Pesticide residue detection method and system based on big data

Abstract

The invention relates to the technical field of data processing, in particular to a pesticide residue detection method and system based on big data. The method comprises the following steps: acquiring a surface spectrum curve of a target crop and pesticide related parameters; determining a dynamic adjustment coefficient of the surface spectrum curve; for a target wave band in each wave band in the surface spectrum curve, determining a time decay factor of the target wave band; determining the spectral sensitivity of the target wave band; determining the weight of the target wave band; dimensionality reduction is carried out on the surface spectrum curve, and a pesticide residue detection model is established, so that pesticide residue detection based on big data is realized. According to the method, the characteristics of each wave band are extracted and the weight is calculated, so that key information most related to the pesticide residues can be kept in the dimension reduction process, then the wave bands are screened according to the weight, noise information irrelevant to or relatively low in correlation with the pesticide residues is removed, the risk of misinformation and missing report is reduced, and the accuracy of the pesticide residues is improved. The accuracy of pesticide residue detection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a pesticide residue detection method and system based on big data. Background Art

[0002] As one of the major challenges in the field of global food safety, the research and application of pesticide residue detection technology is of great significance. Among many detection methods, hyperspectral data analysis technology has become the mainstream technical means in modern agricultural detection and analysis due to its advantages of being non-destructive, non-polluting and requiring no sample pretreatment. The core of this technology is to achieve qualitative detection and quantitative evaluation of pesticide residues by analyzing the spectral features in hyperspectral data.

[0003] However, hyperspectral data usually contains hundreds to thousands of bands, which brings significant challenges to data processing. First, hyperspectral data without dimensionality reduction will take up a lot of computing resources and storage space, affecting detection efficiency. Second, the large number of features in high-dimensional data may cause model overfitting, that is, the model pays too much attention to the noise or irrelevant information in the data, thereby reducing detection accuracy.

[0004] Therefore, dimensionality reduction of hyperspectral data becomes a key link to improve detection performance. Among the traditional dimensionality reduction methods, the principal component analysis (PCA) algorithm and the t-distributed stochastic neighbor embedding (t-SNE) algorithm are more commonly used, but both have certain limitations. The PCA algorithm is based on the linear assumption and may not be able to fully retain important information in the data when processing pesticide residue hyperspectral data with nonlinear characteristics; and although the t-SNE algorithm can better process nonlinear data, its dimensionality reduction results are less interpretable, which is not conducive to subsequent decision analysis for pesticide residue detection that requires clear feature information. Summary of the invention

[0005] In order to solve the problem that when the traditional SMOTE algorithm randomly selects K nearest neighbor points, more minority class samples near the majority class samples are selected as neighbors, making the newly generated samples more inclined to the feature space of the majority class samples, thereby aggravating the data imbalance, resulting in a decrease in the recognition ability of minority class samples when the data set is subsequently used to train the classification model, thereby reducing the accuracy of the classification model for pesticide residue detection, the present invention provides a pesticide residue detection method and system based on big data.

[0006] In the first aspect, the present invention provides a pesticide residue detection method based on big data, which adopts the following technical solution: A pesticide residue detection method based on big data includes: obtaining a surface spectral curve of a target crop among a plurality of crops with known pesticide residue concentrations, and the spraying interval, half-life, actual spraying amount, calibrated reference dose, optimal degradation temperature of the pesticide most recently used in each band within the surface spectral curve, and the collection temperature of the target crop; determining a dynamic adjustment coefficient of the surface spectral curve according to the standard deviation and the second-order derivative of each band within the surface spectral curve, and the spraying interval and the half-life; for a target band among each band within the surface spectral curve, determining the target band according to the spraying interval, half-life, actual spraying amount and calibrated reference dose of the pesticide most recently used in the target band. time attenuation factor; determine the spectral sensitivity of the target band according to the standard deviation and second-order derivative of the target band, the maximum value of the standard deviation and the maximum value of the second-order derivative of each band, the half-life and optimal degradation temperature of the pesticide most recently used in the target band, and the collection temperature of the target crop; determine the weight of the target band according to the dynamic adjustment coefficient, the time attenuation factor and the spectral sensitivity of the target band; according to the size of the weight, reduce the dimension of the surface spectral curve to obtain the reduced-dimensional surface spectral curve, use the target crops with known pesticide residue concentration as the training set of the machine learning algorithm, and establish a pesticide residue detection model based on the reduced-dimensional surface spectral curve to realize pesticide residue detection based on big data.

