A grain toxin intelligent detection system and method

Abstract

The invention discloses an intelligent detection system and method for grain toxins, relates to the technical field of grain detection, and is used to improve the problem of grain secondary pollution caused by ignoring fungal competition. The system and method include collecting sample data, collecting Fusarium graminearum proliferation information and calculating the Fusarium graminearum proliferation rate, detecting the content of vomitoxin and zearalenone in the sample and determining the degree of sample damage, using a ridge regression method to comprehensively consider the degree of sample damage and the Fusarium graminearum proliferation rate to determine the optimal sample detection time, screening the samples according to the optimal sample detection time and counting the Fusarium graminearum content, when the Fusarium graminearum content is low, collecting grain damage data and environmental data to evaluate the degree of other fungal growth, determining the growth rate of other fungi according to the evaluation result, determining their competitive relationship in combination with the Fusarium graminearum proliferation rate, and providing different prediction algorithm models for Fusarium graminearum proliferation through their competitive relationship to reduce the risk of grain secondary pollution and improve food safety.

Description

Technical Field

[0001] The present invention relates to the field of grain detection technology, and more specifically, to a grain toxin intelligent detection system and method. Background Art

[0002] Grain testing technology refers to the technology that uses various groups of technical means to identify, classify, count, and evaluate the quality of grains. It plays an important role in agricultural production, food processing, trade circulation and other fields. When grain testing technology is used to test grains, it can predict the trend of fungal growth, thereby improving grain safety.

[0003] The prior art has the following deficiencies:

[0004] In the past, the prediction of the proliferation trend of Fusarium graminearum was mostly carried out by detecting its historical content or its main products vomitoxin and zearalenone for analysis and processing. The competition between Fusarium graminearum and other fungi and the growth inhibition of other fungi by vomitoxin and zearalenone were not taken into account. When the content of Fusarium graminearum is low, it will cause other fungi to grow faster, leading to secondary contamination of grains and posing a food safety hazard. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a system and method for intelligent detection of grain toxins, which predicts Fusarium graminearum by screening the optimal sample conditions and analyzing the competition relationship between fungi to select different algorithm models 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] A method for intelligent detection of grain toxins comprises the following steps:

[0008] Step S1, obtaining multiple samples, detecting and calculating the proliferation rate of Fusarium graminearum respectively, and collecting the contents of vomitoxin and zearalenone in the samples to determine the degree of sample damage;

[0009] Step S2, using the sample damage degree and the proliferation rate of Fusarium graminearum to generate the optimal sample detection time, screening the samples according to the sample detection time and counting the content of Fusarium graminearum;

[0010] Step S3, when the content of Fusarium graminearum is low, the overall grain damage rate is detected by a capacitance measurement method, the historical data of the enterprise database is accessed to obtain a damage comparison table, and the grain damage rate caused by the content of Fusarium graminearum is determined according to the damage comparison table. The growth rate of other fungi is analyzed according to the overall grain damage rate and the grain damage rate caused by Fusarium graminearum, and the competitive relationship between the growth rates of other fungi and the value-added rate of Fusarium graminearum is determined by comparing the relationship between the growth rates of other fungi and the value-added rate of Fusarium graminearum;

[0011] Step S4, providing a linear regression model to predict the value-added of Fusarium graminearum based on the competitive relationship.

[0012] In a preferred embodiment, in step S1, when the system obtains samples, it selects equal amounts of grains from the same grain pile as analysis samples multiple times, and randomly sets two time points with different time intervals in each analysis sample to detect the content of Fusarium graminearum. The system uses the ratio of the content difference to the corresponding time interval as the Fusarium graminearum proliferation rate and records the time interval data of each analysis sample.

[0013] In a preferred embodiment, in step S1, the system detects the content of vomitoxin and zearalenone at the midpoint of each analysis sample and normalizes them respectively and then sums them as the sample damage degree of the corresponding analysis sample.

