Grain mold early warning method, system, device and storage medium thereof

Abstract

The invention discloses a method, system, device and storage medium for warning mildew of grains, and specifically relates to the technical field of grain warning monitoring, and is used to solve the problem of ignoring the potential influence of mildew of grains on the accuracy of near-infrared spectrum monitoring and warning in the prior art; the mildew of grains is obtained from a near-infrared spectrum monitoring system to determine whether there is a mildew trend; by analyzing the water binding state in the grains, it is identified whether the change of water absorption peak caused by mildew is normal; when there is a mildew trend and the change of water absorption peak is normal, a dual fractal analysis algorithm is used to model the internal optical characteristics of the grains, the abnormal degree of the light transmission path is evaluated, and surface microstructure features are extracted by a sparse representation classification algorithm to evaluate the change of scattering characteristics; an interference model is established based on the abnormal light transmission path and the change of scattering characteristics to quantify the potential adverse effect of mildew on the monitoring system; and a self-adaptive optimization algorithm is used to dynamically adjust the monitoring strategy to improve the monitoring accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of grain early warning monitoring, and more specifically, to a grain mildew early warning method, system, equipment and storage medium thereof. Background Art

[0002] Existing grain quality monitoring systems usually analyze the ingredients in grains through near-infrared spectroscopy to detect whether the grains are moldy. Specifically, they generate spectral data by measuring the absorption and reflection characteristics of grain samples to near-infrared light. Different grain components (such as water, fat, and protein) will produce different absorption peaks for near-infrared light. The system determines whether the grains are moldy by comparing the changes in these absorption peaks. Near-infrared spectroscopy technology is widely used in grain quality monitoring because of its rapid, non-destructive and real-time analysis characteristics.

[0003] However, existing technologies often focus on how to timely warn of the trend of grain mold, ignoring the potential impact of grain mold on the monitoring and early warning accuracy of the grain quality monitoring system based on near-infrared spectroscopy. This will cause the monitoring system itself to be unable to identify the risk of reduced system accuracy when it is disturbed in the early stages of mold, which will lead to inaccurate monitoring results and affect the timeliness and accuracy of early mold warning.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a grain mildew early warning method, system, device and storage medium thereof to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The method for early warning of grain mildew comprises the following steps:

[0008] S1: Obtain the grain mildew situation from the near infrared spectrum monitoring system to determine whether there is a grain mildew trend;

[0009] S2: Analyze the changes in the water binding state in grains through low-field nuclear magnetic resonance to identify whether the changes in the water absorption peak caused by mildew are normal;

[0010] S3: When there is a trend of grain mildew and the change of moisture absorption peak caused by mildew is normal, the internal optical properties of grain are modeled by dual fractal analysis algorithm to evaluate the abnormality of light transmission path caused by grain mildew; the feature extraction of grain surface microstructure data is performed by sparse representation classification algorithm to evaluate the change of scattering characteristics caused by grain mildew;

[0011] S4: Based on the degree of abnormality in the light transmission path and the degree of change in scattering characteristics caused by grain mildew, an interference model of mildew on the accuracy of grain mildew monitoring is established to quantify the potential adverse impact of grain mildew on the accuracy of near-infrared spectroscopy monitoring system in monitoring grain mildew;

[0012] S5: When the potential adverse effect of grain mildew on the accuracy of grain mildew monitoring by the near infrared spectroscopy monitoring system exceeds a preset threshold, a monitoring warning signal is issued;

[0013] Low-field nuclear magnetic resonance is used to analyze the changes in the water binding state in grains and identify whether the changes in the water absorption peak caused by mildew are normal. Specifically:

[0014] S201: Acquire longitudinal relaxation time and transverse relaxation time signal data of nuclear magnetic resonance of a grain sample from a low-field nuclear magnetic resonance device;

[0015] S202: Performing frequency domain conversion on the acquired signal by Fourier transform to remove background noise and abnormal signals, and obtaining a nuclear magnetic signal curve;

[0016] S203: Calculate the ratio of free water to bound water according to the relaxation time distribution of the longitudinal relaxation time and the transverse relaxation time, and extract the characteristic value of the water binding state;

[0017] S204: Compare the extracted characteristic value of the water binding state with the grain water binding pattern under normal conditions to determine whether the current water absorption peak change is within a normal range.

[0018] In a preferred embodiment, the mildew condition of grain is obtained from the near infrared spectrum monitoring system to determine whether there is a mildew trend of grain, specifically:

[0019] S101: acquiring real-time spectral data from a near infrared spectral monitoring system and storing spectral response information in a food environment;

[0020] S102: preprocessing the acquired spectral data to remove noise and outliers to obtain a standardized spectral data set;

[0021] S103: extracting features from the standardized spectral data using a preset spectral analysis algorithm to identify key spectral absorption peaks;

[0022] S104: Compare the extracted spectral features with a preset mildew trend database to determine whether there is a mildew trend in grain.

[0023] In a preferred embodiment, the internal optical properties of grains are modeled by a dual fractal analysis algorithm to evaluate the degree of abnormality in the light transmission path caused by grain mildew, specifically:

[0024] Obtain light transmission path data for grain samples, recording the propagation time and scattering pattern of light in different areas;

[0025] Divide the acquired optical transmission path data into multiple sub-segments according to the temporal and spatial distribution, each segment corresponding to a different optical propagation behavior;

[0026] Apply dual fractal analysis to the data of each segment to extract the self-similar features in the light transmission path and generate multi-dimensional fractal parameters:

[0027] For each subsegment’s light travel time and scattering angle data, dual fractal analysis generates a set of fractal dimensions;

[0028] The calculation formula of fractal dimension is: ;in, For the The fractal dimension of the sub-segments, For the In the sub-segment The difference in light travel time for each data point is For the In the sub-segment The difference in scattering angles of the data points is is the length of the sub-segment, Indicates different sub-segment numbers; The numbers representing the different data points in a particular sub-segment; Indicates The total number of data points contained in the sub-segments;

[0029] Based on the extracted self-similar features, a multi-scale fractal model of the light transmission path inside the grain is constructed:

[0030] A multi-scale fractal model is constructed based on the fractal dimension of each sub-segment; the construction formula of the multi-scale fractal model is: ;in, is a multi-scale fractal model. is the number of sub-segments, For the The weight of each sub-segment;

[0031] The multi-scale fractal model was compared with the standard light transmission path to quantify the degree of light transmission anomaly caused by mildew:

[0032] Calculate the optical transmission path anomaly index, the expression is: ;in, is the optical transmission path anomaly index, It is the standard light transmission path model.

