A method for automatic analysis of whole algae based on fluorescent quantum dots

Abstract

The invention relates to the technical field of algae analysis, and discloses a method for automatic analysis of whole algae based on fluorescent quantum dots, comprising the following steps: step S101, obtaining hyperspectral data and water quality parameters of an experimental water sample; step S102, preprocessing the hyperspectral data and water quality parameters of the experimental water sample through a sliding window to generate a feature sequence; step S103, obtaining a fluorescent signal of the experimental water sample through a fluorescent quantum detection method, and training a generation model based on the feature sequence and the fluorescent signal; step S104, obtaining a measured value of algae concentration of the experimental water sample, and training a prediction model based on the fluorescent signal and the measured value of the algae concentration; the invention utilizes hyperspectral data and water quality parameters to generate a fluorescent signal approximating that obtained through the fluorescent quantum detection method through a generation model, and establishes a nonlinear mapping relationship between the fluorescent signal and the algae concentration through a prediction model, thereby improving the accuracy of measuring the algae concentration.

Description

Technical Field

[0001] The invention relates to the technical field of algae analysis, and more specifically, to a whole algae automatic analysis method based on fluorescent quantum dots. Background Art

[0002] Existing algae analysis methods mainly include: 1. Microscope counting method: directly observe the morphological and structural characteristics of algae through a microscope, and identify and count the species; 2. Spectrophotometry: based on the absorption characteristics of chlorophyll a to light of a specific wavelength, chlorophyll a has strong absorption peaks in the blue and red spectral regions and weak absorption in the green spectral region. The absorbance at a specific wavelength (usually 665 nanometers) is measured by a spectrophotometer or a portable spectrophotometer to calculate the concentration of chlorophyll a; 3. Fluorescence method: chlorophyll a will emit fluorescence under the irradiation of excitation light of a specific wavelength, and the fluorescence intensity is proportional to the concentration of chlorophyll a. The fluorescence intensity of the water sample is measured by a fluorescence photometer or a portable fluorometer, and the concentration of chlorophyll a is calculated based on the relationship between the fluorescence intensity and the concentration of chlorophyll a; 4. Fluorescence quantum detection method: fluorescent quantum dots are used as fluorescent probes to mark specific components in algae cells, and the fluorescence signal is detected by appropriate equipment (such as a fluorescence microscope, flow cytometer, etc.), and the algae concentration is calculated based on the intensity of the fluorescence signal.

[0003] However, both the microscope counting method and the fluorescence quantum detection method need to be carried out in an experimental environment, and there is a serious time lag, and they cannot measure the algae concentration in real time. Although the spectrophotometry and fluorescence methods can measure the algae concentration in real time, they are easily affected by the fluorescence produced by other substances in the water sample, resulting in low measurement accuracy. Summary of the invention

[0004] The present invention provides a whole algae automatic analysis method based on fluorescent quantum dots, which solves the technical problems in the above-mentioned background technology.

[0005] The present invention provides a method for automatic analysis of whole algae based on fluorescent quantum dots, comprising the following steps:

[0006] Step S101, obtaining hyperspectral data and water quality parameters of experimental water samples;

[0007] Hyperspectral data are represented by spectral reflectance from 350 nm to 850 nm;

[0008] Water quality parameters included: chlorophyll a concentration, total phosphorus content, suspended matter concentration, and turbidity;

[0009] Step S102, preprocessing the hyperspectral data and water quality parameters of the experimental water sample through a sliding window to generate a feature sequence;

[0010] The feature sequence includes K sequence units, each of which corresponds to a combination vector of a sliding window;

[0011] Step S103, obtaining the fluorescence signal of the experimental water sample by a fluorescence quantum detection method, and training and generating a model based on the characteristic sequence and the fluorescence signal;

[0012] The input of the generative model is the feature sequence, and the output is the fluorescence signal;

[0013] The fluorescence signal is represented by the fluorescence intensity at a wavelength of M nanometers to N nanometers;

[0014] Step S104, obtaining the measured value of algae concentration of the experimental water sample, and training a prediction model based on the fluorescence signal and the measured value of algae concentration;

[0015] The input of the prediction model is the fluorescence signal, and the output is the predicted value of algae concentration;

[0016] Step S105, generating a feature sequence of the water sample to be detected according to step S101 and step S102 and inputting it into the trained generation model, outputting a fluorescence signal, and inputting the fluorescence signal into the trained prediction model, and outputting a predicted value of the algae concentration of the water sample to be detected.