[0014] In a preferred embodiment, in step S2, the system uses the Fusarium graminearum proliferation rate and time interval data of each analyzed sample as the influencing feature, and uses the sample loss degree as the screening feature to construct a ridge regression model, and uses the ridge regression model to find the optimal detection time interval to detect the Fusarium graminearum content and count it;

[0015] The influencing features are the parameters that affect the accuracy of Fusarium graminearum content detection, and the screening features are the real labels used to screen samples.

[0016] In a preferred embodiment, in step S2, the ridge regression model searches for the optimal detection time interval as follows:

[0017] In step S2, the ridge regression model searches for the optimal detection time interval as follows:

[0018] Data preparation: The proliferation rate and time interval data of Fusarium graminearum were used as the influencing feature input, and the sample loss degree was used as the screening feature input;

[0019] Add regularization term: construct loss function to determine the influence weight of Fusarium graminearum proliferation rate and time interval data on the accuracy of Fusarium graminearum content analysis;

[0020] Solving for optimal weights: Using the gradient descent method to solve the optimal weights of Fusarium graminearum proliferation rate and time interval data in the loss function according to the screening features;

[0021] Output prediction results: Determine the optimal detection time interval based on the optimal weight of the Fusarium graminearum proliferation rate and time interval data.

[0022] In a preferred embodiment, in step S3, the Fusarium graminearum content detected at the optimal detection time interval is compared with a preset qualified threshold value, and when the Fusarium graminearum content is lower than the preset qualified threshold value, it is judged that the Fusarium graminearum content is low.

[0023] In a preferred embodiment, in step S3, the system collects grain damage rate and storage humidity before other fungal growth rates to determine the error of analyzing the competitive relationship, calibrates the analysis results of other fungal growth rates according to the error of the competitive relationship judgment, and increases the error tolerance of analyzing the competitive relationship.

[0024] In a preferred embodiment, in step S4, a comparison threshold is set. When the proliferation rate of Fusarium graminearum and the growth rate of other fungi are outside the comparison threshold, the system provides a linear regression model to predict the proliferation of Fusarium graminearum; when the proliferation rate of Fusarium graminearum and the growth rate of other fungi are within the comparison threshold, the system provides a machine learning model to predict the proliferation of Fusarium graminearum.

[0025] A grain toxin intelligent detection system, used to implement the above-mentioned grain toxin intelligent detection method, including a data acquisition module, a sample screening module, a competition analysis module and a selection prediction module;

[0026] The data collection module is used to collect and analyze the data of Fusarium graminearum in the sample and send it to the sample screening module;

[0027] The sample screening module is used to determine the optimal detection time interval and screen the analysis samples;

[0028] The competition analysis module is used to analyze the competition relationship between fungi in the screened analysis samples and transmit the analysis results to the selection prediction module;

[0029] The prediction module selects different algorithm models to predict the proliferation of Fusarium graminearum based on the analysis results of the competition relationship between fungi.

[0030] The technical effects and advantages of the grain toxin intelligent detection system and method of the present invention are as follows:

[0031] The present invention collects sample data, collects the proliferation situation of Fusarium graminearum and calculates the proliferation rate of Fusarium graminearum, detects the content of vomitoxin and zearalenone in the sample and determines the degree of sample damage, comprehensively determines the optimal sample detection time based on the sample damage degree and the proliferation rate of Fusarium graminearum, screens the samples according to the optimal sample detection time and counts the content of Fusarium graminearum, collects grain damage data and environmental data to evaluate the degree of other fungi when the content of Fusarium graminearum is low, determines the competitive relationship between them according to the evaluation result and the proliferation rate of Fusarium graminearum, and provides different prediction algorithm models for the proliferation of Fusarium graminearum through their competitive relationship to reduce the risk of secondary contamination of grains and improve food safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1This is a schematic diagram of an intelligent detection method for grain toxins according to the present invention.

[0033] Figure 2 This is a flow chart of an intelligent grain toxin detection system of the present invention. DETAILED DESCRIPTION

[0034] 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.

