A method and system for diagnosing multiple nutrients in tea plants based on transmission spectroscopy

By setting registration sampling positions in the vein-avoiding areas of tea leaves and performing optical density spectrum correction, the problems of spatial registration and structural interference in the nutrient diagnosis of tea leaf transmission spectra were solved, and the positional consistency and accuracy of multi-nutrient diagnosis results were achieved.

CN123111979APending Publication Date: 2026-11-17TEA RESEARCH INSTITUTE CHINESE ACADEMY OF AGRICULTURAL SCIENCES
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611626883.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing methods for nutrient diagnosis of tea leaves by transmission spectroscopy have several drawbacks. These include a lack of spatial registration at different detection locations, the mixing of nutrient responses with differences in leaf structure, coupling interference between leaf water content changes and tissue scattering on spectral absorption amplitude and baseline morphology, and difficulty in distinguishing the contributions of each nutrient response when multiple nutrient-related response bands overlap.

Method used

Registration sampling positions were set in the vein-avoiding area of ​​tea leaves to obtain dark reference spectrum, empty light path reference spectrum and transmission spectrum of leaves at each position. These were converted into optical density spectrum and structural interference was corrected based on water absorption reference band and scattering reference band. Nutrient response vector was constructed and multi-nutrient diagnostic values ​​were calculated through response band overlap constraint and position consistency constraint.

Benefits of technology

It achieves spatial correspondence between the transmission spectrum detection position and the optical reference, reduces the interference of leaf vein coverage and longitudinal position mismatch in nutrient spectral response judgment, maintains the positional consistency of multi-nutrient diagnostic results, reduces the mixing of structural optical changes with nutrient response, and improves the accuracy of multi-nutrient calculation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123111979A_ABST
    Figure CN123111979A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for multi-nutrient diagnosis of tea plants based on transmission spectroscopy, belonging to the field of spectral detection technology. It includes setting registration sampling positions in the vein-avoiding area of ​​tea leaves to acquire dark reference spectra, empty optical path reference spectra, and transmission spectra of leaves at various positions. The leaf transmission spectra are converted into optical density spectra, and structural interference is corrected based on water absorption reference bands and scattering reference bands. Nutrient response vectors are constructed from the corrected optical density spectra, and multi-nutrient diagnostic values ​​are calculated according to response band overlap constraints and position consistency constraints. This invention, through registration transmission sampling, optical density structural interference correction, and joint calculation of response band overlap and position consistency, forms a continuous and traceable data chain of optical acquisition positions, reference spectra, spectral responses, and the multi-nutrient diagnosis process of tea leaves, enabling multi-nutrient diagnosis results to be established on specific transmission spectral measurements and independent calibration relationships.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of spectral detection technology, specifically to a method and system for diagnosing multiple nutrients in tea plants based on transmission spectroscopy. Background Technology

[0002] The content of mineral nutrients such as nitrogen, phosphorus, and potassium in tea leaves is closely related to leaf physiological state, growth level, and cultivation management. Current nutrient detection methods typically include laboratory assays such as digestion, colorimetry, and elemental analysis. These methods provide relatively direct chemical results, but usually involve sampling, pretreatment, and analysis processes. With the development of spectroscopic detection technology, characterizing the internal composition of leaves by utilizing their absorption and transmission responses to incident light of different wavelengths has become a promising technical direction for plant leaf component detection. Transmission spectroscopy, in particular, allows detection light to pass through leaf tissue and be received. The resulting spectrum simultaneously contains information on leaf pigments, water content, tissue structure, and optical responses related to nutrient changes, thus providing a basis for leaf nutrient detection.

[0003] However, tea leaves are not homogeneous optical media. The midrib, primary lateral veins, leaf thickness, and the structure of the upper and lower epidermis and cell tissues all alter the light propagation path within the leaf. In existing transmission spectral nutrient detection processes, if different sampling locations lack a corresponding relationship along the leaf length and vein distribution, the spectral differences obtained from different locations on the same leaf may simultaneously reflect nutrient differences and structural differences, resulting in a lack of a stable spatial basis for comparison between the detected objects. On the other hand, leaf water absorption and tissue scattering can respectively alter the absorption amplitude of spectral bands and the morphology of the transmission baseline, easily superimposed with nutrient-related local spectral responses, leading to different spectral amplitudes or baselines for the same nutrient state at different locations. For simultaneous diagnosis of multiple nutrients, there may be band overlap in the associated response intervals of different nutrients. If overlapping bands are directly attributed to a single nutrient, it is difficult to distinguish the contribution of different nutrients to the same spectral response. If only the spectral response of a single location is used for calculation, it is easily affected by local tissue structure and spatial heterogeneity. Therefore, how to reduce the interference of sampling location, water absorption and tissue scattering on nutrient response while maintaining the traceability of the physical detection process of transmission spectroscopy, and how to form multi-nutrient detection results with location consistency when response bands overlap, still needs further research. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is that existing methods for nutrient diagnosis of tea leaves by transmission spectroscopy have the following problems: lack of spatial registration at different detection locations leads to differences in leaf structure mixing into nutrient responses; changes in leaf water content and tissue scattering cause coupling interference to spectral absorption amplitude and baseline morphology; when multiple nutrient-related response bands overlap, it is difficult to distinguish the contributions of each nutrient response; and how to use transmission spectra from different registration locations of the same leaf to form multiple nutrient diagnostic values ​​with positional consistency.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a multi-nutrient diagnosis method for tea trees based on transmission spectroscopy, which includes setting registration sampling positions in the vein-avoiding area of ​​tea leaves to obtain dark reference spectrum, empty light path reference spectrum and transmission spectrum of leaves at each position.