[0007] The present invention can accurately determine the dynamic adjustment coefficient, time decay factor and spectral sensitivity by combining multiple factors such as the standard deviation, second-order derivative, spraying interval, half-life, actual spraying amount and other factors of the surface spectral curve, so as to more accurately simulate the time decay process of pesticide residues and better capture the changes in the surface spectra of different crops; the traditional SMOTE algorithm easily makes the generated samples biased towards the feature space of the majority class, while the present invention avoids the data bias caused by simple oversampling or downsampling by comprehensively considering environmental factors such as the interval of pesticide spraying and degradation temperature, especially in the detection of crops with pesticide residues, it can better capture the characteristics of the minority class (i.e., samples with low or no pesticide residues); by calculating the weight of each band, it can achieve the highlighting of important information and enhance the adaptability of the algorithm. The weight of the band is determined by multiple factors. The method is stable, so that the spectral curve can still retain key features after dimension reduction, avoiding excessive information loss; in the process of pesticide residue detection, the collection temperature, degradation temperature and spraying history of crops may have different effects on the pesticide residue. Through accurate spectral sensitivity calculation, the above factors are comprehensively considered, which is conducive to effective residue detection under different environmental conditions; since dynamic factors and environmental influences are taken into account, the present invention can more accurately identify the pesticide residue concentration in the target crops than the traditional method, which is conducive to improving the detection accuracy of the pesticide residue detection system, and thus enhancing the monitoring and management capabilities of pesticide residue problems; key features are retained in the process of dimensionality reduction, and the trained pesticide residue detection model has more generalization ability, can accurately identify the pesticide residue concentration of different types of crops in practical applications, and is not easily affected by overfitting.

[0008] Furthermore, the dynamic adjustment coefficient satisfies: ; In the formula, is the dynamic adjustment coefficient of the surface spectral curve of the target crop, is the number of bands in the surface spectral curve, The surface spectrum curve The standard deviation of the band, The surface spectrum curve The second-order derivative of the band, and They are the surface spectral curves. The spraying interval and half-life of the most recent pesticide used in each band, is the mean function, is a natural exponential function.

[0009] By combining standard deviation, second-order derivative, spraying interval and half-life, the dynamic adjustment coefficient of the present invention can sensitively capture changes related to pesticide residues in the surface spectral curve, thereby improving the detection accuracy of pesticide residues; the exponential function takes into account the nonlinear process of pesticide attenuation, so that the dynamic adjustment coefficient can automatically adjust the weight of the band according to time changes, so as to better adapt to different time and environmental conditions; the mean function averages the adjustment coefficients of multiple bands, which helps to avoid the influence of abnormal data of a single band on the model, reduce overfitting, and improve the robustness and stability of the detection system; by comprehensively utilizing information from different bands, the sensitivity of the model to pesticide residue detection can be enhanced, thereby providing more accurate results in different agricultural environments and different types of pesticide residue detection.

[0010] Furthermore, the time attenuation factor satisfies: ; In the formula, The first The time attenuation factor of each band, and They are the surface spectral curves. The spraying interval and half-life of the most recent pesticide used in each band, and They are the surface spectral curves. The actual spraying amount of pesticides used in the last time in each band and the calibrated reference dose, is a natural exponential function.

[0011] The time attenuation factor of the present invention combines the spraying interval, half-life, actual spraying amount and calibrated reference dose, and can more accurately reflect the attenuation process of pesticides in the surface spectrum of target crops. The exponential decay models the time attenuation process of pesticide residues, which helps to accurately capture the attenuation law of pesticide residues over time. By introducing the actual spraying amount and the calibrated reference dose, it can adapt to the changes in the use amount of different pesticides, and then more accurately adjust the calculation of the attenuation factor to ensure that the model can make a more appropriate response to the actual spraying situation. By comprehensively considering the changes in time, spraying amount and reference dose, the time attenuation factor can provide a more accurate attenuation model, thereby improving the accuracy and stability of pesticide residue detection and avoiding errors or instability that may occur under different conditions.

[0012] Furthermore, the spectral sensitivity satisfies: ; In the formula, The first The spectral sensitivity of each band, The surface spectrum curve The standard deviation of the band, The surface spectrum curve The second-order derivative of the band, is the maximum value of the standard deviation of each band in the surface spectral curve, is the maximum value of the second-order derivative of each band in the surface spectrum curve, is the collection temperature of the target crop, and They are the surface spectral curves. The half-life and optimal degradation temperature of the pesticide most recently used in each band, is the absolute value symbol.

