Fluoride resistance evaluation system for tea trees based on spectral analysis and machine learning

Abstract

The present invention belongs to the field of spectral analysis technology, and specifically relates to a tea tree fluoride resistance evaluation system based on spectral analysis and machine learning. The system includes: a spectral data acquisition and processing unit, a spectral feature extraction and analysis unit, and a fluorine resistance evaluation unit; the spectral data acquisition and processing unit is used to collect spectral information of the target tea tree area through a hyperspectral imaging instrument; the spectral feature extraction and analysis unit is used to suppress noise and enhance signals on the calculated spectral reflectance to obtain a preprocessing result; the wavelet decomposition result and the chlorophyll fluorescence response index are fused using nonlinear features based on machine learning to obtain a feature fusion result; the fluorine resistance evaluation unit is used to calculate the fluorine resistance index based on the feature fusion result. The present invention effectively solves the problems existing in the prior art, such as high data processing complexity, large limitations of analysis methods, and insufficient dynamic monitoring.

Description

Technical Field

[0001] The invention belongs to the technical field of spectral analysis, and in particular relates to a tea tree fluorine resistance evaluation system based on spectral analysis and machine learning. Background Art

[0002] Tea trees are one of the most widely planted economic crops in the world. Their leaves are rich in a variety of secondary metabolites, such as tea polyphenols, caffeine, amino acids, etc., which have many benefits to human health. However, during the growth of tea trees, various stress factors in the environment, such as heavy metals, pollutants, and pests and diseases, will affect their growth and yield. Among them, fluorine, as a common pollutant, has become an important factor affecting the growth and quality of tea trees, especially in some industrialized areas, due to industrial emissions, soil pollution, etc. The toxicity of fluorine to plants is mainly manifested in leaf yellowing, decreased photosynthesis efficiency, physiological dysfunction, etc., which directly affect the yield and quality of tea. Therefore, how to accurately evaluate the tolerance of tea trees to fluorine stress and then guide the planting and management of tea trees has become an important topic in tea cultivation research.

[0003] In past studies, scientists have used a variety of methods to evaluate the fluorine resistance of tea trees. Traditional methods mainly include chemical analysis and physiological indicator measurement. For example, by measuring physiological parameters such as fluorine content, chlorophyll content, and photosynthesis rate in tea leaves, the physiological state of tea trees under fluorine stress is evaluated. However, these methods often require destructive sampling, which is complicated and time-consuming to operate, and it is difficult to achieve real-time monitoring of large-scale tea gardens. In addition, since fluorine has many effects on plants, it is difficult to fully reflect the comprehensive tolerance of tea trees under fluorine stress with only a single physiological or chemical indicator. Although these traditional methods have achieved certain results in laboratory studies, they have great limitations in practical applications, especially in the monitoring and management of large-scale tea gardens.

[0004] With the development of spectral technology, non-destructive spectral analysis methods have gradually become an effective tool for evaluating plant stress responses. Hyperspectral imaging technology is widely used in monitoring plant physiological status and stress assessment because it can obtain spectral information of plant leaves in multiple wavelength ranges. Hyperspectral imaging technology can obtain the spectral reflectance of plant leaves at different wavelengths and analyze the changes in their spectral characteristics under stress conditions, thereby achieving real-time and rapid monitoring of plant health status. In the study of fluorine resistance of tea trees, spectral analysis technology has also gradually been applied. By analyzing the changes in the spectral characteristics of tea leaves in different bands, the fluorine resistance of tea trees can be reflected to a certain extent.

[0005] However, the existing spectral analysis technology still faces some challenges when applied to the evaluation of fluorine resistance of tea plants. First, although hyperspectral imaging technology can obtain rich spectral data, these data are often high-dimensional and complex. How to extract features closely related to the fluorine resistance of tea plants is still a difficulty. Traditional spectral analysis methods mostly use a single band or a simple spectral index for analysis. Although this method is intuitive, it is difficult to capture the changes in the multidimensional spectral characteristics of tea plants under complex stress conditions. In addition, spectral data are usually affected by a variety of external factors such as ambient light conditions, atmospheric conditions, and observation angles. These interference factors may cause the accuracy of spectral data to decrease, thereby affecting the reliability of the evaluation results. Secondly, when processing spectral data, existing spectral analysis methods mostly use traditional statistical methods such as linear regression and principal component analysis. Although these methods can simplify the data processing process to a certain extent, they show great limitations when facing nonlinear and complex data patterns. The spectral response of tea plants under fluorine stress often shows nonlinear changes, which are not only reflected in specific bands of the spectrum, but also involve interactions between multiple bands. Therefore, it is difficult to fully capture the changes in the spectral characteristics of tea leaves under fluorine stress using only the linear analysis method, resulting in insufficient accuracy and sensitivity in fluorine resistance assessment. Summary of the invention

[0006] The main purpose of the present invention is to provide a tea tree fluorine resistance evaluation system based on spectral analysis and machine learning. The present invention effectively solves the problems existing in the prior art, such as high complexity of data processing, large limitations of analysis methods, insufficient dynamic monitoring, etc., and significantly improves the accuracy and practicality of tea tree fluorine resistance evaluation, providing strong technical support for tea tree planting and management.

[0007] In order to solve the above technical problems, the present invention provides a tea tree fluorine resistance evaluation system based on spectral analysis and machine learning, the system comprising: a spectral data acquisition and processing unit, a spectral feature extraction and analysis unit and a fluorine resistance evaluation unit; the spectral data acquisition and processing unit is used to collect spectral information of a target tea tree area through a hyperspectral imaging instrument, and calculate the spectral reflectance of the target tea tree area at different wavelengths; the spectral feature extraction and analysis unit is used to perform noise suppression and signal enhancement on the calculated spectral reflectance to obtain a preprocessing result, and based on the preprocessing result, perform feature extraction on the fluorine-sensitive band of the target tea tree area to obtain a fluorine-sensitive feature, and calculate the chlorophyll fluorescence response index of the target tea tree area according to the wavelength of chlorophyll; after performing spectral integral transformation on the preprocessing result, perform multi-scale wavelet decomposition to obtain a wavelet decomposition result; the wavelet decomposition result and the chlorophyll fluorescence response index are fused using nonlinear features based on machine learning to obtain a feature fusion result; the fluorine resistance evaluation unit is used to calculate the fluorine resistance index based on the feature fusion result.

[0008] Furthermore, the hyperspectral imaging instrument is: an airborne hyperspectral sensor or a satellite-borne hyperspectral sensor.

[0009] Furthermore, the spectral data acquisition and processing unit calculates the spectral reflectance of the target tea tree area at different wavelengths using the following formula:

[0010]

[0011] Where S(λ) is the spectral reflectance at wavelength λ; L(λ) is the radiation brightness received by the hyperspectral imaging instrument at wavelength λ; d is the distance between the sun and the earth; E0(λ) is the solar spectral irradiance at wavelength λ; θ s is the solar zenith angle; θ v is the observation zenith angle; τ(λ) is the atmospheric optical thickness.

[0012] Furthermore, the spectral feature extraction and analysis unit performs noise suppression and signal enhancement on the calculated spectral reflectance through the following formula to obtain a preprocessing result:

[0013]

[0014] Among them, S′(λ) is the preprocessing result; G(λ,σ) is the Gaussian filter function, σ is the standard deviation; α is the enhancement coefficient based on the adaptive adjustment of the signal-to-noise ratio, and its value range is 0.2 to 1.5; It means to calculate the second derivative of S(λ).

