Leafy vegetable cleaning and packaging process monitoring system and method based on big data

Abstract

This invention discloses a big data-based monitoring system and method for the cleaning and packaging process of leafy vegetables, relating to the fields of image data processing, agricultural product processing, and food safety. The method acquires real-time pesticide residue spectra and three-dimensional morphology data from the surface of leafy vegetables, analyzing the concentrations of water-soluble and fat-soluble pesticide residues and the leaf wrinkling index. Based on this, it assesses the dependence on physical rinsing and chemical cleaning, and optimizes the pressure and coverage duration parameters of high-pressure water mist by combining the leaf structure vulnerability index. Finally, it generates and executes a collaborative cleaning scheme that integrates cleaning needs and leaf protection, achieving personalized and precise control and adaptive optimization of the cleaning process, significantly improving cleaning effectiveness, efficiency, and food safety.

Description

Technical Field

[0001] This invention relates to the fields of image data processing, agricultural product processing and food safety technology, and specifically to a big data-based monitoring system and method for the cleaning and packaging process of leafy vegetables. Background Technology

[0002] As consumers' demands for food safety and quality continue to rise, the effectiveness of pesticide residue removal during the post-harvest washing of leafy vegetables has become a focus of attention. Existing leafy vegetable washing processes typically employ fixed procedures and parameters, such as high-pressure water mist rinsing for a uniform duration or soaking in surfactants at a constant concentration, which have significant shortcomings.

[0003] On the one hand, existing methods neglect the crucial impact of individual differences in leafy vegetables on cleaning effectiveness. Different batches and varieties of leafy vegetables exhibit significant variations in surface wrinkling, pesticide adhesion types, and initial residue concentrations. Cleaning processes with fixed parameters cannot accommodate these differences, potentially leading to incomplete removal of pesticides deep within the wrinkles or over-cleaning that damages the leafy vegetable tissue and causes nutrient loss. On the other hand, monitoring and adjustment of the cleaning process are disconnected. Traditional methods primarily rely on final sampling and testing to determine cleaning effectiveness, lacking real-time monitoring and feedback control of pesticide residue decay rates during the cleaning process. This results in low cleaning efficiency, uneconomical consumption of water and chemical reagents, and potential food safety risks.

[0004] Therefore, there is a need for an intelligent monitoring method that can sense the individual characteristics of leafy vegetables and the status of pesticide residues in real time, and dynamically optimize the cleaning strategy based on the data to achieve precise, efficient and safe cleaning. Summary of the Invention

[0005] To overcome the aforementioned deficiencies in the existing technology, this invention provides a big data-based monitoring system and method for the leafy vegetable washing and packaging process, aiming to achieve personalized and precise control and process optimization of the leafy vegetable washing process.

[0006] In a first aspect, the present invention provides a method for monitoring the leafy vegetable washing and packaging process based on big data, comprising the following steps:

[0007] S1: Real-time acquisition of pesticide residue spectral data and leaf morphology data of leafy vegetables to be cleaned, and analysis to obtain water-soluble pesticide residue concentration data, fat-soluble pesticide residue concentration data and leaf wrinkling index;

[0008] S2: Calculate the degree of dependence of leafy vegetables on physical rinsing and washing based on water-soluble pesticide residue concentration data and leaf wrinkling index; calculate the degree of dependence of leafy vegetables on chemical solvent washing based on fat-soluble pesticide residue concentration data and leaf wrinkling index.

[0009] S3: Determine the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit based on the degree of dependence on physical rinsing cleaning; determine the initial concentration parameters and action duration parameters of the surfactant cleaning unit based on the degree of dependence on chemical solvent cleaning.

[0010] S4: Based on the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit, and the initial concentration parameters and action duration parameters of the surfactant cleaning unit, generate a collaborative execution plan for the composite cleaning process.

[0011] S5: Based on leaf surface morphology data and leaf wrinkle index, calculate the leaf structure vulnerability index, and optimize the parameters of the high-pressure water mist cleaning unit in the collaborative execution scheme according to the degree of dependence on physical scouring and cleaning and the leaf structure vulnerability index, to obtain optimized pressure parameters and optimized coverage duration parameters.

[0012] S6: Update the collaborative execution plan using optimized pressure parameters and optimized coverage duration parameters, and control the cleaning equipment to execute the updated collaborative execution plan to complete the cleaning.

[0013] According to the above technical solution, the real-time acquisition of pesticide residue spectral data and leaf surface morphology data of the leafy vegetables to be washed, and the analysis to obtain water-soluble pesticide residue concentration data, fat-soluble pesticide residue concentration data, and leaf wrinkling index, includes:

[0014] The surface of leafy vegetables was scanned using a near-infrared spectrometer to obtain pesticide residue spectral data. Based on the pesticide residue spectral data, a matching analysis was performed with a preset characteristic spectral library of water-soluble pesticides and a characteristic spectral library of fat-soluble pesticides to extract the concentration data of water-soluble pesticide residues and fat-soluble pesticide residues. The surface morphology data of the leafy vegetables was obtained using a three-dimensional contour scanner, and the leaf wrinkle index was calculated based on the surface morphology data.

[0015] According to the above technical solution, the calculation of the dependence of leafy vegetables on physical rinsing and cleaning based on water-soluble pesticide residue concentration data and leaf wrinkling index, and the calculation of the dependence of leafy vegetables on chemical solvent cleaning based on fat-soluble pesticide residue concentration data and leaf wrinkling index, include:

[0016] The degree of physical rinsing and cleaning requirement is determined based on water-soluble pesticide residue concentration data, and then corrected by leaf wrinkling index to obtain the degree of dependence on physical rinsing and cleaning. The degree of chemical solvent cleaning requirement is determined based on fat-soluble pesticide residue concentration data, and then corrected by leaf wrinkling index to obtain the degree of dependence on chemical solvent cleaning.

[0017] According to the above technical solution, the determination of the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit based on the degree of dependence on physical rinsing cleaning, and the determination of the initial concentration parameters and action duration parameters of the surfactant cleaning unit based on the degree of dependence on chemical solvent cleaning, include:

[0018] The dependence on physical rinsing is mapped to a preset physical cleaning parameter space, and the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit are obtained by interpolation calculation. The dependence on chemical solvent cleaning is input into a preset multi-objective parameter optimization model, and the model is solved with the objectives of maximizing cleaning efficiency and minimizing chemical residue to obtain the initial concentration parameters and action duration parameters of the surfactant cleaning unit.

[0019] According to the above technical solution, the collaborative execution scheme for generating a composite cleaning process based on the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit, and the initial concentration parameters and action duration parameters of the surfactant cleaning unit, includes:

[0020] A process scheduling model is established with the goal of minimizing the total cleaning time. The initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit, as well as the initial concentration parameters and action duration parameters of the surfactant cleaning unit, are used as inputs to the process scheduling model. The process execution sequence is solved to generate a collaborative execution scheme for the composite cleaning process.

[0021] According to the above technical solution, the leaf surface morphology data and leaf wrinkle index are used to calculate the leaf structure vulnerability index. Based on the dependence on physical scouring and cleaning and the leaf structure vulnerability index, the parameters of the high-pressure water mist cleaning unit in the collaborative execution scheme are optimized to obtain optimized pressure parameters and optimized coverage duration parameters, including:

[0022] First, based on the triangular mesh model of the leafy vegetable surface contained in the leaf surface morphology data, the distance from each vertex to the main plane of the leaf is calculated to obtain a distance set. The values ​​in the distance set are sorted, and the values ​​located at the preset high percentile position after sorting are selected as the leaf thickness parameter. Second, based on the leaf wrinkling index and the leaf thickness parameter, the leaf structure vulnerability index is calculated by weighted summing the quotients obtained by dividing the leaf wrinkling index by the baseline thickness value and the leaf thickness parameter. Finally, the physical scouring and cleaning dependence degree and the leaf structure vulnerability index are used as joint input conditions to query a predefined cleaning parameter optimization mapping table. The cleaning parameter optimization mapping table stores the optimized pressure parameters and optimized coverage duration parameters corresponding to different combinations of physical scouring and cleaning dependence degree and different leaf structure vulnerability indices. By matching the physical scouring and cleaning dependence degree and leaf structure vulnerability index combination corresponding to the current input conditions, the corresponding optimized pressure parameters and optimized coverage duration parameters are obtained.

[0023] According to the above technical solution, the step of updating the collaborative execution scheme using optimized pressure parameters and optimized coverage duration parameters, and controlling the cleaning equipment to execute the updated collaborative execution scheme to complete the cleaning, includes:

[0024] The optimized pressure parameters and optimized coverage duration parameters are used to replace the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit in the collaborative execution scheme; the cleaning equipment is then controlled to execute the updated collaborative execution scheme to complete the cleaning.

[0025] Secondly, this invention provides a big data-based monitoring system for the leafy vegetable washing and packaging process, used to implement the above method, the system comprising:

[0026] The data acquisition module is used to acquire pesticide residue spectral data and leaf morphology data of the leafy vegetables to be washed in real time, and analyze them to obtain water-soluble pesticide residue concentration data, fat-soluble pesticide residue concentration data and leaf wrinkling index.

[0027] The dependency assessment module is used to calculate the degree of dependence of leafy vegetables on physical rinsing and washing based on water-soluble pesticide residue concentration data and leaf wrinkling index; and to calculate the degree of dependence of leafy vegetables on chemical solvent washing based on fat-soluble pesticide residue concentration data and leaf wrinkling index.