[0013] The present invention can more accurately describe and predict the spectral sensitivity of different bands by considering the standard deviation and second-order derivative of different bands, as well as the influence of temperature and pesticide degradation, so as to make pesticide detection more precise; the use of pesticides often affects the spectral characteristics of crops, and temperature changes may accelerate or slow down the degradation process of pesticides. By introducing the parameters of pesticide half-life and optimal degradation temperature, the influence of these factors on spectral sensitivity can be quantified, thereby more accurately reflecting pesticide residues and crop health.

[0014] Furthermore, the weights satisfy: ; In the formula, The first The weight of the band, is the dynamic adjustment coefficient of the surface spectral curve of the target crop, The first The time attenuation factor of each band, The first Spectral sensitivity of each band.

[0015] By combining spectral sensitivity and time attenuation factor, the present invention can more accurately reflect the changing characteristics of each band in the growth cycle of crops, thereby improving the reliability of the analysis results; the introduction of dynamic adjustment coefficient enables the system to automatically adjust the weight of each band according to different time points or environmental conditions, adapt to the spectral characteristics of crops at different growth stages, and increase flexibility; the introduction of time attenuation factor helps to eliminate the influence of external environmental changes (such as climate factors, observation conditions, etc.), making the crop surface spectral data more stable and accurate.

[0016] Furthermore, the method of reducing the dimension of the surface spectral curve according to the weight of each band to obtain the reduced-dimensional surface spectral curve includes: in response to the weight being not higher than a preset dimensionality reduction threshold, removing the target band from the surface spectral curve to obtain the reduced-dimensional surface spectral curve.

[0017] Furthermore, the machine learning algorithm adopts a support vector machine.

[0018] Furthermore, the machine learning algorithm adopts BP neural network.

[0019] Furthermore, the pesticide residue detection based on big data includes: for the crops to be detected, obtaining the surface spectral curve of the crops to be detected after dimensionality reduction, inputting the surface spectral curve after dimensionality reduction into the pesticide residue detection model, obtaining the pesticide residue concentration of the crops to be detected, and completing the pesticide residue detection based on big data.

[0020] In a second aspect, the present invention provides a pesticide residue detection system based on big data, which adopts the following technical solution: A pesticide residue detection system based on big data comprises: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned pesticide residue detection method based on big data is implemented.

[0021] By adopting the above technical solution, the above-mentioned pesticide residue detection method based on big data is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a terminal device based on the memory and the processor for easy use.

[0022] The present invention has the following technical effects: Different from the disadvantage that the random selection of K nearest neighbor points in the traditional SMOTE algorithm easily leads to the new sample biased towards the feature space of the majority class sample and aggravates the data imbalance, the present invention performs data processing by determining the dynamic adjustment coefficient, the time attenuation factor, the spectral sensitivity, etc., thereby avoiding the problem of aggravated data imbalance due to unreasonable sample selection, and helping the subsequent classification model to better identify the minority class samples; since the aggravation of the data imbalance problem is effectively avoided, the training data is more reasonable, thereby being able to improve the recognition ability of the classification model trained with the data set for the minority class samples, overcoming the defect of the traditional SMOTE algorithm that reduces the recognition ability of the classification model for the minority class samples; the surface spectral curve is comprehensively considered. The weight of each band is determined by using multiple relevant factors of the line (such as standard deviation, second-order derivative, spraying interval, half-life, actual spraying amount, calibrated benchmark dose, optimal degradation temperature, collection temperature, etc.), and a pesticide residue detection model is established after dimensionality reduction of the surface spectral curve. Compared with the traditional method, it considers more comprehensive factors and processes data more scientifically, which can improve the accuracy of pesticide residue detection by the classification model and make the detection results more reliable. The surface spectral curves of target crops and other data are analyzed and processed from multiple dimensions. The pesticide residue detection model is established by determining various factors and weights. The advantages of big data can be fully utilized to make the model more scientific and reliable, and meet the requirements of accuracy and scientificity for modern pesticide residue detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of a method for detecting pesticide residues based on big data in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0025] The present invention discloses a method for detecting pesticide residues based on big data. Figure 1 , comprising steps S1 to S6: S1: Obtain the surface spectral curve of the target crop and pesticide related parameters.