[0007] The transmission spectrum of the leaf is converted into an optical density spectrum, and structural interference is corrected based on the water absorption reference band and the scattering reference band.

[0008] Nutrient response vectors are constructed from the calibrated optical density spectrum, and multi-nutrient diagnostic values ​​are calculated based on response band overlap constraints and position consistency constraints.

[0009] As a preferred embodiment of the multi-nutrient diagnosis method for tea trees based on transmission spectroscopy described in this invention, the step of setting registration sampling positions in the vein-avoiding area of ​​the tea leaf includes: establishing a leaf length reference axis with the extension direction of the main vein of the tea leaf; determining paired candidate positions in the leaf mesophyll area on both sides of the main vein according to the same projection position along the leaf length reference axis; eliminating candidate positions where the measurement spot covers the first-order lateral vein; and determining the remaining paired candidate positions as registration sampling positions; and assigning position group identifiers to the same position group based on the same projection position along the leaf length reference axis.

[0010] As a preferred embodiment of the multi-nutrient diagnosis method for tea trees based on transmission spectroscopy described in this invention, the acquisition of the dark reference spectrum, the empty optical path reference spectrum, and the transmission spectrum of leaves at each location includes: keeping the light source spectrum, incident light path, receiving light path, and detection conditions unchanged; blocking the incident light to obtain the dark reference spectrum; and removing tea leaves to obtain the empty optical path reference spectrum. The measurement spot is then covered at each registration sampling position sequentially to acquire the leaf transmission spectrum. The dark reference spectrum, the empty optical path reference spectrum, and the leaf transmission spectrum are aligned with sampling points of the same wavelength and associated with position group identifiers to form a registration transmission spectrum group.

[0011] As a preferred embodiment of the multi-nutrient diagnosis method for tea trees based on transmission spectroscopy described in this invention, the step of converting the leaf transmission spectrum into an optical density spectrum includes: for each wavelength sampling point of the registered transmission spectrum group, subtracting the dark reference intensity from the leaf transmission intensity to obtain the net transmission intensity, and subtracting the dark reference intensity from the empty optical path reference intensity to obtain the net reference intensity. The ratio of the net transmission intensity to the net reference intensity is determined as the corrected transmittance, and a negative logarithmic transformation is performed on the corrected transmittance to obtain the optical density value. The optical density values ​​are arranged according to the wavelength sampling points to form an optical density spectrum corresponding to the registration sampling position and position group identifier.

[0012] As a preferred embodiment of the multi-nutrient diagnosis method for tea trees based on transmission spectroscopy described in this invention, the method for correcting structural interference based on water absorption reference bands and scattering reference bands includes separating the water absorption reference band from the target nutrient-related response band used to construct the nutrient response vector, and ensuring that the scattering reference band avoids the target nutrient-related response peak. For the optical density spectrum at each registration sampling location, a local baseline is formed within the water absorption reference band by connecting the two ends of the band, and the band area of ​​the spectral line relative to the local baseline is calculated. The slope of the spectral line baseline is calculated within the scattering reference band. A moisture reference value is formed using the median of the band area at all registration sampling locations of the same tea leaf, and a scattering reference value is formed using the median of the spectral line baseline slope. An absorption amplitude correction factor and a baseline morphology correction factor are formed based on the deviation of the band area relative to the moisture reference value and the deviation of the spectral line baseline slope relative to the scattering reference value at each registration sampling location, respectively. The absorption amplitude correction factor is used to scale the spectral band amplitude of the corresponding optical density spectrum, and the baseline morphology correction factor is used to compensate for the tilted baseline of the scale-corrected optical density spectrum, resulting in a corrected optical density spectrum that retains the location group identifier. The water absorption reference band characterizes the changes in absorption amplitude caused by leaf water content, while the scattering reference band characterizes the changes in transmission baseline caused by leaf tissue structure.

[0013] As a preferred embodiment of the multi-nutrient diagnosis method for tea trees based on transmission spectroscopy described in this invention, the construction of the nutrient response vector from the calibrated optical density spectrum includes: independently measuring the nutrient content of calibrated leaf samples; dividing the local absorption peaks and their adjacent continuous wavelength ranges in the calibrated optical density spectrum into candidate response bands; and calculating the integral response of the calibrated leaf samples for each candidate response band. The calibrated leaf samples are sorted according to the independently measured nutrient content of each target nutrient, and candidate response bands with consistent integral response direction in each position group are retained as corresponding target nutrient-related response bands. When there are overlapping bands in the target nutrient-related response bands of two or more target nutrients, the overlapping bands are recorded as shared overlapping bands, and non-overlapping shoulder bands corresponding to each target nutrient and not coinciding with other target nutrient-related response peaks are determined on both sides of the shared overlapping band. For each position group of the tea leaves to be diagnosed, the integral response of each target nutrient-related response band relative to the local baselines at both ends is calculated, and the integral response of the shared overlapping band and the shoulder band response of the corresponding non-overlapping shoulder band relative to the local baseline are also calculated. The integral response quantities corresponding to the same target nutrient, the integral response quantities of the shared overlapping zone, and the shoulder band response quantities are combined in a fixed order to form the location nutrient response components. Then, all location nutrient response components are combined in a fixed order according to the target nutrient to form a nutrient response vector that retains the location group identifier. The correspondence between the target nutrient associated response band, the shared overlapping zone, and the non-overlapping shoulder band is recorded as the response band overlap relationship.

