are designed for spectral, biosensor and electrochemical data respectively, wherein the spectral branch adopts one-dimensional convolutional neural network to capture the local correlation between wavelengths, the biosensor branch adopts LSTM network to process time-series impedance changes, and the electrochemical branch adopts multilayer perceptron to process discretized peak features; the second level is an adaptive fusion layer, which dynamically calculates the weight coefficients of the three types of data through attention mechanism to avoid the limitations of traditional fixed weight; the third level is a deep residual analysis network, which extracts complete feature pedigree from low-level features to high-level semantic features through the cascade of four residual blocks.

[0098] The feature fusion layer is composed of a principal component analysis unit, a feature selector and a fusion processor. The principal component analysis unit performs PCA dimensionality reduction on the preprocessed spectral data, extracts the first 10 principal components, and explains 98% of the variance; the feature selector selects key features of impedance data and electrochemical data based on Fisher discriminant ratio, and selects 12 and 6 feature quantities respectively; the fusion processor weights and fuses the three features according to the weight coefficients of 0.4:0.3:0.3, and combines the sample type information and the environmental temperature compensation parameter to finally generate a 128-dimensional feature vector.

[0099] The deep residual network layer adopts a four-layer residual block structure. Each residual block contains two 3x3 convolution layers, batch normalization layers and ReLU activation functions, and introduces a skip connection structure to avoid the problem of gradient disappearance. The first residual block mainly processes spectral features, outputting 64 feature maps; the second residual block focuses on impedance data features, outputting 128 feature maps; the third residual block emphasizes electrochemical features, outputting 256 feature maps; the fourth residual block integrates the outputs of the first three blocks, and generates a 512-dimensional feature representation through a global average pooling layer. The network uses the Dropout(0.5) mechanism to prevent overfitting, and uses the attention mechanism to adaptively adjust the weights of different features.

[0100] The multi-task output layer contains three parallel task branches: the freshness classification branch divides the freshness of aquatic products into four grades of excellent (90-100 points), good (75-89 points), qualified (60-74 points) and unqualified (<60 points) through the Softmax activation function; the drug residue prediction branch calculates the probability of containing various drug residues such as antibiotics and preservatives in the sample through the Sigmoid activation function, and labels the risk level of exceeding the standard; the heavy metal content regression branch predicts the specific content of Pb 2+ , Cd 2+ and other heavy metals in μg / kg through a linear output layer.

[0101] The model optimization unit comprises an online learning module and an adaptive calibration module. The online learning module supports incremental learning function, and periodically updates the model parameters through comparison of historical detection data and laboratory verification results, thereby improving the generalization ability of the model on different types of aquatic products. The adaptive calibration module adjusts the prediction deviation of the model in real time according to the test results of the standard samples of the self-calibration system, thereby ensuring the accuracy and reliability of long-term use.

[0102] The data processing procedure is as follows:

[0103] The multi-source data receiving stage: the data processing unit synchronously receives the original data of the three detection modules through the first communication interface, the second communication interface and the third communication interface, and performs timestamp alignment to ensure the consistency of the data;

[0104] The data preprocessing stage: the preprocessing layer performs noise reduction, baseline correction and signal enhancement operations on various types of data to improve the signal-to-noise ratio. The spectral data are enhanced by differential transformation to highlight the characteristics of small changes, the impedance data are analyzed by phase analysis to eliminate the interference of capacitance, and the electrochemical data are corrected by baseline drift to improve the accuracy of the peak value;

[0105] The feature extraction and fusion stage: the system extracts key feature parameters from the preprocessed data, such as reflectivity at the spectral feature wavelength, change slope of the impedance curve, symmetry of the heavy metal elution peak, etc. After dimensionality reduction by principal component analysis, the features are weighted and fused based on prior knowledge to generate a unified feature vector;