[0017] Furthermore, M and N are both custom parameters, and K is determined by the size and step size of the sliding window, and its calculation formula is as follows:

[0018] ;

[0019] Where length represents the wavelength length of the hyperspectral data, which is 850 nanometers minus 350 nanometers, equal to 500 nanometers. Size and step represent the size and step length of the sliding window, respectively. Both the size and step length of the sliding window are custom parameters.

[0020] Furthermore, the absorbance values ​​of the experimental water samples at wavelengths of 663 nanometers and 635 nanometers were measured by a portable spectrophotometer, and the chlorophyll a concentration of the experimental water samples was obtained by the chlorophyll a concentration calculation formula. The chlorophyll a concentration calculation formula is as follows:

[0021] ;

[0022] Where Chla represents the chlorophyll a concentration, Indicates the absorbance value at a wavelength of 663 nanometers. It represents the absorbance value at a wavelength of 635 nanometers, Vol represents volume, and Weight represents weight.

[0023] Furthermore, the hyperspectral data and water quality parameters of the experimental water samples are preprocessed through a sliding window to generate a feature sequence, including the following steps:

[0024] Step S201, extracting characteristic values ​​of hyperspectral data within the sliding window;

[0025] The characteristic values ​​of hyperspectral data include: the maximum value, minimum value, mean value, standard deviation, median, skewness and kurtosis of the spectral reflectance of the hyperspectral data;

[0026] The calculation formula for the skewness of the spectral reflectance of hyperspectral data is as follows:

[0027] ;

[0028] The calculation formula of Kurtosis of spectral reflectance of hyperspectral data is as follows:

[0029] ;

[0030] Where size represents the size of the sliding window. represents the spectral reflectance corresponding to the i-th wavelength of the hyperspectral data in the sliding window, and They represent the mean and standard deviation of the spectral reflectance of the hyperspectral data within the sliding window, respectively;

[0031] Step S202, normalizing the characteristic values ​​of the hyperspectral data and the water quality parameters in the sliding window by using the minimum-maximum normalization method;

[0032] Step S203, concatenating the normalized eigenvalues ​​of the hyperspectral data in the sliding window and the normalized water quality parameters to obtain a combined vector.

[0033] Furthermore, the generative model includes K hidden layers, the kth hidden layer inputs the kth sequence unit of the feature sequence, and outputs an update vector, where 1≤k≤K;

[0034] The update vector output by the Kth hidden layer is input to Q classifiers, and the classification spaces of the Q classifiers respectively represent the fluorescence intensities corresponding to the Q wavelengths of the fluorescence signal, where Q=MN.

[0035] Furthermore, the calculation formula of the generated model includes:

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] ;

[0041] in and denote the update vectors of the kth and k-1th hidden layer outputs, respectively. Each dimension of is assigned a value of 0. The size and The size of is the same, when k-1=0, then Each dimension of is 0. represents the kth sequence unit of the feature sequence input to the kth hidden layer, , , and They represent the correlation coefficient, hidden vector, reset gate and update gate of the kth hidden layer respectively. , , , , , and denote the first weight parameter, the second weight parameter, the third weight parameter, the fourth weight parameter, the fifth weight parameter, the sixth weight parameter and the seventh weight parameter of the k-th hidden layer, respectively. , , and represent the first bias parameter, the second bias parameter, the third bias parameter and the fourth bias parameter of the kth hidden layer respectively, Represents point-by-point multiplication, T represents transpose operation, concat represents concatenation operation, pile represents stacking operation, sigmoid represents sigmoid activation function, tanh represents hyperbolic tangent activation function, and ReLU represents ReLU activation function.

[0042] Further, the feature sequence generated by repeating step S101 to step S102 is used as training data of training samples for training the generation model, and the fluorescence signal obtained by the fluorescence quantum detection method is used as the sample label for training the generation model. A discriminator is connected during the training process of the generation model, and the value output by the discriminator represents the probability value of the fluorescence signal output by the generation model belonging to the sample label of the training sample;

[0043] Loss function of the discriminator The calculation formula is as follows:

[0044] ;

[0045] in represents the training data of the u-th training sample, represents the sample label of the u-th training sample, Indicates that the sample label of the u-th training sample is input to the discriminator, and outputs the probability value of the sample label of the u-th training sample belonging to the sample label of the u-th training sample, represents the fluorescence signal output by the training data input of the u-th training sample to the generated model, Represents the probability value of the fluorescence signal output by the generation model from the training data input of the u-th training sample belonging to the sample label of the u-th training sample.