[0033] In a preferred embodiment, the sparse representation classification algorithm is used to extract features from the grain surface microstructure data to evaluate the degree of change in scattering characteristics caused by grain mildew, specifically:

[0034] Obtain light scattering intensity distribution data on the grain surface and record the reflection and scattering patterns of light by the surface microstructure;

[0035] The obtained light scattering intensity distribution data is processed by multi-dimensional vectorization, and the scattering information at different angles and wavelengths is integrated into an analyzable multi-dimensional vector data set:

[0036] Discretize the continuous coordinate position, wavelength and scattering angle into a finite set of values ​​to form a three-dimensional grid;

[0037] For each sampling point, a multidimensional vector is constructed, and each component of the vector corresponds to the light scattering intensity value at different wavelengths and scattering angles;

[0038] Combine the vectors of all sampling points into a data matrix;

[0039] The sparse representation classification algorithm is used to process the converted multidimensional vector data set to extract sparse features that can reflect the changes in the scattering characteristics of the grain surface:

[0040] Using the pre-collected data on the surface scattering characteristics of grains in normal conditions, a training dictionary matrix is ​​constructed;

[0041] Solve for the sparse coefficient vector such that: ;in, is a sparse coefficient vector, is the residual vector, It is Multidimensional vector representation of sampling points;

[0042] By solving the optimization problem, the sparse coefficient vector of each sample is obtained;

[0043] Use the sparse coefficient vector as a feature to construct a feature matrix;

[0044] Compare the sparse features extracted within a preset time period with the standard scattering pattern, identify the deviation of scattering characteristics caused by mildew, and evaluate the degree of scattering change caused by grain mildew:

[0045] For the feature matrix, calculate the average sparse coefficient vector within a preset time period;

[0046] Get the standard sparse feature vector under normal conditions;

[0047] The relative change rate is used to evaluate the degree of change in the scattering characteristics and the mildew scattering change index is calculated. The expression is: ;in, is the mildew scattering variation index, is the average sparse coefficient vector extracted within a preset time period, is the standard sparse feature vector in the normal state.

[0048] In a preferred embodiment, based on the degree of abnormality of the light transmission path and the degree of change in the scattering characteristics caused by grain mildew, an interference model of mildew on the accuracy of grain mildew monitoring is established to quantify the potential adverse impact of grain mildew on the accuracy of grain mildew monitoring by the near infrared spectroscopy monitoring system, specifically:

[0049] The optical transmission path anomaly index and the mildew scattering change index are normalized, and weight coefficients are assigned to the normalized optical transmission path anomaly index and the mildew scattering change index, respectively, to calculate the mildew impact monitoring coefficient.

[0050] In a preferred embodiment, when the potential adverse effect of grain mildew on the accuracy of grain mildew monitoring by the near infrared spectroscopy monitoring system exceeds a preset threshold, a monitoring warning signal is issued, specifically:

[0051] Setting preset thresholds ; The mold impact monitoring coefficient and the preset threshold For comparison:

[0052] When the mold impact monitoring coefficient is greater than the preset threshold When the monitoring warning signal is issued,

[0053] When the mold impact monitoring coefficient is less than or equal to the preset threshold When , it is determined that the monitoring normal signal is sent.

[0054] On the other hand, the present invention provides a grain mildew early warning system, including a mildew trend judgment module, a moisture combination recognition module, a model parameter acquisition module, an interference model establishment module and a monitoring early warning judgment module;

[0055] Mildew trend judgment module: obtains grain mildew conditions from the near-infrared spectrum monitoring system to determine whether there is a grain mildew trend;

[0056] Moisture binding identification module: Analyze the changes in moisture binding state in grains through low-field nuclear magnetic resonance to identify whether the changes in moisture absorption peaks caused by mildew are normal;

[0057] Model parameter acquisition module: When there is a trend of grain mildew and the moisture absorption peak caused by mildew changes normally, the internal optical properties of the grain are modeled through the dual fractal analysis algorithm to evaluate the degree of abnormality of the light transmission path caused by grain mildew; the sparse representation classification algorithm is used to extract features from the grain surface microstructure data to evaluate the degree of change in scattering characteristics caused by grain mildew;

[0058] Interference model building module: Based on the degree of abnormality in the light transmission path and the degree of change in scattering characteristics caused by grain mildew, an interference model of mildew on the accuracy of grain mildew monitoring is established to quantify the potential adverse impact of grain mildew on the accuracy of near-infrared spectroscopy monitoring system in monitoring grain mildew;

[0059] Monitoring and early warning judgment module: When the potential adverse impact of grain mold on the accuracy of grain mold monitoring by the near-infrared spectroscopy monitoring system exceeds the preset threshold, a monitoring early warning signal is issued.

[0060] On the other hand, the present invention provides a device, comprising: a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction implements a grain mold warning method when executed by the processor.

[0061] On the other hand, the present invention provides a computer-readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, a method for early warning of grain mold is implemented.

[0062] The technical effects and advantages of the grain mildew early warning method, system, device and storage medium of the present invention are as follows:

[0063] 1. By introducing the analysis of the water binding state into the grain mildew early warning method, it is cleverly judged whether it is necessary to establish an interference model for the mildew on the accuracy of grain mildew monitoring. When the mildew trend of grain is detected, but the change of moisture absorption peak is still within the normal range, it means that the mildew is in the early stage and has not yet had a significant impact on the moisture structure of the grain. In this case, traditional near-infrared spectroscopy monitoring may be interfered by the changes in the light transmission path and scattering characteristics caused by mildew, resulting in inaccurate monitoring results. By identifying this specific state, it is possible to accurately determine when it is necessary to establish an interference model to evaluate the potential impact of mildew on the monitoring system, thereby avoiding unnecessary model establishment and improving the timeliness and accuracy of early warning.

[0064] 2. In the process of establishing the interference model, the dual fractal analysis algorithm and the sparse representation classification algorithm were used to conduct in-depth analysis of the internal optical properties and surface microstructure of grains. By evaluating the degree of abnormality in the light transmission path and the degree of change in the scattering characteristics caused by mildew, the interference effect of mildew on the near-infrared spectroscopy monitoring system was quantified. Subsequently, based on these quantitative indicators, a model that can reflect the degree of interference caused by mildew was constructed. When the degree of interference exceeds the preset threshold, the system will issue a warning signal and dynamically adjust the monitoring strategy using an adaptive optimization algorithm, which not only improves the sensitivity of the monitoring system in the early stage of mildew, but also ensures the reliability and accuracy of the monitoring results through real-time adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is a schematic diagram of the method for early warning of grain mildew of the present invention;

[0066] Figure 2 It is a structural schematic diagram of the grain mildew early warning system of the present invention. DETAILED DESCRIPTION

[0067] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] Example 1

[0069] Figure 1 The present invention provides a method for early warning of grain mildew, which comprises the following steps:

[0070] S1: Obtain the grain mildew situation from the near-infrared spectroscopy monitoring system to determine whether there is a grain mildew trend.