[0014] As a preferred embodiment of the multi-nutrient diagnosis method for tea trees based on transmission spectroscopy described in this invention, the calculation of multi-nutrient diagnostic values ​​according to response band overlap constraints and position consistency constraints includes: using the nutrient response vector of calibrated leaf samples and independent nutrient content measurements, determining the response contribution relationship of each target nutrient to the target nutrient associated response band, shared overlapping band, and non-overlapping shoulder band, and forming a nutrient response contribution table according to the correspondence order of target nutrients and response bands. Based on the response band overlap relationship, for each position group of tea leaves to be diagnosed, the position nutrient response component is represented as a combination of the response contributions of each target nutrient using the nutrient response contribution table. The reconstructed response of the shared overlapping band is composed of the response contributions of each target nutrient corresponding to the shared overlapping band, while the reconstructed response of the non-overlapping shoulder band is composed only of the response contributions of the target nutrient corresponding to the non-overlapping shoulder band, thereby forming a response band overlap constraint. For registered sampling positions located on both sides of the main vein within the same location group, the same set of target nutrient values ​​is shared. The positional residuals between the nutrient response components and the corresponding reconstructed responses at each of the two locations are calculated, thus forming a positional consistency constraint. The reconstruction deviations of the shared overlapping zone, the reconstruction deviations of the non-overlapping shoulder zone, and the positional residuals are normalized and summed according to the variation range of the corresponding response components in the calibrated leaf samples. The target nutrient value with the smallest total deviation is selected under the condition that all target nutrient values ​​are non-negative. The median of the target nutrient values ​​obtained from all location groups is taken for each target nutrient, forming a multi-nutrient diagnostic value.

[0015] As a preferred embodiment of the tea tree multinutrient diagnostic system based on transmission spectroscopy described in this invention, it includes a vein-avoidance registration transmission module, a dual-reference spectrum correction module, and an overlapping constraint nutrient solution module.

[0016] The vein-avoidance registration and transmission module is used to set registration sampling positions in the vein-avoidance area of ​​tea leaves to obtain dark reference spectrum, empty light path reference spectrum and transmission spectrum of leaves at each position.

[0017] The dual-reference spectrum correction module is used to convert the leaf transmission spectrum into an optical density spectrum and correct structural interference based on the water absorption reference band and the scattering reference band.

[0018] The overlapping constraint nutrient solution module is used to construct a nutrient response vector from the corrected optical density spectrum and to solve multiple nutrient diagnostic values ​​according to the overlapping constraint and position consistency constraint of the response band.

[0019] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement a method for multi-nutrient diagnosis of tea plants based on transmission spectroscopy.

[0020] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a multi-nutrient diagnostic method for tea plants based on transmission spectroscopy.

[0021] The beneficial effects of this invention are:

[0022] By establishing paired candidate positions on both sides of the main vein according to the same projection position of the leaf length reference axis, eliminating the position covered by the first-level lateral vein, and registering the dark reference spectrum, the empty optical path reference spectrum, and the leaf transmission spectrum with the position group identifier, the spatial correspondence between the transmission spectrum detection position and the optical reference is realized. This allows the spectra on both sides of the same position group to be compared on approximately the same longitudinal leaf tissue area, thereby reducing the interference caused by mismatch of leaf vein coverage and longitudinal position in subsequent nutrient spectrum response judgment.

[0023] By converting the leaf transmission intensity into an optical density spectrum using dark reference and empty optical path reference, and constructing internal references for the same leaf using the area of ​​the water absorption reference band and the baseline slope of the scattering reference band, it is possible to distinguish and correct the changes in absorption amplitude caused by leaf water content and the changes in transmission baseline caused by tissue structure. This allows the target nutrient-related response region to maintain a more consistent absorption scale and baseline morphology between different registration positions, thereby reducing the mixing of structural optical changes with nutrient response.

[0024] By determining the target nutrient-related response band through independent nutrient measurement of calibrated samples, and retaining the shared overlapping band and corresponding non-overlapping shoulder band when multiple nutrient response bands overlap, and then making the two sides of the main vein of the same position group share a set of non-negative target nutrient unsolved values, the joint constraint of the contribution of multiple nutrients and the contribution of single nutrient shoulder band in the overlapping band is realized. This makes the calculation of multiple nutrients simultaneously limited by the wavelength response structure and the spatial consistency of the leaf, thereby reducing the possibility of overlapping responses substituting for each other and a single local position dominating the diagnostic results. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is an overall flowchart of a multi-nutrient diagnosis method for tea trees based on transmission spectroscopy, provided in Embodiment 1 of the present invention. Detailed Implementation

[0027] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0028] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for diagnosing multiple nutrients in tea plants based on transmission spectroscopy is provided, comprising:

[0029] S1: Set up registration sampling positions in the vein-avoiding area of ​​tea leaves to obtain dark reference spectrum, empty light path reference spectrum and transmission spectrum of leaves at each position.

[0030] A leaf length reference axis was established along the direction of the midrib of the tea leaf. Paired candidate positions were determined in the leaf mesophyll area on both sides of the midrib according to the same projection position along the leaf length reference axis. Candidate positions that covered the first-order lateral veins by the measurement spot were eliminated, and the remaining paired candidate positions were determined as registration sampling positions. Position group labels were assigned to the same position group according to the same projection position along the leaf length reference axis.

