Method and system for detecting carbon nitrogen ratio of sweet potato based on multispectrum and fluorescence sensor

Abstract

The invention relates to the technical field of carbon-nitrogen ratio detection, in particular to a sweet potato carbon-nitrogen ratio detection method and system based on a multispectrum and fluorescence sensor, and the method specifically comprises the following steps: constructing a weight coefficient based on the correlation between the leaf carbon-nitrogen ratio in each growth stage and the reflectivity of each wavelength, and combining the multispectrum and fluorescence spectrum of each leaf to obtain a weight coefficient; acquiring a coupling characteristic value; correcting the coupling characteristic value according to the soil conductivity; analyzing the multi-spectral characteristics and the fluorescence spectrum characteristics of the sweet potatoes, and constructing a multi-spectral index and a fluorescence spectrum index; predicting the carbon-nitrogen ratio of the to-be-detected sweet potato leaf based on the features of the leaf, and calculating the carbon-nitrogen ratio of the to-be-detected sweet potato leaf by combining the difference between the feature vectors of the to-be-detected sweet potato leaf and the leaves in all growth stages and introducing the carbon-nitrogen ratio of the previous growth stage; certain continuity and inertia of the change of the carbon-nitrogen ratio are considered, so that the method adapts to the change of the sweet potatoes in different growth stages, and the detection precision of the carbon-nitrogen ratio is improved.

Description

Technical Field

[0001] This application relates to the technical field of carbon-nitrogen ratio detection, and specifically relates to a detection method and system for the carbon-nitrogen ratio of sweet potatoes based on multispectral and fluorescence sensors. Background Art

[0002] As an important food and cash crop, sweet potatoes are different from other crops in that the harvested organ of sweet potatoes is the underground tuberous root, and the carbon-nitrogen ratio directly affects its growth and quality. A too high carbon-nitrogen ratio may lead to excessive growth and affect tuberous root development, while a too low ratio may affect photosynthesis and disease resistance. Detecting the carbon-nitrogen ratio can help optimize fertilization strategies and improve yield and quality. Traditional laboratory detection has the defects of destructiveness and lag, while carbon-nitrogen ratio detection based on spectral technology can achieve in-situ field monitoring and provide a basis for precise regulation during the growth cycle.

[0003] In the related technology "CN103048276A A method for constructing a spectral index for detecting the carbon-nitrogen ratio of crop canopy leaves", relative reflectance is calculated through hyperspectral data, and the spectral slope is calculated in three sub-bands to construct a spectral index, which can eliminate the influence of light differences and reduce the correlation between adjacent wavelengths. However, this method only relies on single spectral data, and since the carbon-nitrogen ratio requirements of sweet potatoes are different at different growth stages. For example, at the seedling stage, the demand for nitrogen is relatively high, which helps to promote the growth and development of plants; at the tuberous root swelling stage and the mature stage, the demand for carbon in sweet potatoes gradually increases, which helps the swelling of tuberous roots and nutrient accumulation. And this method uses a fixed spectral index, without considering the spectral response differences at different growth stages of sweet potatoes and the continuity of carbon-nitrogen ratio changes, and cannot adapt to the spectral response differences at different growth stages of sweet potatoes, resulting in a decrease in detection accuracy. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a detection method and system for the carbon-nitrogen ratio of sweet potatoes based on multispectral and fluorescence sensors, and the specific technical solutions adopted are as follows: In the first aspect, an embodiment of this application provides a detection method for the carbon-nitrogen ratio of sweet potatoes based on multispectral and fluorescence sensors, and this method includes the following steps: At each growth stage, collect the multispectral, fluorescence spectrum, carbon-nitrogen ratio and environmental data of each leaf of the sweet potato; obtain the fluorescence lifetime corresponding to each wavelength in the fluorescence spectrum of each leaf; the environmental data includes soil conductivity; Based on the correlation between the carbon-nitrogen ratio of the leaves and the reflectance of each wavelength at each growth stage, construct the weight coefficient of each wavelength at each growth stage; based on the weight coefficient, and the multispectral and fluorescence spectra of each leaf, construct the coupling eigenvalue of the spectral data of each leaf; Correct the coupling eigenvalue of the spectral data of each leaf according to the soil conductivity of each leaf; Feature combinations are performed based on the reflectance of characteristic wavelengths in the multispectral data of each leaf to construct the multispectral index of each leaf; feature combinations are performed based on the fluorescence intensity and corresponding fluorescence lifetime of characteristic wavelengths in the fluorescence spectrum of each leaf to construct the fluorescence spectrum index of each leaf. Based on the corrected value of the coupled eigenvalue of the spectral data of each leaf, the multispectral index, the fluorescence spectrum index, and the environmental data, the feature vector of each leaf is constructed; according to the feature vectors and carbon-nitrogen ratios of all leaf samples at each growth stage, a prediction model is constructed, and in combination with the feature vector of the sweet potato leaf to be detected, the predicted value of the carbon-nitrogen ratio of the sweet potato leaf to be detected is obtained. Based on the difference between the feature vector of the sweet potato leaf to be detected and the feature vectors of all leaves at all growth stages, and the carbon-nitrogen ratio of the leaves in the previous growth stage of the growth stage where the sweet potato leaf to be detected is located, in combination with the predicted value, the carbon-nitrogen ratio of the sweet potato leaf to be detected is constructed.

[0005] In one embodiment, the process of obtaining the weight coefficient of each wavelength in each growth stage is as follows: The reflectances of the i-th wavelength in the spectral data of all leaves at each growth stage are combined into a set, denoted as the i-th wavelength reflectance set of each growth stage; the mutual information between the elements in the i-th wavelength reflectance set and the corresponding leaf carbon-nitrogen ratio is calculated, denoted as the first mutual information value. The sum value of the first mutual information values of all wavelengths in each growth stage is calculated, denoted as the first sum value; the ratio of the first mutual information value of the i-th wavelength in each growth stage to the first sum value is used as the weight coefficient of the i-th wavelength in each growth stage.