[0015] Furthermore, according to the preprocessing results, the fluorine-sensitive band of the target tea tree area is extracted by the following formula to obtain the fluorine-sensitive feature F(λ i ):

[0016] F(λ i )=α·NDFI(λ i )+β·FSRI(λ i )+γ·FABI(λ i );

[0017] Among them, NDFI(λ i ) is the wavelength λ i FSRI(λ i ) is the wavelength λ i Fluoride stress ratio index; FABI (λ i ) is the wavelength λ i The integral index of the fluorine absorption band at the time of NDFI (λ) is the first weight coefficient; α is the second weight coefficient; γ is the third weight coefficient; i ) is calculated using the following formula:

[0018]

[0019] Among them, λ ref is the reference wavelength not affected by fluorine; λ F is the main characteristic absorption wavelength of fluorine, equal to 680 nanometers; σ F is the fluorine absorption bandwidth, equal to 20 nm; FSRI(λ i ) is calculated using the following formula:

[0020]

[0021] Among them, λ R is the red light band; NIR is the near infrared band; C F is the fluorine content of leaves; k F is the fluorine concentration response coefficient, equal to 0.5; FABI(λ i ) is calculated using the following formula:

[0022]

[0023] Among them, [λ1, λ2] represents the integral range of the fluorine absorption band, λ1 is the lower limit of the integral range of the fluorine absorption band; λ2 is the upper limit of the integral range of the fluorine absorption band.

[0024] Furthermore, the first weight coefficient α is calculated using the following formula:

[0025]

[0026] Among them, C F，threshold is the fluorine content threshold, which is equal to 50ppm; k1 is the first adjustment coefficient, which ranges from 8 to 12; the second weight coefficient β is calculated using the following formula:

[0027] β = 1 - exp (-k2 RES);

[0028] Wherein, RES is the spectral resolution; k2 is the second adjustment coefficient, ranging from 0.09 to 0.13; the third weight coefficient γ is calculated using the following formula:

[0029]

[0030] Furthermore, the chlorophyll fluorescence response index (CFI) of the target tea tree area was calculated by the following formula:

[0031]

[0032] Among them, F(740) represents the fluorine-sensitive characteristic at a wavelength of 740 nanometers; F(690) represents the fluorine-sensitive characteristic at a wavelength of 690 nanometers; Φ PSII is the quantum yield of photosystem II; Φ PSII，max is the maximum quantum yield of photosystem II; PAR is photosynthetically active radiation; PAR0 is the reference photosynthetically active radiation; t represents the hour of the day, with a value ranging from 0 to 24; T is the leaf temperature; T opt is the optimal temperature for photosynthesis; k T is the temperature response coefficient; NPQ is the non-photochemical quenching value; NPQ max is the maximum non-photochemical quenching value; Φ PSII Calculated using the following formula:

[0033]

[0034] Among them, F′ m F is the maximum fluorescence output under light adaptation state; s is the steady-state fluorescence yield; NPQ is calculated using the following formula:

[0035]

[0036] Among them, F m is the maximum fluorescence output under dark adaptation condition.

[0037] Furthermore, the preprocessing result is transformed into a spectral integral by the following formula:

[0038]

[0039] Where I(a, b) is the result of spectrum integration transformation; [a, b] is the integration interval; λ c is the central wavelength; σ c is the integration window width; use the following formula to perform multi-scale wavelet decomposition and obtain the wavelet decomposition result:

[0040]

[0041] Among them, W ψ [S′](λ, a) is the wavelet decomposition result; ψ(u) is the Morlet wavelet mother function; a is the scale parameter; u is the integral variable; the wavelet decomposition result and the chlorophyll fluorescence response index are subjected to nonlinear feature fusion based on machine learning. When the feature fusion result is obtained, nonlinearity is introduced through the sigmoid function. The weight matrix of the wavelet decomposition result, the weight matrix and bias of the chlorophyll fluorescence response index are predefined. A single-layer neural network is used to perform nonlinear feature fusion on the wavelet decomposition result and the chlorophyll fluorescence response index to obtain the feature fusion result Φ(λ).

[0042] Furthermore, the fluorine resistance evaluation unit calculates the fluorine resistance index Y according to the feature fusion result by the following formula:

[0043]

[0044] Among them, FTI is the uncorrected fluorine resistance index; PRI m is the modified photochemical vegetation index; PRI0 is the standard photochemical vegetation index; σ PRI is the sensitivity of the photochemical vegetation index, which is equal to 0.1; FTI is calculated using the following formula:

[0045]

[0046] Among them, A F (λ) is the absorption coefficient function of fluorine at different wavelengths; C is the red edge position response coefficient; PRI m Calculated using the following formula:

[0047]

[0048] The tea tree fluorine resistance evaluation system based on spectral analysis and machine learning of the present invention has the following beneficial effects: the present invention utilizes hyperspectral imaging technology to realize refined monitoring of tea leaves in multiple wavelength ranges. Hyperspectral imaging instruments can capture spectral reflectance information from visible light to near-infrared bands, providing extremely rich spectral data. These data are subjected to steps such as preprocessing and spectral integral transformation to obtain smoother and more representative spectral features. By analyzing the spectral reflectance at different wavelengths, the present invention can accurately capture the spectral response of tea trees under fluorine stress. This high-resolution spectral analysis provides a solid data foundation for fluorine resistance evaluation. Compared with traditional single-band or spectral index analysis, the multi-band comprehensive analysis method of the present invention can more comprehensively reflect the physiological state of tea trees, especially the changes under complex environmental conditions. Secondly, the present invention further improves the feature extraction capability of spectral data by introducing multi-scale wavelet decomposition technology. Wavelet decomposition is a technology that can simultaneously analyze signals in both time (or space) and frequency domains, and can capture subtle changes and overall trends in signals at different scales. By performing multi-scale wavelet decomposition on the pre-processed spectral data, the present invention can identify the multi-level spectral features of tea leaves under fluorine stress, which reflect the spectral response of tea trees at different scales, thereby improving the accuracy of fluorine resistance assessment. Wavelet decomposition also has good noise resistance, which can effectively suppress noise interference in spectral data, making the extracted features more reliable. In addition, the present invention combines the wavelet decomposition results with the chlorophyll fluorescence response index through nonlinear feature fusion technology, further enhancing the expressiveness of the features. The spectral response of tea trees under fluorine stress often shows a complex nonlinear relationship, and traditional linear analysis methods are difficult to capture such changes. The present invention successfully captures the complex correlation between the fluorine resistance characteristics of tea trees by introducing a single-layer neural network to perform nonlinear feature fusion on the wavelet decomposition results and the chlorophyll fluorescence response index. The introduction of the neural network not only improves the accuracy of feature fusion, but also can adaptively optimize the feature weights through the training process, so that the evaluation results are closer to the actual situation. This nonlinear feature fusion method has significant advantages over traditional methods and can more accurately reflect the physiological state of tea trees under different fluorine stress conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0050] Figure 1A schematic diagram of the system structure of a tea tree fluorine resistance evaluation system based on spectral analysis and machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The method of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments of the present invention.

[0052] Example 1: Reference Figure 1 , a tea tree fluorine resistance evaluation system based on spectral analysis and machine learning, the system comprising: a spectral data acquisition and processing unit, a spectral feature extraction and analysis unit and a fluorine resistance evaluation unit; the spectral data acquisition and processing unit is used to collect spectral information of a target tea tree area through a hyperspectral imaging instrument, and calculate the spectral reflectance of the target tea tree area at different wavelengths; the spectral feature extraction and analysis unit is used to perform noise suppression and signal enhancement on the calculated spectral reflectance to obtain a preprocessing result, and based on the preprocessing result, perform feature extraction on the fluorine-sensitive band of the target tea tree area to obtain a fluorine-sensitive feature, and calculate the chlorophyll fluorescence response index of the target tea tree area according to the wavelength of chlorophyll; after performing spectral integral transformation on the preprocessing result, perform multi-scale wavelet decomposition to obtain a wavelet decomposition result; use nonlinear feature fusion based on machine learning to obtain a feature fusion result for the wavelet decomposition result and the chlorophyll fluorescence response index; the fluorine resistance evaluation unit is used to calculate the fluorine resistance index based on the feature fusion result.

