A method and device for detecting nutrients in cod sausage based on machine learning

Abstract

The present invention discloses a method and device for detecting nutrients in cod sausage based on machine learning, which relates to the technical field of food detection. It includes sample acquisition: obtaining a cod sausage sample and sample information, and detecting the nutrient concentration of the cod sausage sample by gas chromatography. According to the user's needs, the present invention provides a target process for detecting target substances that meets the user's needs through a matching adjustment method. After detecting the target chromatogram by matching the target process with gas chromatography to detect the concentration of the target substance, the concentration of the target substance is obtained from the target chromatogram by the set matching recognition model and fed back to the user. It is judged that the concentration of the target substance does not require manual analysis by researchers, reducing the error of the detection result and improving the efficiency. By actively outputting the detection process and automatically identifying the concentration of the target substance, the time for substance detection is shortened, so as to improve the efficiency of detecting nutrients in cod sausage and the accuracy of the detection result.

Description

Technical Field

[0001] The present invention relates to the technical field of food detection, and specifically provides a method and device for detecting the nutrients in cod sausage based on machine learning. Background Art

[0002] Cod sausage is rich in various nutrients such as fatty acids and amino acids, and has high nutritional value. In order to support the product development of cod sausage, meet regulatory requirements, study functional characteristics, monitor raw material quality, and promote scientific research, it is often necessary to detect cod sausage.

[0003] A method for predicting the concentration of water pollutants based on the coupling of IML-GIS, with the patent publication number CN118378731A, includes obtaining water quality data at precise points in the study area; obtaining a shared dataset of natural geographical parameters in the study area; using ArcGIS to fuse all the bands of the prediction variable datasets; converting the fused dataset image into an array recognizable by the Python language; establishing an integrated machine learning model; training the integrated machine learning model to establish a mathematical relationship between the measured dataset and the fused dataset; and predicting the concentration values of pollutants continuously in the entire geographical space of the study area. This method eliminates the cumbersome and complex processes of field sampling and conventional water quality detection in the laboratory, and is also well applicable to predicting the concentration of different types of water pollutants in large-scale basin spaces. Especially for complex and special water bodies such as groundwater that are difficult to observe in the field, it can effectively predict the trend of water quality evolution at the spatial scale.

[0004] The detection of the composition of substances based on the above and similar principles is carried out in a picture-based manner. This way of detecting the composition of substances depends on picture acquisition. If there are errors or inaccuracies in picture acquisition, it will lead to the drawback that the detection results are not precise enough. When in a food health research center, researchers need to regularly monitor the nutritional components of cod sausage to evaluate its adaptability to a specific diet, based on the above scenario, using the above and similar principles of substance detection methods cannot accurately provide the results of food nutritional component detection. In the prior art, when using gas chromatography to detect food nutrients, when analyzing the content or concentration of the target component through the chromatogram, it is usually manual analysis, which takes a long time and has subjective judgment and human errors, easily resulting in errors in the detection results and low efficiency in analyzing the detection results, and is not conducive to improving the efficiency and accuracy of the overall nutritional component detection. Therefore, the present invention is proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and device for detecting the nutrients in cod sausage based on machine learning, so as to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solution: A method for detecting nutrients in cod fish sausage based on machine learning, the method comprising:

[0007] Sample acquisition: Obtain a cod fish sausage sample and sample information, and use gas chromatography to detect the nutrient concentration of the cod fish sausage sample;

[0008] Information processing: Based on the sample information, obtain the actual weight of the sample, preset the detection items and the sample weight, the detection items are nutrients, and based on the detection items, cooperate with gas chromatography to optimize the pretreatment process of the traditional detection process through the detection process to obtain the detection process and data relationship;

[0009] Detection directory establishment: Based on the detection process and data relationship, cooperate with the sample weight to generate the detection process under different sample weights through the detection derivation method to obtain the derivation process, integrate the detection items, detection process, sample weight and derivation process to obtain the detection set, and establish a detection directory for storing the detection set;

[0010] Process output: Obtain user requirements and user address information, obtain the target substance based on the user requirements, and based on the user requirements, cooperate with the actual sample weight and data relationship to select the corresponding process in the detection directory through the matching adjustment method and adjust to obtain the target process;