[0006] In one embodiment, the expression of the coupled eigenvalue of the spectral data of each leaf is: , where represents the coupled eigenvalue of the spectral data of the current leaf; n represents the number of wavelengths in any spectral data; is the weight coefficient of the i-th wavelength in the growth stage where the current leaf is located; represents the reflectance of the i-th wavelength in the multispectral data of the current leaf; represents the fluorescence lifetime corresponding to the i-th wavelength in the fluorescence spectrum data of the current leaf; represents the preset reference fluorescence lifetime; represents the exponential function with the natural constant e as the base.

[0007] In one embodiment, the correction of the coupled eigenvalue of the spectral data of each leaf is expressed as: where is the correction value of the spectral data coupling eigenvalue of the current leaf, represents the coupling eigenvalue of the spectral data of the current leaf, is the compensation coefficient corresponding to the soil conductivity of the current leaf, is the soil conductivity corresponding to the current leaf, and A and B respectively represent the preset intercept and slope, is the tanh function, C represents the reference value of the soil conductivity, and E represents the preset response coefficient.

[0008] In one embodiment, the expression of the multispectral index of each leaf is: wherein, is the multispectral index of the current leaf, , , , respectively represent the reflectance at wavelengths 820nm, 680nm, 950nm, and 760nm in the multispectral data of the current leaf, is the exponential function with the natural constant e as the base; among them, the wavelength 680nm is the wavelength corresponding to the strong absorption peak of chlorophyll a, the wavelength 820nm is the wavelength corresponding to the near-infrared plateau region, the wavelength 760nm is the wavelength corresponding to the photosystem II sensitive region, and the wavelength 950nm is the wavelength corresponding to the water absorption valley.

[0009] In one embodiment, the expression of the fluorescence spectral index of each leaf is: wherein, F is the fluorescence spectral index of the current leaf, , are respectively the fluorescence intensities at wavelengths 685nm and 740nm in the fluorescence spectral data of the current leaf, , are respectively the fluorescence lifetimes corresponding to wavelengths 685nm and 740nm in the fluorescence spectral data of the current leaf; among them, the wavelength 685nm is the wavelength corresponding to the emission peak of chlorophyll a, and the wavelength 740nm is the wavelength corresponding to the far-red light emission peak.

[0010] In one embodiment, the process of obtaining the eigenvector of each leaf is: Calculate the sum of the normalized values of temperature, humidity, and light intensity in the environmental data of each leaf as the environmental index of each leaf; Take the vector composed of the correction value of the spectral data coupling eigenvalue, the multispectral index, the fluorescence spectral index, and the environmental index of each leaf as the eigenvector of each leaf.

[0011] In one embodiment, the process of obtaining the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be detected is as follows: Take the eigenvectors and carbon-nitrogen ratios of all leaf samples in each growth stage as the input of the regression prediction model, train the regression prediction model, and obtain the trained regression prediction model for each growth stage; Take the eigenvector of the sweet potato leaves to be detected as the input of the regression prediction model corresponding to its growth stage, and the output is the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be detected.

[0012] In one embodiment, the process of obtaining the carbon-nitrogen ratio of the sweet potato leaves to be detected is as follows: Randomly select the eigenvectors of a preset number of leaves in each growth stage, and use the set of eigenvectors of the preset number of leaves in all growth stages as the growth stage data set; Calculate the Euclidean distance between the eigenvector of the sweet potato leaves to be detected and each eigenvector in the growth stage data set, denoted as the first distance; take the normalized value of the sum of all the first distances of the sweet potato leaves to be detected as the dynamic growth stage weight of the sweet potato leaves to be detected; Denote the carbon-nitrogen ratio of the sweet potato leaves to be detected as , The expression of In the formula, is the carbon-nitrogen ratio of the sweet potato leaves to be detected, is the dynamic growth stage weight of the sweet potato leaves to be detected, represents the eigenvector of the sweet potato leaves to be detected, represents the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be detected, represents the mean value of the carbon-nitrogen ratios of all leaf samples in the previous growth stage of the growth stage where the sweet potato leaves to be detected are located, where represents the sweet potato leaves to be detected.

[0013] Second, the embodiment of the present application also provides a sweet potato carbon-nitrogen ratio detection system based on a multispectral and fluorescence sensor, which implements the method described in the first aspect. The system includes: Data fusion layer: In each growth stage, collect the multispectral, fluorescence spectrum, carbon-nitrogen ratio and environmental data of each sweet potato leaf; obtain the fluorescence lifetime corresponding to each wavelength in the fluorescence spectrum of each leaf; the environmental data includes soil conductivity; construct the weight coefficient of each wavelength based on the correlation between the carbon-nitrogen ratio of the leaf and the reflectance of each wavelength in each growth stage; based on the weight coefficient, and the multispectral and fluorescence spectra of each leaf, construct the coupling eigenvalue of the spectral data of each leaf; correct the coupling eigenvalue of the spectral data of each leaf according to the soil conductivity of each leaf; Feature engineering layer: Feature combination is performed based on the reflectance of characteristic wavelengths in the multispectral data of each leaf to construct the multispectral index of each leaf; feature combination is performed based on the fluorescence intensity and corresponding fluorescence lifetime of characteristic wavelengths in the fluorescence spectrum of each leaf to construct the fluorescence spectrum index of each leaf; based on the corrected value of the coupled eigenvalue of the spectral data of each leaf, the multispectral index, the fluorescence spectrum index, and the environmental data, the feature vector of each leaf is constructed. Temporal sequence modeling layer: A prediction model is constructed based on the feature vectors and carbon-nitrogen ratios of all leaf samples at each growth stage. Combining the feature vectors of the sweet potato leaves to be detected, the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be detected is obtained; based on the difference between the feature vectors of the sweet potato leaves to be detected and the feature vectors of all leaves at all growth stages, and the carbon-nitrogen ratio of the leaves in the previous growth stage of the growth stage where the sweet potato leaves to be detected are located, combining the predicted value, the carbon-nitrogen ratio of the sweet potato leaves to be detected is constructed.