[0028] The parameter determination module is used to determine the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit based on the degree of dependence on physical rinsing cleaning; and to determine the initial concentration parameters and action duration parameters of the surfactant cleaning unit based on the degree of dependence on chemical solvent cleaning.

[0029] The scheme generation module is used to generate a collaborative execution scheme for the composite cleaning process based on the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit, and the initial concentration parameters and action duration parameters of the surfactant cleaning unit.

[0030] The execution monitoring module is used to calculate the leaf structure vulnerability index based on leaf surface morphology data and leaf wrinkle index, and optimize the parameters of the high-pressure water mist cleaning unit in the collaborative execution scheme according to the degree of dependence on physical rinsing and cleaning and the leaf structure vulnerability index, so as to obtain optimized pressure parameters and optimized coverage duration parameters.

[0031] The dynamic calibration module is used to update the collaborative execution scheme with optimized pressure parameters and optimized coverage duration parameters, and to control the cleaning equipment to execute the updated collaborative execution scheme to complete the cleaning.

[0032] Thirdly, the present invention provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the program to implement the above-mentioned big data-based monitoring method for leafy vegetable washing and packaging processes.

[0033] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the above-mentioned big data-based monitoring method for leafy vegetable washing and packaging processes.

[0034] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0035] The innovation of this invention lies in introducing a dynamic decision-making and closed-loop control mechanism that incorporates the perception of leafy vegetable characteristics into the leafy vegetable cleaning process. By acquiring and analyzing the type, concentration, and surface morphology of pesticide residues on leafy vegetables in real time, the traditional fixed-parameter cleaning process is transformed into a personalized cleaning plan tailored to the specific conditions of each batch of leafy vegetables. By quantitatively assessing the dependence of physical rinsing on chemical solvent cleaning, and accordingly precisely setting the initial parameters of high-pressure water mist and surfactants, on-demand allocation of cleaning resources is achieved. More importantly, by monitoring the decay rate of pesticide residues during the cleaning process online and comparing it with safety thresholds, a closed-loop control system of perception, decision-making, execution, feedback, and calibration is formed, ensuring that the cleaning process always proceeds in a highly efficient and safe direction. This not only significantly improves the pesticide residue removal rate and ensures food safety, but also effectively avoids damage to leafy vegetables and resource waste caused by over-cleaning, improving the economy and intelligence level of the cleaning process. Attached Figure Description

[0036] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0037] Figure 1 This is a schematic diagram of the overall process of the leafy vegetable cleaning and packaging process monitoring method based on big data provided in the embodiments of this application;

[0038] Figure 2 This is a detailed diagram of the data acquisition and parsing process provided in the embodiments of this application;

[0039] Figure 3 This is a detailed diagram of the cleaning dependency assessment process provided in the embodiments of this application;

[0040] Figure 4 This is a detailed flowchart of the cleaning parameter determination process provided in the embodiments of this application;

[0041] Figure 5 This is a detailed diagram of the collaborative solution generation process provided in the embodiments of this application;

[0042] Figure 6 This is a detailed flowchart of the cleaning parameter optimization process provided in the embodiments of this application;

[0043] Figure 7 This is a detailed diagram of the optimized cleaning execution process provided in the embodiments of this application;

[0044] Figure 8 This is a schematic diagram of the structure of the leafy vegetable cleaning and packaging process monitoring system based on big data provided in the embodiments of this application. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail and completely below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this invention.

[0046] Please see Figure 1 , Figure 1 This is a schematic diagram of the overall process of the leafy vegetable washing and packaging process monitoring method based on big data provided in the embodiments of this application, which specifically includes the following steps:

[0047] S1: Real-time acquisition of pesticide residue spectral data and leaf morphology data of leafy vegetables to be cleaned, and analysis to obtain water-soluble pesticide residue concentration data, fat-soluble pesticide residue concentration data and leaf wrinkling index;

[0048] In this embodiment, step S1 includes the following specific contents, and the process can be found in the attached document. Figure 2 , Figure 2 This is a detailed diagram of the data acquisition and parsing process provided in the embodiments of this application:

[0049] S110: Obtain pesticide residue spectral data by scanning the surface of leafy vegetables with a near-infrared spectrometer;

[0050] In this embodiment, the specific steps include:

[0051] A1. An online near-infrared spectroscopy scanning device is installed above the initial station of the cleaning production line. When the conveyor belt carrying leafy vegetables passes through the sensing area directly below at a uniform speed, the online near-infrared spectroscopy scanning device starts to work. The halogen tungsten lamp light source inside the online near-infrared spectroscopy scanning device emits near-infrared light covering wavelengths from 1,000 nanometers to 2,500 nanometers. This beam is first modulated by an interferometer and then irradiates the surface of the leafy vegetables. The light reflected back from the surface of the leafy vegetables is collected by the optical probe and guided to the indium gallium arsenide detector.

[0052] A2. The indium gallium arsenide detector converts the intensity of the received light signal into a weak analog current signal proportionally, completing the photoelectric conversion process. This analog current signal is then sent to a high-precision analog-to-digital converter. The high-precision analog-to-digital converter samples the continuous current signal at a fixed frequency of 100,000 times per second, measures the instantaneous voltage value of the current at each sampling moment, and converts this instantaneous voltage value into a discrete integer value between zero and 167,721. This series of integer values ​​generated in time sequence constitutes the digital sequence of the original reflected light signal.

[0053] A3. Perform digital filtering on the digital sequence obtained in step A2 to eliminate noise: Apply a pre-designed low-pass filter with a cutoff frequency of 100 Hz, convolve the digital sequence with the coefficients of the filter to attenuate high-frequency random fluctuations in the signal; convolve the original digital sequence with the coefficients of the low-pass filter to attenuate high-frequency random fluctuations in the signal, and obtain the filtered digital sequence.

[0054] A4. Apply the standard normal variable transformation algorithm to the filtered digital sequence obtained in step A3 for scattering correction;

[0055] A5. First, calculate the arithmetic mean of all data points in the entire filtered digital sequence;

[0056] A6. At the same time, calculate the standard deviation of the number sequence;

[0057] A7. Next, for each data point in the number sequence, subtract the arithmetic mean calculated in step A5 from its value.

[0058] A8. Then, divide the difference obtained in step A7 by the standard deviation calculated in step A6 to generate a new, standardized value for each data point.

[0059] A9. Arrange the series of standardized values ​​generated in the above steps according to their original time sequence to form a corrected number sequence.

[0060] A10. Map the sequence number of each data point in this corrected digital sequence to the corresponding wavelength value according to the sampling rate of the analog-to-digital converter and the wavelength scanning range of the spectrometer.

[0061] A11. Using the standardized numerical sequence obtained in step A9 as the ordinate and the wavenumber sequence obtained in step A10 as the abscissa, together they form a discrete spectral data.

[0062] A12. Perform linear interpolation and smoothing on this discrete spectral data to generate a continuous, smooth near-infrared spectral curve; output this spectral curve as pesticide residue spectral data characterizing the chemical composition of leafy vegetable surfaces for subsequent analysis.

[0063] S120: Based on pesticide residue spectral data, the water-soluble pesticide residue concentration data and the fat-soluble pesticide residue concentration data are analyzed by matching the preset water-soluble pesticide concentration prediction model and the fat-soluble pesticide concentration prediction model.

[0064] In this embodiment, the collected spectral data are analyzed using two independently running quantitative analysis models. The first model is a water-soluble pesticide concentration prediction model, specifically used to predict the total residual concentration of water-soluble pesticides. The second model is a fat-soluble pesticide concentration prediction model, specifically used to predict the total residual concentration of fat-soluble pesticides.

[0065] The process of constructing the water-soluble pesticide concentration prediction model is as follows:

[0066] B1. Ten representative water-soluble pesticides were selected as target analytes, and standard working solutions with five concentration gradients of 0.1 mg / kg, 0.5 mg / kg, 1 mg / kg, 5 mg / kg and 10 mg / kg were prepared.

[0067] B2. Select fresh spinach leaves with intact surfaces as the substrate. After cleaning and confirming that there is no pesticide residue, use a micro sprayer to evenly spray the above-mentioned standard working solutions of different types and concentrations on both sides of the leaves. Dry them at room temperature in the dark to prepare 50 standardized training samples.

[0068] B3. Using the same online near-infrared spectroscopy scanning device as the production line, collect the spectral data of these 50 training samples with consistent parameters in a constant temperature and humidity dark room;

[0069] B4. After each sprayed leaf sample was cut, ground, extracted and purified, the residues of ten water-soluble pesticides were determined by gas chromatography-mass spectrometry and the results were summed to obtain the reference value of the total concentration of water-soluble pesticides for each sample.

[0070] B5. A training set is formed by combining the spectral data matrix of fifty training samples with the corresponding total concentration reference value vector, and then input into a partial least squares regression algorithm for training. This least squares regression algorithm extracts features from the full spectrum through an iterative calculation process: First, the spectral data and concentration reference values ​​are centered and standardized; then, ten iterations are performed. In each iteration, the least squares regression algorithm calculates a weight vector, projects the spectral data onto a new direction to obtain a score vector, which can explain the covariance between the spectral data and the concentration reference value to the greatest extent; at the same time, the least squares regression algorithm calculates the loading vector and regression coefficients; after each iteration, the explained part is subtracted from the original spectral data, and the residuals are updated; after ten iterations, ten latent variables are obtained, along with the linear regression equations between the scores of these latent variables and the concentration reference values, which together constitute the water-soluble pesticide concentration prediction model.