[0011] Reference establishment and target substance detection: Based on the detection items, establish a reference library for matching and identifying the nutrient concentration through the reference establishment method, and based on the target substance, cooperate with the target process to detect the target substance in the cod fish sausage sample by gas chromatography to obtain a chromatogram;

[0012] Matching output: Find the chromatographic peak corresponding to the target substance in the chromatogram to obtain the target chromatographic peak, mark the target chromatographic peak in the chromatogram to obtain the target chromatogram, import the target chromatogram into the matching recognition model to obtain the concentration of the target substance, and feedback the concentration of the target substance to the user based on the user address information;

[0013] The reference establishment method includes: preset the target concentration, obtain a cod fish sausage sample with the detection item as the target concentration to obtain a reference sample, detect the concentration of the detection item of the reference sample by gas chromatography to obtain a reference chromatogram, establish the correlation between the target concentration and the chromatogram, integrate the target concentration and the chromatogram to obtain a sub-reference, and establish a reference library for storing the sub-reference.

[0014] Furthermore, the method for obtaining the detection process includes: obtaining the traditional process obtained from the process of detecting detection items by gas chromatography, extracting the part that inaccurately expresses parameter data in the traditional process to obtain the processing process and data relationship, where the data relationship is the relationship between the sample weight and the parameter data, supplementing the details of the parameter data in the processing process through an optimization method based on the processing process to obtain the optimized process, and replacing the processing process in the traditional process with the optimized process to obtain the detection process corresponding to the detection item.

[0015] Furthermore, the optimization method includes: splitting the data relationship to obtain several sub-data relationships, splitting the processing process to obtain several sub-processing processes, extracting the parameters in the sub-processing processes to obtain the parameters to be optimized, establishing the association relationship between the sub-data relationships and the sub-processing processes, obtaining the target parameters based on the sample weight and the sub-data relationships, determining the relationship between the target parameters and the parameters to be optimized based on the association relationship, replacing the parameters to be optimized with the target parameters correspondingly in the sub-processing processes based on the relationship between the target parameters and the parameters to be optimized to obtain the sub-optimized processes, and integrating the sub-optimized processes based on the order of splitting the processing process to obtain the optimized process.

[0016] Furthermore, the detection derivation method includes: presetting the derivation unit, obtaining the weight of the sample in the previous sample detection information to obtain the previous weight information, judging whether the previous weight information is an integer to obtain the judgment result, when the judgment result feedback is that the previous weight information is an integer, the derivation unit is an integer, cumulatively increasing or reducing the derivation unit based on the sample weight to obtain the derived sample weight, obtaining the derived parameters corresponding to the derived sample based on the derived sample weight and the data relationship, and replacing the optimized parameters in the detection process with the derived parameters to obtain the derived process; when the judgment result feedback is that the previous weight information is not an integer, the derivation unit is a decimal, cumulatively increasing or reducing the derivation unit based on the sample weight to obtain the derived sample weight, obtaining the derived parameters corresponding to the derived sample based on the derived sample weight and the data relationship, and replacing the optimized parameters in the detection process with the derived parameters to obtain the derived process; when the judgment result feedback is that part of the previous weight information is an integer, then perform the processes of the judgment result feedback that the previous weight information is an integer and the judgment result feedback that the previous weight information is not an integer and integrate them to obtain the derived process.

[0017] Further, the matching adjustment method includes: retrieving a detection set corresponding to the target substance from the detection catalog based on the target substance to obtain a target set, selecting a detection process consistent with the actual sample weight from the target set based on the actual sample weight to obtain a selection result. When the selection result indicates that there is a detection process consistent with the actual sample weight in the target set, the detection process consistent with the actual sample weight is extracted from the target set to obtain a target process. When the selection result indicates that there is no detection process consistent with the actual sample weight in the target set, the detection process with the closest distance to the actual sample weight in the target set is extracted to obtain a preliminary process. The difference between the sample weight in the preliminary process and the actual sample weight is obtained, and the optimization parameters in the preliminary process are adjusted based on the difference and the data relationship to obtain a target process.