[0053] Specifically, the hyperspectral imaging instrument obtains detailed spectral data by detecting the spectral reflectance of tea leaves in multiple wavelength ranges from visible light to near-infrared bands. This spectral reflectance is actually the ratio of the intensity of reflected light to the intensity of incident light after the tea leaves are irradiated with light of different wavelengths. It can reflect the health status of the tea tree and its interaction with the environment. Especially when subjected to fluorine stress, the spectral reflectance of tea leaves will undergo specific changes, and these changes are particularly obvious in specific bands. Therefore, by collecting and recording these spectral reflectance data through hyperspectral imaging instruments, first-hand basic data can be provided for the fluorine resistance assessment of tea trees. After obtaining the spectral data, the next step of the spectral data acquisition and processing unit is to process these data, specifically including noise suppression and signal enhancement. Due to its high dimensionality and complexity, hyperspectral data often contains various noises, such as interference signals caused by changes in ambient light, changes in leaf angles, or instrument noise. If these noises are not effectively suppressed, the extraction of spectral features may be inaccurate, thereby affecting the accuracy of fluorine resistance assessment. To this end, noise suppression processing usually uses techniques such as wavelet transform or filter to separate noise from effective signals, thereby improving the purity of the signal. In addition, in order to further enhance the characteristic expression ability of the signal, the processing unit will also perform signal enhancement processing on the spectral data after noise suppression. This step usually uses methods such as principal component analysis (PCA) to reduce the dimension or enhance the characteristics of multidimensional spectral data, so that the characteristics closely related to the fluorine resistance of tea trees are more prominent. The spectral data after noise suppression and signal enhancement can more realistically reflect the reflection characteristics of tea trees at different wavelengths, especially the bands related to fluorine stress. At this point, the task of the spectral data acquisition and processing unit has not been completed. In order to make the data more in line with the requirements of subsequent analysis, the system will further calculate the processed spectral data, such as spectral integral transformation. Through spectral integral transformation, some unnecessary fluctuations in spectral data can be eliminated, the consistency of spectral features in different bands can be enhanced, and more stable data input can be provided for feature extraction and evaluation units. Finally, after a series of processing, the spectral data acquisition and processing unit converts the original hyperspectral data into preprocessing results that can be used for subsequent analysis. These results not only include the spectral reflectance characteristics of tea leaves at various wavelengths, but also undergo optimization processing such as noise suppression, signal enhancement, and spectral integration, making the data more representative and analyzable. It is precisely because of these sophisticated processing that the system can accurately identify the response of tea trees under fluorine stress, and then combine machine learning algorithms to evaluate fluorine resistance. Therefore, the spectral data acquisition and processing unit plays an important role in connecting the upper and lower parts of the entire system. It is not only the initial acquirer of data, but also the guardian of data quality, providing a solid foundation for the accuracy and reliability of the entire system.

[0054] Although preliminary noise suppression has been performed in the previous stage, in order to further improve the analysis quality of the data, the system adopts more refined noise processing techniques, such as denoising methods based on multi-scale wavelet transform. With this method, the system can analyze spectral signals at multiple scales, distinguish useful components from noise components in the signal, and effectively suppress irrelevant information, so that the accuracy of feature extraction is significantly improved. Signal enhancement processing uses techniques such as principal component analysis (PCA) to further enhance the features of the processed spectral data. Principal component analysis projects high-dimensional spectral data into low-dimensional space through linear transformation, while retaining the main feature information of the data, making those important features that can reflect the fluorine resistance of tea trees more prominent. After completing signal enhancement, the spectral feature extraction and analysis unit begins to extract features from the spectral data of tea trees. This process mainly focuses on identifying bands that are closely related to fluorine stress, which are called fluorine-sensitive bands. By analyzing the changes in the spectral reflectance of tea trees in these bands, the system can infer the physiological response of tea trees under fluorine stress. At the same time, the system also calculates chlorophyll-related features, such as the chlorophyll fluorescence response index (CFRI), because changes in chlorophyll are often one of the first physiological changes that tea trees show when they are under stress. In this way, the system can not only capture the direct spectral response of tea trees under fluorine stress, but also further deepen the understanding of tea trees' fluorine resistance through changes in chlorophyll fluorescence response. In order to ensure that the extracted spectral features can fully reflect the fluorine resistance of tea trees, the spectral feature extraction and analysis unit further performs multi-scale analysis on these features. Multi-scale analysis usually uses wavelet decomposition technology to decompose spectral data into sub-signals of multiple scales, and each sub-signal on a scale represents the changing characteristics of spectral data in different frequency ranges. By analyzing these multi-scale wavelet decomposition results, the system can extract spectral features reflecting tea trees at various scales, thereby obtaining more comprehensive feature information. This multi-scale feature extraction method makes up for the shortcomings of single-scale analysis, ensuring that the system can capture subtle changes in tea trees at different spectral scales and improving the accuracy of fluorine resistance assessment. After extracting the spectral features of the tea tree, the spectral feature extraction and analysis unit fuses these features with the chlorophyll fluorescence response index. The feature fusion process uses nonlinear methods based on machine learning, such as support vector machines (SVM) or neural networks (NN), etc., to integrate information in the multidimensional feature space to construct a comprehensive feature that can effectively characterize the fluorine resistance of tea trees. This feature fusion not only takes into account the importance of each individual feature, but also captures the complex relationships between different features through nonlinear combinations, providing richer and more accurate data support for subsequent fluorine resistance assessments.

[0055] The spectral features and chlorophyll fluorescence response index generated by the spectral feature extraction and analysis unit form a series of feature data after multi-scale wavelet decomposition and signal processing. These data contain information about the response of tea trees to different fluorine stress levels. However, this information is often complex and high-dimensional, and it is difficult to extract useful evaluation indicators from it by manual analysis alone. Therefore, the fluorine resistance evaluation unit integrates and analyzes these multidimensional feature data by introducing machine learning algorithms, especially nonlinear feature fusion technology, to form a comprehensive fluorine resistance evaluation model. In the specific evaluation process, the fluorine resistance evaluation unit first processes the input data such as spectral features and chlorophyll fluorescence response index through a trained machine learning model. These models are trained with a large amount of sample data and can identify the complex nonlinear relationship between spectral features and fluorine resistance of tea trees. Through machine learning methods such as support vector machines (SVM), random forests (RF) or neural networks (NN), the system can automatically extract the most representative feature combinations from the input data and map these combinations to the fluorine resistance index. The fluorine resistance index is a quantitative indicator used to reflect the tolerance of tea trees under fluorine stress. The higher the value, the stronger the fluorine resistance of the tea trees. Through continuous learning and optimization, the model can accurately predict the performance of tea trees under different fluorine concentrations, providing a scientific basis for the evaluation. The work of the fluorine resistance evaluation unit is not limited to simple feature combination analysis. It also ensures the accuracy and reliability of the evaluation results through continuous model optimization and feature selection. In practical applications, different tea varieties, cultivation conditions and environmental factors will affect the spectral characteristics and chlorophyll fluorescence response index, which requires the evaluation model to be highly adaptable. Through the adaptive ability of the machine learning algorithm, the system can automatically adjust the model parameters according to different input data to adapt to different tea varieties and growth environments. At the same time, the system also has the ability to update and optimize the model in real time. With the input of new data, the system can continuously update its knowledge base and improve the accuracy of fluorine resistance evaluation. This dynamic update mechanism enables the fluorine resistance evaluation unit to not only process current spectral data, but also improve the prediction ability of future data by learning new data patterns. In addition, the fluorine resistance evaluation unit also has a certain degree of interpretation ability. Although machine learning models have significant advantages in dealing with complex nonlinear relationships, their results are often difficult to understand directly. To address this problem, the system introduces explanatory models, such as methods based on feature importance analysis, to help users understand the role of different spectral features and chlorophyll fluorescence response index in fluoride resistance assessment. In this way, the system not only provides accurate evaluation results, but also provides users with in-depth insights into the fluoride resistance of tea trees, helping agricultural managers to better understand the response mechanism of tea trees under fluoride stress and formulate more effective planting and management strategies accordingly.In general, the fluorine resistance evaluation unit plays a vital role in the present invention. It uses machine learning technology to deeply integrate and analyze spectral features and chlorophyll fluorescence response index to generate a quantitative fluorine resistance index to help users evaluate the tolerance of tea trees under fluorine stress. Compared with traditional fluorine resistance evaluation methods, the system of the present invention not only has significant advantages in accuracy and efficiency, but also ensures the accuracy and real-time nature of the evaluation results through the model's adaptive and optimization capabilities.