[0046] Furthermore, the fluorescence signal is divided into P bands at equal wavelength intervals, and the average value of the fluorescence intensity corresponding to all wavelengths in each band is calculated as the input of the prediction model, where P is a custom parameter. The calculation formula of the prediction model is as follows:

[0047] ;

[0048] Where P represents the number of bands, Con represents the predicted value of algae concentration output by the prediction model, represents the average value of the fluorescence intensity corresponding to all wavelengths in the pth band, It represents the p-1 power of the average value of the fluorescence intensity corresponding to all wavelengths in the p-th band. Represents the weight parameter corresponding to the p-th band.

[0049] Furthermore, the prediction model was obtained by performing nonlinear fitting on the fluorescence signal and the measured values ​​of algae concentration using the least squares method.

[0050] The present invention provides a whole algae automatic analysis system based on fluorescent quantum dots, comprising:

[0051] A data acquisition module, which is used to obtain hyperspectral data and water quality parameters of experimental water samples;

[0052] A feature sequence generation module is used to pre-process the hyperspectral data and water quality parameters of the experimental water samples through a sliding window to generate a feature sequence;

[0053] A generation model building module, which is used to obtain the fluorescence signal of the experimental water sample by fluorescence quantum detection method, and train the generation model based on the feature sequence and the fluorescence signal;

[0054] The prediction model building module is used to obtain the measured value of the algae concentration of the experimental water sample and train the prediction model based on the fluorescence signal and the measured value of the algae concentration.

[0055] The beneficial effects of the present invention are as follows: the present invention utilizes hyperspectral data and water quality parameters to generate a fluorescence signal approximating that obtained by the fluorescence quantum detection method through a generation model, and establishes a nonlinear mapping relationship between the fluorescence signal and the algae concentration through a prediction model, thereby improving the accuracy of measuring the algae concentration, and the hyperspectral data and water quality parameters can be collected in real time, thereby realizing real-time measurement of the algae concentration. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flow chart of a whole algae automatic analysis method based on fluorescent quantum dots of the present invention;

[0057] Figure 2 It is a flow chart of the present invention for preprocessing the hyperspectral data and water quality parameters of the experimental water sample through a sliding window to generate a feature sequence;

[0058] Figure 3 It is a schematic diagram of a whole algae automatic analysis system based on fluorescent quantum dots of the present invention.

[0059] In the figure: data collection module 301, feature sequence generation module 302, generation model construction module 303, prediction model construction module 304. DETAILED DESCRIPTION

[0060] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0061] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0062] like Figure 1 to Figure 3 As shown, a whole algae automatic analysis method based on fluorescent quantum dots comprises the following steps:

[0063] Step S101, obtaining hyperspectral data and water quality parameters of experimental water samples;

[0064] Hyperspectral data are represented by spectral reflectance from 350 nm to 850 nm;

[0065] Water quality parameters included: chlorophyll a concentration, total phosphorus content, suspended matter concentration, and turbidity;

[0066] Step S102, preprocessing the hyperspectral data and water quality parameters of the experimental water sample through a sliding window to generate a feature sequence;

[0067] The feature sequence includes K sequence units, each of which corresponds to a combination vector of a sliding window;

[0068] Step S103, obtaining the fluorescence signal of the experimental water sample by a fluorescence quantum detection method, and training and generating a model based on the characteristic sequence and the fluorescence signal;

[0069] The input of the generative model is the feature sequence, and the output is the fluorescence signal;

[0070] The fluorescence signal is represented by the fluorescence intensity at a wavelength of M nanometers to N nanometers;

[0071] Step S104, obtaining the measured value of algae concentration of the experimental water sample, and training a prediction model based on the fluorescence signal and the measured value of algae concentration;

[0072] The input of the prediction model is the fluorescence signal, and the output is the predicted value of algae concentration;

[0073] Step S105, generating a feature sequence of the water sample to be detected according to step S101 and step S102 and inputting it into the trained generation model, outputting a fluorescence signal, and inputting the fluorescence signal into the trained prediction model, and outputting a predicted value of the algae concentration of the water sample to be detected.