[0018] Further, the process of finding the chromatographic peak corresponding to the target substance in the chromatogram is as follows: obtaining a comparison chromatogram based on the reference library for the chromatogram with known detection item concentrations, recording the peak residence time in the comparison chromatogram to obtain a comparison time, establishing a correspondence between the comparison time and the detection item, establishing a comparison library for storing the comparison time, the detection item, and the correspondence. Recording the residence time of the chromatographic peak in the chromatogram to obtain a target time, retrieving the detection item consistent with the target substance in the comparison library to obtain a comparison item, extracting the comparison time corresponding to the comparison item based on the correspondence to obtain a reference time, and selecting the chromatographic peak with the same target time and reference time to obtain a target chromatographic peak.

[0019] Further, the method for establishing the matching recognition model includes:

[0020] Training data acquisition: Splitting the sub-reference in the reference library to obtain the target concentration and chromatogram, marking the chromatographic peak corresponding to the target concentration in the chromatogram, and integrating the target concentration and the marked chromatogram to obtain training data;

[0021] Model selection and training: Selecting a suitable deep learning model as the model matrix, and importing the training data into the model matrix for the model to learn to obtain an initial model;

[0022] Model adjustment and output: Presetting test data and test concentration, where the test data is a chromatogram with marked chromatographic peaks. Importing the test data into the initial model to obtain a result concentration, judging the difference between the result concentration and the test concentration to obtain result information, and optimizing and adjusting the initial model based on the result information to obtain a matching recognition model.

[0023] A cod sausage nutrient detection device based on machine learning uses the above-mentioned cod sausage nutrient detection method based on machine learning.

[0024] Compared with the prior art, the beneficial effects of the present invention are:

[0025] The method and device for detecting nutrients in cod sausage based on machine learning listen to user needs, and according to the user needs, cooperate with a matching adjustment method to provide a target process for detecting target substances that meets the user needs. Through the target process, gas chromatography is used to detect the concentration of the target substance. After obtaining the target chromatogram, the set matching recognition model is used to obtain the concentration of the target substance from the target chromatogram and feedback it to the user. The process of judging the concentration of the target substance does not require manual analysis by researchers, reducing the errors in the detection results and improving the efficiency. By actively outputting the detection process and automatically identifying the concentration of the target substance, the time for substance detection is shortened, so as to improve the efficiency of detecting nutrients in cod sausage and the accuracy of the detection results.

[0026] At the same time, through the set monitoring process acquisition method, the specific process of using gas chromatography to detect the concentration of detection items can be obtained, and the specific process is optimized to obtain the detection process, so as to facilitate the operation of the matching adjustment method. Through the set detection derivation method, the detection process under different sample weights is derived to obtain the derivation process, and a detection directory is established to store the detection set, so as to increase the data volume of the detection set in the detection directory, facilitate the subsequent rapid output of the target process that meets the user needs according to the user needs, and extract the parameters in the sub-processing process to obtain the parameters to be optimized. In this process, a single parameter is extracted, so as to facilitate multi-threading to separately supplement and optimize the parameters to be optimized and integrate them after optimization, so as to improve the efficiency of the optimization process.

[0027] At the same time, the detection derivation method calculates the possible unit forms of the weight of future samples based on the weight of samples in the past sample detection information and provides corresponding countermeasures. When the judgment result feedback is that the past weight information is an integer, but the actual weight of the subsequent sample is not an integer, it is fed back to the detection derivation method for continued derivation and supplementation. One of the three methods can also be selected as the main derivation process of the detection derivation method according to actual usage requirements, or the above situation can be used as the main derivation process of the detection derivation method. The specific selection needs to be determined by researchers according to actual usage conditions. Brief Description of the Drawings

[0028] Figure 1 It is a schematic diagram of the overall upper half process structure of the present invention;

[0029] Figure 2 It is a schematic diagram of the overall lower half process structure of the present invention;

[0030] Figure 3 It is a schematic diagram of the establishment and operation structure of the matching recognition model of the present invention;

[0031] Figure 4 It is a schematic diagram of the detection derivation method structure of the present invention. Detailed Description of the Invention

[0032] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0033] Detecting the nutrients in cod sausage helps to evaluate its nutritional value, ensure food safety, and meet the needs of consumers. Detecting the nutritional components such as protein, fat, carbohydrates, vitamins, and minerals in cod sausage ensures that the nutritional information on the product label is accurate, avoids false publicity. The detection results of the nutrients in cod sausage provide important scientific basis for production enterprises, consumers, and scientific research institutions, and have broad application value.