[0014] The embodiments of the present application have at least the following beneficial effects: The present application constructs the weight coefficient of each wavelength in each growth stage based on the correlation between the carbon-nitrogen ratio of the leaves and the reflectance of each wavelength in each growth stage; based on the weight coefficient and the multispectral and fluorescence spectra of each leaf, the coupled eigenvalue of the spectral data of each leaf is obtained, avoiding the problem that single-modal data is difficult to comprehensively characterize the physiological state of crops, resulting in insufficient detection accuracy; the coupled eigenvalue is corrected according to the soil conductivity, so that the calculation result of the coupled eigenvalue is more in line with the actual growth characteristics of sweet potato leaves; by analyzing the multispectral characteristics and fluorescence spectral characteristics of sweet potatoes, the multispectral index and fluorescence spectrum index are constructed, which can more comprehensively characterize the physiological state of sweet potatoes and increase the detection accuracy; based on the above characteristics of the leaves, a regression prediction model for each growth stage is constructed, and the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be detected is obtained. Combining the difference between the feature vectors of the sweet potato leaves to be detected and the feature vectors of all leaves at all growth stages, and introducing the carbon-nitrogen ratio of the previous growth stage, the carbon-nitrogen ratio of the sweet potato leaves to be detected is calculated, considering that the change of the carbon-nitrogen ratio has a certain continuity and inertia, so as to adapt to the changes in different growth stages of sweet potatoes and improve the detection accuracy of the carbon-nitrogen ratio. Description of the drawings

[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0016] Figure 1 It is a step flow chart of a method for detecting the carbon-nitrogen ratio of sweet potatoes based on multispectral and fluorescence sensors provided by an embodiment of the present application; Figure 2 Schematic diagram of the acquisition process of the weight coefficient for each wavelength; Figure 3 Schematic diagram of the acquisition process of the weight in the dynamic growth stage. Detailed implementation manners

[0017] In order to further elaborate on the technical means and effects adopted by this application to achieve the intended invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features, and effects of the sweet potato carbon-nitrogen ratio detection method and system based on multi-spectral and fluorescence sensors proposed according to this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs.

[0019] The following specifically describes the specific solutions of the sweet potato carbon-nitrogen ratio detection method and system based on multi-spectral and fluorescence sensors provided by this application in combination with the accompanying drawings.

[0020] Please refer to Figure 1 , which shows a flowchart of the steps of the sweet potato carbon-nitrogen ratio detection method based on multi-spectral and fluorescence sensors provided by an embodiment of this application. The method includes the following steps: Step S1, in each growth stage, collect the multi-spectral, fluorescence spectrum, carbon-nitrogen ratio, and environmental data of each sweet potato leaf; obtain the fluorescence lifetime corresponding to each wavelength in the fluorescence spectrum of each leaf; the environmental data includes soil conductivity.

[0021] The growth stages of sweet potato include the seedling stage (0 - 30 days after planting), the branching and tuber formation stage (30 - 60 days), the vigorous growth stage of stems and leaves (60 - 90 days), the tuber swelling stage (90 - 120 days), and the mature harvest stage (120 - 150 days). In different growth stages of sweet potato, various spectral data of sweet potato leaves are collected. Specifically: In the seedling stage, the sweet potato begins to develop roots and form stems and leaves. In this stage, the spectral data of sweet potato leaves are collected every three days; in the branching and tuber formation stage, the tubers begin to swell, and the spectral data of sweet potato leaves are collected every two days; in the vigorous growth stage of stems and leaves, the photosynthesis is the strongest and the nutrient competition is fierce, so the spectral data of sweet potato leaves are collected every day; in the tuber swelling stage, the tuber volume grows rapidly, and the spectral data of sweet potato leaves are also collected every day; in the mature harvest stage, the starch accumulation is completed, and the spectral data of sweet potato leaves are collected once a week.

[0022] Among them, each time when collecting the spectral data of sweet potato leaves, N sweet potato leaves are randomly selected, the multispectral data of each sweet potato leaf is collected by a spectrometer, and the fluorescence spectral data of each sweet potato leaf is collected by a fluorescence sensor. In the embodiments of the present application, the value of N is set to 100. As other embodiments of the present application, the implementer can set the value of N according to the actual situation.

[0023] It should be noted that when collecting the two spectral data, ensure that the two spectral data are synchronously collected in the same leaf area; When using a spectrometer to collect multispectral data, in the embodiments of the present application, the spectral range is set to 500 - 1000 nm to cover the carbon-nitrogen sensitive band, and the field of view angle is set to , the resolution is set to 5 nm, and the values of each wavelength in the obtained multispectral data are the reflectance of the leaf to the light of each wavelength; When using a fluorescence sensor to collect fluorescence spectral data, in the embodiments of the present application, the LED blue light with a wavelength of 470 nm is used as the excitation light, and the fluorescence spectral data within the wavelength range of 500 - 1000 nm is collected. The values of each wavelength in the obtained fluorescence spectral data are the fluorescence emission intensities of each wavelength in the fluorescence excited by the leaf, and the fluorescence lifetime of each wavelength is obtained. Among them, the fluorescence lifetime refers to the time required for the fluorescence intensity of a fluorescent substance to decay to 1 / e of its initial value after the excitation light is removed, and its acquisition method is a well-known technology, and the specific process will not be elaborated.

[0024] As other embodiments of the present application, the implementer can set various parameters of the spectrometer and the fluorescence sensor according to the actual situation. Perform standard normal variate transformation (SNV) on the collected spectral data to eliminate the influence of leaf curvature.

[0025] Furthermore, after collecting the spectral data of each leaf, detect the carbon-nitrogen ratio of each leaf. In the present application, the carbon-nitrogen ratio of each leaf is detected by elemental analysis. Among them, elemental analysis is a well-known technology, and the specific process will not be elaborated.

[0026] It should be noted that for the detection of the carbon-nitrogen ratio of leaves, the present application only provides a method for detecting the carbon-nitrogen ratio. There are many existing methods for detecting the carbon-nitrogen ratio. The implementer can also use other methods for detecting the carbon-nitrogen ratio to detect the carbon-nitrogen ratio of leaves, and the present application does not make specific restrictions.

[0027] Furthermore, each time when collecting spectral data, also collect the growth environment data of sweet potatoes. Among them, collect the temperature and humidity of the air through a temperature and humidity sensor, collect the light intensity through a light intensity sensor, and collect the soil electrical conductivity (EC) through a four-electrode sensor. Normalize and then sum the values of temperature, humidity, and light intensity obtained each time when collecting spectral data, and the obtained result is used as the environmental index of all leaves for this time.