[0056] Embodiment 2: The hyperspectral imaging instrument is: an airborne hyperspectral sensor or a satellite-borne hyperspectral sensor.

[0057] Specifically, airborne hyperspectral sensors are usually installed on drones or airplanes, and can obtain hyperspectral images of ground tea gardens at a lower flight altitude. Due to the low flight altitude, airborne sensors can capture higher-resolution spectral data, which is especially suitable for small-area tea gardens or the growth status of tea trees in specific areas that require fine monitoring. In this way, the system can record the reflectance changes of tea trees at different wavelengths with high spatial resolution, and then identify the spectral characteristics of tea trees under fluorine stress. On the other hand, spaceborne hyperspectral sensors are installed on satellites and can monitor the spectral information of ground tea gardens over a larger range. Although their spatial resolution is low, their coverage is wide, which is suitable for the overall fluorine resistance assessment of large-area tea gardens or multiple tea gardens. This sensor can perform regular or continuous spectral monitoring of ground tea gardens, provide long-term spectral data sequences, and help identify the spectral change trends of tea trees at different growth stages or under different fluorine stress conditions. The advantage of spaceborne hyperspectral sensors is that they can monitor a large area simultaneously, and can realize cross-regional and cross-time spectral data comparison and analysis, which is of great practical value for those widely distributed tea garden managers. In the present invention, both airborne and satellite-borne hyperspectral sensors can accurately capture the spectral reflectance of tea leaves in different bands. These reflectance data will be processed by the spectral data acquisition and processing unit and transmitted to the spectral feature extraction and analysis unit for in-depth analysis. The combined use of airborne and satellite-borne sensors can achieve a balance between high resolution and wide coverage in space, so that the system can perform both fine local analysis and wide-area overall evaluation, thereby providing a more comprehensive evaluation result for the fluorine resistance of tea trees.

[0058] Embodiment 3: The spectral data acquisition and processing unit calculates the spectral reflectance of the target tea tree area at different wavelengths using the following formula:

[0059]

[0060] Where S(λ) is the spectral reflectance at wavelength λ; L(λ) is the radiation brightness received by the hyperspectral imaging instrument at wavelength λ; d is the distance between the sun and the earth; E0(λ) is the solar spectral irradiance at wavelength λ; θ s is the solar zenith angle; θ v is the observation zenith angle; τ(λ) is the atmospheric optical thickness.

[0061] Specifically, the spectral reflectance S(λ) is an important indicator to measure the reflectivity of tea leaves under the irradiation of light of a specific wavelength. The spectral reflectance of tea leaves is closely related to their physiological state. Especially when subjected to fluorine stress, the spectral reflectance of tea leaves will change significantly in certain specific bands. Therefore, by accurately calculating the spectral reflectance at different wavelengths, the spectral response characteristics of tea trees under fluorine stress can be effectively captured, and the fluorine resistance of tea trees can be scientifically evaluated. The core idea of ​​the formula is to correct the original radiance data L(λ) obtained from the hyperspectral imaging instrument to eliminate the influence of the external environment and atmospheric conditions, so as to obtain the true spectral reflectance. L(λ) in the formula represents the radiance received by the hyperspectral imaging instrument at wavelength λ. This radiance is the superposition result of the light reflected by tea leaves and the light scattered by the atmosphere, so directly using L(λ) to calculate the reflectance may lead to inaccurate results. In order to eliminate these effects, the formula introduces correction parameters such as solar spectral irradiance E0(λ) and the distance between the sun and the earth d. Solar spectral irradiance E0(λ) refers to the energy that the sun radiates directly outside the earth's atmosphere at wavelength λ without the influence of the atmosphere. This parameter provides a standard reference value for each wavelength. By comparing L(λ) with E0(λ), the error caused by changes in solar radiation intensity can be eliminated, thereby more accurately reflecting the actual reflection characteristics of tea leaves. The distance d between the sun and the earth affects the intensity of solar radiation reaching the earth's surface. Since the earth's orbit around the sun is elliptical, the distance between the sun and the earth changes with the seasons, which affects the radiation brightness received by the spectral imaging instrument. The formula further corrects the calculation of radiation brightness by taking this distance change into account, making the calculation of the spectral reflectance of tea leaves in different seasons and sun-earth distances more accurate. The effect of atmospheric conditions on radiation brightness is also reflected in the formula through the atmospheric optical thickness τ(λ) and the solar zenith angle θ s , observation zenith angle θ v The atmospheric optical thickness τ(λ) describes the absorption and scattering of light by the atmosphere. It varies with wavelength and is related to factors such as aerosols, clouds, and water vapor in the atmosphere. When light passes through the atmosphere, light of different wavelengths will be scattered and absorbed by particles and molecules in the atmosphere, thereby reducing the amount of radiation reaching the ground and being reflected back to the hyperspectral imaging instrument. In order to accurately restore the reflective characteristics of tea leaves, the formula is used to correct for the radiation attenuation caused by the atmosphere. Here θs represents the solar zenith angle, θ v Represents the observation zenith angle, which are the angles between the sun's ray and the observation ray and the normal to the earth's surface, respectively. The introduction of the solar zenith angle and the observation zenith angle further corrects the changes in radiation intensity caused by the incident and exit angles of the light, thereby more accurately estimating the spectral reflectance of tea leaves. Through the calculation of this formula, the true spectral reflectance of tea leaves at different wavelengths can be restored, providing high-precision input data for subsequent spectral feature extraction and fluorine resistance evaluation. This calculation process not only takes into account the direct measurement results of the radiation brightness, but also greatly improves the accuracy of the reflectance calculation through corrections of complex environmental and geometric conditions. Especially in the application of tea tree fluorine resistance assessment, this precise spectral reflectance calculation can help the system capture the subtle effects of fluorine stress on tea trees, thereby better understanding and evaluating the tolerance of tea trees under different fluorine exposure conditions.

[0062] Embodiment 4: The spectral feature extraction and analysis unit performs noise suppression and signal enhancement on the calculated spectral reflectance through the following formula to obtain a preprocessing result:

[0063]

[0064] Among them, S′(λ) is the preprocessing result; G(λ,σ) is the Gaussian filter function, σ is the standard deviation; α is the enhancement coefficient based on the adaptive adjustment of the signal-to-noise ratio, and its value range is 0.2 to 1.5; It means to calculate the second derivative of S(λ).