[0035] The present invention collects sample data, collects the proliferation situation of Fusarium graminearum and calculates the proliferation rate of Fusarium graminearum, detects the contents of vomitoxin and zearalenone in the sample and determines the degree of sample damage; determines the optimal sample detection time based on the degree of sample damage and the proliferation rate of Fusarium graminearum, screens the samples according to the optimal sample detection time and counts the content of Fusarium graminearum, and when the content of Fusarium graminearum is low, collects grain damage data and environmental data to evaluate the degree of other fungi breeding, determines their competitive relationship according to the evaluation result combined with the proliferation rate of Fusarium graminearum, and provides different prediction algorithm models for the proliferation of Fusarium graminearum through their competitive relationship to reduce the risk of secondary contamination of grains and improve food safety.

[0036] Example 1, a method for intelligent detection of grain toxins, such as Figure 1 As shown, the following steps are included:

[0037] Step S1, obtaining multiple samples, detecting and calculating the proliferation rate of Fusarium graminearum respectively, and collecting the contents of vomitoxin and zearalenone in the samples to determine the degree of sample damage;

[0038] Step S2, using the sample damage degree and the proliferation rate of Fusarium graminearum to generate the optimal sample detection time, screening the samples according to the sample detection time and counting the content of Fusarium graminearum;

[0039] Step S3, when the content of Fusarium graminearum is low, collecting the grain damage rate and storage humidity to evaluate the growth rate of other fungi and determine their competitive relationship in combination with the growth rate of Fusarium graminearum;

[0040] Step S4, providing an optimal prediction algorithm for Fusarium graminearum value-added based on the competitive relationship.

[0041] The specific implementation is as follows:

[0042] In step S1, equal amounts of grains from the same grain pile are selected multiple times as analysis samples for analysis, and the content of Fusarium graminearum is detected at two time points with different time intervals randomly set in each analysis sample. After obtaining the content of Fusarium graminearum, the system subtracts the content of Fusarium graminearum detected at the latter time point from the content of Fusarium graminearum detected at the previous time point, and divides the calculated result by the corresponding time interval to obtain the Fusarium graminearum proliferation rate.

[0043] The system combines the Fusarium graminearum growth rates calculated for each analysis sample into a Fusarium graminearum growth rate data set, and combines the time intervals for calculating the Fusarium graminearum growth rates in each analysis sample into a detection time data set.

[0044] The system marks the data in the Fusarium graminearum growth rate data set with different marks, selects the midpoint of the analysis sample time interval corresponding to the different marked data as the detection time point of the vomitoxin and zearalenone content, collects the vomitoxin and zearalenone content in each analysis sample at the detection time point, and normalizes and adds the vomitoxin and zearalenone content in each analysis sample collected as the sample damage degree of the corresponding analysis sample. The sample loss degree calculated for each analysis sample and the different marked data in the Fusarium graminearum growth rate data set are merged into a loss degree data set in sequence.

[0045] It should be noted that the number of analysis samples can be selected according to actual conditions. When normalizing the vomitoxin and zearalenone contents in each analysis sample, the maximum and minimum values ​​of the vomitoxin content or zearalenone content in the analysis sample can be selected, and after summing the maximum value and the minimum value, the vomitoxin or zearalenone content in each analysis sample is used as the numerator, and the corresponding summation result is used as the denominator to perform a ratio operation, and the obtained ratio is used as the normalized result of the vomitoxin content or zearalenone content in the corresponding analysis sample. For example, the number of analysis samples is 5 groups, and the vomitoxin contents in the 5 groups are 800µg / kg, 900µg / kg, 1000µg / kg, 1100µg / kg, and 1200µg / kg, respectively. Taking the vomitoxin content of 800µg / kg as an example, the normalized result is 800 / (800+1200)=0.4. The zearalenone content is normalized in the same way, which is not described here.

[0046] In step S2, the system uses the Fusarium graminearum growth rate dataset and the detection time dataset as influencing features, and the loss degree dataset as a screening feature to construct a ridge regression model. The ridge regression model is used to screen and analyze samples, and the analysis samples with the best detection time interval are found to detect and count the Fusarium graminearum content. The specific steps of using the ridge regression model to screen and analyze samples are as follows:

[0047] Data preparation: When the system is looking for the best detection time with the lowest loss degree, the Fusarium graminearum growth rate dataset, detection time dataset and corresponding loss degree dataset calculated for each analysis sample are used as input.