[0074] It should be noted that chlorophyll a concentration has a strong absorption peak between wavelengths of 650 nanometers and 750 nanometers, total phosphorus content has a strong absorption peak between wavelengths of 380 nanometers and 780 nanometers, suspended matter concentration has a strong absorption peak between wavelengths of 400 nanometers and 700 nanometers, and turbidity has a strong absorption peak between wavelengths of 400 nanometers and 700 nanometers. Therefore, hyperspectral data is represented by the spectral reflectance at a wavelength of 350 nanometers to 850 nanometers, which can indirectly reflect the water quality characteristics of the water sample and can reduce the amount of data calculation to a certain extent, thereby improving the calculation speed of the generation model and the prediction model.

[0075] In one embodiment of the present invention, hyperspectral data is obtained by a handheld or fixed hyperspectral instrument.

[0076] In one embodiment of the present invention, M and N are both custom parameters. Preferably, M is set to 500, N is set to 750, and K is determined by the size and step size of the sliding window. The calculation formula is as follows:

[0077] ;

[0078] Wherein, length represents the wavelength length of the hyperspectral data, and the wavelength length of the hyperspectral data is 850 nanometers minus 350 nanometers, which equals 500 nanometers. Size and step represent the size and step length of the sliding window, respectively, and the size and step length of the sliding window are both custom parameters. Preferably, the size of the sliding window is set to 25, and the step length of the sliding window is set to 25, then K is 20.

[0079] It should be noted that M and N are set according to the emission wavelengths of different substances in the algae in the fluorescence signal. For example, the emission wavelength of chlorophyll a in the algae in the fluorescence signal is about 650 to 670 nanometers, the emission wavelength of carotene in the algae in the fluorescence signal is about 500 to 600 nanometers, the emission wavelength of lutein in the algae in the fluorescence signal is about 500 to 600 nanometers, the emission wavelength of phycoerythrin in the algae in the fluorescence signal is about 560 to 570 nanometers, and the emission wavelength of phycocyanin in the algae in the fluorescence signal is about 630 to 650 nanometers.

[0080] In one embodiment of the present invention, the absorbance value of the experimental water sample at wavelengths of 663 nanometers and 635 nanometers is measured by a portable spectrophotometer, and the chlorophyll a concentration of the experimental water sample is obtained by a chlorophyll a concentration calculation formula;

[0081] The calculation formula of chlorophyll a concentration is as follows:

[0082] ;

[0083] Where Chla represents the chlorophyll a concentration, Indicates the absorbance value at a wavelength of 663 nanometers. It represents the absorbance value at a wavelength of 635 nanometers, Vol represents volume, and Weight represents weight.

[0084] In one embodiment of the present invention, Figure 2 As shown in the figure, the hyperspectral data and water quality parameters of the experimental water samples are preprocessed through a sliding window to generate a feature sequence, including the following steps:

[0085] Step S201, extracting characteristic values ​​of hyperspectral data within the sliding window;

[0086] The characteristic values ​​of hyperspectral data include: the maximum value, minimum value, mean value, standard deviation, median, skewness and kurtosis of the spectral reflectance of the hyperspectral data;

[0087] The calculation formula of the skewness of the spectral reflectance of hyperspectral data is as follows:

[0088] ;

[0089] The calculation formula of Kurtosis of spectral reflectance of hyperspectral data is as follows:

[0090] ;

[0091] Where size represents the size of the sliding window. represents the spectral reflectance corresponding to the i-th wavelength of the hyperspectral data in the sliding window, and They represent the mean and standard deviation of the spectral reflectance of the hyperspectral data within the sliding window, respectively;

[0092] Step S202, normalizing the characteristic values ​​of the hyperspectral data and the water quality parameters in the sliding window by using the minimum-maximum normalization method;

[0093] Step S203, concatenating the normalized eigenvalues ​​of the hyperspectral data in the sliding window and the normalized water quality parameters to obtain a combined vector.

[0094] It should be noted that the combination vector of a sliding window is expressed as:

[0095] ;

[0096] in They represent the characteristic values ​​of the hyperspectral data in the normalized sliding window, namely the maximum, minimum, mean, standard deviation, median, skewness and kurtosis of the spectral reflectance of the hyperspectral data. They represent the normalized water quality parameters, namely chlorophyll a concentration, total phosphorus content, suspended matter concentration and turbidity.