[0028] All the collected leaves above are used as samples for the training of the subsequent regression prediction model.

[0029] Traditional fluorescence intensity is susceptible to interference from environmental light, leaf thickness, etc., while fluorescence lifetime is more sensitive to the microenvironment, such as polarity and viscosity, and has a higher correlation with the carbon-nitrogen ratio. Changes in the carbon-nitrogen ratio directly affect the chlorophyll molecular structure and electron transfer efficiency. When nitrogen is deficient, chlorophyll synthesis is blocked, and the energy transfer efficiency between molecules decreases, resulting in an extended fluorescence lifetime. In the detection of the carbon-nitrogen ratio of sweet potatoes, by measuring the change in fluorescence lifetime, the dynamic change of the carbon-nitrogen ratio can be indirectly quantified.

[0030] Step S2: Construct the weight coefficient of each wavelength in each growth stage based on the correlation between the carbon-nitrogen ratio of the leaves and the reflectance of each wavelength in each growth stage; based on the weight coefficient and the multi-spectral and fluorescence spectra of each leaf, construct the coupled eigenvalue of the spectral data of each leaf.

[0031] Multi-spectral and fluorescence sensors are complementary in the detection of the carbon-nitrogen ratio of sweet potatoes, and it is difficult for single-modal data to comprehensively characterize the physiological state of crops. The multi-spectral reflectance can effectively reflect the leaf pigment content and structural characteristics, but it is susceptible to environmental interference such as soil background and light intensity; although the fluorescence lifetime parameter is sensitive to changes in the carbon-nitrogen ratio, the information content is limited when used alone, and the fluorescence intensity is susceptible to factors such as leaf thickness and moisture. The multi-spectral data provides macroscopic structural information, and the fluorescence lifetime reveals the carbon-nitrogen metabolic differences at the molecular level. The combination of the two can construct a more complete carbon-nitrogen ratio characterization system.

[0032] Therefore, couple the multi-spectral data and fluorescence spectral data of each leaf to obtain the coupled eigenvalue of the spectral data of each leaf. The expression is: In the formula, represents the coupled eigenvalue of the spectral data of the current leaf; n represents the number of wavelengths in any spectral data; is the weight coefficient of the i-th wavelength in the growth stage where the current leaf is located; represents the reflectance of the i-th wavelength in the multi-spectral data of the current leaf; represents the fluorescence lifetime corresponding to the i-th wavelength in the fluorescence spectral data of the current leaf; represents the preset reference fluorescence lifetime. In this embodiment, is set to the average value of the fluorescence lifetimes of all wavelengths of healthy sweet potato leaves; represents the exponential function with the natural constant e as the base, and the purpose is to convert the linear change of the fluorescence lifetime into a non-linear modulation factor to enhance the sensitive response to the carbon-nitrogen ratio.

[0033] Among them, the process of obtaining the weight coefficient of the \(i\)-th wavelength in the current growth stage of the leaf is as follows: Obtain the reflectance of the \(i\)-th wavelength in all leaf spectral data collected in this growth stage, and denote the set composed of them as the \(i\)-th wavelength reflectance set of this growth stage; calculate the mutual information between the elements in this reflectance set and the corresponding leaf carbon-nitrogen ratio, and denote it as the first mutual information value of the \(i\)-th wavelength in this growth stage; among them, the calculation of mutual information is a well-known technology, and the specific process will not be elaborated here; Calculate the sum value of the first mutual information values of all wavelengths in this growth stage, and denote it as the first sum value; use the ratio of the first mutual information value of the \(i\)-th wavelength in this growth stage to the first sum value as the weight coefficient of the \(i\)-th wavelength in this growth stage.

[0034] The traditional fixed weight method assumes that the contributions of all wavelengths to the carbon-nitrogen ratio (CNR) are constant. However, in actual detection, the spectral responses in different growth stages are significantly different, and the spectral sensitive bands in different growth stages are different. For example, the importance of the nitrogen-sensitive band (680nm) of the leaf in the seedling stage is 2-3 times that in the mature stage, while the weight of the sensitive band (950nm) needs to be increased during the tuber swelling stage. In addition, environmental interferences, such as changes in leaf water content, will cause the information content of some wavelengths to decrease, and the fixed weight cannot compensate dynamically. Therefore, the non-linear correlation between the reflectance of each wavelength in the spectrum and the carbon-nitrogen ratio is quantified by mutual information, and the adaptive weight coefficient of each wavelength is constructed.

[0035] Furthermore, the weighted sum of the coupling results of the corresponding wavelength data in the two spectral data is performed through the weight coefficient, deeply combining the macroscopic structure information of the multi-spectral and the microscopic metabolic information of the fluorescence, constructing a composite feature that is sensitive to the carbon-nitrogen ratio and resistant to environmental interference, solving the problem of insufficient detection accuracy of the traditional single modality, and providing higher-quality input data for subsequent modeling.

[0036] Because when nitrogen is deficient, the molecular structure of chlorophyll changes and the fluorescence lifetime is prolonged. When When, the exponential term decays rapidly, suppressing the contribution of high When, the exponential term approaches 1, retaining low When, the exponential term approaches 1, retaining low value reflectance signal. Compared with linear weighting, the exponential function can significantly suppress extreme values and improve the correlation.

[0037] Step S3, correct the coupling eigenvalue of the spectral data of each leaf according to the soil conductivity of each leaf.

[0038] Soil conductivity (EC) is an important environmental interference factor affecting the spectral response of plants. An excessively high EC value will cause the stomata of plant leaves to close, change the arrangement of chlorophyll molecules, and shift the reflectance curve. For example, the absorption valley at a wavelength of 760nm becomes shallower.