[0048] Add regularization term: Construct the loss function and add regularization term to set the influence weight coefficient of the growth rate and time interval of Fusarium graminearum in each analysis sample. The loss function calculation formula is: ,in is the loss function, is the weight coefficient vector, represents the transpose of the weight coefficient vector, is the corresponding loss degree of the i-th analysis sample, is the feature vector of the i-th analysis sample, is the regularization coefficient, is the weight coefficient vector The L2 norm square of the weight coefficient vector The sum of the squares of all elements in .

[0049] It needs to be explained that It is a function of the weight coefficient vector. Its value represents the prediction error of the model under the current weight coefficient. The ridge regression model determines the weight coefficient of each feature by finding the optimal weight coefficient vector to minimize the loss function. The weight coefficient vector is an array containing all the weights of the features that affect the model, the feature vector of the i-th analysis sample is an array containing all the influencing features in the i-th analysis sample, that is, the growth rate and time interval of Fusarium graminearum in the i-th analysis sample. For example, if the growth rate of Fusarium graminearum in the i-th analysis sample is 1.2 / h, that is, the growth rate of Fusarium graminearum in the i-th analysis sample is 1.2 times per hour, the time interval is 2h, and the sample loss degree is 0.9, then is [1.2,2], is 0.9, by transposing the weight coefficient vector And the preset regularization coefficient can be used to calculate the loss function. The product of the regularization coefficient and the square of the L2 norm of the weight coefficient vector is the regularization term. The role of the regularization term is to penalize the modulus of the weight coefficient vector, so that the ridge regression model is more inclined to select a model with a smaller weight value. The influencing feature is the influencing parameter on the accuracy of the detection of Fusarium graminearum content, and the screening feature is the real label used to screen the sample.

[0050] Solving for the optimal weight coefficient: Using the gradient descent method to solve the minimum weight coefficient of the loss function has the following steps:

[0051] Step 1: Calculate the gradient of the loss function and convert the loss function For the weight coefficient vector Perform a derivative and use the derivative result as the gradient of the loss function.

[0052] Step 2: Initialize the weight coefficient vector. Initialized to 1, that is, the data in the array of all influencing features in the analyzed sample is set to 1.

[0053] Step 3, iteratively update the weight coefficient, and update the weight coefficient through the weight coefficient iterative update formula, and the iterative update formula can be: ,in is the learning rate, That is, the gradient of the loss function can be used to set different learning rates according to the degree of change of different influencing features, thereby changing the weight coefficient vector , by repeating the iterative process until the loss function converges to the minimum value, that is, the gradient of the loss function is close to zero, and the corresponding weight coefficient vector is taken as the optimal weight coefficient vector.

[0054] Output prediction results: take the average value in the Fusarium graminearum growth rate data set as quantitative 1 and mark it as Y, take the minimum value in the loss degree data set as quantitative 2 and mark it as Z, and use the prediction formula to calculate the time interval. The prediction formula is: Z=w0+w1Y+w2X, where w0, w1 and w2 are the three coefficient weights in the optimal weight coefficient vector, w1 and w2 are the corresponding optimal weight coefficients in the Fusarium graminearum growth rate data set and the detection time data set respectively, w0 is the constant weight coefficient obtained when calculating the gradient of the loss function, and X is the optimal detection time interval output by the ridge regression model.

[0055] In the same grain pile, the optimal detection time interval and the same amount of grain as the analysis sample were selected as the detection sample. For the detection sample, multiple time points were set to detect the content of Fusarium graminearum, and the average value was calculated as the actual content of Fusarium graminearum. The optimal detection time interval calculated by the ridge regression model was used to detect the content of Fusarium graminearum, which improved the data accuracy and provided a data basis for the subsequent analysis of the competitive relationship between fungi.

