Method and system for rapidly detecting pesticide residues of traditional Chinese medicine safflower carthamus

Abstract

The invention provides a rapid detection method and system for pesticide residues of traditional Chinese medicine safflower, and relates to the technical field of medicinal material detection. According to the method, the characteristics most related to the pesticide residue concentration are screened out through the multi-source spectral data, the characteristics are further screened based on the correlation of the auxiliary information, the accuracy of the characteristics is improved through the comprehensive characteristics, the recognition capability on the pesticide residues is effectively improved, the detection precision and speed are further remarkably improved, the detection cost is reduced, and then the detection efficiency is improved. The application of a real-time monitoring and automatic alarm mechanism ensures that abnormity in the detection process can be found and processed in time, and powerful technical support is provided for quality control and safety guarantee of the traditional Chinese medicine safflower carthamus.

Description

Technical Field

[0001] The invention relates to the technical field of medicinal material detection, and in particular to a method and system for rapid detection of pesticide residues in a traditional Chinese medicine, safflower. Background Art

[0002] In the traditional Chinese medicine industry, safflower is a traditional and widely used Chinese herbal medicine, and its quality and safety are directly related to clinical efficacy and patient health. However, with the rapid development of modern agriculture, the widespread use of pesticides has increased crop yields to a certain extent, but it has also brought about the problem of pesticide residues. For Chinese herbal medicines such as safflower, pesticide residues may not only reduce their efficacy, but also pose a potential threat to human health.

[0003] Traditional pesticide residue detection methods, such as chromatography and mass spectrometry, have high detection accuracy, but often have disadvantages such as long detection cycle, complex operation, and high cost, which are difficult to meet the needs of rapid detection of Chinese herbal medicines. In particular, for scenarios such as Chinese herbal medicine planting bases and Chinese herbal medicine processing companies that require large-scale and rapid screening, traditional detection methods are particularly inadequate.

[0004] Therefore, it is necessary to provide a method and system for rapid detection of pesticide residues in the traditional Chinese medicine safflower to solve the above technical problems. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a method and system for rapid detection of pesticide residues in the traditional Chinese medicine safflower, which are used to solve the problem of low detection efficiency and difficulty in meeting the demand for rapid detection of traditional Chinese medicines.

[0006] The present invention provides a method for rapid detection of pesticide residues in a traditional Chinese medicine safflower, the method comprising: Collect and preprocess the multi-source spectral data of the current safflower sample to obtain a multi-source spectral data set; The feature selection algorithm is trained based on sample data with known pesticide residue concentrations to obtain a feature screening model; The feature screening model is used to extract primary features from the preprocessed multi-source spectral data set, and the extraction process is monitored to determine abnormal extraction results to trigger an alarm and re-extraction; Collect auxiliary information of safflower samples and perform joint analysis with multi-source spectral datasets to obtain secondary features; Combining the primary features and the secondary features, a feature collection is obtained; Based on the correlation between the residue concentration and characteristics of known safflower samples as the training set, the pesticide residue concentration model was trained; The feature collection is input into the pesticide residue concentration model to obtain the pesticide residue concentration detection result of the current safflower sample.

[0007] Preferably, the method of collecting the multi-source spectral data of the current safflower sample and preprocessing it to obtain the multi-source spectral data set specifically includes the following steps: The spectral data of the current safflower sample are collected by using a variety of spectrometers under different measurement conditions and at different time points; The collected spectral data are preprocessed to obtain a multi-source spectral data set, wherein the preprocessing specifically includes denoising processing, correction processing and standardization processing.

[0008] Preferably, the training of the feature selection algorithm based on the sample data of known pesticide residue concentration to obtain the feature screening model specifically comprises the following steps: Collect safflower sample data with known pesticide residue concentrations, including spectral data of the safflower sample data and corresponding pesticide residue concentration values; Based on the spectral data and pesticide residue concentration values ​​of safflower sample data, the features of the spectral data are extracted, and the correlation between the features and the pesticide residue concentration values ​​is identified; A feature selection algorithm model was selected, and the correlation between features and pesticide residue concentration values ​​was used as a training set to train the feature selection algorithm model to obtain a feature screening model.