[0039] Therefore, a compensation coefficient is constructed according to the magnitude of the EC, and the coupling characteristics are compensated according to the compensation coefficient. First, the expression of the compensation coefficient is as follows: In the formula, is the compensation coefficient corresponding to the soil conductivity of the current leaf; A and B respectively represent the preset intercept and slope. Among them, the value range of A is 0.4 to 0.6, and the value range of B is 0.2 to 0.4. Preferably, in the embodiments of the present application, the values of A and B are respectively set to 0.5 and 0.3; is the tanh function; is the soil conductivity of the current leaf; C represents the reference value of the soil conductivity. In the embodiments of the present application, the value of C is set to , which is the optimal EC value for the growth of sweet potatoes. When it is lower than this value, the soil fertility is insufficient, and when it is higher than this value, there is a risk of salt stress; E represents the preset response coefficient, and the value in this embodiment is , which controls the slope of the tanh function, so that when the soil conductivity is larger, the change trend of the tanh function is slower, and when the soil conductivity is closer to 0, the change trend of the tanh function is faster, which conforms to the physiological characteristics of the plant's response to salt. The sensitivity is high at low concentrations and the adaptability decreases at high concentrations.

[0040] Thus rapidly adjusts within the range of , and tends to be saturated outside this interval. That is, when , (no compensation); for every deviation of the EC from the reference value of , changes by , ensuring that the compensation intensity is positively correlated with the interference degree. When , the weight of the multispectral reflectance is reduced to avoid overestimating the carbon-nitrogen ratio due to poor soil. When , the weight of the fluorescence lifetime parameter is enhanced, and the anti-interference ability of the fluorescence characteristics is utilized, such as the response of the fluorescence lifetime to salt stress lags behind the spectrum.

[0041] Furthermore, based on the soil conductivity obtained when collecting spectral data for the current leaf, and the compensation coefficient corresponding to the soil conductivity, the coupling characteristic value of the spectral data of the current leaf is corrected, and the expression is: In the formula, is the corrected value of the coupling characteristic value of the spectral data of the current leaf, represents the coupling characteristic value of the spectral data of the current leaf, is the compensation coefficient corresponding to the soil conductivity of the current leaf, is the soil conductivity corresponding to the current leaf.

[0042] Step S4: Perform feature combination based on the reflectance of characteristic wavelengths in the multispectral data of each leaf to construct the multispectral index of each leaf; perform feature combination based on the fluorescence intensity and corresponding fluorescence lifetime of characteristic wavelengths in the fluorescence spectrum of each leaf to construct the fluorescence spectrum index of each leaf.

[0043] Through dynamic weight allocation and environmental interference compensation, the multispectral reflectance and fluorescence lifetime parameters are efficiently coupled. However, the exponential form is difficult to describe the high-order non-linear relationship between the spectrum and the carbon-nitrogen ratio, such as the exponential growth of starch accumulation during the tuber swelling period. The dynamic weight may fail in extreme environments, such as drought stress, resulting in abnormal contribution degrees of the fluorescence lifetime parameters. Moreover, a single index cannot retain the detailed information in the original data, such as the subtle fluctuations of the spectral curve. Therefore, it is necessary to extract the features of multispectral data and fluorescence spectral data as supplements.

[0044] In this embodiment, the characteristic wavelengths of the multispectral data are selected as follows: the wavelength corresponding to the strong absorption peak of chlorophyll a (680 nm), the reflectance at this wavelength reflects the chlorophyll concentration and is positively correlated with the nitrogen content; the wavelength corresponding to the near-infrared plateau region (820 nm), the reflectance at this wavelength characterizes the leaf structure complexity and is positively correlated with carbon accumulation; the wavelength corresponding to the photosystem II sensitive region (760 nm), the reflectance at this wavelength is affected by photochemical quenching and indirectly reflects the light utilization efficiency; the wavelength corresponding to the water absorption valley (950 nm), the reflectance at this wavelength indicates the leaf water content and is negatively correlated with the carbon-nitrogen ratio. The characteristic wavelengths of the fluorescence spectral data are selected as follows: the wavelength corresponding to the emission peak of chlorophyll a (685 nm), the fluorescence intensity at this wavelength reflects the total amount of chlorophyll and is positively correlated with the nitrogen content; the wavelength corresponding to the far-red light emission peak (740 nm), the fluorescence intensity at this wavelength is affected by the size of the photosystem II antenna and is related to the light energy distribution efficiency; the fluorescence lifetime at the wavelength of 685 nm, when the carbon-nitrogen ratio decreases, the intermolecular energy transfer efficiency decreases and the lifetime prolongs; the fluorescence lifetime at the wavelength of 740 nm reflects the redox state of the photosystem II reaction center and is negatively correlated with the electron transfer rate.

[0045] Furthermore, perform feature combination based on the selected characteristic wavelengths of the multispectral data to construct the multispectral index. In the embodiment of the present application, the expression of the multispectral index can be: In the formula, is the multispectral index of the current leaf, , , , respectively represent the reflectances at the wavelengths of 820 nm, 680 nm, 950 nm and 760 nm in the multispectral data of the current leaf, is an exponential function with the natural constant e as the base. The first part selects 820 nm (carbon-sensitive) and 680 nm (nitrogen-sensitive) to strengthen the carbon-nitrogen ratio response, and the exponential function part enhances the non-linear relationship between moisture (950 nm) and the photosystem (760 nm). The role of the exponential function is to enhance the spectral data characteristics through non-linear transformation, amplify small differences or compress high-value intervals. When the moisture stress intensifies, the ratio increases, and the exponential term amplifies this change, enhancing the sensitivity to drought-induced abnormal carbon-nitrogen ratios, and capturing the complex interaction between the carbon-nitrogen ratio and environmental factors through exponential transformation; According to the characteristic wavelengths of the selected fluorescence spectral data, characteristic combinations are made to construct a fluorescence spectral index. In the embodiments of the present application, the expression of the fluorescence spectral index can be: In the formula, F is the fluorescence spectral index of the current leaf, 、 are the fluorescence intensities at wavelengths 685 nm and 740 nm in the fluorescence spectral data of the current leaf respectively, 、 are the fluorescence lifetimes corresponding to wavelengths 685 nm and 740 nm in the fluorescence spectral data of the current leaf respectively. The fluorescence intensity ratio is a traditional carbon-nitrogen ratio index, but it is easily interfered by ambient light. In this scheme, through the time-resolved characteristic, that is, the fluorescence lifetime, the carbon-nitrogen ratio specificity is enhanced, and then through the geometric mean lifetime parameter, the influence of single-wavelength lifetime fluctuations is eliminated.

[0046] As other embodiments of the present application, the implementer can set the characteristic combination method of the characteristic wavelengths according to the actual situation to construct a multi-spectral index and a fluorescence spectral index.