[0056] It should be noted that the construction of the loss function and the calculation method of the loss function in the ridge regression model are not unique. The regularization coefficient involved in the loss function and the learning rate used to iteratively update the weight coefficient when calculating the loss function are set by professionals in this field and are not analyzed here.

[0057] In step S3, the actual content of Fusarium graminearum is compared with a preset qualified threshold value. When the actual content of Fusarium graminearum is lower than the preset qualified threshold value, it is judged that the content of Fusarium graminearum is low. When the content of Fusarium graminearum is low, the detection sample is analyzed in the same way to calculate the Fusarium graminearum growth rate and mark it as g.

[0058] According to the actual content of Fusarium graminearum, the enterprise data is accessed to obtain a damage comparison table, and the grain damage rate range corresponding to the actual content of Fusarium graminearum is queried in the damage comparison table, and the average value of the grain damage rate range is taken as the grain damage rate caused by the actual content of Fusarium graminearum.

[0059] Since there are too many types of fungi in grains, it is time-consuming and laborious to test them separately. The growth rate of other fungi can be evaluated based on the grain damage rate. For the test samples, the capacitance measurement method can be used to detect the overall grain damage rate. The grain damage rate caused by other fungi can be preliminarily obtained by subtracting the detected overall grain damage rate from the grain damage rate caused by the actual content of Fusarium graminearum.

[0060] The system uses the grain damage rate caused by the actual content of Fusarium graminearum, the Fusarium graminearum growth rate, and the grain damage rate caused by other fungi to calculate the growth rate of other fungi. The calculation formula can be: , where G is the grain damage rate caused by the actual content of Fusarium graminearum, g is the proliferation rate of Fusarium graminearum, k is the error coefficient, Q is the grain damage rate caused by other fungi, and q is the growth rate of other fungi.

[0061] By comparing the grain damage rate caused by other fungi with the grain damage rate caused by the actual content of Fusarium graminearum, the relationship between the growth rate of other fungi and the growth rate of Fusarium graminearum can be reflected, thereby determining their competitive relationship. The error coefficient is used to reduce the assessment error of the growth rate of other fungi. The system obtains the detected grain storage humidity through the humidity sensor and calculates the grain damage rate based on the analysis sample to determine the error coefficient.

[0062] In order to increase the tolerance of the error coefficient, the system takes the analysis sample with the smallest Fusarium graminearum proliferation rate from multiple analysis samples for min marking, and uses the min-marked analysis sample to detect and calculate the grain damage rate using the optimal detection time interval. The grain damage rate of the min-marked sample is detected at the detection time points before and after the optimal detection time interval, and the grain damage rate is obtained by subtracting the grain damage rate of the latter detection point from the grain damage rate of the previous detection point. Since the grain damage rate is between 0 and 1, and the grain damage rate of the latter detection point will not be lower than the grain damage rate of the previous detection point, it is obvious that the grain damage rate is between 0 and 1.

[0063] After obtaining the grain storage humidity using a humidity sensor, access the historical data in the enterprise database to obtain the grain damage rate corresponding to different humidity levels, determine the humidity range in which the grain storage humidity is located, calculate the ratio of the grain damage rate in the corresponding humidity range to the average grain damage rate corresponding to different humidity levels, and use the calculated ratio as the humidity influence coefficient.

[0064] The error coefficient k is obtained by summing the grain damage rate and the humidity influence coefficient. When the grain damage rate is greater, or the humidity influence coefficient is greater, the error coefficient is greater, and the growth rate q of other fungi is greater.

[0065] The competition relationship between fungi is determined by comparing the growth rate g of Fusarium graminearum with the growth rates q of other fungi. The comparison threshold j is set. When |gq|>j, the competition relationship between fungi is judged to be weak; when |gq|<=j, the competition relationship between fungi is judged to be strong. |gq| is the absolute value of the difference between the growth rate g of Fusarium graminearum and the growth rates q of other fungi.