[0071] B6. Evaluate the model's prediction accuracy using an independent validation set and adjust parameters to prevent overfitting. Specifically, prepare fifteen leafy vegetable samples with known reference values ​​for total water-soluble pesticide concentrations as a validation set. Input the spectral data of the validation set samples into the trained model to obtain predicted concentration values. Calculate the coefficient of determination and root mean square error between the predicted concentration values ​​and the true reference values. If the evaluation metrics show that the prediction error is too large, adjust the model parameters, reduce the number of extracted latent variables, and then re-execute the training and validation process until the model achieves stable and reliable prediction accuracy on the validation set.

[0072] The construction of a concentration prediction model for fat-soluble pesticides follows the same technical principles and procedures, but uses a completely independent dataset. The specific steps are as follows:

[0073] B7. Select eight representative lipid-soluble pesticides as target analytes and prepare standard working solutions with five concentration gradients independently according to the method described in B1.

[0074] B8. Using the same matrix and spraying method as B2, 40 fat-soluble pesticide training samples were prepared independently.

[0075] B9. Using the same spectroscopic apparatus and parameters as in B3, collect spectral data for these 40 samples;

[0076] B10. Following the same extraction and detection procedure described in B4, determine the residue levels of eight fat-soluble pesticides in each sample and sum them to obtain a reference value for the total concentration of fat-soluble pesticides.

[0077] B11. The spectral data and concentration reference values ​​of forty samples are combined into an independent training set. The partial least squares regression algorithm is input and the same training process described in B5 is executed independently. Eight latent variables are extracted and corresponding linear regression equations are established to form a lipid-soluble pesticide concentration prediction model.

[0078] B12. Following the same method described in B6, validate and fine-tune the model using an independent validation set of lipid-soluble pesticides to ensure that its prediction accuracy meets the requirements.

[0079] On a real-time cleaning production line, the process of analyzing spectral data is performed according to the following steps:

[0080] B13. Input the preprocessed near-infrared spectral data into the two pre-trained and deployed prediction models mentioned above.

[0081] B14. After receiving the input spectrum, the water-soluble pesticide concentration prediction model first projects the spectral data onto the ten weight vectors trained and stored in step B5 to obtain ten projection scores. Then, it multiplies these ten scores by the corresponding regression coefficients calculated and stored in step B5, sums all the product results, and finally adds the constant term obtained from model training to calculate and output a value, which is the water-soluble pesticide residue concentration data on the surface of the currently detected leafy vegetables.

[0082] B15. Simultaneously, the fat-soluble pesticide concentration prediction model runs in parallel. This fat-soluble pesticide concentration prediction model receives the same spectral data, but uses the eight weight vectors and their corresponding regression coefficients and constant terms that were independently trained and stored in step B11 to perform the same projection, weighted summation and constant term calculation process as in B14, and finally outputs another independent value, which represents the current fat-soluble pesticide residue concentration data on the surface of the leafy vegetables.

[0083] Finally, data on the concentrations of water-soluble pesticide residues and fat-soluble pesticide residues were obtained.

[0084] S130: Obtain leafy vegetable surface morphology data through a 3D contour scanner, and calculate the leaf wrinkle index based on the leafy vegetable surface morphology data;

[0085] In this embodiment, the specific steps include:

[0086] C1. At the same station of spectral scanning, a three-dimensional contour scanning device consisting of a line laser generator, an industrial area array camera, and a control unit is synchronously triggered; the line laser generator emits a beam of 650 nanometers red laser, which is spread into a thin laser line by a cylindrical lens and projected onto the surface of the leafy vegetable passing by at a uniform speed at an incident angle of 60 degrees; the lens axis of the industrial camera is at a 30-degree angle to the laser plane, and synchronously captures the laser stripe image reflected from the surface of the leafy vegetable and deformed by its shape at a rate of 100 frames per second.

[0087] C2. Perform preprocessing on each acquired image frame: First, perform grayscale conversion, then apply a mean value filtering algorithm to remove image noise; use the grayscale centroid method to extract the laser center line from the preprocessed image: for each column of pixels in the image, calculate the weighted average of the grayscale values ​​of all pixels in that column and the row coordinates, thereby determining the sub-pixel precision coordinates of the laser stripe center in that column, and obtaining the two-dimensional coordinate sequence of the entire center line;

[0088] C3. Using the camera's internal parameter matrix, distortion coefficients, and laser plane equations obtained in advance through high-precision calibration, the two-dimensional coordinate sequence obtained in step C2 is corrected to eliminate the influence of lens distortion; based on the principle of triangulation, three-dimensional coordinates are calculated: each corrected two-dimensional pixel is connected to the camera's optical center to form a line-of-sight vector, and the line-of-sight vector is spatially intersected with the known laser plane equation to calculate the three-dimensional spatial coordinates corresponding to each pixel.

[0089] C4. Perform the calculation of step C3 on all pixels on the laser center line in a single frame image to obtain a three-dimensional point set that constitutes a contour line on the surface of the leafy vegetable; register and fuse the point sets calculated in multiple consecutive frames according to the conveyor belt speed to finally form dense three-dimensional point cloud data covering the scanned area.

[0090] C5. The dense 3D point cloud data obtained in step C4 is converted into a continuous surface model using the Poisson surface reconstruction algorithm: First, the outward normal vector of each data point is estimated by principal component analysis, and then an octree spatial grid covering the point cloud is constructed; the position and normal vector information of all data points are converted into the influence weights of the surrounding grid vertices, thereby defining a vector field pointing into the point cloud on all vertices of the octree grid.

[0091] C6. By solving a large sparse linear system of equations, find a scalar function defined on the grid such that its gradient value at each vertex is closest to the vector field defined in step C5; after solving, obtain the scalar function value of each grid vertex; select an appropriate isosurface threshold in the three-dimensional data field composed of the scalar values ​​of the grid vertices, extract all points equal to the threshold to form isosurfaces, and output them as a smooth, water-tight triangular mesh model of the leafy vegetable surface as the final surface morphology data;

[0092] C7. Identify wrinkled regions based on triangular mesh model: Calculate the average curvature of each triangular facet and mark all facets with an absolute curvature value greater than a set threshold of 0.15 mm to the power of -1 as wrinkled regions.

[0093] C8. Calculate key geometric parameters: sum up the areas of all folded facets to get the total surface area of ​​the folded region; project the overall model vertically onto the horizontal plane and calculate the area enclosed by the projected outline as the apparent projected area; calculate the average height of all vertices in the folded region to the bottom plane as the average fold depth; calculate the average dimension of the model in the vertical direction as the average thickness of the blade.

[0094] C9. Comprehensive calculation of leaf wrinkling index: Divide the total surface area of ​​the wrinkled area by the apparent projected area to obtain the area expansion ratio; divide the average wrinkle depth by the average leaf thickness to obtain the depth-thickness ratio; multiply the area expansion ratio and the depth-thickness ratio to obtain a dimensionless value that comprehensively characterizes the complexity of leaf wrinkles, namely the leaf wrinkling index.

[0095] S2: Calculate the degree of dependence of leafy vegetables on physical rinsing and washing based on water-soluble pesticide residue concentration data and leaf wrinkling index; calculate the degree of dependence of leafy vegetables on chemical solvent washing based on fat-soluble pesticide residue concentration data and leaf wrinkling index.

[0096] In this embodiment, step S2 includes the following specific details, which can be found in the flowchart below. Figure 3 , Figure 3 Here is a detailed diagram of the cleaning dependency assessment process provided in the embodiments of this application:

[0097] S210: Determine the degree of physical rinsing and cleaning requirement based on water-soluble pesticide residue concentration data, and correct the degree of physical rinsing and cleaning requirement by combining the leaf wrinkle index to obtain the degree of dependence on physical rinsing and cleaning.

[0098] To quantify the dependence on high-pressure physical flushing by combining the concentration of water-soluble pesticide residues and the leaf wrinkling index, this embodiment constructs a physical flushing dependence conversion model. The construction and calculation method of this physical flushing dependence conversion model are as follows:

[0099] D1. Extract all high-pressure water mist cleaning operation records completed within the past year from the production management database. Each record includes the operation number, the initial concentration of water-soluble pesticides detected before cleaning, the pressure gauge reading of the water pump used for cleaning, and the duration of the cleaning process in seconds. Link the data to the quality inspection database to obtain the pesticide removal rate test report after cleaning for each record. This report is issued by the quality inspection department according to standard testing methods. Preprocess the extracted data to remove invalid data, specifically including records with missing removal rate data and records with pressure values ​​exceeding the upper or lower limits of the equipment's normal operating range.

[0100] D2. Input the preprocessed data into a statistical analysis module; this statistical analysis module identifies all cleaning records where the pesticide removal rate reaches more than 95% for each initial concentration value of water-soluble pesticide that has appeared in the historical data.

[0101] D3. For all successful cleaning records with the same initial concentration found in step D2, calculate the product of the water pump pressure gauge reading and the cleaning duration in seconds for each record to obtain a comprehensive effect value that characterizes the energy input of the cleaning.

[0102] D4. For each initial concentration value, calculate the median of the combined effect value set of all successful cleaning records; this median represents a typical cleaning energy value required to achieve the standard removal rate at that specific initial concentration.