[0009] Preferably, the method of extracting primary features from the preprocessed multi-source spectral data set by the feature screening model, monitoring the extraction process, and judging abnormal extraction results to trigger an alarm for re-extraction specifically includes the following steps: The multi-source spectral data set is input into the feature screening model to extract features and obtain primary features; Preset monitoring thresholds, including feature abnormal fluctuation thresholds and data integrity thresholds; Monitor the data situation during the feature extraction process in real time, and use the preset monitoring threshold to determine if the data in the extraction process exceeds the monitoring threshold, then trigger an alarm and re-execute the extraction operation.

[0010] Preferably, the collecting of auxiliary information of safflower samples and performing joint analysis with multi-source spectral data sets to obtain secondary features specifically includes the following steps: Obtain auxiliary information of safflower samples, including origin, planting method and growth cycle data; Through statistical analysis and data mining, the correlation between auxiliary information and spectral data of safflower samples was identified; Based on the correlation between the auxiliary information and spectral data of safflower samples, secondary features related to the pesticide residue concentration were extracted.

[0011] Preferably, the step of integrating the primary features and the secondary features to obtain a feature collection specifically includes the following steps: Integrate the obtained primary features and secondary features, remove duplicate features, and obtain a preliminary integrated feature set; The mutual information MI feature selection technology is used to screen the preliminary integrated feature set, remove useless features, and obtain a feature collection.

[0012] Preferably, the method of using the correlation between the residue concentration and the feature of the known safflower sample as a training set to train a pesticide residue concentration model specifically comprises the following steps: Obtaining data on the residual concentration and characteristics of known safflower samples, and obtaining the corresponding relationship between the residual concentration and characteristics of the safflower samples by a correlation analysis method; The neural network (NN) algorithm was selected as the initial model, and the correspondence between the residue concentration and characteristics of safflower samples was used as the training set for training to obtain the pesticide residue concentration model.

[0013] A rapid detection system for pesticide residues in a Chinese medicinal herb, safflower, comprising: An acquisition and processing module is used to acquire multi-source spectral data of the current safflower sample and perform preprocessing to obtain a multi-source spectral data set; The first model building module is used to train a feature selection algorithm based on sample data with known pesticide residue concentrations to obtain a feature screening model; The feature extraction and monitoring module is used to extract primary features from the preprocessed multi-source spectral data set through the feature screening model, monitor the extraction process, and judge abnormal extraction results to trigger alarm re-extraction; The joint analysis module is used to collect auxiliary information of safflower samples and conduct joint analysis with multi-source spectral datasets to obtain secondary features; The sorting module is used to integrate the primary features and the secondary features to obtain a feature collection; The second model building module is used to train a pesticide residue concentration model based on the correlation between the residue concentration and the characteristics of the known safflower samples as a training set; The detection module is used to input the feature set into the pesticide residue concentration model to obtain the pesticide residue concentration detection result of the current safflower sample.

[0014] Compared with the related art, the method and system for rapid detection of pesticide residues in the Chinese medicinal material safflower provided by the present invention have the following beneficial effects: The present invention screens out the features most relevant to the pesticide residue concentration through multi-source spectral data, and further screens the features based on the correlation of auxiliary information. The accuracy of the features is improved by comprehensive features, which effectively enhances the recognition ability of pesticide residues, thereby significantly improving the accuracy and speed of detection and reducing the detection cost. Secondly, the application of real-time monitoring and automatic alarm mechanism ensures that anomalies in the detection process can be discovered and handled in time, providing strong technical support for the quality control and safety assurance of the traditional Chinese medicine safflower. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of a method for rapid detection of pesticide residues in a Chinese medicinal herb, safflower, according to the present invention; Figure 2 The present invention is a system block diagram of a rapid detection system for pesticide residues in the traditional Chinese medicine safflower. DETAILED DESCRIPTION