[0026] It should be noted that a representative detection area should be selected on the target crops (such as corn, wheat, etc.) to ensure that the area can reflect the overall pesticide residue situation, and suitable hyperspectral data equipment should be selected to ensure that its spectral range and resolution meet the detection requirements. According to the characteristics and detection requirements of the target crops, the parameters of the hyperspectral data system should be set, such as exposure time (such as 0.05s), scanning speed (such as 0.6cm / s), etc., and the hyperspectral data system should be used to obtain the surface spectral curve of the detection area of ​​the target crops.

[0027] The surface spectral curve of the target crop among multiple crops with known pesticide residue concentrations, as well as the most recent spraying interval, half-life, actual spraying amount, calibrated reference dose, optimal degradation temperature of the pesticide used for the target crop, and the collection temperature of the target crop are obtained.

[0028] S2: Determine the dynamic adjustment coefficient of the surface spectral curve.

[0029] It should be noted that the dynamic adjustment coefficient of the surface spectral curve is used to comprehensively consider the characteristics of the spectral curve itself and the influence of the pesticide spraying time factor on the spectrum; the standard deviation reflects the discreteness of the spectral data of each band, and the second-order derivative reflects the rate of change of the spectral curve in the band. These two parameters describe the characteristics of different bands from the perspective of the characteristics of the spectral curve itself, while the spraying interval and half-life (for example, 7 for organophosphorus) are considered from the perspective of the degradation of pesticides on the surface of crops. Over time, the residual amount of pesticides will change, thereby affecting the spectral curve; therefore, this step calculates the dynamic adjustment coefficient based on the above parameters.

[0030] The dynamic adjustment coefficient of the surface spectrum curve is determined according to the standard deviation and the second-order derivative of each band in the surface spectrum curve, as well as the spraying interval and the half-life.

[0031] Specifically, the dynamic adjustment coefficient satisfies: ; In the formula, is the dynamic adjustment coefficient of the surface spectral curve of the target crop, is the number of bands in the surface spectral curve, The surface spectrum curve The standard deviation of the band, The surface spectrum curve The second-order derivative of the band, and They are the surface spectral curves. The spraying interval and half-life of the most recent pesticide used in each band, is the mean function, is a natural exponential function.

[0032] If the standard deviation and second-order derivative of a certain band are large, it means that the spectral data of this band fluctuates greatly and has a high rate of change. The value of will be relatively large, which makes this band account for a larger proportion in the calculation of the dynamic adjustment coefficient, which means that this band has a greater impact on the overall characteristics of the spectral curve; The bigger, The smaller the value, the more time the pesticide takes to degrade and the less residual amount. The value of will increase, This indicates that with the decrease of pesticide residue, the influence of the spectral curve's own characteristics is relatively enhanced.

[0033] S3: For a target band in each band in the surface spectrum curve, determine a time attenuation factor of the target band.

[0034] It should be noted that pesticide residues will decay exponentially over time (for example, 7 for organophosphorus), and temperature changes will accelerate or delay the degradation process. The exponential decay of pesticide residues over time and temperature changes can be used to quantify the impact of the environment on the residues to avoid inflated test values ​​due to time lags. At the same time, the difference between the actual spray dose and the standard dose directly affects the residual concentration. Adjusting the time decay factor through the ratio of the two can reduce errors caused by uneven application operations.

[0035] The time attenuation factor of the target band is determined based on the spraying interval, half-life, actual spraying amount and calibrated reference dose of the pesticide most recently used in the target band.

[0036] Specifically, the time attenuation factor satisfies: ; In the formula, The first The time attenuation factor of each band, and They are the surface spectral curves. The spraying interval and half-life of the most recent pesticide used in each band, and They are the surface spectral curves. The actual spraying amount of pesticides used in the last time in each band and the calibrated reference dose, is a natural exponential function.

[0037] in, The bigger, The smaller the value, the more the pesticide residue decreases with the passage of time, and the time decay factor will also decrease accordingly; if the actual spraying amount Greater than the reference dose ,but Greater than 1 will cause the time decay factor Increase, on the contrary, if the actual spraying amount Less than the standard dose , then the time decay factor It will decrease, which indicates that the actual amount of pesticide used will affect the attenuation of pesticide residues.

[0038] S4: Determine the spectral sensitivity of the target band.