[0066] It should be noted that the capacitance measurement method uses the difference in capacitance between damaged grains and undamaged grains to count the proportion of grains with different capacitance values ​​and determine the overall grain damage rate of the sample; the damage comparison table is a comparison table that is compiled by counting different grain damage rate intervals corresponding to different Fusarium graminearum contents in the historical data saved by the company in the past. The damage comparison table can be used to determine the grain damage caused by the actual content of Fusarium graminearum; in this example, the use of grain damage rate to evaluate the growth rate of other fungi greatly saves the time spent on detecting the growth rates of different fungi. Setting the error coefficient by the damage status and environmental conditions improves the accuracy of calculating the growth rates of other fungi, which adds a lot of fault tolerance space for the subsequent analysis of the competitive relationship between fungi. The competition coefficient in the above steps can be set according to the actual situation, for example, the comparison threshold j is set to 0.1.

[0067] In step S4, when the competition between fungi is weak, the system provides a linear regression model to predict the value-added status of Fusarium graminearum; when the competition between fungi is strong, the system provides a machine learning model to predict the value-added status of Fusarium graminearum.

[0068] When the competition between fungi is weak, the interaction between fungi is negligible, and the linear regression model algorithm is relatively simple, which can quickly predict the value-added of Fusarium graminearum. When the competition between fungi is strong, the interaction between fungi is large, and the machine learning algorithm can process multiple variables and consider the interaction between different fungi, and the accuracy of predicting Fusarium graminearum is higher.

[0069] When the competition between fungi is weak, the system uses the stepwise regression model in the linear regression model to predict the proliferation of Fusarium graminearum. The system randomly selects N groups of equal amounts of grains in the grain pile as valid samples at different experimental temperatures at the optimal detection time interval, and detects the Fusarium graminearum content of the valid samples at two time points before and after the time interval. The Fusarium graminearum content detected at the latter time point of each group is subtracted from the Fusarium graminearum content detected at the previous time point and divided by the optimal detection time interval to obtain the Fusarium graminearum proliferation rate of each group.

[0070] The proliferation rate of Fusarium graminearum in each group was used as input to predict the value-added situation of Fusarium graminearum using the stepwise regression method. The specific steps are as follows:

[0071] Initialization: Build an empty model that does not contain any data or independent variables.

[0072] Add data: Calculate the average of the proliferation rate of Fusarium graminearum in each group and add it to the model as the first data.

[0073] Set the deletion threshold: Set the first deletion threshold to the first data. Each time a data is added, the deletion threshold is updated to the average value of all data in the model.

[0074] Delete data: Before adding new data, determine whether the new data exceeds the deletion threshold. If it exceeds the deletion threshold, it will not be added to the model. When the new data is lower than the deletion threshold, it will be added to the model and the deletion threshold will be recalculated.

[0075] Repeat the operation: determine the proliferation rates of Fusarium graminearum that have not been added to the model one by one and add them to the model.

[0076] Predicting the proliferation of Fusarium graminearum: After all the proliferation rates of Fusarium graminearum have been judged, the deletion threshold at this time is output as the predicted proliferation rate of Fusarium graminearum.

[0077] It should be noted that in the above steps, the setting rules of the deletion threshold can be adjusted according to the actual situation. The stepwise regression method is a method for automatically selecting variables and can predict data with lower complexity.

[0078] When the competition between fungi is strong, the system uses the support vector machine model in the linear regression model to predict the proliferation of Fusarium graminearum. The system randomly selects n groups of grains of different qualities in the grain pile at different experimental temperatures at the optimal detection time interval as effective samples, and merges different experimental temperatures into experimental temperature data sets, and different grain qualities into grain quality data sets. The system uses the experimental temperature data set and grain quality data set as input variables to construct a support vector machine model. The support vector machine model selects the best data in the experimental temperature data set and grain quality data set through training to predict the proliferation of Fusarium graminearum. The specific steps are as follows:

[0079] Step A1: Normalize the data in the experimental temperature dataset and the grain quality dataset and use them as input features.

[0080] Step A2: Select the radial basis function as the kernel function to transform the input features and merge them into an array. The feature transformation formula can be: ,in is the calculation result after the support vector machine is trained on the normalized average of the two data sets. is the average value of the normalized experimental temperature data in the experimental temperature data set, is the average value of normalized grain quality data in the grain quality data set, is the natural base, is a hyperparameter.