[0034] As Figures 1 - 4 shown, the present invention provides a technical solution: a method for detecting the nutrients in cod sausage based on machine learning, the method includes:

[0035] Sample acquisition: Obtain a cod sausage sample and sample information, and use gas chromatography to detect the nutrient concentration of the cod sausage sample;

[0036] Information processing: Based on the sample information, obtain the actual weight of the sample, preset the detection items and the sample weight, the detection items are nutrients, and based on the detection items, cooperate with gas chromatography to optimize the pretreatment process of the traditional detection process through the detection process to obtain the detection process and data relationship;

[0037] Detection directory establishment: Based on the detection process and data relationship, cooperate with the sample weight to generate the detection process under different sample weights through the detection derivation method to obtain the derivation process, integrate the detection items, detection process, sample weight, and derivation process to obtain the detection set, and establish a detection directory for storing the detection set;

[0038] Process output: Obtain the user requirements and user address information, obtain the target substance based on the user requirements, and based on the user requirements, cooperate with the actual weight of the sample and the data relationship to select the corresponding process in the detection directory through the matching adjustment method and adjust it to obtain the target process;

[0039] Reference establishment and target substance detection: Based on the detection items, establish a reference library for matching and identifying the nutrient concentration through the reference establishment method, and based on the target substance, cooperate with the target process to detect the target substance in the cod sausage sample by gas chromatography to obtain a chromatogram;

[0040] Matching output: Locate the chromatographic peak corresponding to the target substance in the chromatogram to obtain the target chromatographic peak, mark the target chromatographic peak in the chromatogram to obtain the target chromatogram, import the target chromatogram into the matching recognition model to obtain the concentration of the target substance, and feedback the concentration of the target substance to the user based on the user address information;

[0041] It should be noted that in the sample acquisition stage, the cod sausage sample is directly obtained from the cod sausage to be tested. The sample information includes but is not limited to the actual weight of the sample. For example, the sample information may also include the content information of the cod sausage, so as to facilitate subsequent comparison to determine whether the cod sausage is qualified or whether there is false publicity, etc. By detecting the nutrient concentration of the cod sausage sample through gas chromatography, the components in the complex mixture can be distinguished, meeting the high-throughput detection requirements, ensuring the detection rate while obtaining accurate detection results. That is, through the optimized pretreatment techniques in gas chromatography (such as derivatization, solid-phase extraction), the interference substances such as proteins and salts in the cod sausage can be effectively removed, improving the detection accuracy of the target substance. For example, the fatty acids in the cod sausage are converted into fatty acid methyl esters to eliminate matrix interference and improve the detection sensitivity. In the information processing stage, the detection items and the sample weight are set by the researchers according to the actual use situation. The detection items need to cover all the nutrients present in the cod sausage. Through the set monitoring process acquisition method, the specific process of detecting the concentration of the detection items by gas chromatography can be obtained, and the detection process and data relationship are optimized. The data relationship is the relationship between the sample weight and the parameter data. Specifically, for example, a specific dose of solvent needs to be added when the sample weight reacts. In the detection catalog establishment stage, the detection process under different sample weights is derived through the set detection derivation method to obtain the derived process, and a detection catalog is established to store the detection set, so as to increase the data volume of the detection set in the detection catalog, facilitating the subsequent rapid output of the target process that meets the user's needs according to the user's needs. In the process output stage, both the user's needs and the user's address information can be obtained by the user. The user's needs are the specific nutrients to be detected, and the user's address information is the address information for feedback of the detection results to the user, including but not limited to email and phone number. Through the set matching adjustment method, the corresponding detection process is selected from the detection catalog according to the user's needs and adjusted to obtain the target process. In the reference establishment and target substance detection stage, a reference library for matching and identifying the nutrient concentration is established through the set reference establishment method, so as to facilitate the subsequent identification of the target substance concentration. The target substance concentration of the cod sausage sample is detected by the target process in cooperation with gas chromatography to obtain a chromatogram. In the matching output stage, the chromatographic peak corresponding to the target substance in the chromatogram is found and the chromatographic peak is marked to obtain the target chromatogram. The target chromatogram is imported into the matching recognition model to obtain the concentration of the target substance. The concentration of the target substance is feedback to the user based on the user's address information. The present invention listens to the user's needs, provides the user with the target process for detecting the target substance that meets the user's needs according to the user's needs in cooperation with the matching adjustment method, detects the concentration of the target substance by the target process in cooperation with gas chromatography, and after obtaining the target chromatogram by detection, obtains the concentration of the target substance according to the target chromatogram through the set matching recognition model and feedbacks it to the user. The process of judging the concentration of the target substance does not require manual analysis by the researcher, reducing the error in the detection results and improving the efficiency.Shorten the time for substance detection by means of an active output detection process and automatically identifying the concentration of the target substance, so as to improve the efficiency of detecting the nutrients in codfish sausage and the accuracy of the detection results.