[0071] S2: Analyze the changes in the water binding state in grains through low-field nuclear magnetic resonance to identify whether the changes in the water absorption peak caused by mold are normal.

[0072] S3: When there is a trend of grain mildew and the changes in the moisture absorption peak caused by mildew are normal, the internal optical properties of the grain are modeled through the dual fractal analysis algorithm to evaluate the degree of abnormality in the light transmission path caused by grain mildew; the sparse representation classification algorithm is used to extract features from the grain surface microstructure data to evaluate the degree of change in the scattering characteristics caused by grain mildew.

[0073] S4: Based on the degree of abnormality in the light transmission path and the degree of change in scattering characteristics caused by grain mildew, an interference model of mildew on the accuracy of grain mildew monitoring is established to quantify the potential adverse impact of grain mildew on the accuracy of near-infrared spectroscopy monitoring system in monitoring grain mildew.

[0074] S5: When the potential adverse effect of grain mold on the accuracy of grain mold monitoring by the near-infrared spectroscopy monitoring system exceeds a preset threshold, a monitoring warning signal is issued.

[0075] Obtain the grain mildew situation from the near infrared spectrum monitoring system to determine whether there is a grain mildew trend, specifically:

[0076] S101: Obtain real-time spectral data from the near-infrared spectral monitoring system and store spectral response information in the food environment:

[0077] In the monitoring system in the granary, the near-infrared spectral sensor is arranged above the grain stacking area. The sensor emits a near-infrared beam to the grain surface and receives the reflected spectral signal. The sensor obtains real-time spectral data of the grain surface by detecting reflected light of different wavelengths. The collected spectral data includes multiple spectrums with a wavelength range of 800 nm to 2500 nm. The data is transmitted in real time through the central processing unit (CPU) of the monitoring system and stored in the built-in database. The data collection frequency can be set to, for example, once every 15 minutes to ensure that the acquired spectral data has temporal continuity. The stored data includes spectral intensity information of each band, recorded in the form of a multi-dimensional matrix, reflecting the spectral response under different environmental conditions.

[0078] S102: Preprocess the acquired spectral data to remove noise and outliers to obtain a standardized spectral data set:

[0079] The acquired spectral data is subjected to a multi-step preprocessing operation. First, a low-pass filter is used to remove high-frequency noise signals to ensure that the spectral data is smooth and free of interference. Next, a median filter method is used to remove occasional abnormal data points to prevent extreme values ​​from misleading subsequent analysis. Subsequently, the spectral data is standardized using a normalization method, and the spectral intensity value corresponding to each wavelength is normalized to a range of 0 to 1 to eliminate fluctuations and systematic errors that may occur under different acquisition conditions. Finally, after the preprocessing is completed, a standardized spectral data set is generated, which will serve as the basic input for subsequent analysis.

[0080] S103: Use a preset spectrum analysis algorithm to extract features from the standardized spectrum data and identify key spectrum absorption peaks:

[0081] The preset spectral analysis algorithm is applied to the generated standardized spectral data set. The algorithm used is a feature extraction algorithm based on principal component analysis (PCA). The algorithm first performs dimensionality reduction on the standardized spectral data set and converts the original multidimensional spectral data into several principal component eigenvectors. Then, the characteristic variables in each principal component are analyzed, and the most critical spectral absorption peaks in the grain mold process are identified by the contribution rate of the characteristic variables. Usually, the absorption peaks related to mold appear in the moisture absorption bands near 1400 nm and 1900 nm. The significant changes in these two bands are identified through feature extraction and recorded as mold characteristic signals for subsequent comparative analysis.

[0082] S104: Compare the extracted spectral features with a preset mildew trend database to determine whether there is a mildew trend in grains:

[0083] The identified key spectral absorption peaks are compared with the pre-stored mildew trend database. The mildew trend database is based on historical data and records the spectral change patterns of different mildew stages, including the spectral characteristic information of early, middle and severe mildew.

[0084] If the comparison result shows that the similarity between the current spectral feature and the moldy pattern exceeds the corresponding preset threshold, it is judged that the current grain has a moldy trend.

[0085] The preset threshold here is usually set based on the statistical results of a large amount of historical data. When establishing the mold trend database, the spectral characteristics of different mold stages were analyzed, including the spectral patterns of the early stage of mold, mild mold, moderate mold, and severe mold. By analyzing the similarity of the spectral characteristics of these different stages, the difference between each stage and the mildew-free state is determined, and a reasonable threshold range is derived.

[0086] Among them, by comparing the current spectral features with the mildew patterns in the database, a support vector machine (SVM) classification algorithm, for example, may be used to classify and identify the mildew trend.

[0087] Low-Field Nuclear Magnetic Resonance (LF-NMR) is a device used to detect the internal microstructure and dynamic characteristics of materials, and is particularly good at analyzing the binding state and distribution of water in materials. In the detection of mildew in granaries, low-field nuclear magnetic resonance can accurately reflect the changes in the water absorption peak during the mildew process by analyzing the ratio of free water and bound water in grains. The near-infrared spectroscopy monitoring system is good at detecting changes in the chemical composition of the grain surface, and LF-NMR supplements the fine analysis of the internal moisture state of grains, which helps to improve the accuracy and comprehensiveness of mildew detection.

[0088] Low-field nuclear magnetic resonance is used to analyze the changes in the water binding state in grains and identify whether the changes in the water absorption peak caused by mildew are normal. Specifically:

[0089] S201: Obtaining the longitudinal relaxation time and transverse relaxation time signal data of the NMR of the grain sample from the low-field NMR device:

[0090] Water in grains can be divided into two types: free water and bound water. Free water is water that is not bound to particles and has a faster relaxation time, while bound water is water that is adsorbed or wrapped in the grain cell structure and has a slower relaxation time.