[0016] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0017] Embodiment 1 like Figure 1 As shown, a method for rapid detection of pesticide residues in safflower, a traditional Chinese medicine, specifically comprises the following steps: S100, collecting multi-source spectral data of the current safflower sample and performing preprocessing to obtain a multi-source spectral data set; In the specific implementation process, the specific steps of step S100 are: S101, collecting spectral data of the current safflower sample by using a variety of spectrometers under different measurement conditions and at different time points; Specifically, the spectral data of the current safflower sample is collected by an ultraviolet-visible spectrometer, a near-infrared spectrometer and a fluorescence spectrometer under specifically set measurement conditions and time points, respectively, wherein the set measurement conditions include but are not limited to setting wavelength range and resolution.

[0018] Exemplarily, safflower samples were measured using a UV-visible spectrometer in the range of 200-800 nm, and three measurements were performed immediately, 24 hours later, and 48 hours later after collection.

[0019] S102, preprocessing the collected spectral data to obtain a multi-source spectral data set, wherein the preprocessing specifically includes denoising, correction and standardization; Specifically, the moving average or Savitzky-Golay filter in the smoothing algorithm is used to remove random noise in the collected spectral data, improve the signal-to-noise ratio, identify and eliminate outliers caused by measurement errors or sample contamination, adjust the baseline of the spectral data to make it more stable for subsequent analysis, and normalize the spectral data to the same scale so that the spectral data between different samples can be compared.

[0020] S200, training a feature selection algorithm based on sample data with known pesticide residue concentrations to obtain a feature screening model; In the specific implementation process, the specific steps of step S200 are: S201, collecting safflower sample data with known pesticide residue concentrations, including spectral data of the safflower sample data and corresponding pesticide residue concentration values; Specifically, safflower samples with known pesticide residue concentrations are obtained from reliable sources, including safflower sample data of different pesticide types and residue concentration ranges, and spectral data are collected. It should be noted that each sample is measured under the same conditions and the data is obtained through precise chemical analysis methods.

[0021] S202, based on the spectral data and the pesticide residue concentration value of the safflower sample data, extracting features of the spectral data, and identifying the correlation between the features and the pesticide residue concentration value; Specifically, various features are extracted from the spectral data and pesticide residue concentration values ​​of the safflower sample data, including but not limited to the peak value, valley value, slope, area, wavelength position, etc. of the spectrum, and regression analysis in statistical methods is used to analyze the correlation between each feature and the pesticide residue concentration.

[0022] S203, selecting a feature selection algorithm model, using the correlation between the feature and the pesticide residue concentration value as a training set, and training the feature selection algorithm model to obtain a feature screening model; Specifically, according to the characteristics of the problem and the scale of the data, the filtering method in the feature selection algorithm is selected, and the features identified in step S202 and the corresponding pesticide residue concentration values ​​are used as training sets to train the feature selection algorithm model. After verification, a feature screening model that can screen out the features most relevant to the pesticide residue concentration is obtained.

[0023] S300, extracting primary features from the preprocessed multi-source spectral data set through a feature screening model, monitoring the extraction process, and judging abnormal extraction results to trigger an alarm for re-extraction; In the specific implementation process, the specific steps of step S300 are: S301, inputting the multi-source spectral data set into the feature screening model to perform feature extraction and obtain primary features; Specifically, the preprocessed multi-source spectral data set is input into the feature screening model trained in step S203, and the feature screening model processes the spectral data according to the learned feature importance, and extracts the primary features most relevant to the pesticide residue concentration.

[0024] S302, preset monitoring thresholds, specifically including feature abnormal fluctuation thresholds and data integrity thresholds; Specifically, a feature abnormal fluctuation threshold is preset based on historical data or experience to determine whether the extracted feature value fluctuates within a reasonable range, and a data integrity threshold is preset, specifically a threshold such as the number of data points, the proportion of missing data, etc., to determine whether the extracted data meets the analysis requirements.