[0039] It should be noted that the specific absorption peaks of pesticide residues (such as the characteristic response of organophosphorus pesticides in the near-infrared band) may be masked by environmental noise. By calculating the second-order derivative of the spectrum, the weak absorption characteristics can be amplified and the detection sensitivity can be improved. Different types of pesticides (such as carbamates and pyrethroids) have different sensitive bands in the spectrum. Screening high-resolution bands through standard deviation can reduce redundant data interference. Therefore, the spectral sensitivity of the band is calculated here based on the standard deviation and second-order derivative of the band.

[0040] The spectral sensitivity of the target band is determined based on the standard deviation and second-order derivative of the target band, the maximum value of the standard deviation and the maximum value of the second-order derivative of each band, the half-life and optimal degradation temperature of the pesticide most recently used in the target band, and the collection temperature of the target crop.

[0041] Specifically, the spectral sensitivity satisfies: ; In the formula, The first The spectral sensitivity of each band, The surface spectrum curve The standard deviation of the band, The surface spectrum curve The second-order derivative of the band, is the maximum value of the standard deviation of each band in the surface spectral curve, is the maximum value of the second-order derivative of each band in the surface spectrum curve, is the collection temperature of the target crop, and They are the surface spectral curves. The half-life and optimal degradation temperature of the pesticide most recently used in each band, is the absolute value symbol.

[0042] in, The larger the value is, the more discrete the data in this band is, which may contain stronger characteristic signals of pesticides, that is, the band with larger standard deviation is more sensitive to residues. The larger the value is, the more obvious the characteristic peak of the band is, and the higher the sensitivity to pesticide residues is. The larger it is, the more it indicates that the degradation rate of pesticides has changed and the stronger the environmental interference is, and the residual error needs to be dynamically compensated through a proportional term.

[0043] S5: Determine the weight of the target band.

[0044] It should be noted that the weight of the target band is obtained by comprehensively considering the dynamic adjustment coefficient, time attenuation factor and spectral sensitivity; the dynamic adjustment coefficient reflects the overall dynamic characteristics of the surface spectral curve, the time attenuation factor reflects the change in the residual amount of pesticides in the target band over time, and the spectral sensitivity describes the sensitivity of the target band to pesticide residues and environmental factors. By weighted combination of these three parameters, the weight obtained can more comprehensively evaluate the importance of the target band in pesticide residue detection.

[0045] The weight of the target band is determined according to the dynamic adjustment coefficient, the time attenuation factor and the spectral sensitivity of the target band.

[0046] Specifically, the weights satisfy: ; In the formula, The first The weight of the band, is the dynamic adjustment coefficient of the surface spectral curve of the target crop, The first The time attenuation factor of each band, The first Spectral sensitivity of each band.

[0047] Among them, when the band spectral sensitivity is high, tends to 1, at this time, the spectral sensitivity is emphasized to obtain the weight; when the time attenuation factor is large, It tends to 0, and the time decay factor is emphasized to obtain the weight.

[0048] S6: Reduce the dimension of the surface spectral curve and establish a pesticide residue detection model to realize pesticide residue detection based on big data.

[0049] According to the size of the weight, the surface spectral curve is reduced in dimension to obtain the reduced-dimensional surface spectral curve. The target crops with known pesticide residue concentrations are used as the training set of the machine learning algorithm. A pesticide residue detection model is established based on the reduced-dimensional surface spectral curve to realize pesticide residue detection based on big data.

[0050] Specifically, the dimension reduction of the surface spectrum curve is performed according to the weight of each band to obtain the surface spectrum curve after dimension reduction, including: In response to the weight being no higher than a preset dimensionality reduction threshold, the target band is removed from the surface spectral curve to obtain a surface spectral curve after dimensionality reduction.

[0051] Implementers can set the dimensionality reduction threshold according to specific implementation circumstances, for example, 0.25.

[0052] Specifically, the machine learning algorithm adopts a support vector machine.

[0053] In another embodiment, the machine learning algorithm uses a BP neural network.

[0054] Specifically, the method for realizing pesticide residue detection based on big data includes: For the crops to be tested (the crops to be tested and the target crops are the same crops), the surface spectral curve after dimensionality reduction of the crops to be tested is obtained, and the surface spectral curve after dimensionality reduction is input into the pesticide residue detection model to obtain the pesticide residue concentration of the crops to be tested, thereby completing the pesticide residue detection based on big data.