[0091] Longitudinal relaxation time refers to the time required for the system to return to thermal equilibrium after a magnetic field is applied in one direction; transverse relaxation time refers to the time required for nuclear spins to lose phase consistency. By collecting these two types of relaxation time data, preliminary information about the water binding state can be obtained. The data acquisition of nuclear magnetic signals is carried out at a specific sampling rate and frequency, usually recorded in Hertz (Hz), and contains the relaxation time distribution corresponding to different areas of the grain sample.

[0092] By controlling the magnetic field strength, pulse sequence and signal acquisition window of the low-field nuclear magnetic resonance equipment, the longitudinal relaxation time and transverse relaxation time at different positions in the grain can be effectively extracted. The signal data is stored in the form of a time series for subsequent frequency domain analysis and feature extraction.

[0093] S202: Perform frequency domain conversion on the acquired signal through Fourier transform to remove background noise and abnormal signals, and obtain a clear nuclear magnetic signal curve:

[0094] Fourier transform is a mathematical tool that converts time domain signals into frequency domain signals. It can decompose the original signal into different frequency components and analyze the main characteristics of the signal through the spectrum.

[0095] For nuclear magnetic resonance signals, the relaxation signal in the time domain is converted into a frequency domain signal through Fourier transform to obtain a frequency distribution curve. Since Fourier transform is a relatively mature method for converting the acquired signal into the frequency domain, it will not be described here.

[0096] Through Fourier transform, the abnormal signal peaks in the frequency domain can be identified and removed, ensuring that the obtained nuclear magnetic signal curve is clear and reliable, and the noise components and abnormal signals of the signal curve have been filtered out.

[0097] S203: Calculate the ratio of free water to bound water based on the relaxation time distribution of the longitudinal relaxation time and the transverse relaxation time, and extract the characteristic value of the water binding state:

[0098] By analyzing the frequency domain signal after Fourier transformation, the relaxation time distribution of free water and bound water can be obtained. When analyzing the relaxation time distribution of free water and bound water, a multi-exponential regression model is usually used to fit the relaxation behavior of different types of water.

[0099] The distribution of the longitudinal relaxation time of free water can be expressed as: ;in, is the longitudinal relaxation time distribution function of free water, is the longitudinal relaxation time, is the strength coefficient of the free water signal, is the average longitudinal relaxation time constant of free water.

[0100] The relaxation time distribution of free water follows an exponential decay law, reflecting the rapid relaxation characteristics of the free water molecular motion.

[0101] For bound water, the relaxation time is longer and different time constants are usually used to describe its relaxation behavior: ;in, is the transverse relaxation time distribution function of bound water, is the transverse relaxation time, is the intensity coefficient of bound water signal, is the average transverse relaxation time constant of bound water.

[0102] By fitting these relaxation time distribution functions, the characteristic distribution of each water type can be obtained. Based on the characteristic distribution of each water type, the proportion of free water is calculated: ;in, is the proportion of free water, and are the minimum and maximum longitudinal relaxation times of free water, and are the minimum and maximum transverse relaxation times of bound water, respectively.

[0103] Calculate the proportion of bound water: ;in, is the proportion of bound water.

[0104] S204: Compare the extracted characteristic value of the water binding state with the grain water binding pattern under normal conditions to determine whether the current water absorption peak change is within a normal range:

[0105] The ratio of free water to bound water extracted is compared with the grain water binding pattern under normal conditions. The water binding pattern under normal conditions is obtained through historical experimental data statistics, including the standard distribution and normal range of different water types. The comparison process determines whether the water absorption peak is normal by calculating the difference between the characteristic value and the normal pattern.

[0106] The difference between the current free water ratio and the normal free water ratio is: ;in, is the difference between the current free water ratio and the normal free water ratio, Normal free water ratio.

[0107] The difference between the current bound water ratio and the normal bound water ratio is: ;in, is the difference between the current bound water ratio and the normal bound water ratio, This is the normal bound water ratio.

[0108] If the difference between the current free water ratio and the normal free water ratio and the difference between the current bound water ratio and the normal bound water ratio are both within the preset allowable range, it is determined that the moisture absorption peak change is normal, and the moisture absorption peak change caused by mildew is normal; otherwise, the moisture absorption peak change is abnormal, which may indicate the early stage of mildew.

[0109] The internal optical properties of grains are modeled by a dual fractal analysis algorithm to evaluate the degree of abnormality in the light transmission path caused by grain mildew. Specifically:

[0110] Obtain light transmission path data for grain samples, recording the light propagation time and scattering pattern in different areas:

[0111] Use optical sensors to obtain data on the light transmission paths of grain samples. When light passes through grain samples, it is scattered, absorbed, and reflected. To ensure the accuracy of data collection, ultrashort pulse lasers can be used as light sources with emission wavelengths between 800 and 1200 nanometers. Optical sensors capture the propagation patterns of light inside grains by recording the time it takes for photons to arrive at different detection points and the attenuation information of these photons on different paths.

[0112] The light transmission path data includes light propagation time and scattering angle. The light propagation time reflects the change in the speed of light when it passes through different materials of the grain, and the scattering angle describes the scattering behavior of light at different positions.

[0113] The acquired optical transmission path data is divided into multiple sub-segments according to the temporal and spatial distribution, and each segment corresponds to different light propagation behaviors:

[0114] The definition of each sub-segment depends on the changes in the propagation speed and scattering pattern of light in different areas. For example, when light passes through a more uniform medium, its propagation time is shorter and the scattering angle is smaller; while when it passes through an inhomogeneous medium, its propagation time is longer and the scattering angle is larger. The data for each sub-segment includes the light propagation time and scattering angle in that area. The division of these segments is based on the temporal and spatial distribution of the data, ensuring that different light propagation behaviors can be analyzed separately.

[0115] Apply dual fractal analysis to the data of each segment to extract the self-similar features in the light transmission path and generate multi-dimensional fractal parameters:

[0116] The dual fractal analysis algorithm is suitable for processing systems with complex, multi-scale structures and can reveal the inhomogeneity and complexity of light propagation inside grain.

[0117] For each sub-segment of light propagation time and scattering angle data, dual fractal analysis generates a set of fractal dimensions. These dimensions describe the complex propagation behavior of light in that sub-segment. The core of dual fractal analysis is to calculate the fractal dimension of each region and identify changes in the light propagation path based on multi-scale distribution characteristics.

[0118] The calculation formula of fractal dimension is as follows: ;in, For the The fractal dimension of the sub-segments is used to describe the complexity of the light transmission path; For the In the sub-segment The difference in light travel time for each data point, in seconds; For the In the sub-segment The difference in scattering angles of the data points, in radians; is the length of the sub-segment in meters; Indicates different sub-segment numbers; The numbers representing the different data points in a particular sub-segment; Indicates The total number of data points contained in the sub-segments.