[0065] Specifically, the formula first uses a Gaussian filter function G(λ, σ) to smooth the original spectral reflectance S(λ). Gaussian filtering is a classic noise suppression method that effectively weakens high-frequency noise in the data through convolution operations with spectral data. The intensity of the Gaussian filter is controlled by the standard deviation σ. The larger the σ, the more significant the smoothing effect, but it may lead to loss of spectral details; while a smaller σ retains more details, but the noise suppression effect is relatively weak. Therefore, choosing a suitable σ value is particularly critical for balancing noise suppression and signal retention. In the application of the present invention, through Gaussian filtering, the system can eliminate most of the noise in the tea tree spectral data, making the spectral curve smoother, thereby providing a purer data basis for subsequent analysis. In addition to noise suppression, signal enhancement is also an important part of this formula. The signal enhancement part in the formula introduces the second-order derivative of the spectral reflectance to achieve this. The role of the second-order derivative is to highlight the drastic changes in the spectral curve, such as the peaks and valleys of the spectrum, which often contain important characteristic information reflecting the fluorine resistance of tea trees. By amplifying these drastic changes in the area, the system can better capture the spectral characteristics of tea trees under fluorine stress, thereby improving the accuracy of fluorine resistance assessment. In order to control the intensity of signal enhancement, an enhancement coefficient α is introduced into the formula, and its value range is 0.2 to 1.5. The enhancement coefficient α is adaptively adjusted based on the signal-to-noise ratio, which means that the system will automatically adjust the value of α according to the signal-to-noise ratio of the actual data to ensure that the signal enhancement effect can effectively highlight important features without introducing too much noise or artifacts.

[0066] Example 5: According to the following formula, the fluorine-sensitive band of the target tea tree area is extracted according to the preprocessing results to obtain the fluorine-sensitive feature F(λ i ):

[0067] F(λ i )=α·NDFI(λ i )+β·FSRI(λ i )+γ·FABI(λ i );

[0068] Among them, NDFI(λ i ) is the wavelength λ i FSRI(λ i ) is the wavelength λ i Fluoride stress ratio index; FABI (λ i ) is the wavelength λ i The integral index of the fluorine absorption band at the time of NDFI (λ) is the first weight coefficient; α is the second weight coefficient; γ is the third weight coefficient; i ) is calculated using the following formula:

[0069]

[0070] Among them, λ ref is the reference wavelength not affected by fluorine; λ F is the main characteristic absorption wavelength of fluorine, equal to 680 nanometers; σ F is the fluorine absorption bandwidth, equal to 20 nm; FSRI(λ i ) is calculated using the following formula:

[0071]

[0072] Among them, λ R is the red light band; NIR is the near infrared band; C F is the fluorine content of leaves; k Fis the fluorine concentration response coefficient, equal to 0.5; FABI(λ i ) is calculated using the following formula:

[0073]

[0074] Among them, [λ1, λ2] represents the integral range of the fluorine absorption band, λ1 is the lower limit of the integral range of the fluorine absorption band; λ2 is the upper limit of the integral range of the fluorine absorption band.

[0075] Specifically, the normalized fluorine difference index (NDFI (λ i )) is one of the core parts of this formula, which reveals the effect of fluorine stress on the spectral characteristics of tea leaves by comparing the spectral reflectance of tea leaves at different wavelengths. Specifically, NDFI(λ i ) by comparing the target wavelength λ i With reference wavelength λ ref The spectral reflectance difference of the leaves captured the spectral changes caused by fluorine stress. This change is at the characteristic absorption wavelength of fluorine (λ F ), and the change gradually weakens as it moves away from this band. In order to more accurately reflect this trend, the exponential term of the Gaussian function is introduced into the formula It gives a higher weight to the spectral reflectance near the fluorine absorption wavelength. This processing method effectively enhances the impact of fluorine stress on spectral characteristics, allowing the system to capture subtle spectral differences in the fluorine-sensitive band, thereby improving the accuracy of tea tree fluorine tolerance assessment. In addition to the normalized fluorine difference index, the fluorine stress ratio index (FSRI (λ i )) is also an important component in extracting fluorine-sensitive features. The FSRI index is obtained by measuring the intensity of tea leaves under red light (λ R ) and near infrared (λ NIR ) bands, revealing the profound impact of fluorine stress on the physiological state of tea trees. Fluorine stress often leads to an increase in the red light reflectance of tea leaves and a decrease in the near-infrared reflectance. This change can be intuitively expressed by the FSRI index. A nonlinear correction term (1-exp(-k F ·C F )), where k F represents the fluorine concentration response coefficient, C F is the fluorine content in the leaves. This correction term takes into account the nonlinear effect of fluorine concentration on the spectral response of tea trees, and can more accurately reflect the changes in the spectral characteristics of leaves under different fluorine concentrations. Through this nonlinear adjustment, the system can more sensitively capture the changes in spectral characteristics under different fluorine stress conditions and provide more accurate fluorine-sensitive feature extraction results. In addition, the fluorine absorption band integral index (FABI (λ iThe introduction of FABI further enriches the extraction methods of spectral features. The FABI index integrates the spectral reflectance curve in the fluorine absorption band (λ1 to λ2), comprehensively considering the reflectance characteristics of tea leaves in this band. FABI not only reflects the overall intensity of spectral reflectance through the integral operation, but also introduces the derivative term The rate of change of the spectral curve, that is, the steepness of the curve, is emphasized. This processing method can highlight the areas of drastic changes in the spectrum, which usually correspond to physiological changes or stress responses of the leaves. By amplifying these drastic change areas, the FABI index can effectively capture the significant spectral changes of tea leaves due to fluorine stress, thereby enhancing the system's sensitivity to fluorine stress. By combining the three spectral indices of NDFI, FSRI and FABI, the system extracts the fluorine-sensitive features F(λ i ). These features not only include the changes in the spectral reflectance of the leaves in specific bands, but also comprehensively consider the ratio relationship between different bands, as well as the overall morphology and change rate of the spectral curve. The weight coefficients α, β and γ in the formula further balance the influence of different indexes in feature extraction, so that the final extracted fluorine-sensitive features can fully reflect the physiological state of the tea tree under fluorine stress. This feature extraction method based on multi-spectral index fusion has significant advantages over the traditional single spectral feature analysis method. Traditional methods usually rely only on a single spectral feature and cannot fully capture the complex effects of fluorine stress on tea trees. The present invention, by combining multiple spectral indices, can not only capture the multi-dimensional effects of fluorine stress on spectral features, but also can flexibly respond to complex situations under different tea varieties and different stress conditions through nonlinear adjustment and weight distribution. This fusion of multiple spectral indices greatly improves the sensitivity and accuracy of the system to fluorine stress, ensuring the scientificity and reliability of fluorine resistance assessment.

[0076] Embodiment 6: The first weight coefficient α is calculated using the following formula:

[0077]

[0078] Among them, C F，threshold is the fluorine content threshold, which is equal to 50ppm; k1 is the first adjustment coefficient, which ranges from 8 to 12; the second weight coefficient β is calculated using the following formula:

[0079] β = 1 - exp (-k2 RES);

[0080] Wherein, RES is the spectral resolution; k2 is the second adjustment coefficient, ranging from 0.09 to 0.13; the third weight coefficient γ is calculated using the following formula:

[0081]