[0031] Furthermore, during the positioning process, the tea leaves to be tested are naturally flattened while ensuring the leaf tissue is not stretched. The center of the midrib near the junction of the petiole and the leaf body is used as the starting reference for the leaf length direction, and the leaf length reference axis is determined by the direction in which the midrib extends to the leaf tip. The normalized position along the leaf length reference axis is then limited to the leaf length... Within this range, the impact of excessively thick midribs near the petiole and insufficient effective transmission area at the leaf tip on sampling can be reduced. This embodiment can be set... There are 1 initial position group, and their normalized vertical positions are respectively: , , , and ,in The leaf length is used. The number and distribution of locations are preset values ​​in the engineering, based on the principle of covering the main mesophyll area in the middle of the leaf without causing significant spatial overlap between adjacent measurement spots. This setting ensures that different location groups have clear longitudinal spatial representativeness.

[0032] Furthermore, considering that there may be local differences in leaf mesophyll thickness and tissue structure on both sides of the midrib, and that if the two comparison positions are offset in the leaf length direction, longitudinal tissue differences will be mixed into the comparison results of the two sides, this embodiment calculates the normalized projection deviation along the leaf length reference axis for each pair of candidate positions, and uses this deviation to limit the degree of longitudinal registration of the two candidate positions within the same position group:

[0033]

[0034] in, Indicates the first The normalized projection deviation of the candidate positions on both sides of the main vein within each position group is obtained by dividing the difference in the projection distance between the centers of the two candidate positions on the leaf length reference axis by the leaf length, and is a dimensionless quantity. Indicates the first The center coordinates of the candidate positions located on one side of the main vein in each position group are directly determined by the sampling position on the blade surface. Indicates the first The center coordinates of the candidate positions located on the other side of the main vein in each position group are directly determined by the sampling position on the blade surface. The origin of the coordinate system represents the reference axis of leaf length. In this embodiment, the center of the midrib is taken near the junction of the petiole and the main body of the leaf blade. This represents the unit direction vector of the leaf length reference axis along the midrib towards the leaf tip, with a magnitude of . . The leaf length of the tea leaf under test, measured along the leaf length reference axis, is determined by the axial distance from the beginning of the leaf outline to the leaf tip, and is expressed in units of 1. . Indicates the position group number, in this embodiment . , These represent the two sides of the main vein, used only to distinguish paired sampling locations, and do not indicate the positive or negative value relationship.

[0035] In this embodiment, the registration determination boundary is set as follows: This boundary stems from limiting the permissible axial offset to the blade length. Engineering sampling constraints within a certain range. For lengths of... The blades, with corresponding axial offsets not exceeding approximately This setup avoids paired positions falling into significantly different leaf tissue segments due to longitudinal misalignment, while retaining the ability to normalize registration for leaves of different sizes. (The formula is obtained...) It is only used to determine whether the positions in the same group meet the axial registration requirements, and is not used as the subsequent nutrient response.

[0036] Furthermore, a near-circular light spot is preferably used for measurement. In this embodiment, the equivalent light spot diameter is... Set as This range is determined based on the available area of ​​the tea leaf mesophyll region and the receiving aperture of the transmitted light path, ensuring that the light spot covers sufficient mesophyll tissue to obtain detectable transmitted energy, while reducing the probability of simultaneously covering multiple leaf veins. First-order lateral veins are lateral veins that extend directly from the main vein towards the leaf margin and form a distinct continuous network within the measurement area. When the projection area of ​​the light spot corresponding to a candidate position overlaps with the area of ​​a first-order lateral vein, the candidate position is discarded. In one specific embodiment, a distance of not less than [amount missing] is maintained between the edge of the light spot and the boundary of the identifiable first-order lateral vein. The interval. The preset safe distance for avoiding pulses in the project is set to reduce the probability of side pulses entering the detection area due to slight positioning errors at the edge of the light spot.

[0037] Furthermore, when any candidate position in an initial position group is eliminated due to coverage by a first-order lateral vein, instead of forming a separate position group with the remaining positions on the other side, the two candidate positions in that position group are moved synchronously along the leaf length reference axis to maintain their projected positions being identical. The synchronous movement step size in this embodiment... Set as The cumulative movement distance of a single location group does not exceed The numerical values ​​are set based on the size of the measured light spot and the spacing of local lateral veins on the leaf, ensuring that the position adjustment can cross the local leaf veins without causing the position group to deviate from the original longitudinal partition. After each movement, the coverage condition of the primary lateral veins is rechecked. The registration conditions. If in If no valid paired positions can be formed within the range, then the initial position group will not be included in the sampling position set.

[0038] Furthermore, after screening, every two sampling locations that satisfy the condition of consistent projection and are both located in the mesophyll region are determined as the same location group, and the location group is assigned an identifier in order from the petiole side to the leaf tip side. One specific implementation uses... The logical identifier stores the order of the position groups, and also stores the correspondence between the positions on both sides of the midrib within each position group. The position group identifier is only used to maintain the correlation between the transmission spectra, optical density spectra, corrected optical density spectra, and nutrient response vectors on both sides of the same longitudinal position, and does not participate in spectral numerical calculations. The resulting registration sampling position set retains the leaf identity, position group identifier, and midrib side identification, providing a fixed spatial index for subsequent transmission spectrum acquisition.

[0039] Keeping the light source spectrum, incident light path, receiving light path, and detection conditions unchanged, the incident light is blocked to obtain a dark reference spectrum, and the tea leaves are removed to obtain an empty light path reference spectrum. The measurement spot is then covered at each registration sampling position sequentially, and the leaf transmission spectrum is obtained. The dark reference spectrum, empty light path reference spectrum, and leaf transmission spectrum are aligned with sampling points of the same wavelength and associated with position group identifiers to form a registration transmission spectrum group.