[0119] Based on the extracted self-similar features, a multi-scale fractal model of the light transmission path inside the grain is constructed:

[0120] After obtaining the fractal dimension of each sub-segment, a multi-scale fractal model is constructed based on the fractal dimension of each sub-segment. The multi-scale fractal model characterizes the complexity of the optical transmission path at different scales by combining the fractal dimensions of different segments.

[0121] The construction formula of the multi-scale fractal model is: ;in, It is a multi-scale fractal model used to describe the complexity of the light transmission path inside the whole grain sample; is the number of sub-segments; For the The weight of each sub-segment is assigned according to the importance of each sub-segment in the overall optical transmission path.

[0122] Among them, the weight setting of the sub-segment is determined based on the importance of light propagation in different areas and its impact on the overall light transmission path. The light propagation time and scattering behavior of each sub-segment are different, so the contribution to the complexity of the overall transmission path is also different. The setting of weights can be evaluated by analyzing the changes in the frequency of light transmission, propagation path length and light scattering angle in each sub-segment. The distribution of weights generally follows the following principle: sub-segments with longer light propagation time and larger scattering angles have relatively higher weights, because these areas often reflect the heterogeneity and complexity of the internal structure of the grain, and have a greater impact on the light transmission path. By quantifying the contribution of each sub-segment, its weight is finally set for the construction of a multi-scale fractal model.

[0123] The multi-scale fractal model was compared with the standard light transmission path to quantify the degree of light transmission anomaly caused by mildew:

[0124] Among them, the standard light transmission path model is established based on grain samples that have not been moldy, and is used to evaluate whether there is any abnormality in the light transmission path. Specifically:

[0125] The standard optical transmission path model is established based on the statistics and analysis of optical transmission experimental data of multiple grain samples that have not been moldy. These experimental data are collected by optical sensors under stable conditions to ensure that there is no mold or other structural changes in the grain samples. The optical transmission path data collected in the experiment include multiple parameters such as light propagation time, scattering angle, and light intensity attenuation. These parameters have stable distribution characteristics when they are not moldy.

[0126] In order to construct a standard light transmission path model, the light transmission data of the non-moldy samples were first preprocessed to remove noise and outliers, and then the multi-scale fractal analysis algorithm was applied to extract its self-similar features and generate fractal dimensions. Then, the fractal dimensions of multiple samples were statistically averaged to construct a multi-scale fractal model that describes the standard light transmission path in the non-moldy state. This model is used as a benchmark. When the actual detected light transmission path data deviates from this standard model, it can be judged whether the light transmission path is abnormal.

[0127] Calculate the optical transmission path anomaly index, the expression is: ;in, is the optical transmission path abnormality index, which is used to evaluate the abnormality degree of the optical transmission path; It is the standard light transmission path model.

[0128] The larger the optical transmission path anomaly index is, the greater the deviation of the optical transmission path from the standard model is, reflecting the higher degree of optical transmission path anomaly caused by mildew, indicating that the difference between the optical transmission path of the current grain sample and the standard non-moldy model is more significant. This is because mildew will change the microstructure inside the grain, causing significant changes in the scattering, absorption and reflection behavior of light in the sample, which in turn affects the complexity and self-similarity of the optical transmission path. Therefore, the increase in the optical transmission path anomaly index directly reflects the increased impact of mildew on the optical transmission path, resulting in increased deviation.

[0129] The sparse representation classification algorithm is used to extract features from the grain surface microstructure data and evaluate the degree of change in scattering characteristics caused by grain mildew. Specifically:

[0130] Obtain light scattering intensity distribution data on the grain surface and record the reflection and scattering patterns of light by the surface microstructure:

[0131] The grain surface is scanned using a high-resolution optical imaging device combined with a light scattering measurement device. The light source uses an adjustable wavelength laser, with a wavelength range set between 650 nanometers and 950 nanometers to cover common visible and near-infrared spectra. The light beam is irradiated perpendicular to the grain surface, and the intensity distribution of the reflected and scattered light is recorded by the detector.

[0132] During the data collection process, the light scattering intensity value of each detection point is recorded as ,in, Indicates the coordinate position The wavelength is , the scattering angle is The light scattering intensity at .

[0133] : The coordinate position of the two-dimensional plane of the grain surface, in meters (m).

[0134] : The wavelength of light, in nanometers (nm).

[0135] : Scattering angle refers to the angle at which light deviates from the direction of incident light, measured in degrees (°).

[0136] By measuring at different positions, wavelengths and scattering angles, a complete data set of the reflection and scattering patterns of light from the grain surface microstructure is obtained.

[0137] The obtained light scattering intensity distribution data is processed by multi-dimensional vectorization, and the scattering information at different angles and wavelengths is integrated into an analyzable multi-dimensional vector data set:

[0138] The continuous coordinate position, wavelength and scattering angle are discretized into a finite set of values ​​to form a three-dimensional grid.

[0139] For each sampling point, a multidimensional vector is constructed. Each component of the vector corresponds to the light scattering intensity value at different wavelengths and scattering angles. The multidimensional vector is expressed as: ;in, It is The multidimensional vector representation of the sampling points is It is The coordinate positions of the sampling points, is selected The wavelength value, is selected The scattering angle value is is the corresponding light scattering intensity value, Represents the index of the sampling point, Indicates the number of selected wavelength values, Indicates the number of selected scattering angle values.

[0140] Combine the vectors of all sample points into a data matrix: ;in, The dimension is The data matrix, is the total number of sampling points, It is The vector transpose of the sampling points, Represents the transpose operation of a matrix or vector.

[0141] The sparse representation classification algorithm is used to process the converted multidimensional vector data set to extract sparse features that can reflect the changes in the scattering characteristics of the grain surface:

[0142] The core idea of ​​sparse representation classification algorithm is to represent high-dimensional data as a linear combination of training dictionaries with only a few non-zero coefficients. This helps to extract key features in the data, reduce dimensionality, and improve classification efficiency.

[0143] Use the pre-collected data on the grain surface scattering properties under normal conditions to construct a training dictionary matrix: ;in, The dimension is Dictionary matrix of ; It is dictionary atoms, representing a scattering characteristic pattern in the normal state; is the number of atoms in the dictionary, Represents the atom index in the training dictionary.