[0025] S303, real-time monitoring of the data during the feature extraction process, and judging through a preset monitoring threshold that the data during the extraction process exceeds the monitoring threshold, triggering an alarm to re-execute the extraction operation; Specifically, during the feature extraction process, the extracted data is monitored in real time, including checking whether each eigenvalue is stable, whether the data points are complete, etc., and then the real-time monitored data is compared with the preset monitoring threshold to determine whether a certain eigenvalue exceeds the abnormal feature fluctuation threshold, or the data integrity is lower than the data integrity threshold, then it is determined to be an abnormal situation. Once an abnormal situation is detected, the existing alarm mechanism is immediately triggered, where the alarm methods include but are not limited to sound, light, and email notification. Finally, after the alarm, the feature extraction operation is automatically or manually re-executed, including steps such as reloading spectral data and re-running the feature screening model.

[0026] S400, collect auxiliary information of safflower samples and perform joint analysis with multi-source spectral datasets to obtain secondary features; In the specific implementation process, the specific steps of step S400 are: S401, obtaining auxiliary information of safflower samples, including origin, planting method and growth cycle data; Specifically, auxiliary information of each sample was collected from the relevant database of growers of safflower samples, including the geographical location where the safflower was grown, including different climate and soil conditions, the planting method including whether organic fertilizers were used, irrigation methods, pest and disease control measures, etc., the growth cycle including the length of time from sowing to harvesting, and information on key growth stages.

[0027] S402, identifying the correlation between the auxiliary information and the spectral data of the safflower sample by means of statistical analysis and data mining; Specifically, necessary preprocessing is performed on the spectral data and auxiliary information, including missing value filling, outlier detection and processing, data standardization, etc., to ensure data quality, and correlation analysis in statistical methods is used to explore the potential relationship between auxiliary information and spectral data. For example, it is possible to analyze whether there are significant differences in the spectral characteristics of safflower from different origins, and then use deeper data mining techniques such as association rule mining to deeply explore the complex relationship between auxiliary information and spectral data, and identify the correlation between the auxiliary information and spectral data of safflower samples.

[0028] S403, extracting secondary features related to pesticide residue concentration based on the correlation between the auxiliary information of the safflower sample and the spectral data; Specifically, based on the results of statistical analysis and data mining, auxiliary information and spectral features that are highly correlated with pesticide residue concentration and rich in information content are selected as candidate secondary features. Then, the selected auxiliary information and spectral features are fused to form new secondary features through feature weighting and feature transformation.

[0029] For example, the spectral intensity of safflower samples grown using traditional planting methods (using chemical pesticides) at specific wavelengths is significantly higher than that of samples grown using organic planting methods. At the same time, data mining results show that there is a certain nonlinear relationship between the growth cycle and spectral characteristics of these samples. Therefore, "planting method" and "growth cycle" can be used as auxiliary information and fused with the specific wavelength intensity in the spectral data to form a new secondary feature.

[0030] S500, integrating the primary features and the secondary features, and obtaining a feature collection; In the specific implementation process, the specific steps of step S500 are: S501, integrating the obtained primary features and secondary features, and removing duplicate features to obtain a preliminary integrated feature set; Specifically, the primary features extracted in step S300 and the secondary features extracted in step S400 are merged to form a preliminary integrated feature set, which includes multiple features from spectral data and auxiliary information, and can more comprehensively reflect the characteristics of safflower samples. The cosine similarity method is used and a threshold is set to screen and remove duplicate features.

[0031] S502, using the mutual information MI feature selection technology to screen the preliminary integrated feature set, remove useless features, and obtain a feature set; Specifically, in feature selection, mutual information is used to evaluate the correlation between features and pesticide residue concentrations. Specifically, for each feature in the preliminary integrated feature set, the mutual information value between it and the pesticide residue concentration is calculated, the features are sorted according to the mutual information value, and the top-ranked features are selected as the final feature collection. It should be noted that when selecting features, a threshold is set according to actual conditions (such as the MI value is greater than a certain value) or a fixed number of features (such as the first N features) is selected. Finally, the screened features are sorted into a feature collection in a certain order (such as MI values ​​from high to low).