[0055] An embodiment of the present invention also discloses a pesticide residue detection system based on big data, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a pesticide residue detection method based on big data according to the present invention is implemented.

[0056] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

[0057] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A pesticide residue detection method based on big data, characterized in that: include: Obtaining the surface spectral curve of the target crop among multiple crops with known pesticide residue concentrations, as well as the spraying interval, half-life, actual spraying amount, calibrated reference dose, optimal degradation temperature of the pesticide most recently used in each band within the surface spectral curve, and the collection temperature of the target crop; Determining a dynamic adjustment coefficient of the surface spectrum curve according to the standard deviation and the second-order derivative of each band in the surface spectrum curve, as well as the spraying interval and the half-life; For the target band in each band in the surface spectral curve, the time attenuation factor of the target band is determined according to the spraying interval, half-life, actual spraying amount and calibration reference dose of the pesticide used in the target band most recently; the spectral sensitivity of the target band is determined according to the standard deviation and second-order derivative of the target band, the maximum value of the standard deviation and the maximum value of the second-order derivative of each band, the half-life and optimal degradation temperature of the pesticide used in the target band most recently, and the collection temperature of the target crop; the weight of the target band is determined according to the dynamic adjustment coefficient, the time attenuation factor and the spectral sensitivity of the target band; According to the size of the weight, the surface spectral curve is reduced in dimension to obtain the reduced-dimensional surface spectral curve. The target crops with known pesticide residue concentrations are used as the training set of the machine learning algorithm. A pesticide residue detection model is established based on the reduced-dimensional surface spectral curve to realize pesticide residue detection based on big data.

2. The pesticide residue detection method based on big data according to claim 1, characterized in that: The dynamic adjustment coefficient satisfies: ； In the formula, is the dynamic adjustment coefficient of the surface spectral curve of the target crop, is the number of bands in the surface spectral curve, The surface spectrum curve The standard deviation of the band, The surface spectrum curve The second-order derivative of the band, and They are the surface spectral curves. The spraying interval and half-life of the most recent pesticide used in each band, is the mean function, is a natural exponential function.

3. The pesticide residue detection method based on big data according to claim 1, characterized in that: The time decay factor satisfies: ； In the formula, The first The time attenuation factor of each band, and They are the surface spectral curves. The spraying interval and half-life of the most recent pesticide used in each band, and They are the surface spectral curves. The actual spraying amount of pesticides used in the last time in each band and the calibrated reference dose, is a natural exponential function.

4. The pesticide residue detection method based on big data according to claim 1, characterized in that: The spectral sensitivity satisfies: ； In the formula, The first The spectral sensitivity of each band, The surface spectrum curve The standard deviation of the band, The surface spectrum curve The second-order derivative of the band, is the maximum value of the standard deviation of each band in the surface spectral curve, is the maximum value of the second-order derivative of each band in the surface spectrum curve, is the collection temperature of the target crop, and They are the surface spectral curves. The half-life and optimal degradation temperature of the pesticide most recently used in each band, is the absolute value symbol.

5. The pesticide residue detection method based on big data according to claim 1, characterized in that: The weights satisfy: ； In the formula, The first The weight of the band, is the dynamic adjustment coefficient of the surface spectral curve of the target crop, The first The time attenuation factor of each band, The first Spectral sensitivity of each band.

6. The pesticide residue detection method based on big data according to claim 1, characterized in that: The surface spectral curve is reduced in dimension according to the weight of each band to obtain the surface spectral curve after dimension reduction, including: In response to the weight being no higher than a preset dimensionality reduction threshold, the target band is removed from the surface spectral curve to obtain a surface spectral curve after dimensionality reduction.

7. The pesticide residue detection method based on big data according to claim 1, characterized in that: The machine learning algorithm adopts support vector machine.

8. The pesticide residue detection method based on big data according to claim 1, characterized in that: The machine learning algorithm adopts BP neural network.

9. The pesticide residue detection method based on big data according to claim 1, characterized in that: The method for realizing pesticide residue detection based on big data includes: For the crops to be tested, the surface spectral curves after dimensionality reduction of the crops to be tested are obtained, and the surface spectral curves after dimensionality reduction are input into the pesticide residue detection model to obtain the pesticide residue concentration of the crops to be tested, thereby completing the pesticide residue detection based on big data.

10. A pesticide residue detection system based on big data, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a pesticide residue detection method based on big data according to any one of claims 1 to 9 is implemented.

Images