[0103] D5. For all different initial concentration values ​​that have appeared in the historical data, repeat steps D3 and D4 to obtain a series of paired data consisting of initial concentration values ​​and corresponding typical cleaning energy values.

[0104] D6. Use the least squares method to perform linear regression analysis on the series of paired data obtained in step D5, and fit a straight line.

[0105] D7. The slope of the fitted line obtained in step D6 is defined as the cleaning energy conversion coefficient of water-soluble pesticides. Its physical meaning is the typical increase in cleaning energy required when the concentration of water-soluble pesticides increases by one unit.

[0106] D8. For any newly input data on the concentration of water-soluble pesticide residues in leafy vegetables, multiply by the cleaning energy conversion coefficient obtained in step D7 to directly calculate a linear physical rinsing cleaning requirement baseline value; combine the leaf wrinkling index to correct the above physical rinsing cleaning requirement baseline value to obtain the physical rinsing cleaning dependence value; the correction is achieved through an exponential function, which has the natural constant e as the base and the exponent term is the leaf wrinkling index minus one; multiply the calculated physical rinsing cleaning requirement baseline value with this exponential function value to obtain the corrected physical rinsing cleaning dependence value;

[0107] D9. Normalize the corrected physical rinsing and cleaning dependence value, mapping it to the range of zero to one, to obtain the final physical rinsing and cleaning dependence. The normalization method is as follows: collect all possible corrected physical rinsing and cleaning dependence values ​​in history, determine the maximum and minimum values; subtract the minimum value from the currently calculated corrected physical rinsing and cleaning dependence value, and then divide by the difference between the maximum and minimum values ​​to obtain the final physical rinsing and cleaning dependence. The closer the physical rinsing and cleaning dependence value is to one, the higher the current dependence of the leafy vegetables on high-pressure water mist cleaning.

[0108] S220: The degree of chemical solvent cleaning requirement is determined based on the concentration data of fat-soluble pesticide residues, and the degree of chemical solvent cleaning requirement is corrected by combining the leaf wrinkling index to obtain the degree of chemical solvent cleaning dependence.

[0109] To quantify the dependence on surfactant-based chemical cleaning by combining the concentration of fat-soluble pesticide residues and the leaf wrinkling index, this embodiment constructs a chemical solvent cleaning dependence conversion model; the construction and calculation method of this chemical solvent cleaning dependence model are as follows:

[0110] E1. Extract all cleaning experiment records for fat-soluble pesticides from the chemical cleaning process experimental database. Each record includes the experiment number, the initial concentration of the fat-soluble pesticide measured before cleaning, the concentration of the surfactant solution used for cleaning, the soaking time, and the removal rate of the fat-soluble pesticide measured after cleaning. Preprocess the extracted data to remove records with abnormal cleaning conditions and obvious outliers. From all experimental records, select successful cleaning records with a final pesticide removal rate of 90%.

[0111] E2. For each initial concentration value of a fat-soluble pesticide that appears in a successful cleaning record, find all successful cleaning records corresponding to this concentration; for these records, calculate the product of the surfactant solution concentration value and the action time value to obtain a value characterizing the energy input of a single chemical cleaning experiment.

[0112] E3. Calculate the arithmetic mean of all successfully recorded chemical cleaning energy values ​​at the same initial concentration in step E2; this arithmetic mean represents a typical chemical cleaning energy required to achieve effective cleaning at that concentration.

[0113] E4. Repeat step E3 for all the initial concentration values ​​that have appeared to obtain a series of paired data of initial concentration values ​​and corresponding typical chemical cleaning energies;

[0114] E5. Perform linear regression analysis on the paired data obtained in step E4 using the least squares method to fit a straight line; the slope of this line, that is, the typical chemical cleaning energy required for each unit increase in the concentration of fat-soluble pesticide, is defined as the cleaning energy conversion coefficient of fat-soluble pesticide.

[0115] E6. For newly input data on the concentration of fat-soluble pesticide residues in leafy vegetables, multiply it by the energy conversion coefficient obtained in step E5 to calculate a linear baseline value for the degree of chemical solvent cleaning required.

[0116] E7. Determine the power function exponent for wrinkle correction against fat-soluble pesticides through a specialized experiment. This experiment selects leafy vegetable samples of the same variety with different wrinkle indices, applies the same type and concentration of fat-soluble pesticide, and then washes them under the same surfactant concentration and soaking time conditions. After washing, the pesticide removal rate of each group of samples is accurately measured. The law of change of removal rate with leaf wrinkle index is analyzed. Through nonlinear curve fitting, it is determined that the degree of decrease in removal rate is inversely proportional to a certain power of the wrinkle index. The power obtained by fitting is the exponent required for the correction model.

[0117] E8. Correct the baseline value of chemical solvent cleaning requirement obtained in step E6 by combining the leaf wrinkle index to obtain the chemical solvent cleaning dependence value; take the leaf wrinkle index as the base and calculate the power of the exponent determined in step E7 to obtain the correction factor; multiply the baseline value of chemical solvent cleaning requirement by this correction factor to obtain the corrected chemical solvent cleaning dependence value.

[0118] E9. Normalize the corrected chemical solvent cleaning dependence value and map it to the range of zero to one to obtain the final chemical solvent cleaning dependence. The normalization method is exactly the same as the method described in step D9. The closer the chemical solvent cleaning dependence value is to one, the higher the dependence of the leafy vegetable on surfactant chemical cleaning.

[0119] S3: Determine the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit based on the degree of dependence on physical rinsing cleaning; determine the initial concentration parameters and action duration parameters of the surfactant cleaning unit based on the degree of dependence on chemical solvent cleaning.

[0120] In this embodiment, step S3 includes the following specific details, which can be found in the flowchart below. Figure 4 , Figure 4 Here is a detailed flowchart of the cleaning parameter determination process provided in the embodiments of this application:

[0121] S310: Map the physical flushing cleaning dependence to a preset physical cleaning parameter space, and obtain the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit through interpolation calculation;

[0122] This embodiment mainly includes the following steps:

[0123] F1. During the process development stage, a physical cleaning parameter relationship table is pre-built and stored. First, dozens of theoretical values ​​of physical rinsing cleaning dependence are uniformly selected within a numerical range from zero to one. For each selected theoretical value of dependence, hundreds of independent high-pressure water mist cleaning experiments are conducted. In each cleaning experiment, the water pump pressure and spray coverage time of the high-pressure water mist cleaning unit are precisely adjusted according to the cleaning requirements represented by the theoretical value of dependence corresponding to the current experiment.

[0124] F2. Immediately after each washing experiment, two parallel tests were performed on the leafy vegetable samples. The first test was the pesticide residue removal rate test: according to national standard methods, the pesticide residue in the leafy vegetable samples after washing was accurately measured using gas chromatography-mass spectrometry, and the removal rate was calculated by comparing it with the residue in the samples before washing. The second test was the assessment of the degree of physical damage to the leaf surface: using a contact conductivity meter, five standard points were selected on the surface of the leafy vegetable samples to measure the conductivity value caused by cell sap exudation; the growth rate of the average conductivity of the five points after washing relative to the average conductivity of the samples before washing was calculated, and this growth rate was used as a quantitative indicator of leaf surface damage.

[0125] F3. By analyzing and comparing the results of hundreds of experiments under the same theoretical value of dependence, a set of optimal parameters is determined. This set of parameters must simultaneously meet the requirements of achieving the pesticide removal rate and minimizing damage to the leafy vegetable surface, resulting in an optimal pressure value and an optimal coverage time value. A mapping relationship is established between each theoretical value of physical rinsing and cleaning dependence and its corresponding set of optimal pressure and optimal coverage time values.

[0126] F4. All the above mapping relationships are structured and stored in the system's process parameter database to form the final physical cleaning parameter relationship table.

[0127] F5. In the real-time cleaning process, the actual value of the physical rinsing and cleaning dependence of the current leafy vegetables calculated in step S2 is used as the input condition for querying the physical cleaning parameter relationship table. The physical cleaning parameter relationship table is queried to locate the two discrete storage nodes that are closest to the input value. The first node is the node with the largest value among all nodes with values ​​less than the input value, and the second node is the node with the smallest value among all nodes with values ​​greater than the input value.

[0128] F6. Extract the optimal pressure value and optimal coverage duration value pre-stored in the database for the first node; extract the optimal pressure value and optimal coverage duration value pre-stored in the database for the second node;

[0129] F7. Calculate the relative position ratio of the input value with respect to the two node intervals; First, subtract the value of the first node from the input value of the physical flushing cleaning dependence to obtain the first difference; then, subtract the value of the first node from the value of the second node to obtain the second difference; finally, divide the first difference by the second difference, and the quotient obtained is the relative position ratio.

[0130] F8. Calculate the initial pressure parameters of the high-pressure water mist cleaning unit; multiply the optimal pressure value corresponding to the first node by a coefficient and subtract the relative position ratio calculated in step F7 to obtain a first product; multiply the optimal pressure value corresponding to the second node by the relative position ratio calculated in step F7 to obtain a second product; add the first product and the second product together to obtain the initial pressure parameters.

[0131] F9. Calculate the initial coverage duration parameter of the high-pressure water mist cleaning unit; multiply the optimal coverage duration value corresponding to the first node by the coefficient and subtract the relative position ratio calculated in step F7 to obtain a third product; multiply the optimal coverage duration value corresponding to the second node by the relative position ratio calculated in step F7 to obtain a fourth product; add the third product and the fourth product together to obtain the initial coverage duration parameter.