[0081] It should be explained that the hyperparameter is a parameter used to adjust the width of the kernel function. Since the system predicts the value-added situation of Fusarium graminearum by considering the competition relationship between fungi, the system sets the hyperparameter as the ratio of the value-added rate g of Fusarium graminearum to the growth rate q of other fungi. When the competition relationship between fungi is strong, the value-added rate g of Fusarium graminearum is closer to the growth rate q of other fungi. It can be seen from the feature conversion formula that when the ratio of the value-added rate g of Fusarium graminearum to the growth rate q of other fungi is larger, that is, the value-added rate of Fusarium graminearum is faster than the growth rate of other fungi, the calculation result The bigger.

[0082] Step A3: Divide the training set and the test set. Randomly divide the experimental temperature data set and the grain quality data set into training sets and test sets with different proportions. Randomly select a set of data from the training set and the test set to perform calculations using the kernel function to obtain the calculation results of the training set and the test set.

[0083] Step A4: Use the loss function to converge the calculation results. The loss function formula can be: ,in, is the kernel function loss, Compute the results for the test set, The calculation results of the training set are converged according to the kernel function loss. The convergence formula is: , by randomly selecting data multiple times in the training set and the test set to calculate the kernel function loss , and the final convergence result is .

[0084] Step A5: predict the proliferation of Fusarium graminearum, take the average value of the normalized experimental temperature data in the experimental temperature data set and the average value of the normalized grain quality data in the grain quality data set as the detection conditions, detect the content of Fusarium graminearum at two time points before and after the optimal detection time interval, mark the average Fusarium graminearum proliferation rate obtained by subtracting the detected Fusarium graminearum content and dividing it by the optimal detection time interval as u, mark the predicted Fusarium graminearum proliferation rate as v, and calculate the predicted Fusarium graminearum proliferation rate v by the average Fusarium graminearum proliferation rate u, and the calculation formula is: ,in The calculation result of the support vector machine is obtained by using the normalized average value of the two data sets. It is the final convergence result of the support vector machine.

[0085] Step A6: Output v as the predicted proliferation rate of Fusarium graminearum.

[0086] By using support vector machines to determine the optimal detection conditions and predict the proliferation of Fusarium graminearum, factories or users can conduct auxiliary analysis and take timely measures.

[0087] It should be noted that the kernel function and loss function in the support vector machine are not unique. The present invention only gives examples among many kernel functions and loss functions. The ratio of dividing the training set and the test set in the above step 3 can be set by oneself. For example, the division ratio of the training set and the test set is set to 7:3, that is, the experimental temperature data set and the grain quality data set are divided according to the division ratio of 7:3 respectively. In addition, the linear regression model and the machine learning model are not unique. The present invention only gives examples with the stepwise regression model and the support vector machine model.

[0088] Example 2, a grain toxin intelligent detection system, such as Figure 2 As shown, it is used to implement the above-mentioned intelligent detection method of grain toxins, including a data acquisition module, a sample screening module, a competition analysis module and a selection prediction module;

[0089] The data collection module is used to collect and analyze the data of Fusarium graminearum in the sample and send it to the sample screening module;

[0090] The sample screening module is used to determine the optimal detection time interval and screen the analysis samples;

[0091] The competition analysis module is used to analyze the competition relationship between fungi in the screened analysis samples and transmit the analysis results to the selection prediction module;

[0092] The prediction module selects different algorithm models to predict the proliferation of Fusarium graminearum based on the analysis results of the competition relationship between fungi.

[0093] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0094] 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 of the technical solution and the invention constraints. 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.

[0095] 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.

[0096] 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.