[0144] for , solve for the sparse coefficient vector such that: ;in, is a sparse coefficient vector with dimension , where most of the elements are zero; is the residual vector, representing the reconstruction error.

[0145] The problem of solving for a sparse coefficient vector can be formulated as an optimization problem: ;in, is the Euclidean norm, which represents the sum of squares of the residuals; yes Norm, which represents the absolute value and of the sparse coefficients; is a regularization parameter that controls the degree of sparsity.

[0146] By solving the optimization problem, the sparse coefficient vector of each sample is obtained, and the sparse coefficient vector reflects the representation of the sample in the dictionary.

[0147] Take the sparse coefficient vector as the feature and construct the feature matrix: ;in, The dimension is The feature matrix of .

[0148] Compare the sparse features extracted within a preset time period with the standard scattering pattern, identify the deviation of scattering characteristics caused by mildew, and evaluate the degree of scattering change caused by grain mildew:

[0149] For the feature matrix , calculate the average sparse coefficient vector within a preset time period.

[0150] The standard sparse feature vector under normal conditions is obtained. The standard sparse feature vector is the average sparse coefficient calculated in the normal grain sample.

[0151] The relative change rate is used to evaluate the degree of change in the scattering characteristics and the mildew scattering change index is calculated. The expression is: ;in, is the mildew scattering variation index; is the Euclidean norm, which is used to calculate the difference between two sparse coefficient vectors; is the average sparse coefficient vector extracted within a preset time period, representing the characteristics of the current grain surface scattering properties; is the standard sparse feature vector under normal conditions, used as a comparison benchmark.

[0152] The larger the mold scattering change index, the greater the degree of change in the scattering characteristics caused by grain mold, which means that the light scattering behavior of the grain surface microstructure has deviated significantly, reflecting the strong impact of mold on the surface structure. Significant changes in scattering characteristics may interfere with the accuracy of spectral data because scattering behavior affects the propagation and absorption characteristics of light inside the grain. The accuracy of near-infrared spectroscopy monitoring system for monitoring grain mold may be reduced. This is because the scattering changes caused by mold will cause distortion of spectral data, which in turn affects the accuracy of the system's detection and analysis results of mold.

[0153] Among them, the preset time period is set according to actual needs and will not be repeated here.

[0154] Changes in moisture binding state are a direct indicator of grain mildew. If there is an obvious moisture abnormality, it means that the mildew process has reached a more serious stage and may have a greater impact on the physical and chemical properties of the grain. In this case, direct warning of mildew itself will be more urgent and effective. When the moisture binding state is normal, mildew is still in the early stage. Although it has little effect on the physical and chemical properties of the grain itself, it has begun to gradually affect the accuracy of the near-infrared spectroscopy monitoring system. In this case, it is more appropriate to establish an interference model of mildew on the monitoring system, because it can detect the potential adverse effects of mildew on the monitoring system in advance and help the system maintain the accuracy of monitoring. If the moisture binding state is already abnormal, the monitoring accuracy of the system may have been greatly disturbed, and it is too late to establish an interference model. Therefore, taking action in the early stage of mildew (normal moisture state) can ensure earlier identification of monitoring errors and adjustments. By early warning of the interference of the monitoring system rather than directly detecting mildew, too late reaction is avoided, thereby ensuring the advance of monitoring and the effectiveness of the system.

[0155] Based on the degree of abnormality in the light transmission path and the degree of change in scattering characteristics caused by grain mildew, an interference model of mildew on the accuracy of grain mildew monitoring was established to quantify the potential adverse impact of grain mildew on the accuracy of near-infrared spectroscopy monitoring system in monitoring grain mildew, specifically:

[0156] The specific steps for establishing the interference model of mildew on the accuracy of grain mildew monitoring are as follows:

[0157] The optical transmission path anomaly index and the mildew scattering change index are normalized, and weight coefficients are assigned to the normalized optical transmission path anomaly index and the mildew scattering change index, respectively, to calculate the mildew impact monitoring coefficient.

[0158] The specific implementation method of calculating the mildew impact monitoring coefficient is not specifically limited here. Any method that can calculate the mildew impact monitoring coefficient by assigning weight coefficients to the normalized optical transmission path anomaly index and the mildew scattering change index respectively is sufficient. In order to implement the technical solution of the present invention, the present invention provides a specific method for calculating the mildew impact monitoring coefficient.

[0159] The expression of the monitoring coefficient of mildew impact is: ;in, is the monitoring coefficient of mildew impact, and are the weight coefficients of the optical transmission path anomaly index and the mildew scattering variation index, respectively, and and Both are greater than 0.

[0160] in, and The relative impact of optical transmission path anomalies and scattering characteristic changes on the monitoring system error can be determined by fitting experimental data.

[0161] The greater the monitoring coefficient of mold, the greater the potential adverse impact of grain mold on the accuracy of the near-infrared spectroscopy monitoring system in monitoring grain mold, which means that the mold has a more significant impact on the internal light transmission and surface scattering properties of grain. These changes will cause distortion of near-infrared spectral data and interfere with the monitoring system's accurate detection of mold.

[0162] When the potential adverse effect of grain mildew on the accuracy of grain mildew monitoring by the near infrared spectroscopy monitoring system exceeds the preset threshold, a monitoring warning signal is issued, specifically:

[0163] Setting preset thresholds , It is set based on historical data analysis, experimental results, and the monitoring system's requirements for mold detection accuracy. First, by performing near-infrared spectroscopy on multiple grain samples that are not moldy and at different moldy stages, the corresponding monitoring error data is collected. Next, by analyzing the impact of mold on the light transmission path and scattering characteristics, the distribution of monitoring errors at each stage is statistically analyzed. Then, a value with a small monitoring error in the experiment and an acceptable impact on monitoring accuracy is selected as the threshold, which is usually set to the upper limit of the small but still acceptable impact of mold on the system. This threshold can be optimized through tuning to ensure that the monitoring system still has sufficient reliability and accuracy within the error range.

[0164] The mold impact monitoring coefficient and the preset threshold For comparison:

[0165] When the mold impact monitoring coefficient is greater than the preset threshold When the temperature reaches 0, it is determined that a monitoring warning signal is issued. At this time, the potential adverse effect of grain mildew on the accuracy of the near-infrared spectroscopy monitoring system in monitoring grain mildew is relatively large, which indicates that mildew has a significant adverse effect on the monitoring results of the near-infrared spectroscopy monitoring system, and the monitoring error is large, which may lead to inaccurate detection of mildew. At this time, the system will remind you to take further measures, such as adjusting the monitoring strategy or conducting more data analysis.