[0047] The construction of the above two spectral indices combines the physiological mechanism of sweet potato carbon-nitrogen metabolism. Through key wavelength combinations and non-linear transformations, the signals related to the carbon-nitrogen ratio are strengthened and environmental interference is suppressed.

[0048] Step S5, construct the characteristic vector of each leaf based on the corrected value of the spectral data coupling eigenvalue of each leaf, the multi-spectral index, the fluorescence spectral index and the environmental data; construct a prediction model according to the characteristic vectors of all leaf samples at each growth stage and the carbon-nitrogen ratio, and combine the characteristic vectors of the sweet potato leaves to be detected to obtain the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be detected.

[0049] Since the carbon-nitrogen ratio ranges vary significantly at different growth stages of sweet potato, the growth stage of sweet potato is of great significance for carbon-nitrogen ratio detection. For example, there may be significant differences in the carbon-nitrogen ratio between the tuber formation stage and the maturity stage. Therefore, regression prediction models for each growth stage are constructed according to the leaf data of each growth stage respectively, specifically: The vector composed of the corrected value of the spectral data coupling eigenvalue, the multispectral index, the fluorescence spectral index, and the environmental index of each leaf is used as the eigenvector of each leaf. Among them, the eigenvector of the t-th leaf is denoted as , where is the corrected value of the spectral data coupling eigenvalue of the t-th leaf, is the multispectral index of the t-th leaf, is the fluorescence spectral index of the t-th leaf, is the vector composed of the environmental index of the t-th leaf; In each growth stage of sweet potato, the eigenvectors and carbon-nitrogen ratios of all leaf samples in this growth stage are used as the input of the LightGBM regression prediction model for training the LightGBM regression prediction model. Among them, the maximum number of leaves of each tree is set to 25, the maximum depth of the tree is set to 6, the number of trees is set to 1000, the feature ratio used by each tree is 0.8, and the learning rate is set to 0.1. The trained LightGBM regression prediction model for each growth stage is obtained. The training process of the LightGBM regression prediction model is a well-known technology, and the specific steps will not be elaborated.

[0050] Based on the multispectral, fluorescence spectral, carbon-nitrogen ratio, and environmental data of the sweet potato leaf to be detected, the eigenvector of the sweet potato leaf to be detected is obtained by using the same acquisition method as the eigenvector of the above-mentioned leaf samples. The eigenvector of the sweet potato leaf to be detected is used as the input of the LightGBM regression prediction model corresponding to the growth stage of this leaf, and the output is the predicted value of the carbon-nitrogen ratio of the sweet potato leaf to be detected.

[0051] Step S6, based on the difference between the eigenvector of the sweet potato leaf to be detected and the eigenvectors of all leaves in all growth stages, and the carbon-nitrogen ratio of the leaves in the previous growth stage of the growth stage where the sweet potato leaf to be detected is located, combined with the predicted value, construct the carbon-nitrogen ratio of the sweet potato leaf to be detected.

[0052] Randomly select the eigenvectors of x leaves in each growth stage, and use the set composed of the eigenvectors of the x leaves in all growth stages as the growth stage data set; preferably, in the embodiment of the present application, the value of x is set to 10. As other embodiments of the present application, the implementer can set the value of x according to the actual situation.

[0053] Calculate the Euclidean distance between the eigenvector of the sweet potato leaf to be detected and each eigenvector in the growth stage data set, denoted as the first distance; use the normalized value of the sum of all the first distances of the sweet potato leaf to be detected as the dynamic growth stage weight of the sweet potato leaf to be detected.

[0054] Calculate the carbon-nitrogen ratio of the sweet potato leaves to be detected according to the carbon-nitrogen ratio of the leaves in the previous growth stage of the growth stage where the sweet potato leaves to be detected are located, combined with the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be detected. The expression is as follows: In the formula, is the carbon-nitrogen ratio of the sweet potato leaves to be detected, is the dynamic growth stage weight of the sweet potato leaves to be detected, represents the feature vector of the sweet potato leaves to be detected, represents the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be detected, represents the average value of the carbon-nitrogen ratios of all leaf samples in the previous growth stage of the growth stage where the sweet potato leaves to be detected are located. Among them, represents the sweet potato leaves to be detected. Since the change of the carbon-nitrogen ratio has a certain continuity and inertia, introducing the carbon-nitrogen ratio of the previous growth stage makes the result smoother and more accurate.

[0055] It should be noted that if the growth stage where the sweet potato leaves to be detected are located is the first growth stage, then .

[0056] Table 1 Comparison table of the effects of the solution of this application and the prior art Among them, the content corresponding to the dynamic mutual information weight in Table 1 is to construct the weight coefficient of each wavelength in each growth stage through the mutual information between the carbon-nitrogen ratio of the leaves and the reflectance of each wavelength in each growth stage, and the content corresponding to the EC compensation function is to construct the compensation coefficient through the magnitude of EC.

[0057] The schematic diagram of the acquisition process of the weight coefficient of each wavelength is as shown in Figure 2 shown, and the schematic diagram of the acquisition process of the dynamic growth stage weight is as shown in Figure 3 shown.

[0058] Based on the same inventive concept as the above method, the embodiment of this application also provides a sweet potato carbon-nitrogen ratio detection system based on a multi-spectral and fluorescence sensor. The system includes: Data fusion layer: In each growth stage, collect the multi-spectral, fluorescence spectrum, carbon-nitrogen ratio and environmental data of each sweet potato leaf; obtain the fluorescence lifetime corresponding to each wavelength in the fluorescence spectrum of each leaf; the environmental data includes soil conductivity; construct the weight coefficient of each wavelength in each growth stage based on the correlation between the carbon-nitrogen ratio of the leaves and the reflectance of each wavelength in each growth stage; construct the coupling eigenvalue of the spectral data of each leaf based on the weight coefficient and the multi-spectral and fluorescence spectra of each leaf; correct the coupling eigenvalue of the spectral data of each leaf according to the soil conductivity of each leaf. Feature engineering layer: Feature combinations are performed based on the reflectance of characteristic wavelengths in the multispectral data of each leaf to construct the multispectral index of each leaf; feature combinations are performed based on the fluorescence intensity and corresponding fluorescence lifetime of characteristic wavelengths in the fluorescence spectrum of each leaf to construct the fluorescence spectrum index of each leaf; based on the corrected value of the coupled eigenvalue, multispectral index, fluorescence spectrum index of each leaf, and environmental data, the feature vector of each leaf is constructed. Temporal sequence modeling layer: A prediction model is constructed based on the feature vectors and carbon-nitrogen ratio of all leaf samples at each growth stage, and combined with the feature vector of the sweet potato leaf to be detected, the predicted value of the carbon-nitrogen ratio of the sweet potato leaf to be detected is obtained; based on the difference between the feature vector of the sweet potato leaf to be detected and the feature vectors of all leaves at all growth stages, and the carbon-nitrogen ratio of the leaves in the previous growth stage of the growth stage where the sweet potato leaf to be detected is located, combined with the predicted value, the carbon-nitrogen ratio of the sweet potato leaf to be detected is constructed.