[0097] 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. An intelligent detection method for grain toxins, It is characterized in that The steps include: Step S1, obtaining multiple samples, detecting and calculating the proliferation rate of Fusarium graminearum respectively, and collecting the contents of vomitoxin and zearalenone in the samples to determine the degree of sample damage; Step S2, using the sample damage degree and the proliferation rate of Fusarium graminearum to generate the optimal sample detection time, screening the samples according to the sample detection time and counting the Fusarium graminearum content; In step S2, the system uses the Fusarium graminearum growth rate and time interval data of each analyzed sample as the influencing feature, and the sample loss degree as the screening feature to construct a ridge regression model, and uses the ridge regression model to find the optimal detection time interval to detect the Fusarium graminearum content and count it; The influencing features are the parameters that affect the accuracy of Fusarium graminearum content detection, and the screening features are the real labels used to screen samples; Step S3, when the content of Fusarium graminearum is low, the overall grain damage rate is detected by a capacitance measurement method, the grain damage rate caused by Fusarium graminearum is determined according to the content of Fusarium graminearum by accessing historical data in the enterprise database, the growth rate of other fungi is analyzed according to the overall grain damage rate and the grain damage rate caused by Fusarium graminearum, and the competitive relationship between the growth rates of other fungi and the value-added rate of Fusarium graminearum is determined by comparing the relationship between the growth rates of other fungi and the value-added rate of Fusarium graminearum; Step S4, setting a comparison threshold. When the growth rate of Fusarium graminearum and the growth rate of other fungi are outside the comparison threshold, the system provides a linear regression model to predict the growth of Fusarium graminearum; when the growth rate of Fusarium graminearum and the growth rate of other fungi are within the comparison threshold, the system provides a machine learning model to predict the growth of Fusarium graminearum.

2. A method for intelligent detection of grain toxins according to claim 1, Features: In step S1, when the system obtains samples, it selects equal amounts of grains from the same grain pile as analysis samples multiple times, and randomly sets two time points with different time intervals in each analysis sample to detect the content of Fusarium graminearum. The system uses the ratio of the content difference to the corresponding time interval as the Fusarium graminearum proliferation rate and records the time interval data of each analysis sample.

3. A method for intelligent detection of grain toxins according to claim 2, Features: In step S1, the system takes the midpoint of each analysis sample to detect the content of vomitoxin and zearalenone, normalizes them respectively, and then sums them as the sample damage degree of the corresponding analysis sample.

4. A method for intelligent detection of grain toxins according to claim 1, Features: In step S2, the ridge regression model searches for the optimal detection time interval as follows: Data preparation: The proliferation rate and time interval data of Fusarium graminearum were used as the influencing feature input, and the sample loss degree was used as the screening feature input; Add regularization term: construct loss function to determine the influence weight of Fusarium graminearum proliferation rate and time interval data on the accuracy of Fusarium graminearum content analysis; Solving for optimal weights: Using the gradient descent method to solve the optimal weights of Fusarium graminearum proliferation rate and time interval data in the loss function according to the screening features; Output prediction results: Determine the optimal detection time interval based on the optimal weight of the Fusarium graminearum proliferation rate and time interval data.

5. A method for intelligent detection of grain toxins according to claim 1, Features: In step S3, the Fusarium graminearum content detected at the optimal detection time interval is compared with a preset qualified threshold value, and when the Fusarium graminearum content is lower than the preset qualified threshold value, it is determined that the Fusarium graminearum content is low.

6. A method for intelligent detection of grain toxins according to claim 1, Features: In step S3, the system collects grain damage rate and storage humidity before other fungal growth rates to determine the error of competitive relationship analysis, calibrates the analysis results of other fungal growth rates according to the error of competitive relationship judgment, and increases the error tolerance of competitive relationship analysis.

7. A grain toxin intelligent detection system, based on a grain toxin intelligent detection method according to any one of claims 1 to 6, It is characterized in that It includes data collection module, sample screening module, competition analysis module and selection prediction module; The data collection module is used to collect and analyze the data of Fusarium graminearum in the sample and send it to the sample screening module; The sample screening module is used to determine the optimal detection time interval and screen the analysis samples; The competition analysis module is used to analyze the competition relationship between fungi in the screened analysis samples and transmit the analysis results to the selection prediction module; The prediction module selects different algorithm models to predict the proliferation of Fusarium graminearum based on the analysis results of the competition relationship between fungi.

Images