[0082] Specifically, the calculation of the first weight coefficient α is based on the fluorine content in tea leaves. Fluorine content is an important indicator for evaluating the fluorine resistance of tea plants. Therefore, the calculation formula of α uses the logistic function, which has the characteristic of nonlinear growth. When the fluorine content approaches or exceeds a certain threshold, the value of the weight coefficient can be rapidly increased, thereby enhancing the system's sensitivity to spectral features under high fluorine content. When the fluorine content C F Close to the set threshold C F，threshold When α increases rapidly, it means that the feature is more important when the fluorine content is high, and the system will pay more attention to the spectral features related to it. When the fluorine content is low, α is close to 0, which weakens the influence of the feature, and the system can automatically ignore the spectral changes that may not be significant at this time. This design enables the system to flexibly respond to the changes in spectral features under different fluorine stress conditions, ensuring the accuracy and reliability of feature extraction. The second weight coefficient β is adjusted according to the spectral resolution. The spectral resolution determines the fineness of the spectral data and directly affects the system's ability to identify and extract spectral features. In order to make full use of high-resolution spectral data, the calculation formula of β adopts an exponential function model, so that when the spectral resolution is high, the value of β is close to 1, thereby giving higher weight to the spectral features under high resolution. This design can give full play to the advantages of high-resolution spectral data in subtle feature identification and improve the system's evaluation ability under high-precision conditions. When the spectral resolution is low, the value of β tends to 0, which means that the system will reduce its reliance on these spectral features and rely on other more reliable feature indicators. This method of dynamically adjusting weights according to spectral resolution ensures that the system can always maintain a high evaluation accuracy under different spectral data qualities. The third weight coefficient γ is based on the wavelength λ i The characteristic absorption wavelength of fluorine F The calculation is based on the relationship between fluorine and fluorine. The characteristic absorption wavelength of fluorine usually corresponds to the band where the spectral response of tea leaves when subjected to fluorine stress is most significant. Therefore, the calculation formula of γ uses a function based on the inverse square of the wavelength difference, so that when the wavelength λ i Close to the characteristic absorption wavelength λ of fluorine F When the wavelength is far from λ, the value of γ is the largest, indicating that the spectral characteristics at these wavelengths play an important role in the evaluation. F When the γ value decreases rapidly, the system will automatically reduce its attention to these bands in this case. This design enables the system to more focus on analyzing the wavelength ranges that are closely related to fluorine stress, improving the pertinence and effectiveness of feature extraction.

[0083] Example 7: The chlorophyll fluorescence response index CFI of the target tea tree area is calculated by the following formula:

[0084]

[0085] Among them, F(740) represents the fluorine-sensitive characteristic at a wavelength of 740 nanometers; F(690) represents the fluorine-sensitive characteristic at a wavelength of 690 nanometers; Φ PSII is the quantum yield of photosystem II; Φ PSII，max is the maximum quantum yield of photosystem II; PAR is photosynthetically active radiation; PAR0 is the reference photosynthetically active radiation; t represents the hour of the day, with a value ranging from 0 to 24; T is the leaf temperature; T opt is the optimal temperature for photosynthesis; k T is the temperature response coefficient; NPQ is the non-photochemical quenching value; NPQ max is the maximum non-photochemical quenching value; Φ PSII Calculated using the following formula:

[0086]

[0087] Among them, F′ m F is the maximum fluorescence output under light adaptation state; s is the steady-state fluorescence yield; NPQ is calculated using the following formula:

[0088]

[0089] Among them, F m is the maximum fluorescence output under dark adaptation condition.

[0090] Specifically, the design of the CFI formula starts with spectral characteristics, especially the difference in reflectance at wavelengths of 740 nanometers and 690 nanometers. These two wavelengths were chosen because they correspond to the areas where chlorophyll fluorescence changes significantly under the red edge effect. The red edge effect refers to the phenomenon that when plants are healthy, chlorophyll absorbs blue and red light (wavelengths of about 400-700 nanometers) while strongly reflecting near-infrared light (wavelengths of about 700-1000 nanometers). This effect is particularly obvious between 740 nanometers and 690 nanometers. Healthy tea leaves have a higher reflectance at 740 nanometers and a relatively low reflectance at 690 nanometers. When tea plants are subjected to fluorine stress, this difference in reflectance changes significantly. Therefore, by calculating the ratio of the reflectance at these two wavelengths, the CFI formula can sensitively capture the slight changes in chlorophyll fluorescence of tea plants under fluorine stress, which is usually an early signal of changes in plant health status. The design of this part is based on the application of spectral analysis technology in plant stress research. Changes in spectral characteristics can directly reflect the physiological response of plants under different stress conditions. In addition, the quantum yield of photosystem II, Φ, is introduced into the CFI formula PSII, this parameter is one of the core indicators for measuring the efficiency of plant photosynthesis. Photosystem II is the main mechanism for absorbing light energy and converting it into chemical energy in plant photosynthesis. By capturing the efficiency of photosystem II in converting light energy in photochemical reactions, Φ PSII It can effectively reflect the photosynthetic efficiency of plants under different light conditions. When the plant is healthy, the quantum yield of photosystem II is close to the maximum value Φ PSII，max , but under fluorine stress conditions, this yield usually decreases because the presence of fluorine affects the function of chlorophyll and thus reduces the efficiency of photosynthesis. The CFI formula can accurately capture the dynamic changes in photosynthesis efficiency by standardizing and comparing the actual quantum yield with the maximum quantum yield of photosystem II. Through this treatment, the CFI formula not only reflects the photosynthesis efficiency of tea trees under normal light conditions, but also reveals the impact of the decreased function of photosystem II on overall photosynthesis under fluorine stress conditions. Photosynthetically active radiation (PAR) is one of the most critical external light sources in the photosynthesis process of plants. The CFI formula dynamically adjusts the CFI value by introducing the ratio of PAR to the reference photosynthetically active radiation (PAR0) to reflect the efficiency of tea trees in utilizing light energy under different light conditions. PAR refers to the part of light radiation that can be used by plants for photosynthesis, and its wavelength range is usually between 400 and 700 nanometers. Unlike total radiation, PAR more accurately reflects the amount of light energy that plants can actually utilize. The formula is passed through , which takes the effect of ambient light intensity on chlorophyll fluorescence into the calculation. This design takes into account the impact of nonlinear changes in light intensity on plant photosynthesis. Under low light intensity conditions, the photosynthesis efficiency increases linearly with the increase of PAR, but when the light intensity reaches a certain level, the photosynthesis of the plant tends to be saturated, and the effect of further increasing PAR on the photosynthesis efficiency is weakened. By incorporating PAR into the calculation of CFI, the formula can dynamically adjust CFI, reflecting the efficiency of tea plants in utilizing light energy under different light conditions, especially under fluorine stress, the effective utilization rate of photosynthetically active radiation may be affected, thereby further reducing the photosynthesis efficiency. The time factor t is also an important part of the CFI calculation that cannot be ignored. The intensity and angle of solar radiation will change at different times of the day, and these changes directly affect the photosynthesis dynamics of plants. The CFI formula introduces a time-based sinusoidal function The periodic changes in photosynthesis efficiency during the day are simulated. The changes in solar radiation intensity during the day roughly present a sinusoidal waveform. The photosynthesis efficiency is usually higher in the morning and afternoon, while at noon, due to excessive light and high temperature, the photosynthesis efficiency decreases. Through this time function, the formula can more accurately reflect the fluctuations in tea tree photosynthesis in different time periods, especially under high fluorine stress conditions, the response of tea trees to light changes may be more obvious. This design not only makes CFI more temporal and dynamic, but also reveals the impact of fluorine stress on the photosynthesis efficiency of plants in different time periods. Leaf temperature T is one of the important parameters affecting CFI. The influence of temperature on photosynthesis is multifaceted, especially in the activity of photosynthetic enzymes and stomatal conductance. The CFI formula introduces an exponential function based on temperature difference The effect of temperature change on photosynthesis efficiency was simulated. The optimal temperature for photosynthesis is T opt It means that the photosynthesis efficiency of plants is the highest at this temperature. When the leaf temperature is close to the optimal temperature, the photosynthesis efficiency reaches the highest and the CFI value increases accordingly; when the temperature deviates from the optimal value, the photosynthesis efficiency decreases and the CFI value decreases accordingly. The effect of temperature on plant photosynthesis is mainly reflected in two aspects: first, temperature affects the activity of photosynthetic enzymes. When the temperature is too high or too low, the activity of the enzymes decreases, resulting in a decrease in photosynthesis efficiency; second, temperature affects the opening and closing of stomata, thereby affecting the absorption of carbon dioxide. When the leaf temperature is too high, the stomata close to reduce water loss, but this also limits the entry of carbon dioxide, further inhibiting photosynthesis. Through this temperature response function, the CFI formula can dynamically reflect the impact of temperature changes on tea tree photosynthesis, especially under fluorine stress, temperature changes may aggravate the decline in photosynthesis efficiency, thereby further affecting the health of tea trees. Non-photochemical quenching (NPQ) is another key physiological parameter, which reflects the ability of plants to release excess light energy through heat dissipation pathways when absorbing too much light energy. The higher the NPQ value, the more effectively the plant can prevent photo-oxidative damage caused by excess light energy. This ratio incorporates NPQ into the calculation of CFI. This processing method allows CFI to reflect how tea trees protect themselves from photooxidative damage by increasing heat dissipation under high light intensity or fluorine stress conditions. When fluorine stress affects plant photosynthesis, NPQ usually rises to increase heat dissipation and protect leaves from damage. Therefore, the NPQ part of the CFI formula not only reflects the adaptability of tea trees in the face of changes in light intensity, but also reveals the impact of fluorine stress on the plant's photosynthesis protection mechanism.