[0106] The deep learning analysis stage: the fused feature vector is sequentially passed through four residual blocks for deep feature extraction and conversion. Each residual block focuses on capturing different levels of feature associations. The original information is preserved through the jump connection structure, while deep abstract features are extracted. Finally, high-dimensional feature representations are outputted;

[0107] The multi-task prediction stage: the system performs three prediction tasks in parallel based on the deep feature representations. Freshness score is obtained by comparison with the historical database; drug residue detection calculates the probability of exceeding the standard based on impedance change characteristics; heavy metal content prediction is quantitatively analyzed by combining elution peak characteristics and standard curve;

[0108] The result integration and output stage: the system integrates the three types of prediction results into a unified evaluation report, and compares them with the national standard threshold to generate a risk level assessment. The final detection results are displayed on the touch display screen of the data output module and uploaded to the blockchain storage platform.

[0109] Performance optimization mechanism:

[0110] Periodic calibration stage: the system performs zero-point calibration every day to adjust the model prediction deviation through standard sample testing;

[0111] Parameter updating stage: monthly verification data is summarized to update the model weight through transfer learning method to adapt to the changing characteristics of different seasons and different batches of aquatic products.

[0112] Abnormality monitoring stage: the system continuously monitors the model prediction variance, and when abnormal fluctuations occur continuously, it triggers the early warning mechanism and suggests comprehensive calibration.

[0113] The data processing module has the following technical advantages:

[0114] First, the multi-technology fusion feature extraction method fully excavates the complementary information of spectral, biosensor and electrochemical data, improving the comprehensiveness of detection;

[0115] Second, the deep residual network architecture effectively solves the problem of deep network training difficulty, and can capture complex nonlinear relationships;

[0116] Finally, the multi-task learning framework realizes the unified prediction of freshness, drug residues and heavy metal content, simplifying the decision-making process. After 2000 groups of actual aquatic product samples are trained and verified, the freshness classification accuracy rate reaches 97.8%, the drug residue detection rate reaches 96.5%, and the relative error of heavy metal content prediction is less than 5%, fully meeting the accuracy requirements of on-site rapid detection.

[0117] The data output module includes a touch display screen and a wireless transmission unit. The touch display screen displays the detection results and quality scores in real time, and the wireless transmission unit uploads the detection data to the blockchain storage platform through an encryption protocol and generates a detection report;

[0118] The data output module includes a touch display screen and a wireless transmission unit. The touch display screen is connected to the data processing unit through an HDMI 2.0 interface and displays the detection results and quality scores in real time; the wireless transmission unit is connected to the data processing unit through a USB Type-C interface and is responsible for uploading the detection data to the blockchain storage platform and generating a PDF format detection report.

[0119] Self-calibration system, the self-calibration system is built-in standard material warehouse, including freshness standard sample, drug residue standard liquid and heavy metal standard solution, through the mechanical arm automatic execution every day zero point calibration, calibration data is used to correct the light intensity drift of the spectral analysis module, the baseline offset of the biosensor module and the background current of the electrochemical detection module.

[0120] As shown in Figure 2 A method for rapid non-destructive quality detection of aquatic products, comprising the following steps:

[0121] In the first stage (sample pretreatment), the operation steps are as follows:

[0122] S1, place the water product to be tested in the detection cabin, the equipment automatically identifies the sample type (fish, shellfish or crustacean) and adjusts the detection parameters;

[0123] S2, the equipment automatically performs pre-checking, including light source stability verification, electrode state detection and system self-checking, to ensure that the equipment is in the best working state;

[0124] S3, the system calls the corresponding detection scheme, sets different detection thresholds and reference standards for different types of water products.