[0040] Furthermore, this embodiment employs a broadband light source and spectral receiving device capable of covering the visible to near-infrared range for transmission detection, and the spectral sampling range can be set to... adjacent wavelength sampling interval Set as This range is used to simultaneously retain absorption, transmission, and baseline variation information in both the visible and near-infrared regions of the leaf, and covers the near-infrared water absorption region subsequently used for structural disturbance correction. Single integration time. Available Pre-selection ensures that the maximum effective intensity of the empty optical path reference spectrum is located within the detector's full scale. Once determined, it remains unchanged throughout all sampling processes on the same leaf. Each spectrum is taken... The average value of the results from each consecutive acquisition is used to reduce noise from a single detection.

[0041] Furthermore, the dark reference spectrum is obtained by blocking the incident light while keeping the detector integration time, gain, and receiving optical path unchanged, and is used to characterize the detector dark current and electronic bias. The empty optical path reference spectrum is obtained when the tea leaf is removed while keeping the light source, incident optical path, and receiving optical path in their original states, and is used to characterize the current light source output and optical path transmission reference. In one embodiment, a dark reference spectrum and an empty optical path reference spectrum are acquired once before leaf sampling begins, and after each leaf sampling is completed... After each position group, the empty optical path reference spectrum is acquired again. The verification cycle for each location group is a preset value for the project. Its purpose is not to change the reference definition, but to verify whether the light source and optical path are still in the same detection state, so as to avoid light source drift during long-term sampling being mistaken for changes in leaf transmission.

[0042] Furthermore, considering that the transmission detection results are directly affected by the light source intensity and the stability of the receiving optical path, the reference drift is calculated for the reference spectra of the empty optical path before and after the two tests to determine whether keeping the light source spectrum, incident optical path, receiving optical path, and detection conditions unchanged meets the acceptable detection boundary.

[0043]

[0044] in, It represents the reference drift between two empty optical path reference spectra. It is a dimensionless quantity and is formed by the median of the relative changes at the effective wavelength sampling points. This indicates that the previous empty optical path reference spectrum was at wavelength The intensity at that location is directly collected by a spectral receiving device. This indicates the reference spectrum of the subsequent empty optical path at wavelength. The intensity at the location was obtained directly from the verification data. Indicates the dark reference spectrum at wavelength Dark reference intensity at that location. Represents the first in the uniform wavelength sampling sequence Each wavelength sampling point. This indicates the wavelength sampling point number. This represents the set of effective wavelength sampling points used for reference quality checks, consisting of samples from wavelengths where detector saturation has not occurred and the net reference intensity in the empty optical path is greater than the detector's full scale. It consists of wavelength points. This indicates that the median of all relative changes within the set is taken to reduce the impact of individual wavelength noise spikes on the determination of optical path stability. This indicates that the larger of the two input values ​​within the parentheses is taken, to avoid the denominator being close to zero when the net reference strength is too low. This represents the minimum intensity denominator in the reference drift calculation, which is set to the detector's full-scale intensity in this embodiment. of ,Right now . The digital full-scale intensity of the spectral receiver is obtained directly from the detector output specifications.

[0045] In this embodiment, the reference drift determination boundary is set as... That is, the median relative change in effective wavelength between two empty optical path references does not exceed This value serves as an engineering preset boundary for optical acquisition stability, preventing significant light source drift from entering the blade spectrum while preserving normal short-term light source fluctuation tolerance. At that time, the empty optical path reference spectrum is reacquired, and the position acquired after the previous valid reference is remeasured.

[0046] Furthermore, when acquiring the transmission spectrum of the leaf, the center of the light spot at the registration sampling position is aligned with the pre-recorded position center, while maintaining the orientation of the leaf's front and back sides, the incident direction, the detector's receiving direction, and the relative distance between the leaf and the optical path. This process is repeated continuously at each position. The transmission spectrum of the blade at that location is formed after each spectral reading. Switching between two adjacent locations only moves the blade to the next registration sampling position, without changing the light source power, integration time, or detector gain. This ensures that the differences in spectral intensity at different locations primarily correspond to changes in the blade's own transmission characteristics, rather than changes in detection conditions. If detector saturation occurs during a sampling, the detection parameters for that single location are not changed; instead, a fixed integration time is selected for the entire blade, and the data is re-acquired to maintain comparability between locations.

[0047] Furthermore, the dark reference spectrum, the empty optical path reference spectrum, and the blade transmission spectrum all use the same unified wavelength sequence. Storage. When the same spectral receiving device directly outputs a fixed sampling grid, it is directly aligned using wavelength indices. If the output wavelength of the device has a rounding deviation of no more than half a sampling interval, it is mapped to the nearest uniform wavelength sampling point. The wavelength matching deviation allowed in this embodiment is... Set as The source of this is the discretization boundary corresponding to the spectral sampling interval, which avoids mispairing adjacent wavelength points. Spectra exceeding this boundary are not subjected to forced cross-point stitching, but instead undergo wavelength calibration or are sampled at the corresponding location again.

[0048] Furthermore, after wavelength alignment, a registered transmission spectrum group is formed by using leaf identity—position group identifier—main vein side—wavelength sampling point as an index to combine the dark reference intensity, empty optical path reference intensity, and leaf transmission intensity. The dark reference spectrum and empty optical path reference spectrum in the registered transmission spectrum group serve as common references for the position spectra during the same acquisition period, and their effective temporal order is preserved. The transmission spectrum of each leaf remains associated with its corresponding position group identifier. This registered transmission spectrum group constitutes the output of S1, enabling subsequent calculations of net transmission intensity, net reference intensity, and optical density at any wavelength sampling point to find the reference intensity and a clear spatial position of the leaf under the same optical acquisition conditions.