[0042] As Figure 1 shown, the reference establishment method includes: presetting a target concentration, obtaining a codfish sausage sample with the detection item being the target concentration to get a reference sample, detecting the concentration of the detection item of the reference sample based on gas chromatography to obtain a reference chromatogram, establishing the correlation between the target concentration and the chromatogram, integrating the target concentration and the chromatogram to get a sub-reference, and establishing a reference library for storing the sub-reference.

[0043] It should be noted that the target concentration is set by the researchers according to the actual usage. The reference sample has the detection item as the target concentration. Here, the detection item refers to various nutrients, that is, one of the various nutrients. When detecting the reference sample to obtain the reference chromatogram, the corresponding detection process can also be matched through a matching adjustment method, and then the reference sample is detected by the detection process in cooperation with gas chromatography to obtain the reference chromatogram. The reference library is used to store the sub-reference. The more the number of sub-references, the better. The number of sub-references is specifically determined according to the actual usage.

[0044] As Figure 1 shown, the method for obtaining the detection process includes: obtaining the process of detecting the detection item by gas chromatography to get a traditional process, extracting the part with inaccurate expression of parameter data in the traditional process to get a processing process and a data relationship. The data relationship is the relationship between the sample weight and the parameter data. Based on the processing process, the details of the parameter data in the processing process are supplemented through an optimization method to get an optimized process, and the processing process in the traditional process is replaced with the optimized process to get the detection process corresponding to the detection item.

[0045] It should be noted that the traditional process can be obtained from the information of previously detecting the detection item by gas chromatography. In the process of extracting the part with inaccurate expression of parameter data in the traditional process to get the processing process and the data relationship, the part with inaccurate expression of parameter data is the part where adjusting the sample weight causes the solvent reacting with the sample or the time and temperature for heating the sample to change. The data relationship is the data on how the change in the reaction sample weight will cause the change in the solvent dosage or the time and temperature for heating the sample. The details of the parameter data are supplemented and optimized through the set optimization method to get the optimized process, so as to directly match the target process corresponding to the actual sample weight according to the actual sample weight, in order to improve the efficiency of detecting the target substance.

[0046] The optimization method includes: splitting data relationships to obtain several sub-data relationships, splitting the processing process to obtain several sub-processing processes, extracting parameters in the sub-processing processes to obtain parameters to be optimized, establishing the association relationship between the sub-data relationships and the sub-processing processes, obtaining target parameters based on the sample weight in combination with the sub-data relationships, determining the relationship between the target parameters and the parameters to be optimized based on the association relationship, replacing the parameters to be optimized with the target parameters correspondingly in the sub-processing processes based on the relationship between the target parameters and the parameters to be optimized to obtain sub-optimized processes, and integrating the sub-optimized processes based on the order of the split processing process to obtain the optimized process.

[0047] It should be noted that the process of extracting parameters in the sub-processing processes to obtain the parameters to be optimized, that is, extracting the parameters with unclear expressions in the processing process to obtain the parameters to be optimized, is convenient for multi-threading to separately supplement and optimize the parameters to be optimized and integrate them after optimization, so as to improve the efficiency of the optimization process.