[0166] When the mold impact monitoring coefficient is less than or equal to the preset threshold When , it is determined that a normal monitoring signal is issued. At this time, the potential adverse effect of grain mildew on the accuracy of near-infrared spectroscopy monitoring system in monitoring grain mildew is normal, indicating that the impact of mildew on the near-infrared spectroscopy monitoring system is within a controllable range, the monitoring error is small, and the monitoring results are still accurate and reliable, indicating that the current degree of mildew impact is within the normal range.

[0167] When a monitoring warning signal is issued, the monitoring strategy is dynamically adjusted using an adaptive optimization algorithm, specifically:

[0168] When the impact of mildew is significant, the frequency of data collection is increased through the adaptive optimization algorithm. By increasing the acquisition frequency, more spectral information can be captured, making the data more refined, thereby reducing random errors and noise interference caused by mildew.

[0169] Automatically start advanced signal processing algorithms, such as filters and denoising algorithms, to filter out noise and abnormal signals introduced by mildew. At the same time, conduct in-depth analysis of the collected data to remove unstable or abnormal data to ensure that the remaining data is of high quality and reliable.

[0170] Dynamically adjust the parameters in the spectral analysis according to the current degree of mildew. For example, you can change the wavelength range of spectral acquisition to select a specific band suitable for detecting mildew, or adjust the spectral resolution to make it more adaptable to the changes in light scattering or absorption caused by mildew.

[0171] Recalibrate the previous monitoring model based on the new collected data. Through the adaptive algorithm, the model will automatically adjust the parameters according to the deviation of the current data, so that the output of the model is more in line with the actual monitoring situation, thereby reducing the error caused by the influence of mold.

[0172] Example 2

[0173] The difference between Example 2 of the present invention and Example 1 is that this example introduces a grain mildew early warning system.

[0174] Figure 2 A structural schematic diagram of a grain mildew early warning system of the present invention is given, which includes a mildew trend judgment module, a moisture combination recognition module, a model parameter acquisition module, an interference model establishment module and a monitoring and early warning judgment module.

[0175] Mildew trend judgment module: obtains grain mildew conditions from the near-infrared spectroscopy monitoring system to determine whether there is a grain mildew trend.

[0176] Moisture binding identification module: Analyze the changes in the moisture binding state in grain through low-field nuclear magnetic resonance to identify whether the changes in the moisture absorption peak caused by mildew are normal.

[0177] Model parameter acquisition module: When there is a trend of grain mildew and the changes in the moisture absorption peak caused by mildew are normal, the internal optical properties of the grain are modeled through the dual fractal analysis algorithm to evaluate the degree of abnormality in the light transmission path caused by grain mildew; the sparse representation classification algorithm is used to extract features from the grain surface microstructure data to evaluate the degree of change in the scattering characteristics caused by grain mildew.

[0178] Interference model establishment module: Based on the abnormal degree of light transmission path and the degree of change in scattering characteristics caused by grain mildew, an interference model of mildew on the accuracy of grain mildew monitoring is established to quantify the potential adverse impact of grain mildew on the accuracy of near-infrared spectroscopy monitoring system in monitoring grain mildew.

[0179] Monitoring and early warning judgment module: When the potential adverse impact of grain mold on the accuracy of grain mold monitoring by the near-infrared spectroscopy monitoring system exceeds the preset threshold, a monitoring early warning signal is issued.

[0180] Example 3

[0181] A device comprises: a processor, a memory, and a program or instruction stored in the memory and executable on the processor. When the program or instruction is executed by the processor, a method for early warning of grain mildew is implemented.

[0182] Example 4

[0183] A computer-readable storage medium stores a program or instruction, and a method for early warning of grain mildew is implemented when the program or instruction is executed by a processor.