[0091] Embodiment 8: The preprocessing result is transformed into a spectrum integral by the following formula:

[0092]

[0093] Where I(a, b) is the result of spectrum integration transformation; [a, b] is the integration interval; λ c is the central wavelength; σ c is the integration window width; use the following formula to perform multi-scale wavelet decomposition and obtain the wavelet decomposition result:

[0094]

[0095] Among them, W ψ [S′](λ, a) is the wavelet decomposition result; ψ(u) is the Morlet wavelet mother function; a is the scale parameter; u is the integral variable; the wavelet decomposition result and the chlorophyll fluorescence response index are subjected to nonlinear feature fusion based on machine learning. When the feature fusion result is obtained, nonlinearity is introduced through the sigmoid function. The weight matrix of the wavelet decomposition result, the weight matrix and bias of the chlorophyll fluorescence response index are predefined. A single-layer neural network is used to perform nonlinear feature fusion on the wavelet decomposition result and the chlorophyll fluorescence response index to obtain the feature fusion result Φ(λ).

[0096] Specifically, spectral integral transform is a technique for smoothing and enhancing spectral data. Its core idea is to extract the overall spectral characteristics within a specific band by weighted integration of the spectral reflectance S′(λ). In the formula, the integral result I(a, b) is obtained by integrating the preprocessed spectral data S′(λ) in the specified integral interval [a, b]. In order to make the integral result more representative, the formula introduces a Gaussian weight function where λ c is the central wavelength, σ c is the width of the integration window. This Gaussian weight function ensures that the spectral data farther away from the central wavelength has a smaller proportion in the integration, thereby focusing on the analysis of spectral features near the central wavelength. This method can effectively filter out the noise in the spectrum while retaining the main features within the band, providing a smooth and representative basic data for subsequent feature extraction. Next, the system uses multi-scale wavelet decomposition to further analyze the pre-processed spectral data. Wavelet decomposition is a powerful signal processing tool that can simultaneously analyze the characteristics of the signal in both the time (or space) and frequency domains. Through the formula W ψ[S′](λ, a), the system performs multi-scale wavelet decomposition on the pre-processed spectral data S′(λ), in which Morlet wavelet is used as the mother wavelet function ψ(u). In wavelet decomposition, the scale parameter a controls the resolution and range of the analysis: when a is small, the decomposition focuses more on capturing the high-frequency characteristics of the signal; when a is large, the decomposition tends to capture the low-frequency trend of the signal. Through multi-scale analysis, the system can extract the features of spectral data at different scales, thereby capturing the spectral changes of tea leaves in different spatial and frequency ranges. This multi-scale feature extraction method can effectively identify local detail changes and overall trends in the spectrum, providing richer information for subsequent fluorine resistance evaluation. After obtaining the wavelet decomposition results, the system further performs nonlinear feature fusion with the chlorophyll fluorescence response index. The core of this step is to achieve the feature fusion of wavelet decomposition results and chlorophyll fluorescence response index through a simple single-layer neural network. The nonlinear feature fusion in the neural network is introduced through the sigmoid function, which can smooth the input data and make the output features more continuous and separable. During the fusion process, the system predefines the weight matrix of the wavelet decomposition results and the weight matrix of the chlorophyll fluorescence response index, and introduces a bias term. These parameters are optimized through the training process so that the complex nonlinear relationship between the two can be better captured when fusing features. Through the feature fusion of this neural network, the system can not only combine the respective advantages of wavelet decomposition and chlorophyll fluorescence response index, but also explore the potential correlation between the two through nonlinear processing. The final feature fusion result Φ(λ) is a composite indicator that integrates multiple feature information. It establishes a strong connection between spectral data and physiological indicators, providing a more comprehensive and discriminating feature expression for subsequent fluorine resistance evaluation.

[0097] Embodiment 9: The fluorine resistance evaluation unit calculates the fluorine resistance index Y according to the feature fusion result by the following formula:

[0098]

[0099] Among them, FTI is the uncorrected fluorine resistance index; PRI m is the modified photochemical vegetation index; PRI0 is the standard photochemical vegetation index; σ PRI is the sensitivity of the photochemical vegetation index, which is equal to 0.1; FTI is calculated using the following formula:

[0100]

[0101] Among them, A F (λ) is the absorption coefficient function of fluorine at different wavelengths; C is the red edge position response coefficient; PRI mCalculated using the following formula:

[0102]

[0103] Specifically, the core idea of ​​FTI is to use the spectral feature Φ(λ) and the fluorine absorption feature A F (λ) to preliminarily evaluate the fluorine resistance of tea trees. In the formula, the spectral data Φ(λ) is integrated from 400 nanometers to 900 nanometers. This band covers the range of visible light and near-infrared light, and is the most critical spectral region for plant photosynthesis. By integrating the spectral features within this band, the formula captures the overall spectral response of tea leaves under fluorine stress. This integration method makes the evaluation not limited to a single wavelength or a narrow band, but comprehensively considers the characteristic changes in the entire spectral range, thereby better reflecting the comprehensive spectral response of tea trees under fluorine stress conditions. After integration, the spectral feature Φ(λ) is further compared with the fluorine absorption feature A F (λ), which is a function that describes the absorption characteristics of fluorine at different wavelengths. Through this operation, the formula combines the spectral characteristics of tea leaves with the specific absorption characteristics of fluorine stress. This combination helps to identify the characteristic spectral changes exhibited by tea plants under fluorine stress. For example, strong absorption of fluorine in specific bands may cause a decrease in reflectivity in these bands. By integrating this spectral feature with the absorption feature, the system can more accurately assess the impact of this change on the fluorine resistance of tea plants. When calculating FTI, the formula also introduces a red edge position response coefficient C and an integral I (680, 740) between specific bands. The red edge is a sharply rising area between red light and near-infrared light in the spectral reflectance curve of plants, usually between 680 and 750 nanometers. The position and morphology of the red edge are highly sensitive to the health of plants, especially when the plant is under stress, the position of the red edge may move. By introducing the red edge position response coefficient C, the formula can more accurately capture the impact of this spectral feature change on the assessment of fluorine resistance. The I(680, 740) integral focuses on the key band of the red edge. By calculating the integral of the spectral characteristics within this band, the formula further strengthens the focus on the red edge effect, making FTI more accurate in reflecting the red edge effect. The uncorrected fluorine tolerance index FTI, as the core part of the formula, can already preliminarily reflect the spectral response characteristics of tea trees under fluorine stress, but in order to further improve the accuracy of the evaluation, the formula also introduces the photochemical vegetation index (PRI) as a correction factor. PRI is an important indicator that reflects the efficiency of plant photosynthesis. It reveals the activity of photochemical reactions of plants under different environmental conditions by comparing the spectral reflectance at specific wavelengths. The corrected photochemical vegetation index PRI in the formula mBy combining PRI with FTI, a more sensitive indicator is formed to correct the performance of FTI under different photochemical conditions. The photochemical vegetation index is usually calculated by the difference in reflectance at two wavelengths, 531 nanometers and 570 nanometers. These two wavelengths are located in the green and yellow light regions, respectively. 531 nanometers corresponds to the sensitive band of photosystem II, while 570 nanometers is the lower absorption area of ​​photosynthetic pigments. The difference in reflectance between these two bands reflects the dynamic changes of plant photosynthetic pigments. Especially under fluorine stress, the efficiency of photosynthesis may be affected, thereby changing the photochemical vegetation index. By introducing PRI m The formula can capture the changes in the spectral characteristics of tea trees under different photochemical states, thereby correcting the FTI, so that the final fluorine resistance index Y more accurately reflects the actual fluorine resistance of tea trees. In addition, the Gaussian function part of the formula further corrects the impact of the photochemical vegetation index on the fluorine resistance assessment. Gaussian function σ PRI is the sensitivity parameter of the photochemical vegetation index, and its value is set to 0.1, indicating that changes in the photochemical vegetation index within a certain range have a significant impact on the fluorine resistance assessment. Through this Gaussian function, the formula can accurately adjust the fluorine resistance index correction caused by changes in the photochemical state, so that the final Y value can better reflect the actual fluorine resistance of the tea tree. The design of this part ensures that the nonlinear relationship between the photochemical vegetation index and the fluorine resistance index can be reasonably captured, thereby improving the accuracy of the assessment. Finally, the formula further strengthens the impact of the red edge effect on the fluorine resistance index Y by introducing the integral ratio between the two bands. Specifically, the formula calculates the ratio of the integral of the spectral features in the 700 to 750 nanometer band to the integral of the spectral features in the 650 to 700 nanometer band. These two bands are located in the rising and stable sections of the red edge, respectively. By comparing their integral results, the formula can more accurately capture the changes in the spectral response of the tea tree in the red edge area. The calculation of this ratio strengthens the role of the red edge effect in fluoride resistance assessment, enabling the formula to better identify changes in the spectral characteristics of tea trees under fluoride stress, thereby providing a more recognizable correction term for the final fluoride resistance index.