[0048] Such as Figure 1 and Figure 4 As shown, the detection and derivation method includes: presetting a derivation unit, obtaining the weight of the sample in the previous sample detection information to obtain the previous weight information, judging whether the previous weight information is an integer to obtain a judgment result. When the judgment result feedback is that the previous weight information is an integer, the derivation unit is an integer. Cumulatively increasing or reducing the derivation unit based on the sample weight to obtain the derived sample weight, obtaining the derived parameters corresponding to the derived sample based on the derived sample weight in combination with the data relationship, and replacing the optimized parameters in the detection process with the derived parameters to obtain the derivation process. When the judgment result feedback is that the previous weight information is not an integer, the derivation unit is a decimal. Cumulatively increasing or reducing the derivation unit based on the sample weight to obtain the derived sample weight, obtaining the derived parameters corresponding to the derived sample based on the derived sample weight in combination with the data relationship, and replacing the optimized parameters in the detection process with the derived parameters to obtain the derivation process. When the judgment result feedback is that part of the previous weight information is an integer, then perform the processes of the judgment result feedback that the previous weight information is an integer and the judgment result feedback that the previous weight information is not an integer and integrate them to obtain the derivation process.

[0049] It should be noted that the derivation unit is an integer or a decimal, and the specific value is formulated by the user according to the actual usage. The detection and derivation method calculates the possible unit forms of the future sample weight based on the weight of the sample in the previous sample detection information and provides corresponding coping methods. When there is a judgment result feedback that the previous weight information is an integer, but the actual weight of the subsequent sample is not an integer, it is fed back to the detection and derivation method for continued derivation and supplementation. At the same time, one of the three methods can also be selected as the main derivation process of the detection and derivation method according to the actual usage requirements, or the above situation can also be used as the main derivation process of the detection and derivation method. The specific selection needs to be formulated by the researchers according to the actual usage.

[0050] As Figure 2 shown, the matching adjustment method includes: retrieving a detection set corresponding to the target substance from the detection catalog based on the target substance to obtain a target set, selecting a detection process consistent with the actual weight of the sample from the target set based on the actual weight of the sample to obtain a selection result. When the selection result feedback indicates that there is a detection process consistent with the actual weight of the sample in the target set, the detection process consistent with the actual weight of the sample is extracted from the target set to obtain a target process. When the selection result feedback indicates that there is no detection process consistent with the actual weight of the sample in the target set, the detection process with the closest distance to the actual weight of the sample in the target set is extracted to obtain a preliminary process, obtaining the difference between the sample weight and the actual weight of the sample in the preliminary process, and adjusting the optimization parameters in the preliminary process based on the difference in combination with the data relationship to obtain a target process.

[0051] It should be noted that through the set matching adjustment method, the substance whose detection concentration needs to be determined according to the user's needs is obtained as the target substance, the target substance is traversed through the detection catalog to obtain a detection set corresponding to the target substance to obtain a target set, and a detection process consistent with the sample weight is selected from the detection set to obtain a target process. In this process, a deviation range can also be added to the actual weight of the sample. When the sample weight is within the deviation range added to the actual weight of the sample, the detection process is selected to obtain a target process. Specifically, whether to select the deviation range or determine the specific value of the deviation range is determined by the researcher according to the actual usage situation. When there is no detection process consistent with the actual weight of the sample in the target set, the sample weight closer to or farther from the actual weight of the sample is selected, and the difference between the two is obtained. The optimization parameters of the preliminary process are adjusted based on the difference between the two in combination with the data relationship to obtain a target process. The specific optimization process is the same as that in the detection process acquisition method.

[0052] The process of finding the chromatographic peak corresponding to the target substance in the chromatogram is as follows: obtaining a comparison chromatogram based on the reference library for the chromatogram with known detection item concentrations, recording the peak residence time in the comparison chromatogram to obtain a comparison time, establishing a correspondence between the comparison time and the detection item, establishing a comparison library for storing the comparison time, the detection item, and the correspondence, recording the residence time of the chromatographic peak in the chromatogram to obtain a target time, retrieving the detection item consistent with the target substance in the comparison library to obtain a comparison item, extracting the comparison time corresponding to the comparison item based on the correspondence to obtain a reference time, and selecting the chromatographic peak with the same target time and reference time to obtain a target chromatographic peak.

[0053] It should be noted that the detection item and the target substance belong to the same substance. Through the comparison chromatogram with known detection item concentrations, the relationship between the chromatographic peak in the chromatogram and the detection item can be inversely inferred, that is, the chromatographic peak corresponding to the detection item is extracted, the residence time of the chromatographic peak is recorded to obtain a comparison time, and then the reference time consistent with the comparison time is extracted, so as to determine the target chromatographic peak corresponding to the target substance in the chromatogram.