[0059] In summary, the embodiment of the present application provides a method for detecting the carbon-nitrogen ratio of sweet potatoes based on multispectral and fluorescence sensors. The weight coefficient of each wavelength in each growth stage is constructed based on the correlation between the carbon-nitrogen ratio of the leaves and the reflectance of each wavelength in each growth stage; based on the weight coefficient, and the multispectral and fluorescence spectra of each leaf, the coupled eigenvalue of the spectral data of each leaf is obtained, avoiding the problem that single-modal data is difficult to comprehensively characterize the physiological state of crops, resulting in insufficient detection accuracy; the coupled eigenvalue is corrected according to the soil conductivity, so that the calculation result of the coupled eigenvalue is more in line with the actual growth characteristics of sweet potato leaves; by analyzing the multispectral characteristics and fluorescence spectral characteristics of sweet potatoes, the multispectral index and fluorescence spectrum index are constructed, which can more comprehensively characterize the physiological state of sweet potatoes and increase the detection accuracy; based on the above characteristics of the leaves, a regression prediction model for each growth stage is constructed, the predicted value of the carbon-nitrogen ratio of the sweet potato leaf to be detected is obtained, combined with the difference between the feature vector of the sweet potato leaf to be detected and the feature vectors of all leaves at all growth stages, and the carbon-nitrogen ratio of the previous growth stage is introduced to calculate the carbon-nitrogen ratio of the sweet potato leaf to be detected, considering that the change of the carbon-nitrogen ratio has a certain continuity and inertia, so as to adapt to the changes in different growth stages of sweet potatoes and improve the detection accuracy of the carbon-nitrogen ratio.

[0060] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the advantages or disadvantages of the embodiments. And the above specific embodiments of the present application have been described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0061] The various embodiments in the present application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments.

[0062] The above are only the preferred embodiments of the present application, and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present application shall be included within the protection scope of the present application.

Claims

1. A method for detecting the carbon-nitrogen ratio of sweet potatoes based on multispectral and fluorescence sensors, characterized in that: The method comprises the following steps: At each growth stage, the multi-spectrum, fluorescence spectrum, carbon-nitrogen ratio and environmental data of each sweet potato leaf are collected; the fluorescence lifetime corresponding to each wavelength in the fluorescence spectrum of each leaf is obtained; the environmental data includes soil conductivity; Based on the correlation between the carbon-nitrogen ratio of the leaves and the reflectivity of each wavelength in each growth stage, a weight coefficient of each wavelength in each growth stage is constructed; based on the weight coefficient, and the multi-spectrum and fluorescence spectrum of each leaf, a coupling characteristic value of the spectral data of each leaf is constructed; The coupled characteristic value of the spectral data of each leaf is corrected according to the soil conductivity of each leaf; The multispectral index of each leaf is constructed by combining the reflectance of the characteristic wavelengths in the multispectral spectrum of each leaf; the fluorescence intensity and the corresponding fluorescence lifetime of the characteristic wavelengths in the fluorescence spectrum of each leaf are combined to construct the fluorescence spectrum index of each leaf; Based on the correction value of the coupled characteristic value of the spectral data of each leaf, the multi-spectral index, the fluorescence spectral index and the environmental data, the characteristic vector of each leaf is constructed; according to the characteristic vectors and carbon-nitrogen ratios of all leaf samples at each growth stage, a prediction model is constructed, and the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be tested is obtained in combination with the characteristic vectors of the sweet potato leaves to be tested; Based on the difference between the characteristic vector of the sweet potato leaf to be detected and the characteristic vector of all leaves in all growth stages, and the carbon-nitrogen ratio of the leaves in the growth stage before the growth stage of the sweet potato leaf to be detected, combined with the predicted value, the carbon-nitrogen ratio of the sweet potato leaf to be detected is constructed.

2. The sweet potato carbon-nitrogen ratio detection method based on multispectral and fluorescence sensor as claimed in claim 1, characterized in that: The process of obtaining the weight coefficient of each wavelength in each growth stage is as follows: The reflectance of the i-th wavelength in all leaf spectral data at each growth stage is formed into a set, which is recorded as the i-th wavelength reflectance set at each growth stage; the mutual information between the elements in the i-th wavelength reflectance set and the corresponding leaf carbon-nitrogen ratio is calculated, which is recorded as the first mutual information value; The sum of the first mutual information values ​​of all wavelengths in each growth stage is calculated and recorded as the first sum value; the ratio of the first mutual information value of the i-th wavelength in each growth stage to the first sum value is used as the weight coefficient of the i-th wavelength in each growth stage.

3. The sweet potato carbon-nitrogen ratio detection method based on multispectral and fluorescence sensor as claimed in claim 1, characterized in that: The expression of the coupling characteristic value of the spectral data of each blade is: , where Indicates the coupling characteristic value of the spectral data of the current blade; n indicates the number of wavelengths in any spectral data; is the weight coefficient of the i-th wavelength of the current leaf growth stage; Represents the reflectance of the i-th wavelength in the multispectral data of the current leaf; Represents the fluorescence lifetime corresponding to the i-th wavelength in the fluorescence spectrum data of the current leaf; Indicates the preset benchmark fluorescence lifetime; Represents an exponential function with the natural constant e as the base.