[0104] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that these specific embodiments are only illustrative, and those skilled in the art may omit, replace, and change the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, merging the above method steps so as to perform substantially the same functions in substantially the same manner to achieve substantially the same results is within the scope of the present invention. Therefore, the scope of the present invention is limited only by the appended claims.

Claims

1. A tea tree fluoride resistance evaluation system based on spectral analysis and machine learning, characterized in that: The system comprises: a spectral data acquisition and processing unit, a spectral feature extraction and analysis unit and a fluorine resistance evaluation unit; the spectral data acquisition and processing unit is used to collect spectral information of a target tea tree area through a hyperspectral imaging instrument, and calculate the spectral reflectance of the target tea tree area at different wavelengths; the spectral feature extraction and analysis unit is used to perform noise suppression and signal enhancement on the calculated spectral reflectance to obtain a preprocessing result, and based on the preprocessing result, perform feature extraction on the fluorine-sensitive band of the target tea tree area to obtain a fluorine-sensitive feature, and calculate the chlorophyll fluorescence response index of the target tea tree area according to the wavelength of chlorophyll; after performing spectral integral transformation on the preprocessing result, perform multi-scale wavelet decomposition to obtain a wavelet decomposition result; the wavelet decomposition result and the chlorophyll fluorescence response index are fused using nonlinear features based on machine learning to obtain a feature fusion result; the fluorine resistance evaluation unit is used to calculate the fluorine resistance index based on the feature fusion result; The spectral data acquisition and processing unit calculates the spectral reflectance of the target tea tree area at different wavelengths using the following formula: ； in, The wavelength Spectral reflectance at ; The wavelength The brightness of the radiation received by the hyperspectral imaging instrument at ; is the distance between the Sun and the Earth; The wavelength The solar spectral irradiance at ; is the solar zenith angle; To observe the zenith angle; is the atmospheric optical thickness; The spectral feature extraction and analysis unit performs noise suppression and signal enhancement on the calculated spectral reflectance through the following formula to obtain the preprocessing result: ； in, is the preprocessing result; is the Gaussian filter function, is the standard deviation; is the enhancement factor adaptively adjusted based on the signal-to-noise ratio, ranging from 0.2 to 1.5; Representation calculation The second derivative of The following formula is used to extract the fluorine-sensitive band of the target tea tree area according to the preprocessing results to obtain the fluorine-sensitive feature : ； in, The wavelength is Normalized fluorine difference index at ; The wavelength is Fluoride stress ratio index at 100 ℃; The wavelength is The integral index of the fluorine absorption band at ; is the first weight coefficient; is the second weight coefficient; is the third weight coefficient; Use the following formula to calculate: ； in, is the reference wavelength not affected by fluorine; It is the main characteristic absorption wavelength of fluorine element, which is equal to 680 nanometers; is the fluorine absorption bandwidth, equal to 20 nm; Use the following formula to calculate: ； in, It is the red light band; It is near infrared band; is the leaf fluorine content; is the fluorine concentration response coefficient, equal to 0.5; Use the following formula to calculate: ； in, represents the integral interval of the fluorine absorption band, is the lower limit of the integral interval of the fluorine absorption band; is the upper integral interval of the fluorine absorption band; The first weight coefficient Use the following formula to calculate: ； in, is the fluorine content threshold, equal to 50ppm; is the first adjustment coefficient, ranging from 8 to 12; the second weight coefficient Use the following formula to calculate: ； Where RES is the spectral resolution; is the second adjustment coefficient, ranging from 0.09 to 0.13; the third weight coefficient Use the following formula to calculate: ； The chlorophyll fluorescence response index of the target tea tree area is calculated by the following formula: : ； in, Indicates the fluorine sensitive characteristics at a wavelength of 740 nanometers; Indicates the fluorine sensitive characteristics at a wavelength of 690 nanometers; is the quantum yield of photosystem II; is the maximum quantum yield of photosystem II; is photosynthetically active radiation; is the reference photosynthetically active radiation; Represents the hour of the day, an integer from 0 to 24; is the leaf temperature; The optimal temperature for photosynthesis; is the temperature response coefficient; is the non-photochemical quenching value; is the maximum non-photochemical quenching value; Calculated using the following formula: ； in, is the maximum fluorescence output under light adaptation state; is the steady-state fluorescence yield; Use the following formula to calculate: ； in, is the maximum fluorescence output under dark adaptation state; The preprocessing results are transformed into spectral integrals using the following formula: ； in, is the result of spectral integration transformation; is the integration interval; is the central wavelength; is the integration window width; use the following formula to perform multi-scale wavelet decomposition and obtain the wavelet decomposition result: ； in, is the wavelet decomposition result; is the Morlet wavelet mother function; is the scale parameter; is the integral variable; nonlinear feature fusion based on machine learning is used for the wavelet decomposition results and the chlorophyll fluorescence response index. When the feature fusion result is obtained, nonlinearity is introduced through the sigmoid function. The weight matrix of the wavelet decomposition result, the weight matrix and bias of the chlorophyll fluorescence response index are predefined. A single-layer neural network is used to perform nonlinear feature fusion on the wavelet decomposition result and the chlorophyll fluorescence response index to obtain the feature fusion result. ; The fluorine resistance evaluation unit calculates the fluorine resistance index according to the feature fusion result by the following formula: : ； in, is the uncorrected fluorine resistance index; is the modified photochemical vegetation index; is the standard photochemical vegetation index; is the sensitivity of the photochemical vegetation index, which is equal to 0.1; Use the following formula to calculate: ； in, is the absorption coefficient function of fluorine at different wavelengths; is the red edge position response coefficient; Calculated using the following formula: 。 2. The tea tree fluorine resistance evaluation system based on spectral analysis and machine learning as claimed in claim 1, characterized in that: The hyperspectral imaging instrument is: an airborne hyperspectral sensor or a satellite-borne hyperspectral sensor.

Images