[0054] As Figure 3 shown, the method for establishing a matching recognition model includes:

[0055] Training data acquisition: splitting the sub - references in the reference library to obtain the target concentration and chromatograms, marking the chromatographic peaks corresponding to the target concentration in the chromatograms, and integrating the target concentration and the marked chromatograms to obtain the training data;

[0056] Model selection and training: selecting a suitable deep - learning model as the model matrix, and importing the training data into the model matrix for the model to learn to obtain the initial model;

[0057] Model adjustment and output: presetting test data and test concentration, where the test data is a chromatogram with marked chromatographic peaks, importing the test data into the initial model to obtain the result concentration, judging the difference between the result concentration and the test concentration to obtain the result information, and optimizing and adjusting the initial model based on the result information to obtain the matching recognition model.

[0058] It should be noted that in the training data acquisition stage, the training data is obtained from the reference library established by the reference establishment method. Thus, the more the number of sub - references in the reference library, the more the training data. The process of marking the chromatographic peaks corresponding to the target concentration in the chromatogram can determine the chromatographic peaks corresponding to the target concentration by finding the chromatographic peaks corresponding to the target substance provided in the matching output stage. By marking the chromatographic peaks and providing the target concentration, the training data can be obtained. In the model selection and training stage, the deep - learning model selects the corresponding model as the model matrix according to the actual usage requirements. For example, a convolutional neural network, etc. Specifically, an image - processing model can be selected as the model matrix. By importing the training data into the model matrix for the model to learn, the initial model is obtained. In the model adjustment and output stage, by presetting the test data and test concentration as the data for testing the initial model, importing the test data into the initial model to obtain the result concentration, and then according to the difference between the result concentration and the test concentration, it is decided whether the initial model is put into use. When using the matching recognition model to recognize the target chromatogram, it is necessary to pre - process the target chromatogram in advance, such as image sharpening, denoising, etc., to improve the accuracy of the output result of the matching recognition model.

[0059] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting nutrients in cod intestines based on machine learning, characterized in that: The method comprises: Sample acquisition: obtain cod intestine samples and sample information, use gas chromatography with matching recognition model to detect the nutrient concentration of cod intestine samples, and obtain the actual weight of the samples based on the sample information; Information processing: preset test items and sample weights, and optimize the test process based on the test items through the test process acquisition method to obtain the test process and data relationship; Establishing a test catalog: Based on the test process and data relationship and the sample weight, the test process under different sample weights is generated through the test derivation method to obtain the derived process. The test items, test process, sample weight and derived process are integrated to obtain the test set, and a test catalog is established to store the test set. Process output: Obtain user needs, obtain target substances based on user needs, select corresponding processes in the detection catalog through matching adjustment methods based on user needs, and adjust to obtain target processes; Reference establishment and target substance detection: A reference library for matching and identifying nutrient concentrations was established through a reference establishment method, and a chromatogram was obtained by performing target substance detection on cod intestine samples by gas chromatography based on a target process; Matching output: Find and mark the chromatographic peaks corresponding to the target substance in the chromatogram to obtain the target chromatogram, and import the target chromatogram into the matching recognition model to obtain the concentration of the target substance; The method for establishing the matching recognition model includes: Acquisition of training data: Split the reference library to obtain the target concentration and chromatogram, mark the chromatographic peaks corresponding to the target concentration in the chromatogram, and integrate the target concentration and the marked chromatogram to obtain training data; Model selection and training: Select a suitable deep learning model as the model matrix, import the training data into the model matrix for the model to learn and obtain the initial model; Model adjustment and output: preset test data and test concentration, the test data is a chromatogram with chromatographic peaks marked, the test data is imported into the initial model to obtain the result concentration, the difference between the result concentration and the test concentration is determined to obtain the result information, and the initial model is optimized and adjusted based on the result information to obtain a matching recognition model; The matching adjustment method includes: based on the target substance, searching the detection set corresponding to the target substance in the detection catalog to obtain the target set; based on the actual weight of the sample, selecting the detection process consistent with the actual weight of the sample in the target set to obtain the selection result; when the selection result feedback is that there is a detection process consistent with the actual weight of the sample in the target set, then extracting the detection process consistent with the actual weight of the sample in the target set to obtain the target process; when the selection result feedback is that there is no detection process consistent with the actual weight of the sample in the target set, extracting the detection process closest to the actual weight of the sample in the target set to obtain the pre-process, obtaining the difference between the sample weight and the actual weight of the sample in the pre-process, and adjusting the optimization parameters in the pre-process based on the difference and data relationship to obtain the target process.