[0132] F10: Through the above steps, complete the parameter decision based on the degree of dependence on physical rinsing and cleaning, and output the initial pressure parameters and initial coverage duration parameters of the high-pressure water mist cleaning unit; S320: Input the degree of dependence on chemical solvent cleaning into the preset multi-objective parameter optimization model, and solve it with the goal of maximizing cleaning efficiency and minimizing chemical residue, to obtain the initial concentration parameters and action duration parameters of the surfactant cleaning unit.

[0133] To minimize chemical residues while achieving high cleaning efficiency, this embodiment constructs a multi-objective optimization model for surfactant cleaning parameters to determine the optimal combination of initial concentration and contact time. This multi-objective optimization model aims to maximize pesticide removal rate and minimize surfactant residue, and is implemented through a solution process integrating a genetic algorithm and a surrogate model. The specific steps for constructing and solving the multi-objective optimization model for surfactant cleaning parameters are as follows:

[0134] G1. The calculated chemical solvent cleaning dependence is used as the core constraint for solving the multi-objective optimization model of surfactant cleaning parameters. This constraint determines the boundary of the feasible region for searching the optimal parameter combination. The decision variables of the multi-objective optimization model of surfactant cleaning parameters are defined. The two decision variables that need to be optimized by the multi-objective optimization model of surfactant cleaning parameters are: the concentration percentage of surfactant cleaning solution and the duration of the soaking cleaning process.

[0135] G2. Set dual optimization objectives for the multi-objective optimization model of surfactant cleaning parameters; the first optimization objective is to maximize cleaning efficiency, that is, to achieve the highest pesticide removal rate on the surface of leafy vegetables; the second optimization objective is to minimize chemical residues after cleaning, that is, to minimize the amount of surfactant residues adsorbed on the surface of leafy vegetables.

[0136] G3. Genetic algorithm is used as the solution framework for the multi-objective optimization model of surfactant cleaning parameters. First, a population initialization operation is performed: hundreds of independent candidate solutions are randomly generated in the parameter space composed of decision variables to form the initial population. Each individual in the population is a complete parameter combination, including a specific concentration value and a specific duration of action.

[0137] G4. Call a pre-trained proxy model to quickly evaluate the performance of each individual in the population. The proxy model is a deep neural network trained on massive historical cleaning experimental data. Its function is to receive any set of concentration and time parameters as input, and through internal forward propagation calculation, quickly and accurately predict and output the pesticide removal rate and surfactant residue values ​​corresponding to the parameter combination.

[0138] G5. Perform a non-dominated sorting operation on all individuals in the current population. The purpose of this operation is to stratify all individuals based on bi-objective values. The specific rule is: compare any two individuals one by one. If individual A has a pesticide removal rate that is not lower than that of individual B, and its surfactant residue is strictly lower than that of individual B, then individual A is said to dominate individual B. First, find all individuals in the current population that are not dominated by any other individual and classify them into the first non-dominated level.

[0139] G6. Temporarily remove individuals that have been assigned to the first level from the comparison set. Repeat the comparison rule of step G5 among the remaining individuals to find new individuals that are not dominated by the remaining individuals and assign them to the second non-dominated level.

[0140] G7. Repeat step G6 until all individuals in the population are assigned to a specific non-dominated level; the smaller the level number, the better the overall performance of the individuals in that level, i.e., closer to the theoretical Pareto optimal frontier.

[0141] G8. Calculate the crowding distance of each individual within the same non-dominated hierarchy to measure the distribution density of individuals in the target space. The specific calculation method is as follows: First, sort all individuals within the hierarchy in ascending order according to the numerical values ​​of the two objective functions: pesticide removal rate and chemical residue. Then, for each individual in the middle position after sorting, calculate the sum of the absolute values ​​of the differences between the preceding and following individuals in the sorting sequence in terms of the two objective function values. This sum is the crowding distance of that individual. The larger the crowding distance, the sparser the solution space around that individual and the better the diversity.

[0142] G9. Based on the non-dominated hierarchy obtained in step G7 and the crowding distance obtained in step G8, a tournament selection mechanism is used to select superior individuals from the current population as parents for subsequent reproduction. When selecting, individuals with smaller non-dominated hierarchy numbers are given priority. When two individuals are at the same hierarchy, individuals with larger crowding distances are given priority.

[0143] G10. Perform genetic operations on the selected parent individuals to generate offspring. First, with a preset crossover probability, perform simulated binary crossover on randomly paired parent individuals: through a specific random number generation mechanism, exchange and mix the encoded information of some concentration and time parameters in the paired individuals to generate new parameter combinations for offspring individuals. Then, with a preset mutation probability, perform polynomial mutation on these offspring individuals: apply a small random perturbation that conforms to a specific probability distribution to their concentration and time parameter values ​​to introduce new genetic characteristics and increase the diversity of the population. Merge the original parent population with the newly generated offspring population to form a temporary, larger set of candidate solutions.

[0144] G11. Based on the elite retention strategy, individuals to form the next generation population are selected from the temporary candidate set obtained in step G10. The selection rules are consistent with the selection mechanism in step G9: individuals with smaller numbers in the non-dominant hierarchy are given priority; within the same hierarchy, individuals with greater crowding distance are given priority, until the number of selected individuals reaches the preset population size.

[0145] G12. Repeat steps G4 to G11 to form a complete iteration cycle of the genetic algorithm. The genetic algorithm repeats this iteration cycle hundreds of times. As the iteration continues, the overall quality of individuals in the population is continuously improved and eventually converges to a stable state. At this time, the set of optimal individuals in the population constitutes the Pareto optimal solution frontier. Each solution on the Pareto optimal solution frontier represents a set of concentration and time parameters that achieve the best trade-off between the two conflicting objectives of cleaning efficiency and chemical residue.

[0146] G13. Based on the preset safety priority decision criteria, select a final implementation scheme from the Pareto optimal solution set obtained by final convergence to determine the specific operating parameters. The specific selection logic is as follows: First, from the entire Pareto optimal solution set, screen out all feasible solutions with pesticide removal rate prediction values ​​of not less than 98%, forming a high-cleaning-efficiency candidate subset; then, in this high-cleaning-efficiency candidate subset, select the solution with the smallest surfactant residue prediction value; the specific concentration value and soaking time value corresponding to this solution are officially determined as the initial concentration parameter and initial action time parameter of the surfactant cleaning unit.

[0147] S4: Based on the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit, and the initial concentration parameters and action duration parameters of the surfactant cleaning unit, generate a collaborative execution plan for the composite cleaning process.

[0148] In this embodiment, step S4 includes the following specific details, which can be found in the flowchart below. Figure 5 , Figure 5 Here is a detailed flowchart of the collaborative solution generation process provided in this application embodiment:

[0149] S410: Establish a process scheduling model with the goal of minimizing the total cleaning time;

[0150] In this embodiment, the specific steps include:

[0151] H1. First, clarify the actual physical configuration and process flow of the cleaning production line: In this embodiment, the production line is set up with two core processing units in sequence. The first unit is a high-pressure water mist cleaning unit, and the second unit is a surfactant immersion cleaning unit. The two cleaning units are arranged in series in space and connected by a conveyor belt to form a production line.

[0152] Define material transfer rules: After a batch of leafy vegetables is processed in one unit, it is automatically transported to the next unit by a conveyor belt. This transfer process takes a fixed and known amount of time.

[0153] H2. Based on the above production configuration, construct a process scheduling model. This process scheduling model defines each batch of leafy vegetables to be processed as an independent operation. This process scheduling model treats two connected washing units as two machines, and each operation must access these two machines sequentially according to the process order. The process scheduling model needs to handle equipment capacity constraints: it stipulates that each washing unit can only process one batch at a time, that is, one machine can only process one operation at a time. The process scheduling model needs to handle process order constraints: it stipulates that the processing order of each operation must strictly follow the fixed process of first passing through the high-pressure water mist washing unit and then through the surfactant soaking washing unit.

[0154] H3. Assign specific processing time parameters to each operation: The processing time of each batch in the high-pressure water mist cleaning unit is directly determined by the initial coverage time parameter of the batch determined in step S310; The processing time of each batch in the surfactant immersion cleaning unit is directly determined by the initial action time parameter of the batch determined in step S320.

[0155] H4. When calculating the overall operation timeline, the fixed transfer time between units for batches as defined in step H1 must be taken into account.

[0156] H5. Based on satisfying all the constraints described in steps H2 and H4, set the optimization objective of the process scheduling model. The optimization objective is to find an optimal job processing sequence and plan the start and end times of each job on each machine. The optimal criterion is to minimize the total process time from the first job entering the production line for cleaning to the last job completing all cleaning processes and leaving the production line.

[0157] H6. Complete the construction of the process scheduling model;

[0158] S420: The initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit, and the initial concentration parameters and action duration parameters of the surfactant cleaning unit are used as inputs to the process scheduling model to obtain the process execution sequence and generate a collaborative execution scheme for the composite cleaning process.

[0159] In this embodiment, the specific steps include:

[0160] I1. The initial pressure parameters and initial coverage duration parameters determined for each batch in step S310, and the initial concentration parameters and initial action duration parameters determined for each batch in step S320, together with the unique identification information of each batch, are compiled into the input data of the process scheduling model.

[0161] I2. Load the above input data into the process scheduling model constructed in step S410.

[0162] 3. The process scheduling model formalizes this problem into a classic flow shop scheduling problem involving two machines, where each machine represents a washing unit and each operation represents a batch of leafy vegetables.