[0097] In one embodiment of the present invention, the generative model includes K hidden layers, the kth hidden layer inputs the kth sequence unit of the feature sequence, and outputs an update vector, where 1≤k≤K;

[0098] The update vector output by the Kth hidden layer is input to Q classifiers, and the classification spaces of the Q classifiers respectively represent the fluorescence intensities corresponding to the Q wavelengths of the fluorescence signal, where Q=MN.

[0099] It should be noted that the wavelength interval between the Q wavelengths of the fluorescence signal can be customized. For example, if M is set to 500 and N is set to 750, then Q is 750-500=250. In this case, the number of classifiers is too large, which increases the computational complexity of the generated model and is not conducive to model convergence. Then the wavelength interval can be set to 5 nanometers, then Q is (750-500) / 5=50, or the wavelength interval can be set to 10 nanometers, then Q is (750-500) / 10=25.

[0100] In one embodiment of the present invention, the calculation formula for generating the model includes:

[0101] ;

[0102] ;

[0103] ;

[0104] ;

[0105] ;

[0106] in and denote the update vectors of the kth and k-1th hidden layer outputs, respectively. Each dimension of is assigned a value of 0. The size and The size of is the same, when k-1=0, then Each dimension of is 0. represents the kth sequence unit of the feature sequence input to the kth hidden layer, , , and They represent the correlation coefficient, hidden vector, reset gate and update gate of the kth hidden layer respectively. , , , , , and denote the first weight parameter, the second weight parameter, the third weight parameter, the fourth weight parameter, the fifth weight parameter, the sixth weight parameter and the seventh weight parameter of the k-th hidden layer, respectively. , , and represent the first bias parameter, the second bias parameter, the third bias parameter and the fourth bias parameter of the kth hidden layer respectively, Represents point-by-point multiplication, T represents transpose operation, concat represents concatenation operation, pile represents stacking operation, sigmoid represents sigmoid activation function, tanh represents hyperbolic tangent activation function, and ReLU represents ReLU activation function.

[0107] It should be noted that the weight parameters and bias parameters of the generative model are all learnable parameters. For example, the kth sequence unit of the feature sequence input to the kth hidden layer, that is, the combination vector corresponding to the kth sequence unit, has a size of 1×11. The size of the update vector output by the kth and k-1th hidden layers can be designed to be 1×16. Then, the vector size obtained by concatenating the combination vector and the update vector is 1×27. The sixth and seventh weight parameters of the kth hidden layer can be designed to be a vector of size 27×1. The reset gate and update gate corresponding to the multiplication of the two are both real numbers.

[0108] It should be noted that The vector size is The vector sizes must be the same, otherwise the two cannot be stacked. For example, if the two are stacked into a 2×64 matrix, the third weight parameter can be designed as a 64×1 matrix. The 2×64 matrix is ​​multiplied by the 64×1 matrix to obtain a 2×1 matrix, and then the transposition operation is performed to obtain a 1×2 matrix. The fourth weight parameter can be designed as a 2×1 matrix. The correlation coefficient obtained by multiplying the two is a real number.

[0109] In one embodiment of the present invention, the feature sequence generated by repeating step S101 to step S102 is used as training data of a training sample for training a generation model, and the fluorescent signal obtained by the fluorescent quantum detection method is used as a sample label for training the generation model. A discriminator is connected during the training process of the generation model, and the value output by the discriminator represents the probability value of the fluorescent signal output by the generation model belonging to the sample label of the training sample;

[0110] Loss function of the discriminator The calculation formula is as follows:

[0111] ;

[0112] in represents the training data of the u-th training sample, represents the sample label of the u-th training sample, Indicates that the sample label of the u-th training sample is input to the discriminator, and outputs the probability value of the sample label of the u-th training sample belonging to the sample label of the u-th training sample, represents the fluorescence signal output by the training data input of the u-th training sample to the generated model, Represents the probability value of the fluorescence signal output by the generation model from the training data input of the u-th training sample belonging to the sample label of the u-th training sample.