[0049] It should be noted that S1 does not simply involve repeatedly acquiring spectra by arbitrarily selecting multiple measurement points on the leaf surface. Instead, it first establishes a spatial correspondence using the projection relationship of the same leaf length on both sides of the midrib, then excludes cases where first-order lateral veins directly enter the measurement spot, and integrates this spatial relationship into the data indexes of the dark reference, the empty optical path reference, and the leaf transmission spectrum. This solves the prerequisite problem in transmission spectrum detection of whether different spectra originate from comparable leaf tissue regions, ensuring that the consistency constraint between the positions on both sides is established on the real spatial registration relationship, rather than mathematically pairing arbitrary spectra after acquisition.

[0050] S2: Convert the leaf transmission spectrum into an optical density spectrum and correct structural interference based on the water absorption reference band and scattering reference band.

[0051] The registered transmission spectrum obtained in step S1 is used as the input for this step, performing optical density conversion and structural interference correction. For each wavelength sampling point in the registered transmission spectrum, the net transmission intensity is obtained by subtracting the dark reference intensity from the leaf transmission intensity, and the net reference intensity is obtained by subtracting the dark reference intensity from the empty optical path reference intensity. The ratio of net transmission intensity to net reference intensity is determined as the corrected transmittance, and the optical density value is obtained by performing a negative logarithmic transformation on the corrected transmittance. The optical density values ​​are arranged according to the wavelength sampling points to form an optical density spectrum corresponding to the registration sampling position and position group identifier.

[0052] Furthermore, considering that the inconsistency in detector dark current, electron null point, and absolute output of the light source at different wavelengths will directly affect the original transmission intensity, this embodiment does not use the leaf transmission intensity itself as the nutrient response. Instead, it subtracts the dark reference at each uniform wavelength sampling point and normalizes it with the net reference intensity, then performs a negative logarithmic transformation. For the first... The main vein lateral to each location group The light density value of the leaf's transmission spectrum is formed according to the following formula:

[0053]

[0054] in, Indicates the first Position groups, sides At wavelength The optical density value at that location is a dimensionless quantity. This indicates that the corresponding registration sampling position is at wavelength The transmission intensity of the blade at point S1 was obtained from the registered transmission spectrum. Indicates wavelength The dark reference intensity at that point is obtained from the dark reference spectrum of S1. Indicates wavelength The reference intensity of the empty optical path at point S1 is obtained from the reference spectrum of the empty optical path at point S1. This indicates the position group number, following the position group identifier order of S1. Indicates the lateral branch of the main meridian, taken from... . Represents the first wavelength in the unified wavelength sequence Each wavelength sampling point. Indicates the wavelength sampling point number, take . This represents the total number of wavelength sampling points contained in a uniform wavelength sequence, which is determined by the spectral range and the sampling interval. Indicated by The logarithmic operation with base 1 transforms the multiplicative change in transmittance into an additive change in the optical density domain.

[0055] Furthermore, to ensure that the aforementioned negative logarithm calculation has definite numerical boundaries, this embodiment requires a net reference strength. Not less than This lower limit is derived from the detector's effective dynamic range constraint and is used to avoid amplifying electronic noise where the reference light is too weak. The corrected transmittance is denoted as... When generated by sampling noise Take the upper boundary at time .when Remove the boundary at time The corresponding maximum optical density is If the net transmitted intensity is not greater than zero, the wavelength point is marked as invalid and the acquisition status at the corresponding location is rechecked, instead of directly taking the logarithm of the non-positive number. The boundary is derived from the requirements of numerical stability and detection dynamic range, and its function is to prevent unbounded optical density caused by extremely low transmitted intensity.

[0056] Furthermore, for each registration sampling location, after completing the optical density calculation at all effective wavelength points, the optical density spectrum is arranged in the order of the unified wavelength sequence of S1 to form a one-dimensional optical density spectrum, which continues to carry leaf identity, position group identifier, and midrib side identification. The optical density spectrum is not implemented in this step as it would change the wavelength shift of local peaks, nor is the average spectrum between different locations used to replace the spectrum of a single location. This preserves the independent positional response required for establishing positional consistency constraints using the spectra on both sides of the same position group. If the average transmission spectrum at the same location is formed by multiple consecutive acquisitions from S1, the optical density conversion is applied to the averaged leaf transmission intensity, ensuring that the reference subtraction and transmittance calculations use the same calculation order at all locations.

[0057] Furthermore, during the formation of the optical density spectrum, arbitrary interpolation is performed on invalid wavelength sampling points without spanning a wide interval. This embodiment allows for a limited number of consecutive invalid wavelength points. When the adjacent valid sampling points are linearly filled in, it is used. At that time, the spectrum at the corresponding location is reacquired. The boundary originates from this embodiment. The wavelength sampling interval keeps the width of the padded interval at approximately [value missing]. It is smaller than the typical width of the subsequent response band, which can handle individual detection anomalies and avoid artificially forming absorption peaks or shoulder bands through large-range interpolation.