[0163] I4. Calculate the processing time for each batch on the first virtual machine: this time is the initial coverage duration parameter of the batch in the high-pressure water mist cleaning unit; calculate the processing time for each batch on the second virtual machine: this time is the initial action duration parameter of the batch in the surfactant immersion cleaning unit.

[0164] I5. Between the processing time of the above two processes, a fixed inter-unit transfer time determined by the conveyor belt running time must also be included;

[0165] I6. A heuristic optimization algorithm is used to solve the problem. First, the improved Johnson rule is applied to perform a preliminary sorting of all batches to generate an initial job sequence. The specific process of applying the Johnson rule is as follows: compare the relative processing time of each batch in the first process with the processing time in the second process. Based on the comparison results, prioritize the batches with shorter processing times that are arranged earlier, thereby generating an initial job sequence that can effectively reduce the waiting time between processes.

[0166] I7. After obtaining the initial job sequence, a tabu search algorithm is further used for refined optimization to find a sequence with a shorter total completion time. The tabu search algorithm takes the initial job sequence as the starting point and defines its neighborhood as all new sequences that can be obtained by swapping the positions of any two batches in the current sequence. In each iteration, the tabu search algorithm evaluates the total completion time corresponding to all neighborhood sequences of the current sequence. From all neighborhood sequences, a sequence that can shorten the total completion time to the greatest extent and whose swap operation performed in generating the sequence has not been tabu recently is selected as the new current solution. The batch swap operation that caused this state transition is recorded in a tabu list, and its reverse operation is prohibited from happening again within the subsequent preset number of iterations, so as to avoid the search process getting trapped in local optima.

[0167] I8. Repeat step I7 for multiple iterations until the tabu search algorithm converges to a final batch processing sequence whose total completion time cannot be significantly optimized further.

[0168] I9. Based on the final optimized batch processing sequence, combined with the specific processing time parameters of each batch and the fixed transfer time, calculate the start and end times of each batch on each machine with millisecond precision.

[0169] I10. Integrate all information to generate a detailed process execution plan. This plan specifies the complete spatiotemporal trajectory of each leafy vegetable batch on the production line, including: the moment of entering the production line, the moment of arriving at the high-pressure water mist cleaning unit, the moment of starting high-pressure water mist cleaning, the pressure value used during cleaning, the moment of ending high-pressure water mist cleaning, the moment of starting to transfer to the soaking unit, the moment of arriving at the surfactant soaking unit, the moment of starting soaking, the concentration value used during soaking, the moment of ending soaking, and the moment of finally leaving the production line.

[0170] I11. The generated detailed process execution plan is formally confirmed as the collaborative execution scheme for the composite cleaning process; this collaborative execution scheme is then sent to the central control system of the production line to drive subsequent physical execution.

[0171] S5: Based on leaf surface morphology data and leaf wrinkle index, calculate the leaf structure vulnerability index, and optimize the parameters of the high-pressure water mist cleaning unit in the collaborative execution scheme according to the degree of dependence on physical scouring and cleaning and the leaf structure vulnerability index, to obtain optimized pressure parameters and optimized coverage duration parameters.

[0172] In this embodiment, step S5 includes the following specific details, which can be found in the flowchart below. Figure 6 , Figure 6 This is a detailed flowchart of the cleaning parameter optimization process provided in the embodiments of this application:

[0173] S510: Based on the leaf surface morphology data containing the leaf vegetable surface triangular mesh model, calculate the set of distances from each vertex to the main plane of the leaf, and extract the leaf thickness parameter representing the thickness of the main leaf tissue from the set of distances.

[0174] In this embodiment, the specific steps include:

[0175] J1. Obtain the triangular mesh model of the leafy vegetable surface obtained through the three-dimensional contour scanning and reconstruction in step S130; extract the three-dimensional coordinates of all vertices from the triangular mesh model of the leafy vegetable surface to form a vertex coordinate set;

[0176] J2. Perform principal component analysis on the vertex coordinate set; first, calculate the arithmetic mean of the coordinates of all vertices in the set to determine the coordinates of the center point of the point cloud;

[0177] J3. Subtract the center point coordinates obtained in step J2 from the three-dimensional coordinates of each vertex to obtain a decentralized coordinate data set;

[0178] J4. Based on the decentralized coordinate data set, calculate the covariance matrix; solve for all eigenvalues ​​and corresponding eigenvectors of the covariance matrix; among all eigenvectors, select the eigenvector with the smallest eigenvalue, and define the direction indicated by the eigenvector as the thickness direction of the blade; among all eigenvectors, select the two eigenvectors with the largest eigenvalues, and define the plane spanned by these two eigenvectors as the principal plane, which best represents the two-dimensional extension direction of the blade.

[0179] J5. Using the main plane defined in step J4 as a reference, calculate the vertical distance from each vertex in the mesh model to the plane; summarize the vertical distance values ​​calculated for all vertices to obtain a set of distance values ​​reflecting the relative thickness of each point on the leaf surface; to eliminate the interference caused by leaf edge curling and individual extremely deep folds on the overall thickness characterization, sort the set of distance values ​​in ascending order; in the sorted distance value sequence, find the distance value corresponding to the 85th percentile position; use the 85th percentile distance value as the final leaf surface thickness parameter used to characterize the physical thickness of the leaf mesophyll body.

[0180] S520: Based on the leaf wrinkle index and the leaf thickness parameter, the leaf structure fragility index is obtained by weighted calculation;

[0181] In this embodiment, the specific steps are as follows:

[0182] K1. A pre-defined evaluation model for calculating the leaf structure vulnerability index is used. The evaluation model is calculated by the sum of two weighted terms: the first term is the product of the leaf wrinkle index and a first coefficient, and the second term is the product of a baseline thickness value divided by the leaf thickness parameter and a second coefficient.

[0183] K2. The first coefficient, the second coefficient, and the reference thickness value in the evaluation model need to be determined through specialized experiments and data analysis. The specific determination process is as follows;

[0184] K3. Prepare experimental samples for model calibration; select a large number of leafy vegetables with different surface characteristics to ensure that there are significant differences in leaf wrinkling index and leaf thickness parameters among the samples; conduct standardized water mist rinsing experiments on each experimental sample, keeping the experimental conditions strictly consistent; immediately after each rinsing experiment, use the conductivity detection method to measure the conductivity of the sample surface and calculate the growth rate relative to before the experiment, and record this growth rate as the measured damage quantification value of the sample; summarize all experimental data to form a dataset; each record in the dataset contains the leaf wrinkling index, leaf thickness parameter and its corresponding measured damage quantification value of a sample.

[0185] K4. Using the leaf wrinkling index and leaf thickness parameter of all samples in the dataset as two independent variables and the measured damage quantification value as the dependent variable, perform multiple linear regression analysis; extract the coefficient corresponding to the leaf wrinkling index from the mathematical relationship obtained from the regression analysis as the original value of the first coefficient; extract the coefficient corresponding to the leaf thickness parameter from the same mathematical relationship as the original value of the second coefficient; normalize the original values ​​of the first and second coefficients to satisfy a specific mathematical relationship, and define the two new values ​​obtained after processing as the first and second coefficients used to evaluate the model.

[0186] K5. The method for determining the baseline thickness value is as follows: measure the leaf thickness parameters of a large number of undamaged fresh leafy vegetables, calculate the arithmetic mean of all these thickness parameters, and set the arithmetic mean as the baseline thickness value.

[0187] K6. When calculating the leaf structure vulnerability index online in real time, first obtain two input parameters for the current leafy vegetable: the leaf wrinkle index from step S130 and the leaf thickness parameter from step S510; firstly, calculate the first term of the evaluation model: multiply the obtained leaf wrinkle index by the first coefficient to obtain the result of the first term; secondly, calculate the second term of the evaluation model: firstly, divide the baseline thickness value by the obtained leaf thickness parameter to obtain a quotient; then, multiply this quotient by the second coefficient to obtain the result of the second term;

[0188] K7. Add the result of the first item obtained in step K6 to the result of the second item. The sum is the leaf structure vulnerability index of the current leafy vegetable.

[0189] K8. This vulnerability index is a dimensionless quantitative indicator. Its value directly reflects the leaf surface's ability to resist physical erosion. The larger the value, the more fragile the leaf surface structure is, and the easier it is to be damaged during the washing process.

[0190] S530: Using the physical scouring and cleaning dependence degree and the leaf structure vulnerability index as joint inputs, query the predefined cleaning parameter optimization mapping table to obtain the corresponding optimized pressure parameter and optimized coverage duration parameter;

[0191] In this embodiment, the specific steps are as follows:

[0192] L1. In the process development stage, in order to construct a cleaning parameter optimization mapping table, the entire numerical range of the physical rinsing cleaning dependence is first divided into multiple continuous intervals. For each divided dependence interval, a minimum cleaning efficiency target value is preset. The target value is set based on the typical pesticide residue level corresponding to the interval and the national food safety standards. A basic compliance removal rate is determined through the previous process verification experiment, and a safety margin percentage is added on this basis.

[0193] L2. For each dependence interval, prepare multiple sets of leafy vegetable experimental samples; for each set of samples, the leaf structure vulnerability index calculated by the K7 step has a gradient distribution from low to high.