[0032] For example, in the preliminary integrated feature set, there are two features F1 and F2 that have a certain correlation with the pesticide residue concentration. However, by calculating the mutual information value, it is found that the MI value of F1 is significantly higher than that of F2, and the similarity between F1 and F2 is also high (that is, there is repeated information). Therefore, in the feature selection process, it is chosen to retain F1 and remove F2 to reduce feature redundancy and improve the efficiency of the model.

[0033] S600, based on the correlation between the residue concentration and the characteristics of the known safflower samples as a training set, a pesticide residue concentration model is trained; In the specific implementation process, the specific steps of step S600 are: S601, obtaining data on the residual concentration and characteristics of known safflower samples, and obtaining the corresponding relationship between the residual concentration and the characteristics of the safflower samples by a correlation analysis method; Specifically, a set of safflower sample data with known pesticide residue concentrations was collected, including the spectral characteristics of the samples, auxiliary information, and pesticide residue concentration values ​​obtained by standard detection methods (such as chromatographic analysis). The collected data were sorted to ensure that the characteristics and residue concentration values ​​of each sample corresponded correctly. The Pearson correlation coefficient in statistical methods was used to analyze the correlation between the residue concentration of the safflower samples and each feature, and then the correspondence between the residue concentration and the feature was obtained through correlation analysis.

[0034] S602, selecting a neural network NN algorithm as an initial model, using the correspondence between the residue concentration and the characteristics of the safflower sample as a training set for training, and obtaining a pesticide residue concentration model; Specifically, based on the analysis result of step S601, a neural network (NN) algorithm is selected as the initial model, and then the correspondence between the residue concentration and the characteristics of the safflower sample is used as a training set for training. The model will continuously adjust the network weights and biases according to the input features (i.e., spectral features and auxiliary information) and the corresponding residue concentration values ​​to minimize the error between the predicted value and the actual value. The model is tuned through the validation set data to finally obtain the pesticide residue concentration model.

[0035] S700: Input the feature set into the pesticide residue concentration model to obtain the pesticide residue concentration detection result of the current safflower sample.

[0036] Embodiment 2 like Figure 2 As shown, a rapid detection system for pesticide residues of safflower, a traditional Chinese medicine, is applied to a rapid detection method for pesticide residues of safflower, specifically comprising: An acquisition and processing module is used to acquire multi-source spectral data of the current safflower sample and perform preprocessing to obtain a multi-source spectral data set; Specifically, the acquisition and processing module is used to collect spectral data of the current safflower sample through an ultraviolet-visible spectrometer, a near-infrared spectrometer and a fluorescence spectrometer, and respectively collect spectral data under specifically set measurement conditions and time points, and preprocess the collected spectral data to obtain a multi-source spectral data set, wherein the preprocessing specifically includes denoising processing, correction processing and standardization processing.

[0037] The first model building module is used to train a feature selection algorithm based on sample data with known pesticide residue concentrations to obtain a feature screening model; Specifically, the first model building module is used to obtain safflower samples with known pesticide residue concentrations from reliable sources, including safflower sample data of different pesticide types and residue concentration ranges, and collect spectral data thereof. It should be noted that each sample is ensured to be measured under the same conditions and its data is obtained through precise chemical analysis methods; and based on the spectral data and pesticide residue concentration values ​​of the safflower sample data, the features of the spectral data are extracted, and the correlation between the features and the pesticide residue concentration values ​​is identified, and a feature selection algorithm model is selected. The correlation between the features and the pesticide residue concentration values ​​is used as a training set to train the feature selection algorithm model to obtain a feature screening model.