[0058] Furthermore, each optical density spectrum obtained after this processing still corresponds one-to-one with the registration sampling position in S1. In the subsequent water absorption and scattering structure correction, instead of averaging the spectra of all positions on the same leaf first, we use all positions to form an internal reference for the leaf before returning to each position for correction. This allows the spectral response of individual positions to be preserved, and ensures that subsequent water and scattering reference values ​​originate from multiple real sampling positions within the same leaf, rather than from uncontrollable absolute optical density differences between different leaves.

[0059] The water absorption reference band is separated from the target nutrient correlation response band used to construct the nutrient response vector, and the scattering reference band avoids the target nutrient correlation response peak. For the optical density spectrum at each registration sampling location, a local baseline is formed within the water absorption reference band by connecting the two ends of the band, and the band area of ​​the spectral line relative to the local baseline is calculated. The slope of the spectral line baseline is calculated within the scattering reference band. A moisture reference value is formed using the median of the band area at all registration sampling locations for the same tea leaf, and a scattering reference value is formed using the median of the spectral line baseline slope. An absorption amplitude correction factor and a baseline morphology correction factor are formed based on the deviation of the band area relative to the moisture reference value and the deviation of the spectral line baseline slope relative to the scattering reference value at each registration sampling location, respectively. The spectral amplitude of the corresponding optical density spectrum is scaled using the absorption amplitude correction factor, and the tilt baseline is compensated for using the baseline morphology correction factor to obtain a corrected optical density spectrum that retains the location group identifier. The water absorption reference band characterizes the change in absorption amplitude caused by leaf water content, and the scattering reference band characterizes the change in transmission baseline caused by leaf tissue structure.

[0060] Furthermore, the target nutrient-related response bands are predetermined during the calibration phase and maintained fixed during the treatment of the leaves to be diagnosed. Therefore, the selection of the water absorption reference band and the scattering reference band can be completed before entering the current calibration process. In this embodiment, the water absorption reference band can be... Set as And require any target nutrient to be associated with the response band and Retain no less than Wavelength spacing. Scattering reference band. Can be set to And require the peak position of the already labeled target nutrient correlation response peak to be consistent with... The boundary is at least separated The above range is set according to the leaf water absorption characteristics and relatively flat transmission zone within the detection band of this embodiment. Its technical function is to make the two reference quantities mainly bear the correction role of absorption amplitude and baseline morphology, respectively, rather than directly using the target nutrient-related response peak itself to form a correction reference.

[0061] Furthermore, considering that the water content at different locations on a single leaf can cause variations in absorption amplitude, and that tissue thickness and cell interface scattering can alter the transmission baseline slope, and that these two types of effects would be mixed if only a single scaling factor were used, this embodiment calculates the area of ​​the water absorption reference band and the baseline slope of the scattering reference band separately, and uses the median of all locations on the same leaf to form two internal references, then sequentially generates the absorption amplitude correction factor and the baseline morphology correction factor:

[0062]

[0063] in, Indicates the first Position groups, sides The local baseline in the water absorption reference band is formed by linearly connecting the optical density values ​​at the two ends of the water absorption reference band. This embodiment represents the low-wavelength endpoint of the water absorption reference band. . This embodiment represents the high-wavelength endpoint of the water absorption reference band. . The area of ​​the water absorption reference band at the corresponding registration sampling location is obtained by integrating the optical density spectrum relative to the local baseline, and its dimensions are... . This represents the scattering reference band, which is [not specified in the original text]. . This represents the baseline slope of the spectral line at the corresponding location within the scattering reference band, in units of... . The average wavelength of all wavelength sampling points within the scattering reference band is represented by... The actual sampling wavelength within the range is directly calculated. Indicates the first Position, side The average value of all optical density values ​​within the scattering reference band. The median of the water absorption reference zone area at all valid registration sampling locations of the same tea leaf is the moisture reference value. The median of the baseline slope of the scattering reference band represents all valid registration sampling positions of the same tea leaf, i.e., the scattering reference value. This represents the absorption amplitude correction factor, which is formed by the deviation of the water absorption reference zone area at the current location from the water reference value, and is a dimensionless quantity. This represents the smallest denominator for area calculations. In this embodiment, we take... This is used to avoid the scale factor becoming unbounded due to the area of ​​the current location being close to zero. Indicates that the input will be entered. Limited to the range Inside. Below Time to take higher than Time to take . , These represent the lower and upper boundaries of the absorption amplitude correction factor, respectively, serving as engineering stability boundaries to ensure that the scale correction at a single location does not exceed the range of obvious abnormalities, thus avoiding excessive amplification caused by local wrinkles, residual leaf veins, or abnormal transmission points. This represents the baseline morphology correction factor, in units of It is formed based on the difference between the current position baseline slope after scale correction and the scattering reference value. The wavelength pivot for tilt baseline compensation is taken in this embodiment as the center wavelength of the scattering reference band, i.e. . This indicates the corrected optical density spectrum after absorption amplitude scale correction and tilt baseline compensation, while retaining the location group identifier and main vein side identification.

[0064] Furthermore, both moisture and scattering reference values ​​are calculated only within the same tea leaf, while the median is calculated for all registered sampling locations on that leaf, not across different leaves. The median is used because a single location may deviate due to minute leaf curvature, local tissue thickness, or optical path contact conditions; the median will not be directly affected by a single extreme location. When the number of effective registered sampling locations is less than... One-sided position, that is, less than When constructing a complete set of positions, this embodiment does not utilize too few positions to construct an internal reference for the blade, but instead re-examines the sampling positions of S1. This quantity boundary stems from the requirement to form a robust median and retain at least three longitudinal position sets, the advantage of which is to avoid defining the structural reference of the entire blade based on only one or two local positions.