[0125] In the second stage (multi-module parallel detection), each functional module is started synchronously, and the following detection steps are performed:

[0126] S4, the spectral analysis module performs full-band spectral scanning (400-2500nm) to collect water product surface reflectivity distribution data. Specifically, the LED light source array irradiates the sample surface at an incident angle of 45°, the reflective probe captures the reflected light, and the grating spectrometer system decomposes the light signal into different wavelengths, and the multi-spectral sensor obtains the complete spectral curve;

[0127] S5, the biosensor module automatically injects 50μL of sample extraction solution into the microfluidic chip, and the liquid is fully mixed in the serpentine mixing channel before entering the reaction chamber and contacting the antibody-modified electrode. The piezoelectric micropump drives the sample flow at a flow rate of 0.1μL / s, and the impedance measurement unit monitors the electrode interface impedance change rate in the differential pulse voltammetry mode at a scan rate of 50mV / s, generating a drug residue detection curve;

[0128] S6, the electrochemical detection module applies a constant potential of-1.2V to the working electrode for 300s in the pre-concentration stage, so that the target heavy metal ions are reduced and deposited on the electrode surface; then in the dissolution scanning stage, a square wave frequency of 25Hz, a potential increment of 4mV, and an amplitude of 50mV are used to perform three times of square wave voltammetry scanning in the potential range of-1.2V to +0.5V, and the heavy metal dissolution peak integral area is recorded.

[0129] In the third stage (data fusion analysis), the data processing unit integrates the three-way detection data and performs the following processing steps:

[0130] S7, the data fusion sub-module applies the Savitzky-Golay filtering algorithm to smooth and denoise the spectral reflectivity data, uses the FFT algorithm to analyze the frequency domain characteristics of the impedance change rate data, and uses the wavelet denoising technology (db4 wavelet basis, decomposition level 5) to remove environmental electromagnetic interference from the dissolution peak current data;

[0131] S8, the pre-processed data is reduced in dimension by principal component analysis algorithm, 28-dimensional principal component features are extracted, and then combined with environmental temperature compensation parameters and sample type features to generate a 128-dimensional feature vector;

[0132] S9, the machine learning analysis submodule inputs the feature vector into a pre-trained deep residual network model, performs multi-index joint analysis through 4 residual blocks, and obtains freshness grade score, drug residue type and probability, and heavy metal content prediction value.

[0133] In the fourth stage (result output), the system generates a comprehensive evaluation report and displays the test results:

[0134] S10, the report generation engine automatically generates a test report containing freshness score (0-100 points), drug residue over-standard warning (normal / mild over-standard / serious over-standard), and heavy metal content threshold comparison based on the model output results;

[0135] S11, the system automatically compares the test results with the GB 2762-2017 and GB 2733-2015 national food safety standards thresholds to give a pass / fail determination;

[0136] S12, the data output module displays the test results and quality score in real time through the touch display screen, and the wireless transmission unit uploads the test data to the blockchain storage platform through an encryption protocol and generates a PDF format test report containing a risk prompt two-dimensional code.

[0137] After the test is completed, the device will also automatically perform cleaning and maintenance procedures:

[0138] S13, the biological sensing module is sequentially injected with 0.1M NaOH solution, deionized water and nitrogen for three flushing cycles to ensure that the residual rate is less than 0.5%;

[0139] S14, the electrochemical detection module is sequentially injected with 0.1M HNO3 solution, deionized water and anhydrous ethanol for electrode surface regeneration treatment to make the metal residual amount less than 0.1ng / cm 2 ;

[0140] S15, the device automatically records the test data for this time, updates the calibration parameters, and prepares for the next test.