[0194] L3. For each experimental sample, various combinations of high-pressure water mist pressure parameters and coverage duration parameters were tested on the cleaning experimental platform. After each cleaning test, two detection processes were executed in parallel. The first process was pesticide removal efficiency evaluation: the pesticide residue of the sample after cleaning was measured using gas chromatography-mass spectrometry, the pesticide removal percentage relative to the sample before cleaning was calculated, and this percentage was compared with the preset minimum efficiency target value of the dependence range to which the current sample belongs to determine whether it meets the standard. The second process was leaf surface physical damage determination: the conductivity of multiple standard points on the leaf surface after cleaning was measured using a contact conductivity probe, and the average growth rate relative to the sample before cleaning was calculated. This growth rate needs to be compared with a predetermined damage determination threshold.

[0195] L4. The damage judgment threshold is determined through a special calibration experiment. The calibration experiment method is as follows: apply gradually increasing physical stress to a known intact leaf surface until the first visible damage point appears, record the conductivity change data at each stress level, and establish the average conductivity growth rate corresponding to the first visible damage as the damage judgment threshold.

[0196] L5. If the conductivity growth rate of the sample does not exceed the damage judgment threshold determined in step L4 after a certain cleaning test, it is determined that the test did not cause physical damage to the leaf surface.

[0197] L6. For all tests conducted on multiple experimental samples within the same dependence range and with the same or similar leaf structure vulnerability index, select all test records that simultaneously meet the pesticide removal rate requirement of step L3 and the requirement of not causing physical damage to the leaf surface in step L5.

[0198] L7. Gather all the pressure parameters and coverage duration parameters corresponding to the test records selected in step L6 to form a feasible parameter set for the current dependence level and current vulnerability index combination; calculate a technical evaluation index for each parameter combination in the feasible parameter set; the calculation method of the technical evaluation index is as follows: add a unit time constant coefficient calculated according to the power consumption and operating cost model of the production line equipment to the coverage duration value in the parameter combination to obtain an intermediate result one; at the same time, add a unit pressure constant coefficient calculated according to the same model to the pressure value in the parameter combination to obtain an intermediate result two; multiply intermediate result one and intermediate result two, and the product is the technical evaluation index of the parameter combination; the technical evaluation index comprehensively characterizes the relative energy consumption load when using this set of parameters for cleaning;

[0199] L8. Compare the technical evaluation index values ​​of all parameter combinations in the current feasible parameter set, and select the parameter combination with the smallest technical evaluation index value.

[0200] L9. Record the specific pressure values ​​in the parameter combination selected in step L8 as the recommended pressure parameters corresponding to the current dependence range and the current leaf structure vulnerability index combination.

[0201] L10. Record the specific coverage duration value in the parameter combination selected in step L8 as the recommended coverage duration parameter corresponding to the current dependence interval and the current leaf structure vulnerability index combination.

[0202] L11. For all the divided physical scouring and cleaning dependence intervals, and for all leaf structure vulnerability index gradients involved in each interval, repeat the sample preparation, experimental testing, detection evaluation, screening and calculation process described in steps L2 to L10.

[0203] L12. By repeating step L11, a set of recommended stress parameters and recommended coverage duration parameters are determined for all possible combinations of dependency and vulnerability index.

[0204] L13. All the correspondences obtained in step L12, namely the mapping relationship between the physical scouring and cleaning dependence range, the leaf surface structure vulnerability index, the recommended pressure parameter, and the recommended coverage duration parameter, are structured and stored in the system database to form the final cleaning parameter optimization mapping table.

[0205] L14. When applying online in real time, the specific values ​​of the physical scouring and cleaning dependence obtained by the current leafy vegetable through step S2 and the specific values ​​of the leaf structure vulnerability index obtained by step K7 are used as joint query conditions. Using these joint query conditions, the cleaning parameter optimization mapping table is queried, and the corresponding dependence interval and vulnerability index range are found through matching. The specific values ​​of the recommended pressure parameter and the recommended coverage duration parameter bound to the matching condition are read from the cleaning parameter optimization mapping table.

[0206] L15. Output the specific values ​​of the recommended pressure parameters read in step L14 as the optimized pressure parameters for the high-pressure water mist cleaning unit for the current leafy vegetables; output the specific values ​​of the recommended coverage duration parameters read in step L14 as the optimized coverage duration parameters for the high-pressure water mist cleaning unit for the current leafy vegetables.

[0207] S6: Update the collaborative execution plan using optimized pressure parameters and optimized coverage duration parameters, and control the cleaning equipment to execute the updated collaborative execution plan to complete the cleaning;

[0208] In this embodiment, step S6 includes the following specific details, which can be found in the flowchart below. Figure 7 , Figure 7Here is a detailed diagram of the optimized cleaning execution process provided in the embodiments of this application:

[0209] S610: Replace the initial pressure parameter and coverage duration parameter of the high-pressure water mist cleaning unit in the collaborative execution scheme with the optimized pressure parameter and the optimized coverage duration parameter;

[0210] The obtained optimized pressure parameters and optimized coverage duration parameters are written into the generated collaborative execution scheme, directly replacing the original initial pressure parameters and initial coverage duration parameters of the high-pressure water mist cleaning unit in the collaborative execution scheme; the initial concentration parameters and initial action duration parameters of the surfactant cleaning unit in the collaborative execution scheme remain unchanged.

[0211] S620: Controls the cleaning equipment to perform operations;

[0212] Based on the updated collaborative execution plan, the central control unit of the production line sends precise control commands to each equipment unit. The high-pressure water mist cleaning unit starts first, and the frequency conversion drive system adjusts the water pump pressure to the optimized pressure parameter value and continues to run for the time specified by the optimized coverage duration parameter. After this process is completed, the conveying system automatically transfers the leafy vegetable batch to the surfactant soaking cleaning unit. The surfactant cleaning unit then starts, and the precision metering system adjusts the cleaning solution concentration to the initial concentration parameter value determined in the plan and soaks the leafy vegetables at this concentration for the time specified by the initial action duration parameter.

[0213] S630: Cleaning complete;

[0214] Once the two core processes of high-pressure water mist cleaning and surfactant immersion cleaning are accurately completed according to the updated collaborative execution plan, the personalized cleaning process for the current batch of leafy vegetables is deemed complete, and the leafy vegetables will automatically enter the subsequent water rinsing, dehydration and packaging stages.

[0215] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of the leafy vegetable cleaning and packaging process monitoring system based on big data provided in the embodiments of this application;

[0216] This embodiment demonstrates the overall system architecture for implementing the above method; the system, through the synergy of six core functional modules, constructs a complete intelligent cleaning technology chain, from precise perception of individual characteristics of leafy vegetables, intelligent decision-making of cleaning strategies, multi-process collaborative scheduling to process closed-loop dynamic optimization;

[0217] The data acquisition module, as the front-end sensing and analysis unit of the system, is responsible for controlling the near-infrared spectrometer and the three-dimensional contour scanner to simultaneously collect pesticide residue spectral data and morphological data on the surface of leafy vegetables, and performing spectral preprocessing, feature extraction, concentration inversion and three-dimensional reconstruction, accurately outputting water-soluble pesticide residue concentration data, fat-soluble pesticide residue concentration data and leaf wrinkle index.

[0218] The dependency assessment module, as the analysis unit for cleaning mode and intensity requirements, is responsible for receiving the output of the data acquisition module. Through the built-in statistical model and correction function, it calculates and quantifies the comprehensive dependency of the leafy vegetables on high-pressure physical rinsing and surfactant chemical cleaning, and generates the dependency on physical rinsing and chemical solvent cleaning.

[0219] The parameter determination module, as the decision-making unit for the initial cleaning operation parameters, is responsible for determining the initial set values ​​of pressure and duration of the high-pressure water mist cleaning unit, as well as the initial set values ​​of concentration and duration of the surfactant cleaning unit, based on two degrees of dependence by querying the pre-experimental parameter relationship table for interpolation calculation and running a multi-objective optimization algorithm for solution.

[0220] The scheme generation module, as the scheduling and planning unit of the production process, is responsible for integrating the output of the parameter determination module and the production line resource status. It generates a collaborative execution scheme containing the process execution sequence, precise timing, and detailed parameter instructions for each process through the process scheduling optimization algorithm, and sends it to the production line execution control system.

[0221] The execution monitoring module serves as the monitoring unit for the execution of the scheme and process data. It is responsible for driving the cleaning equipment to operate according to the collaborative scheme, while controlling the online spectral sensor to collect data at preset key nodes. It also analyzes and calculates the decay rate data of pesticide residues in real time through a lightweight model.

[0222] The dynamic calibration module, as a closed-loop control unit that ensures the final cleaning effect and process economy, is responsible for continuously comparing real-time decay rate data with preset safety thresholds. It dynamically adjusts the cleaning parameters in operation through proportional-integral control algorithm, forming a closed loop of perception, decision-making, execution, feedback, and correction until the cleaning efficiency is confirmed to meet the standard, thus completing the closed-loop optimization control of the cleaning process.

[0223] Embodiments of the present invention also provide an electronic device, including a memory, a processor, and a communication bus; the memory and the processor are connected via the communication bus. The memory stores a method for monitoring the leafy vegetable washing and packaging process based on big data, which can be loaded by the processor and executed as provided in the above embodiments.

[0224] The memory can be used to store instructions, programs, code, code sets, or instruction sets. The memory may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function, and instructions for implementing the big data-based leafy vegetable washing and packaging process monitoring method provided in the above embodiments. The data storage area may store data involved in the big data-based leafy vegetable washing and packaging process monitoring method provided in the above embodiments.