[0038] The feature extraction and monitoring module is used to extract primary features from the preprocessed multi-source spectral data set through the feature screening model, monitor the extraction process, and judge abnormal extraction results to trigger alarm re-extraction; Specifically, the feature extraction and monitoring module is used to input the multi-source spectral data set into the feature screening model for feature extraction, obtain the primary feature, and preset a feature abnormal fluctuation threshold based on historical data or experience to determine whether the extracted feature value fluctuates within a reasonable range, and preset a data integrity threshold, specifically the number of data points, the proportion of missing data and other thresholds, to determine whether the extracted data meets the analysis requirements. During the feature extraction process, the extracted data is monitored in real time; the real-time monitored data is compared with the preset monitoring threshold to determine that a certain feature value exceeds the feature abnormal fluctuation threshold, or the data integrity is lower than the data integrity threshold, then it is judged as an abnormal situation. Once an abnormal situation is detected, the existing alarm mechanism is immediately triggered, and the feature extraction operation is automatically or manually re-executed.

[0039] The joint analysis module is used to collect auxiliary information of safflower samples and conduct joint analysis with multi-source spectral datasets to obtain secondary features; Specifically, auxiliary information of safflower samples is obtained, including origin, planting method and growth cycle data. The correlation between the auxiliary information and spectral data of safflower samples is identified through statistical analysis and data mining. Based on the results of statistical analysis and data mining, auxiliary information and spectral features that are highly correlated with pesticide residue concentration and rich in information content are selected as candidate secondary features. Then, the selected auxiliary information and spectral features are fused to form new secondary features through feature weighting and feature transformation.

[0040] The sorting module is used to integrate the primary features and the secondary features to obtain a feature collection; The second model building module is used to train a pesticide residue concentration model based on the correlation between the residue concentration and the characteristics of the known safflower samples as a training set; Specifically, the second model building module obtains the data of the residue concentration and characteristics of the known safflower samples, and obtains the correspondence between the residue concentration and the characteristics of the safflower samples through the correlation analysis method, and selects the neural network NN algorithm as the initial model, and uses the correspondence between the residue concentration and the characteristics of the safflower samples as the training set for training to obtain the pesticide residue concentration model.

[0041] The detection module is used to input the feature set into the pesticide residue concentration model to obtain the pesticide residue concentration detection result of the current safflower sample.

[0042] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0043] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0044] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A rapid detection method for pesticide residues in the Chinese medicinal herb safflower. It is characterized in that The detection method comprises: Collect and preprocess the multi-source spectral data of the current safflower sample to obtain a multi-source spectral data set; The feature selection algorithm is trained based on sample data with known pesticide residue concentrations to obtain a feature screening model; The feature screening model is used to extract primary features from the preprocessed multi-source spectral data set, and the extraction process is monitored to determine abnormal extraction results to trigger an alarm and re-extraction; Collect auxiliary information of safflower samples and perform joint analysis with multi-source spectral datasets to obtain secondary features; Combining the primary features and the secondary features, a feature collection is obtained; Based on the correlation between the residue concentration and characteristics of known safflower samples as the training set, the pesticide residue concentration model was trained; The feature collection is input into the pesticide residue concentration model to obtain the pesticide residue concentration detection result of the current safflower sample.

2. A method for rapid detection of pesticide residues in safflower, a traditional Chinese medicine, according to claim 1, It is characterized in that The method of collecting the multi-source spectral data of the current safflower sample and preprocessing it to obtain a multi-source spectral data set specifically includes the following steps: The spectral data of the current safflower sample are collected by using a variety of spectrometers under different measurement conditions and at different time points; The collected spectral data are preprocessed to obtain a multi-source spectral data set, wherein the preprocessing specifically includes denoising processing, correction processing and standardization processing.