[0141] The working principle is as follows:

[0142] First, place the water product to be tested in the detection cabin; the spectral analysis module starts the LED light source array, collects the water product surface reflection spectrum through the reflection probe, and acquires the spectral data in the 400-2500nm wave band through the multi-spectral sensor, and transmits it to the data processing unit through the USB3.0 interface;

[0143] Meanwhile, the biosensor module automatically injects 50 muL of sample extraction solution into the microfluidic chip, the antibody-modified electrode specifically binds to the drug residues in the sample, the impedance measurement unit monitors the change of electrode interface impedance, and transmits the data to the data processing unit through the RS485 protocol;

[0144] The electrochemical detection module performs square wave voltammetry scanning in the potential range of-1.2V to +0.5V, the three-electrode system detects the heavy metal elution peak current, and transmits the data to the data processing unit through the Modbus RTU protocol; the data fusion submodule of the data processing unit pre-processes and extracts features from the three-way data to generate a 128-dimensional feature vector, and the machine learning analysis submodule performs multi-index joint analysis through a deep residual network model;

[0145] Finally, the analysis results are displayed on the touch display screen of the data output module and uploaded to the blockchain storage platform through the wireless transmission unit to generate a PDF format detection report. The entire detection process is completed within 30 seconds, realizing rapid and non-destructive detection of the freshness, drug residues and heavy metal content of aquatic products.

[0146] The scheme of the present application has been verified by actual scene (sampling detection in coastal aquatic product market), the single detection error rate is less than 3%, the consistency with the laboratory detection result is 98.5%, and the present application can be widely applied to aquatic product wholesale, logistics and market supervision links.

[0147] Therefore, the present application adopts the above-mentioned method and device for rapid and non-destructive quality detection of aquatic products. Firstly, multi-technology parallel detection greatly shortens the detection period, and the single complete detection time is not more than 30 seconds. Secondly, multi-dimensional data fusion improves the accuracy and comprehensiveness of detection, and realizes one-stop evaluation of the freshness, drug residues and heavy metal content of aquatic products. Finally, automatic sample processing and data analysis reduce the operation complexity, so that non-professionals can easily complete professional-level quality detection. The method is particularly suitable for on-site rapid screening and quality grading in aquatic product circulation links, and provides an efficient and reliable technical means for ensuring aquatic product food safety.

[0148] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application but not to limit them, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that: the technical solutions of the present application can still be modified or replaced by equivalents, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A device for rapid non-destructive quality detection of aquatic products, characterized in that, The application relates to a freshness and drug residue detection system for aquatic products, which comprises the following modules: a spectrum analysis module for scanning the surface reflection spectrum of aquatic products, acquiring spectrum data in a 400-2500nm wave band and transmitting the spectrum data to a data processing unit; a biosensor module comprising a microfluidic chip, an antibody modified electrode and an impedance measurement unit, which is used for automatically injecting sample extraction liquid into the microfluidic chip, detecting the impedance change rate of the electrode interface, generating a drug residue detection curve, and transmitting the detected drug residue detection curve data to the data processing unit; an electrochemical detection module for performing pre-concentration and dissolution scanning, recording the integral area of a heavy metal dissolution peak, detecting the dissolution peak current of the heavy metal, and transmitting the detection data to the data processing unit; a data processing module for processing the data transmitted from the spectrum analysis module, the biosensor module and the electrochemical detection module, and outputting the processed data through a data output module; a data output module comprising a touch display screen and a wireless transmission unit, the touch display screen displays the analysis results in real time, and the wireless transmission unit uploads the analysis results to a block chain storage platform through an encryption protocol and generates a detection report; a self-calibration system, the self-calibration system is internally provided with a standard substance bin comprising a freshness standard sample, a drug residue standard liquid and a heavy metal standard solution, and a mechanical arm is used for automatically performing zero-point calibration every day, and the calibration data are used for correcting the light intensity drift of the spectrum analysis module, the baseline deviation of the biosensor module and the background current of the electrochemical detection module; the data processing module comprises a data fusion sub-module and a machine learning analysis sub-module, the data fusion sub-module performs normalization processing on the spectrum data, the impedance change rate data and the dissolution peak current data and generates a feature vector, and the machine learning analysis sub-module classifies and regresses the feature vector through a pre-trained deep neural network model to obtain analysis results comprising a freshness grade score, a drug residue type and probability and a heavy metal content prediction value.