[0113] In one embodiment of the present invention, the fluorescence signal is divided into P bands at equal wavelength intervals, and the average value of the fluorescence intensity corresponding to all wavelengths in each band is calculated as the input of the prediction model, where P is a custom parameter. Preferably, P is set to 10. The calculation formula of the prediction model is as follows:

[0114] ;

[0115] Where P represents the number of bands, Con represents the predicted value of algae concentration output by the prediction model, represents the average value of the fluorescence intensity corresponding to all wavelengths in the pth band, It represents the p-1 power of the average value of the fluorescence intensity corresponding to all wavelengths in the p-th band. Represents the weight parameter corresponding to the p-th band.

[0116] In one embodiment of the present invention, the prediction model is obtained by performing nonlinear fitting on the fluorescence signal and the measured value of the algae concentration using the least square method.

[0117] In one embodiment of the present invention, Figure 3 As shown, the present invention provides a whole algae automatic analysis system based on fluorescent quantum dots, comprising:

[0118] A data acquisition module 301 is used to obtain the hyperspectral data and water quality parameters of the experimental water samples;

[0119] A feature sequence generation module 302 is used to pre-process the hyperspectral data and water quality parameters of the experimental water sample through a sliding window to generate a feature sequence;

[0120] A generation model building module 303 is used to obtain the fluorescence signal of the experimental water sample by a fluorescence quantum detection method, and to train a generation model based on the characteristic sequence and the fluorescence signal;

[0121] The prediction model building module 304 is used to obtain the measured value of the algae concentration of the experimental water sample and train the prediction model based on the fluorescence signal and the measured value of the algae concentration.

[0122] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are protected by the present embodiment.

Claims

1. A whole algae automatic analysis method based on fluorescent quantum dots, characterized in that: The following steps are involved: Step S101, obtaining hyperspectral data and water quality parameters of experimental water samples; Hyperspectral data are represented by spectral reflectance from 350 nm to 850 nm; Water quality parameters included: chlorophyll a concentration, total phosphorus content, suspended matter concentration, and turbidity; Step S102, preprocessing the hyperspectral data and water quality parameters of the experimental water sample through a sliding window to generate a feature sequence; The feature sequence includes K sequence units, each of which corresponds to a combination vector of a sliding window; Step S103, obtaining the fluorescence signal of the experimental water sample by a fluorescence quantum detection method, and training and generating a model based on the characteristic sequence and the fluorescence signal; The input of the generative model is the feature sequence, and the output is the fluorescence signal; The fluorescence signal is represented by the fluorescence intensity at a wavelength of M nanometers to N nanometers; Step S104, obtaining the measured value of algae concentration of the experimental water sample, and training a prediction model based on the fluorescence signal and the measured value of algae concentration; The input of the prediction model is the fluorescence signal, and the output is the predicted value of algae concentration; Step S105, generating a feature sequence of the water sample to be detected according to step S101 and step S102, inputting the feature sequence into the trained generation model, outputting a fluorescence signal, and inputting the fluorescence signal into the trained prediction model, outputting a predicted value of the algae concentration of the water sample to be detected; M and N are both custom parameters. K is determined by the size and step size of the sliding window. The calculation formula is as follows: ； Where length represents the wavelength length of the hyperspectral data, which is 850 nanometers minus 350 nanometers, equal to 500 nanometers. Size and step represent the size and step length of the sliding window, respectively. Both the size and step length of the sliding window are custom parameters. The hyperspectral data and water quality parameters of the experimental water samples are preprocessed through a sliding window to generate a feature sequence, including the following steps: Step S201, extracting characteristic values ​​of hyperspectral data within the sliding window; The characteristic values ​​of hyperspectral data include: the maximum value, minimum value, mean value, standard deviation, median, skewness and kurtosis of the spectral reflectance of the hyperspectral data; The calculation formula of the skewness of the spectral reflectance of hyperspectral data is as follows: ； The calculation formula of Kurtosis of spectral reflectance of hyperspectral data is as follows: ； Where size represents the size of the sliding window. represents the spectral reflectance corresponding to the i-th wavelength of the hyperspectral data in the sliding window, and They represent the mean and standard deviation of the spectral reflectance of the hyperspectral data within the sliding window, respectively; Step S202, normalizing the characteristic values ​​of the hyperspectral data and the water quality parameters in the sliding window by using the minimum-maximum normalization method; Step S203, concatenating the normalized eigenvalues ​​of the hyperspectral data in the sliding window and the normalized water quality parameters to obtain a combined vector; The generative model includes K hidden layers, the kth hidden layer inputs the kth sequence unit of the feature sequence and outputs an update vector, where 1≤k≤K; The update vector output by the Kth hidden layer is input to Q classifiers, and the classification space of the Q classifiers represents the fluorescence intensity corresponding to the Q wavelengths of the fluorescence signal, where Q=MN; The calculation formula for the generated model includes: ； ； ； ； ； in and denote the update vectors of the kth and k-1th hidden layer outputs, respectively. Each dimension of is assigned a value of 0. The size and The size of is the same, when k-1=0, then Each dimension of is 0. represents the kth sequence unit of the feature sequence input to the kth hidden layer, , , and They represent the correlation coefficient, hidden vector, reset gate and update gate of the kth hidden layer respectively. , , , , , and denote the first weight parameter, the second weight parameter, the third weight parameter, the fourth weight parameter, the fifth weight parameter, the sixth weight parameter and the seventh weight parameter of the k-th hidden layer, respectively. , , and represent the first bias parameter, the second bias parameter, the third bias parameter and the fourth bias parameter of the kth hidden layer respectively, represents point-by-point multiplication, T represents transpose operation, concat represents concatenation operation, pile represents stacking operation, sigmoid represents sigmoid activation function, tanh represents hyperbolic tangent activation function, and ReLU represents ReLU activation function; Repeating step S101 to step S102 to generate the feature sequence as training data of the training sample for training the generation model, the fluorescence signal obtained by the fluorescence quantum detection method is used as the sample label for training the generation model, and a discriminator is connected during the training process of the generation model, and the value output by the discriminator represents the probability value of the fluorescence signal output by the generation model belonging to the sample label of the training sample; Loss function of the discriminator The calculation formula is as follows: ； in represents the training data of the u-th training sample, represents the sample label of the u-th training sample, Indicates that the sample label of the u-th training sample is input to the discriminator, and outputs the probability value of the sample label of the u-th training sample belonging to the sample label of the u-th training sample, represents the fluorescence signal output by the training data input of the u-th training sample to the generated model, Represents the probability value of the fluorescence signal output by the generation model from the training data of the u-th training sample belonging to the sample label of the u-th training sample; The fluorescence signal is divided into P bands at equal wavelength intervals, and the average fluorescence intensity corresponding to all wavelengths in each band is calculated as the input of the prediction model, where P is a custom parameter. The calculation formula of the prediction model is as follows: ； Where P represents the number of bands, Con represents the predicted value of algae concentration output by the prediction model, represents the average value of the fluorescence intensity corresponding to all wavelengths in the pth band, It represents the p-1 power of the average value of the fluorescence intensity corresponding to all wavelengths in the p-th band. Represents the weight parameter corresponding to the p-th band.