[0184] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0185] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or may be transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0186] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0187] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0188] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0189] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0190] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0191] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0192] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0193] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for early warning of grain mildew, characterized in that: The steps include: S1: Obtain the grain mildew situation from the near infrared spectrum monitoring system to determine whether there is a grain mildew trend; S2: Analyze the changes in the water binding state in grains through low-field nuclear magnetic resonance to identify whether the changes in the water absorption peak caused by mildew are normal; S3: When there is a trend of grain mildew and the change of moisture absorption peak caused by mildew is normal, the internal optical properties of grain are modeled by dual fractal analysis algorithm to evaluate the abnormal degree of light transmission path caused by grain mildew; The sparse representation classification algorithm is used to extract features from grain surface microstructure data and evaluate the degree of change in scattering characteristics caused by grain mildew. S4: Based on the degree of abnormality in the light transmission path and the degree of change in scattering characteristics caused by grain mildew, an interference model of mildew on the accuracy of grain mildew monitoring is established to quantify the potential adverse impact of grain mildew on the accuracy of near-infrared spectroscopy monitoring system in monitoring grain mildew; S5: When the potential adverse effect of grain mildew on the accuracy of grain mildew monitoring by the near infrared spectroscopy monitoring system exceeds a preset threshold, a monitoring warning signal is issued; Low-field nuclear magnetic resonance is used to analyze the changes in the water binding state in grains and identify whether the changes in the water absorption peak caused by mildew are normal. Specifically: S201: Acquire longitudinal relaxation time and transverse relaxation time signal data of nuclear magnetic resonance of a grain sample from a low-field nuclear magnetic resonance device; S202: Performing frequency domain conversion on the acquired signal by Fourier transform to remove background noise and abnormal signals, and obtaining a nuclear magnetic signal curve; S203: Calculate the ratio of free water to bound water according to the relaxation time distribution of the longitudinal relaxation time and the transverse relaxation time, and extract the characteristic value of the water binding state; S204: comparing the extracted characteristic value of the water binding state with the grain water binding pattern under normal conditions to determine whether the current water absorption peak change is within a normal range; The internal optical properties of grains are modeled by a dual fractal analysis algorithm to evaluate the degree of abnormality in the light transmission path caused by grain mildew. Specifically: Obtain light transmission path data for grain samples, recording the propagation time and scattering pattern of light in different areas; Divide the acquired optical transmission path data into multiple sub-segments according to the temporal and spatial distribution, each segment corresponding to a different optical propagation behavior; Apply dual fractal analysis to the data of each segment to extract the self-similar features in the light transmission path and generate multi-dimensional fractal parameters: For each subsegment’s light travel time and scattering angle data, dual fractal analysis generates a set of fractal dimensions; The calculation formula of fractal dimension is: ;in, For the The fractal dimension of the sub-segments, For the In the sub-segment The difference in light travel time for each data point is For the In the sub-segment The difference in scattering angles of the data points is is the length of the sub-segment, Indicates different sub-segment numbers; Indicates the number of different data points in the sub-segment; Indicates The total number of data points contained in the sub-segments; Based on the extracted self-similar features, a multi-scale fractal model of the light transmission path inside the grain is constructed: A multi-scale fractal model is constructed based on the fractal dimension of each sub-segment; the construction formula of the multi-scale fractal model is: ;in, is a multi-scale fractal model. is the number of sub-segments, For the The weight of each sub-segment; The multi-scale fractal model was compared with the standard light transmission path to quantify the degree of light transmission anomaly caused by mildew: Calculate the optical transmission path anomaly index, the expression is: ;in, is the optical transmission path anomaly index, It is the standard light transmission path model; The sparse representation classification algorithm is used to extract features from the grain surface microstructure data and evaluate the degree of change in scattering characteristics caused by grain mildew. Specifically: Obtain light scattering intensity distribution data on the grain surface and record the reflection and scattering patterns of light by the surface microstructure; The obtained light scattering intensity distribution data is processed by multi-dimensional vectorization, and the scattering information at different angles and wavelengths is integrated into an analyzable multi-dimensional vector data set: Discretize the continuous coordinate position, wavelength and scattering angle into a finite set of values ​​to form a three-dimensional grid; For each sampling point, a multidimensional vector is constructed, and each component of the vector corresponds to the light scattering intensity value at different wavelengths and scattering angles; Combine the vectors of all sampling points into a data matrix; The sparse representation classification algorithm is used to process the converted multidimensional vector data set to extract sparse features that can reflect the changes in the scattering characteristics of the grain surface: Using the pre-collected data on the surface scattering characteristics of grains in normal conditions, a training dictionary matrix is ​​constructed; Solve for the sparse coefficient vector such that: ;in, is a sparse coefficient vector, is the residual vector, It is Multidimensional vector representation of sampling points; The dimension is The dictionary matrix of is the number of atoms in the dictionary, Indicates the number of selected wavelength values, Indicates the number of selected scattering angle values; By solving the optimization problem, the sparse coefficient vector of each sample is obtained; Use the sparse coefficient vector as a feature to construct a feature matrix; Compare the sparse features extracted within a preset time period with the standard scattering pattern, identify the deviation of scattering characteristics caused by mildew, and evaluate the degree of scattering change caused by grain mildew: For the feature matrix, calculate the average sparse coefficient vector within a preset time period; Get the standard sparse feature vector under normal conditions; The relative change rate is used to evaluate the degree of change in the scattering characteristics and the mildew scattering change index is calculated. The expression is: ;in, is the mildew scattering variation index, is the average sparse coefficient vector extracted within a preset time period, is the standard sparse feature vector in the normal state.

2. The method for early warning of grain mildew according to claim 1, characterized in that: Obtain the grain mildew situation from the near infrared spectrum monitoring system to determine whether there is a grain mildew trend, specifically: S101: acquiring real-time spectral data from a near infrared spectral monitoring system and storing spectral response information in a food environment; S102: preprocessing the acquired spectral data to remove noise and outliers to obtain a standardized spectral data set; S103: extracting features from the standardized spectral data using a preset spectral analysis algorithm to identify key spectral absorption peaks; S104: Compare the extracted spectral features with a preset mildew trend database to determine whether there is a mildew trend in grain.

3. The method for early warning of grain mildew according to claim 2, characterized in that: Based on the degree of abnormality in the light transmission path and the degree of change in scattering characteristics caused by grain mildew, an interference model of mildew on the accuracy of grain mildew monitoring was established to quantify the potential adverse impact of grain mildew on the accuracy of near-infrared spectroscopy monitoring system in monitoring grain mildew, specifically: The optical transmission path anomaly index and the mildew scattering change index are normalized, and weight coefficients are assigned to the normalized optical transmission path anomaly index and the mildew scattering change index, respectively, to calculate the mildew impact monitoring coefficient.

4. The method for early warning of grain mildew according to claim 3, characterized in that: When the potential adverse effect of grain mildew on the accuracy of grain mildew monitoring by the near infrared spectroscopy monitoring system exceeds the preset threshold, a monitoring warning signal is issued, specifically: Setting preset thresholds ; The mold impact monitoring coefficient and the preset threshold For comparison: When the mold impact monitoring coefficient is greater than the preset threshold When the monitoring warning signal is issued, When the mold impact monitoring coefficient is less than or equal to the preset threshold When , it is determined that the monitoring normal signal is sent.

5. A grain mildew early warning system, used to implement the grain mildew early warning method according to any one of claims 1 to 4, characterized in that: It includes a mold trend judgment module, a moisture combination identification module, a model parameter acquisition module, an interference model establishment module, and a monitoring and early warning judgment module; Mildew trend judgment module: obtains grain mildew conditions from the near-infrared spectrum monitoring system to determine whether there is a grain mildew trend; Moisture binding identification module: Analyze the changes in moisture binding state in grains through low-field nuclear magnetic resonance to identify whether the changes in moisture absorption peaks caused by mildew are normal; Model parameter acquisition module: When there is a trend of grain mildew and the moisture absorption peak caused by mildew changes normally, the internal optical properties of the grain are modeled through the dual fractal analysis algorithm to evaluate the degree of abnormality of the light transmission path caused by grain mildew; The sparse representation classification algorithm is used to extract features from grain surface microstructure data and evaluate the degree of change in scattering characteristics caused by grain mildew. Interference model building module: Based on the degree of abnormality in the light transmission path and the degree of change in scattering characteristics caused by grain mildew, an interference model of mildew on the accuracy of grain mildew monitoring is established to quantify the potential adverse impact of grain mildew on the accuracy of near-infrared spectroscopy monitoring system in monitoring grain mildew; Monitoring and early warning judgment module: When the potential adverse impact of grain mold on the accuracy of grain mold monitoring by the near-infrared spectroscopy monitoring system exceeds the preset threshold, a monitoring early warning signal is issued.

6. A grain mildew early warning device, characterized in that: The grain mold early warning device comprises: a processor, a memory, and a program or instruction stored in the memory and executable on the processor. When the program or instruction is executed by the processor, the grain mold early warning method as described in any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium, characterized in that: A program or instruction is stored on a computer-readable storage medium, and when the program or instruction is executed by a processor, the grain mold early warning method as described in any one of claims 1 to 4 is implemented.

Images