2. The device for rapid non-destructive quality detection of aquatic products according to claim 1, characterized in that, The microfluidic chip drives sample liquid flow through a micropump, triggers an antibody-antigen specific binding reaction, and the impedance measurement unit monitors the impedance change rate of the electrode interface in real time, generates drug residue detection curve data and transmits the data to the data processing unit; The microfluidic chip is composed of a PDMS base layer and a glass substrate, and comprises a serpentine mixing channel, a reaction chamber and a waste pool; The antibody modified electrode adopts a nano-gold / graphene composite substrate, and a chloramphenicol monoclonal antibody is fixed on the surface through an EDC / NHS coupling method; The impedance measurement unit is integrated with an electrochemical workstation.

3. The device for rapid non-destructive quality detection of aquatic products according to claim 1, characterized in that, The electrochemical detection module adopts a three-electrode system, and comprises a working electrode, a counter electrode and a reference electrode, the surface of the working electrode is modified with a heavy metal ion selective film, and the heavy metal dissolution peak current is detected through a square wave anodic stripping voltammetry method.

4. The device for rapid non-destructive quality detection of aquatic products according to claim 1, characterized in that, The data processing steps of the data processing module and the data output module include: a baseline calibration stage: PBS buffer liquid is injected through a micropump, an initial impedance spectrum is recorded, and baseline equivalent circuit parameters are established; a dynamic detection stage: sample liquid flows through the surface of the antibody modified electrode, antigen-antibody binding causes double-layer capacitance change, and the system continuously collects impedance amplitude and phase data in multiple cycles. Signal processing stage: Savitzky-Golay filter algorithm is used to eliminate high-frequency noise, and the impedance real part change of characteristic frequency point is extracted by sliding window method to generate three-dimensional impedance fingerprint spectrum; Specificity verification mechanism: a non-specific antibody modified control electrode is set, and the impedance change difference between the main detection site and the control site is used to eliminate matrix interference; Multi-source data receiving stage: the data processing unit synchronously receives the original data of the spectrum analysis module, the biological sensing module and the electrochemical detection module through the first communication interface, the second communication interface and the third communication interface, and performs timestamp alignment to ensure data consistency; The second communication interface adopts RS485 protocol to transmit structured data packets containing timestamp, impedance amplitude spectrum and characteristic frequency point change rate at 115200bps; The microfluidic chip is equipped with an automatic cleaning process, and after detection, 0.1M NaOH solution, deionized water and nitrogen are injected in turn for three flushing cycles, and the residual rate is less than 0.5%.

5. The device for rapid non-destructive quality detection of aquatic products according to claim 3, characterized in that, The working electrode of the three-electrode system is a glassy carbon electrode with a diameter of 3mm, and the surface is modified with a nano platinum / graphene composite layer and a heavy metal ion selective membrane in sequence; The selective membrane is made of chitosan-mercaptoacetic acid copolymer and The ion imprinted polymer is compounded at a mass ratio of 1:

3.

6. A method for rapid non-destructive quality detection of aquatic products, characterized in that, The device as claimed in any one of claims 1-5 is used, comprising the following steps: S1, placing the water product to be tested in the detection cabin, performing full-band spectrum scanning by the spectrum analysis module to obtain surface reflectivity distribution data; S2, the biological sensing module automatically injects sample extraction liquid into the microfluidic chip, monitors the impedance change rate of the antibody modified electrode, and generates a drug residue detection curve; S3, the electrochemical detection module performs three times of square wave voltammetry scanning in the potential range of-1.2V to +0.5V, and records the heavy metal dissolution peak integral area; S4, the data processing module extracts and fuses the spectrum data, impedance change rate data and dissolution peak integral area data, inputs a deep neural network model for multi-index joint analysis; S5, according to the analysis result output by the data processing module, a detection report containing freshness score, drug residue over-limit warning and heavy metal content threshold comparison is generated, and displayed and stored through the data output module.

Images