4. The sweet potato carbon-nitrogen ratio detection method based on multispectral and fluorescence sensor as claimed in claim 1, characterized in that: The coupling characteristic value of the spectral data of each leaf is corrected, and the expression is: In the formula, is the correction value of the spectral data coupling characteristic value of the current leaf, Represents the coupled eigenvalue of the spectral data of the current leaf, is the compensation coefficient corresponding to the soil conductivity of the current leaf, is the soil conductivity corresponding to the current leaf, A and B represent the preset intercept and slope respectively. is the tanh function, C is the reference value of soil conductivity, and E is the preset response coefficient.

5. The sweet potato carbon-nitrogen ratio detection method based on multispectral and fluorescence sensor as claimed in claim 1, characterized in that: The expression of the multispectral index of each leaf is: In the formula, is the multispectral index of the current leaf, , , , They represent the reflectance at wavelengths of 820nm, 680nm, 950nm and 760nm in the multispectral data of the current leaf, respectively. It is an exponential function with the natural constant e as the base; the wavelength of 680nm is the wavelength corresponding to the strong absorption peak of chlorophyll a, the wavelength of 820nm is the wavelength corresponding to the near-infrared platform area, the wavelength of 760nm is the wavelength corresponding to the sensitive area of ​​photosystem II, and the wavelength of 950nm is the wavelength corresponding to the water absorption valley.

6. The sweet potato carbon-nitrogen ratio detection method based on multispectral and fluorescence sensor as claimed in claim 1, characterized in that: The expression of the fluorescence spectrum index of each leaf is: Where F is the fluorescence spectrum index of the current leaf, , are the fluorescence intensities at wavelengths of 685nm and 740nm in the current leaf fluorescence spectrum data, , They are the fluorescence lifetimes corresponding to wavelengths of 685nm and 740nm in the current leaf fluorescence spectrum data, respectively; wherein, the wavelength of 685nm is the wavelength corresponding to the emission peak of chlorophyll a, and the wavelength of 740nm is the wavelength corresponding to the far-red light emission peak.

7. The sweet potato carbon-nitrogen ratio detection method based on multispectral and fluorescence sensor as claimed in claim 1, characterized in that: The process of acquiring the characteristic vector of each leaf is as follows: Calculate the sum of the normalized values ​​of temperature, humidity and light intensity in the environmental data of each leaf as the environmental index of each leaf; The vector composed of the correction value of the spectral data coupling characteristic value of each leaf, the multi-spectral index, the fluorescence spectral index and the environmental index is used as the characteristic vector of each leaf.

8. The sweet potato carbon-nitrogen ratio detection method based on multispectral and fluorescence sensor as claimed in claim 1, characterized in that: The process of obtaining the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be tested is as follows: The characteristic vectors and carbon-nitrogen ratios of all leaf samples in each growth stage are used as inputs of the regression prediction model, and the regression prediction model is trained to obtain the trained regression prediction model for each growth stage; The characteristic vector of the sweet potato leaf to be tested is used as the input of the regression prediction model of its corresponding growth stage, and the output is the predicted value of the carbon-nitrogen ratio of the sweet potato leaf to be tested.

9. The method for detecting the carbon-nitrogen ratio of sweet potatoes based on multispectral and fluorescence sensor as claimed in claim 1, characterized in that: The process of obtaining the carbon-nitrogen ratio of the sweet potato leaves to be detected is: Randomly selecting feature vectors of a preset number of leaves in each growth stage, and taking a set consisting of the feature vectors of the preset number of leaves in all growth stages as a growth stage data set; Calculate the Euclidean distance between the feature vector of the sweet potato leaf to be detected and each feature vector in the growth stage data set, recorded as the first distance; take the normalized value of the sum of all the first distances of the sweet potato leaf to be detected as the dynamic growth stage weight of the sweet potato leaf to be detected; The carbon-nitrogen ratio of the sweet potato leaves to be tested is recorded as , The expression is: In the formula, is the carbon-nitrogen ratio of the sweet potato leaves to be tested, is the dynamic growth stage weight of the sweet potato leaves to be tested, represents the feature vector of the sweet potato leaf to be detected, represents the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be tested, represents the mean carbon-nitrogen ratio of all leaf samples in the growth stage before the growth stage of the sweet potato leaves to be tested, where Represents the sweet potato leaves to be tested.

10. A sweet potato carbon-nitrogen ratio detection system based on multispectral and fluorescence sensors, implementing the method as claimed in claim 1, characterized in that: The system comprises: Data fusion layer: at each growth stage, collect the multi-spectrum, fluorescence spectrum, carbon-nitrogen ratio and environmental data of each sweet potato leaf; obtain the fluorescence lifetime corresponding to each wavelength in the fluorescence spectrum of each leaf; the environmental data includes soil conductivity; construct the weight coefficient of each wavelength in each growth stage based on the correlation between the carbon-nitrogen ratio of the leaves and the reflectivity of each wavelength in each growth stage; construct the coupling characteristic value of the spectral data of each leaf based on the weight coefficient and the multi-spectrum and fluorescence spectrum of each leaf; correct the coupling characteristic value of the spectral data of each leaf according to the soil conductivity of each leaf; Feature engineering layer: Combine features according to the reflectivity of characteristic wavelengths in the multi-spectrum of each leaf to construct the multi-spectral index of each leaf; Combine features according to the fluorescence intensity and corresponding fluorescence lifetime of characteristic wavelengths in the fluorescence spectrum of each leaf to construct the fluorescence spectrum index of each leaf; Construct the feature vector of each leaf based on the correction value of the coupling characteristic value of the spectral data of each leaf, the multi-spectral index, the fluorescence spectrum index and the environmental data; Time series modeling layer: a prediction model is constructed based on the feature vectors and carbon-nitrogen ratios of all leaf samples at each growth stage, and the feature vectors of the sweet potato leaves to be tested are combined to obtain the predicted value of the carbon-nitrogen ratio of the sweet potato leaves to be tested; based on the difference between the feature vectors of the sweet potato leaves to be tested and the feature vectors of all leaves at all growth stages, as well as the carbon-nitrogen ratio of the leaves in the previous growth stage of the sweet potato leaves to be tested, the carbon-nitrogen ratio of the sweet potato leaves to be tested is constructed in combination with the predicted value.

Images