3. A method for rapid detection of pesticide residues in safflower, a traditional Chinese medicine, according to claim 1, It is characterized in that The feature selection algorithm is trained based on sample data with known pesticide residue concentrations to obtain a feature screening model, which specifically includes the following steps: Collect safflower sample data with known pesticide residue concentrations, including spectral data of the safflower sample data and corresponding pesticide residue concentration values; Based on the spectral data and pesticide residue concentration values ​​of safflower sample data, the features of the spectral data are extracted, and the correlation between the features and the pesticide residue concentration values ​​is identified; A feature selection algorithm model was selected, and the correlation between features and pesticide residue concentration values ​​was used as a training set to train the feature selection algorithm model to obtain a feature screening model.

4. A method for rapid detection of pesticide residues in safflower, a traditional Chinese medicine, according to claim 1, It is characterized in that The method of extracting primary features from the preprocessed multi-source spectral data set through the feature screening model, monitoring the extraction process, and judging abnormal extraction results to trigger an alarm for re-extraction specifically includes the following steps: The multi-source spectral data set is input into the feature screening model to extract features and obtain primary features; Preset monitoring thresholds, including feature abnormal fluctuation thresholds and data integrity thresholds; Monitor the data situation during the feature extraction process in real time, and use the preset monitoring threshold to determine if the data in the extraction process exceeds the monitoring threshold, then trigger an alarm and re-execute the extraction operation.

5. A method for rapid detection of pesticide residues in safflower, a traditional Chinese medicine, according to claim 1, It is characterized in that The collecting of auxiliary information of safflower samples and joint analysis with multi-source spectral data sets to obtain secondary features specifically includes the following steps: Obtain auxiliary information of safflower samples, including origin, planting method and growth cycle data; Through statistical analysis and data mining, the correlation between auxiliary information and spectral data of safflower samples was identified; Based on the correlation between the auxiliary information and spectral data of safflower samples, secondary features related to the pesticide residue concentration were extracted.

6. A method for rapid detection of pesticide residues in safflower, a traditional Chinese medicine, according to claim 1, It is characterized in that The step of integrating the primary features and the secondary features to obtain a feature collection specifically includes the following steps: Integrate the obtained primary features and secondary features, remove duplicate features, and obtain a preliminary integrated feature set; The mutual information MI feature selection technology is used to screen the preliminary integrated feature set, remove useless features, and obtain a feature collection.

7. A method for rapid detection of pesticide residues in safflower, a traditional Chinese medicine, according to claim 1, It is characterized in that The method uses the correlation between the residual concentration and the feature of the known safflower sample as a training set to train a pesticide residue concentration model, which specifically includes the following steps: Obtaining data on the residual concentration and characteristics of known safflower samples, and obtaining the corresponding relationship between the residual concentration and characteristics of the safflower samples by a correlation analysis method; The neural network (NN) algorithm was selected as the initial model, and the correspondence between the residue concentration and characteristics of safflower samples was used as the training set to obtain the pesticide residue concentration model.

8. A system for rapid detection of pesticide residues in safflower, a Chinese medicinal herb, applied to a method for rapid detection of pesticide residues in safflower, a Chinese medicinal herb, as claimed in any one of claims 1 to 7. It is characterized in that The detection system comprises: An acquisition and processing module is used to acquire multi-source spectral data of the current safflower sample and perform preprocessing to obtain a multi-source spectral data set; The first model building module is used to train a feature selection algorithm based on sample data with known pesticide residue concentrations to obtain a feature screening model; The feature extraction and monitoring module is used to extract primary features from the preprocessed multi-source spectral data set through the feature screening model, monitor the extraction process, and judge abnormal extraction results to trigger alarm re-extraction; The joint analysis module is used to collect auxiliary information of safflower samples and conduct joint analysis with multi-source spectral datasets to obtain secondary features; The sorting module is used to integrate the primary features and the secondary features to obtain a feature collection; The second model building module is used to train a pesticide residue concentration model based on the correlation between the residue concentration and the characteristics of the known safflower samples as a training set; The detection module is used to input the feature set into the pesticide residue concentration model to obtain the pesticide residue concentration detection result of the current safflower sample.

Images