[0225] The processor may include one or more processing cores. The processor executes instructions, programs, code sets, or instruction sets stored in memory, and calls data stored in memory to perform various functions and process data as described in this application. The processor may be at least one of the following: Application-Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), Central Processing Unit (CPU), controller, microcontroller, and microprocessor. It is understood that, for different devices, the electronic devices used to implement the above-described processor functions may also be other types, and the embodiments of this application do not specifically limit this.

[0226] A communication bus can include a pathway for transmitting information between the aforementioned components. The communication bus can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Communication buses can be categorized as address buses, data buses, control buses, etc.

[0227] This application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described in the above embodiments, a big data-based leafy vegetable washing and packaging process monitoring method.

[0228] In this embodiment, a computer-readable storage medium can be a tangible device that holds and stores instructions used by an instruction execution device. A computer-readable storage medium can be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. Specifically, a computer-readable storage medium can be a portable computer disk, a hard disk, a USB flash drive, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), spoofing random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory stick, floppy disk, optical disk, magnetic disk, mechanical encoding device, or any combination thereof.

[0229] The terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0230] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. A method for monitoring the leafy vegetable washing and packaging process based on big data, characterized in that, Includes the following steps: S1: Real-time acquisition of pesticide residue spectral data and leaf morphology data of leafy vegetables to be cleaned, and analysis to obtain water-soluble pesticide residue concentration data, fat-soluble pesticide residue concentration data and leaf wrinkling index; S2: Calculate the degree of dependence of leafy vegetables on physical rinsing and washing based on water-soluble pesticide residue concentration data and leaf wrinkling index; calculate the degree of dependence of leafy vegetables on chemical solvent washing based on fat-soluble pesticide residue concentration data and leaf wrinkling index. S3: Determine the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit based on the degree of dependence on physical rinsing cleaning; determine the initial concentration parameters and action duration parameters of the surfactant cleaning unit based on the degree of dependence on chemical solvent cleaning. S4: Based on the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit, and the initial concentration parameters and action duration parameters of the surfactant cleaning unit, generate a collaborative execution plan for the composite cleaning process. S5: Based on leaf surface morphology data and leaf wrinkle index, calculate the leaf structure vulnerability index, and optimize the parameters of the high-pressure water mist cleaning unit in the collaborative execution scheme according to the degree of dependence on physical scouring and cleaning and the leaf structure vulnerability index, to obtain optimized pressure parameters and optimized coverage duration parameters. S6: Update the collaborative execution plan using optimized pressure parameters and optimized coverage duration parameters, and control the cleaning equipment to execute the updated collaborative execution plan to complete the cleaning.

2. The method for monitoring the leafy vegetable washing and packaging process based on big data according to claim 1, characterized in that, The process involves real-time acquisition of pesticide residue spectral data and leaf surface morphology data of the leafy vegetables to be washed, and analysis to obtain water-soluble pesticide residue concentration data, fat-soluble pesticide residue concentration data, and leaf wrinkling index, including: S110: Obtain pesticide residue spectral data by scanning the surface of leafy vegetables with a near-infrared spectrometer; S120: Based on pesticide residue spectral data, the water-soluble pesticide residue concentration data and the fat-soluble pesticide residue concentration data are analyzed by matching the preset water-soluble pesticide concentration prediction model and the fat-soluble pesticide concentration prediction model. S130: Obtain leafy vegetable surface morphology data through a three-dimensional contour scanner, and calculate the leaf wrinkle index based on the leafy vegetable surface morphology data.

3. The method for monitoring the leafy vegetable washing and packaging process based on big data according to claim 2, characterized in that, The degree of dependence of leafy vegetables on physical rinsing and cleaning is calculated based on water-soluble pesticide residue concentration data and leaf wrinkling index. Based on the concentration data of fat-soluble pesticide residues and the leaf wrinkling index, the degree of dependence on chemical solvent washing for leafy vegetables was calculated, including: S210: Determine the degree of physical rinsing and cleaning requirement based on water-soluble pesticide residue concentration data, and correct the degree of physical rinsing and cleaning requirement by combining the leaf wrinkle index to obtain the degree of dependence on physical rinsing and cleaning. S220: The degree of chemical solvent cleaning requirement is determined based on the concentration data of fat-soluble pesticide residues, and the degree of chemical solvent cleaning requirement is corrected by combining the leaf wrinkling index to obtain the degree of chemical solvent cleaning dependence.

4. The method for monitoring the leafy vegetable washing and packaging process based on big data according to claim 3, characterized in that, The initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit are determined based on the degree of dependence on physical rinsing; the initial concentration parameters and action duration parameters of the surfactant cleaning unit are determined based on the degree of dependence on chemical solvent cleaning, including: S310: Map the physical flushing cleaning dependence to a preset physical cleaning parameter space, and obtain the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit through interpolation calculation; S320: Input the degree of dependence on chemical solvent cleaning into the preset multi-objective parameter optimization model, and solve it with the goal of maximizing cleaning efficiency and minimizing chemical residue to obtain the initial concentration parameters and action time parameters of the surfactant cleaning unit.

5. The method for monitoring the leafy vegetable washing and packaging process based on big data according to claim 4, characterized in that, The method for generating a collaborative execution plan for the composite cleaning process based on the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit, and the initial concentration parameters and action duration parameters of the surfactant cleaning unit, includes: S410: Establish a process scheduling model with the goal of minimizing the total cleaning time; S420: The initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit, and the initial concentration parameters and action duration parameters of the surfactant cleaning unit are used as inputs to the process scheduling model to obtain the process execution sequence and generate a collaborative execution scheme for the composite cleaning process.

6. The method for monitoring the leafy vegetable washing and packaging process based on big data according to claim 5, characterized in that, Step S5 specifically includes the following steps: S510: Based on the leaf surface morphology data containing the leaf vegetable surface triangular mesh model, calculate the distance from each vertex to the main plane of the leaf to obtain a distance set, sort the values ​​in the distance set, and select the values ​​located at the preset high percentile position after sorting as the leaf thickness parameter. S520: Based on the leaf wrinkle index and the leaf thickness parameter, the leaf structure fragility index is calculated by weighted summing the quotients obtained by dividing the leaf wrinkle index and the reference thickness value by the leaf thickness parameter. S530: Using the physical scouring and cleaning dependence degree and the leaf structure vulnerability index as joint input conditions, query a predefined cleaning parameter optimization mapping table; the cleaning parameter optimization mapping table stores the optimized pressure parameters and optimized coverage duration parameters corresponding to different combinations of physical scouring and cleaning dependence degree and different leaf structure vulnerability indices; by matching the physical scouring and cleaning dependence degree and leaf structure vulnerability index combination corresponding to the current input conditions, obtain the corresponding optimized pressure parameters and optimized coverage duration parameters.

7. The method for monitoring the leafy vegetable washing and packaging process based on big data according to claim 6, characterized in that, The step of updating the collaborative execution scheme using optimized pressure parameters and optimized coverage duration parameters, and controlling the cleaning equipment to execute the updated collaborative execution scheme to complete the cleaning, includes: S610: Replace the initial pressure parameter and coverage duration parameter of the high-pressure water mist cleaning unit in the collaborative execution scheme with the optimized pressure parameter and the optimized coverage duration parameter; S620: Controls the cleaning equipment to execute the updated collaborative execution plan and complete the cleaning process.

8. A big data-based monitoring system for the leafy vegetable washing and packaging process, used to implement the big data-based monitoring method for the leafy vegetable washing and packaging process as described in any one of claims 1 to 7, characterized in that, include: The data acquisition module is used to acquire pesticide residue spectral data and leaf morphology data of the leafy vegetables to be washed in real time, and analyze them to obtain water-soluble pesticide residue concentration data, fat-soluble pesticide residue concentration data and leaf wrinkling index. The dependency assessment module is used to calculate the degree of dependence of leafy vegetables on physical rinsing and washing based on water-soluble pesticide residue concentration data and leaf wrinkling index; and to calculate the degree of dependence of leafy vegetables on chemical solvent washing based on fat-soluble pesticide residue concentration data and leaf wrinkling index. The parameter determination module is used to determine the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit based on the degree of dependence on physical rinsing cleaning; and to determine the initial concentration parameters and action duration parameters of the surfactant cleaning unit based on the degree of dependence on chemical solvent cleaning. The scheme generation module is used to generate a collaborative execution scheme for the composite cleaning process based on the initial pressure parameters and coverage duration parameters of the high-pressure water mist cleaning unit, and the initial concentration parameters and action duration parameters of the surfactant cleaning unit. The execution monitoring module is used to calculate the leaf structure vulnerability index based on leaf surface morphology data and leaf wrinkle index, and optimize the parameters of the high-pressure water mist cleaning unit in the collaborative execution scheme according to the degree of dependence on physical rinsing and cleaning and the leaf structure vulnerability index, so as to obtain optimized pressure parameters and optimized coverage duration parameters. The dynamic calibration module is used to update the collaborative execution scheme with optimized pressure parameters and optimized coverage duration parameters, and to control the cleaning equipment to execute the updated collaborative execution scheme to complete the cleaning.

9. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program that can be called by the processor; the processor executes the big data-based leafy vegetable cleaning and packaging process monitoring method as described in any one of claims 1 to 7 by calling the computer program stored in the memory.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the big data-based monitoring method for leafy vegetable washing and packaging processes as described in any one of claims 1 to 7.

Images