2. The whole algae automatic analysis method based on fluorescent quantum dots according to claim 1 is characterized in that: The absorbance values ​​of the experimental water samples at wavelengths of 663 nanometers and 635 nanometers were measured by a portable spectrophotometer, and the chlorophyll a concentration of the experimental water samples was obtained by the chlorophyll a concentration calculation formula. The chlorophyll a concentration calculation formula is as follows: ； Where Chla represents the chlorophyll a concentration, Indicates the absorbance value at a wavelength of 663 nanometers. It represents the absorbance value at a wavelength of 635 nanometers, Vol represents volume, and Weight represents weight.

3. The whole algae automatic analysis method based on fluorescent quantum dots according to claim 1 is characterized in that: The prediction model was obtained by performing nonlinear fitting on the fluorescence signal and the measured values ​​of algae concentration using the least squares method.

4. A whole algae automatic analysis system based on fluorescent quantum dots, characterized in that: The method for automatic analysis of whole algae based on fluorescent quantum dots according to any one of claims 1 to 3 comprises: A data acquisition module, which is used to obtain hyperspectral data and water quality parameters of experimental water samples; A feature sequence generation module is used to pre-process the hyperspectral data and water quality parameters of the experimental water samples through a sliding window to generate a feature sequence; A generation model building module, which is used to obtain the fluorescence signal of the experimental water sample by fluorescence quantum detection method, and train the generation model based on the feature sequence and the fluorescence signal; The prediction model building module is used to obtain the measured value of the algae concentration of the experimental water sample and train the prediction model based on the fluorescence signal and the measured value of the algae concentration.

Images