2. The method for detecting nutrients in cod intestines based on machine learning according to claim 1, characterized in that: The detection process acquisition method includes: acquiring the process of detecting the detection item using gas chromatography to obtain a traditional process, extracting the part of the traditional process that inaccurately expresses the parameter data to obtain a processing process and a data relationship, the data relationship is the relationship between the sample weight and the parameter data, based on the processing process, supplementing the details of the parameter data in the processing process through an optimization method to obtain an optimized process, and replacing the processing process in the traditional process with the optimized process to obtain a detection process corresponding to the detection item.

3. The method for detecting nutrients in cod intestines based on machine learning according to claim 2, characterized in that: The optimization method includes: splitting a data relationship to obtain a plurality of sub-data relationships, splitting a processing process to obtain a plurality of sub-processes, extracting parameters in the sub-processes to obtain parameters to be optimized, establishing an association relationship between the sub-data relationship and the sub-processes, obtaining a target parameter based on sample weight and the sub-data relationship, determining a relationship between the target parameter and the parameter to be optimized based on the association relationship, replacing the parameter to be optimized with the target parameter in the sub-processing process based on the relationship between the target parameter and the parameter to be optimized to obtain a sub-optimization process, and integrating the sub-optimization processes in sequence based on the split processing processes to obtain an optimization process.

4. The method for detecting nutrients in cod intestines based on machine learning according to claim 1, characterized in that: The detection derivation method includes: presetting a derivative unit, obtaining the weight of the sample in the previous sample detection information to obtain the previous weight information, judging whether the previous weight information is an integer to obtain a judgment result, when the judgment result feedback is that the previous weight information is an integer, the derivative unit is an integer, the derivative unit is cumulatively increased or reduced based on the sample weight to obtain the derived sample weight, based on the derived sample weight and data relationship to obtain the derivative parameters corresponding to the derived sample, the optimization parameters in the detection process are replaced with the derivative parameters to obtain the derivative process, when the judgment result feedback is that the previous weight information is not an integer, the derivative unit is a decimal, the derivative unit is cumulatively increased or reduced based on the sample weight to obtain the derived sample weight, based on the derived sample weight and data relationship to obtain the derivative parameters corresponding to the derived sample, the optimization parameters in the detection process are replaced with the derivative parameters to obtain the derivative process, and when the judgment result feedback is that part of the previous weight information is an integer, the process of judging that the previous weight information is fed back as an integer and the process of judging that the previous weight information is not an integer are performed and integrated to obtain the derivative process.

5. The method for detecting nutrients in cod intestines based on machine learning according to claim 1, characterized in that: The process of finding the chromatographic peak corresponding to the target substance in the chromatogram to obtain the target chromatographic peak is as follows: based on the reference library, a chromatogram with a known detection item concentration is obtained to obtain a comparison chromatogram, the peak residence time in the comparison chromatogram is recorded to obtain the comparison time, a correspondence between the comparison time and the detection item is established, a comparison library is established to store the comparison time, the detection item and the correspondence, the residence time of the chromatographic peak in the chromatogram is recorded to obtain the target time, the detection item consistent with the target substance is retrieved from the comparison library to obtain the comparison item, the comparison time corresponding to the comparison item is extracted based on the correspondence to obtain the reference time, and the chromatographic peak with the same target time and reference time is selected to obtain the target chromatographic peak.

6. The method for detecting nutrients in cod intestines based on machine learning according to claim 1, characterized in that: The reference establishment method includes: presetting a target concentration, obtaining a cod intestine sample with a detection item of the target concentration to obtain a reference sample, detecting the detection item concentration of the reference sample based on gas chromatography to obtain a reference chromatogram, establishing a correlation between the target concentration and the chromatogram, integrating the target concentration and the chromatogram to obtain a sub-reference, and establishing a reference library for storing the sub-reference.

7. A cod intestine nutrient detection device based on machine learning, characterized in that: A method for detecting nutrients in cod intestines based on machine learning as described in any one of claims 1 to 6 is used.

Images