[0065] Furthermore, the absorption amplitude correction and tilt baseline compensation are performed in the order defined in the claims, i.e., first with... Perform spectral amplitude scale correction on the corresponding optical density spectrum, and then... Apply baseline tilt compensation to the scale-corrected spectral shape. If This indicates that the current water absorption reference zone cannot form an effective area reference. Therefore, instead of forcibly scaling up the ratio, we will... It also requires verification of the spectral range, reference spectrum, and blade placement. For slope calculations, if... The number of effective wavelength points within the range is less than If there are insufficient points, the baseline slope will not be fitted with the missing points; instead, the corresponding spectral band will be reacquired. The above boundary is used to prevent division by zero, overcorrection, and a small number of outliers from dominating baseline correction.

[0066] Furthermore, the water absorption reference band targets the overall absorption range difference caused by changes in leaf water content, and its output is dimensionless. The scattering reference band acts on the differences in transmission baseline morphology caused by leaf tissue structure, thickness, and interface scattering; its output is... Neither of these values ​​is directly used as a nutrient diagnostic value, nor is the moisture reference value or scattering reference value interpreted as a certain nutrient content. Instead, they are only used to ensure that the optical density spectra at different registration locations are at comparable absorption scales and baseline morphologies relative to internal references within the same leaf. This avoids substituting undefined comprehensive correction indices for actually calculable spectral correction quantities.

[0067] Furthermore, the final output of this step is a corrected optical density spectrum that retains leaf identity, location group identifier, and midrib side identification. The wavelength order of each corrected optical density spectrum is consistent with that of the S1 registered transmission spectrum group, and the sources of the water absorption reference band, scattering reference band, absorption amplitude correction factor, and baseline morphology correction factor can all be traced back to the measured optical density spectrum of the same leaf. This output is directly used as the input for the integrated quantities of the target nutrient-related response band, the shared overlapping band, and the non-overlapping shoulder band response quantities constructed in S3, so that subsequent multi-nutrient calculations are based on the spectrum that has been corrected for structural interference and still retains spatial correspondence.

[0068] It should be noted that S2 does not apply a fixed global baseline subtraction or a single normalization factor to all leaves. Instead, it uses multiple registration locations on the same leaf to generate water absorption band area and scattering baseline slope, and then performs two-level corrections for amplitude and baseline morphology at each location using the leaf's internal median reference. This processing addresses the different physical effects of water absorption and tissue scattering on the transmission spectrum, improving the comparability of spectral shapes at different locations without using the target nutrient-related response peak itself as a correction reference. It also avoids structural interference correction directly eliminating response information intended for nutrient diagnosis.

[0069] S3: Construct nutrient response vectors from the calibrated optical density spectrum, and solve for multi-nutrient diagnostic values ​​according to the overlap constraint of response bands and the position consistency constraint.

[0070] The corrected optical density spectrum of the retained position group obtained in step S2 is used as the input for this step to construct the nutrient response vector and solve for the multi-nutrient diagnostic value. Independent nutrient content measurements are performed on the calibrated leaf samples. The local absorption peaks in the corrected optical density spectrum and their adjacent continuous wavelength intervals are divided into candidate response bands, and the integrated response of the calibrated leaf sample is calculated for each candidate response band. The calibrated leaf samples are sorted according to the independent nutrient content measurements of each target nutrient, and the candidate response bands with the same sorting direction of the integrated response in each position group are retained as the corresponding target nutrient-related response bands. When there are overlapping bands in the target nutrient-related response bands of two or more target nutrients, the overlapping bands are recorded as shared overlapping bands, and non-overlapping shoulder bands corresponding to each target nutrient and not coinciding with other target nutrient-related response peaks are determined on both sides of the shared overlapping band. For each position group of the tea leaves to be diagnosed, the integrated response of each target nutrient-related response band relative to the local baselines at both ends is calculated, and the integrated response of the shared overlapping band and the shoulder band response of the corresponding non-overlapping shoulder band relative to the local baseline are also calculated. The integral response quantities corresponding to the same target nutrient, the integral response quantities of the shared overlapping zone, and the shoulder band response quantities are combined in a fixed order to form the location nutrient response components. Then, all location nutrient response components are combined in a fixed order according to the target nutrient to form a nutrient response vector that retains the location group identifier. The correspondence between the target nutrient associated response band, the shared overlapping zone, and the non-overlapping shoulder band is recorded as the response band overlap relationship.

[0071] Furthermore, the calibrated leaf samples and the tea leaves to be diagnosed employ the same S1 transmission acquisition and S2 structural interference correction procedures. Target nutrients can be determined based on the diagnostic task. Taking the simultaneous diagnosis of nitrogen, phosphorus, and potassium as an example, independent nutrient content measurements are obtained through chemical or elemental analysis processes independent of transmission spectroscopy detection. Nitrogen content can be obtained through combustion elemental analysis or nitrogen content determination after standardized pretreatment. Phosphorus and potassium contents can be obtained through instrumental analysis with elemental quantification capabilities after leaf sample digestion. Independent measurements are correlated one-to-one with the corresponding calibrated leaf and uniformly converted to the same dry matter mass standard. These independent measurements are only used to establish the calibration correspondence between spectral response and target nutrient content; they are not used as the reverse calibration truth value, thus maintaining a closed-loop calibration data source.

[0072] Furthermore, this embodiment allows setting the number of calibration leaf samples. The sample size is used as an example range for a calibration batch. This range is a calibration scale setting in the embodiment and not a real experimental result. It is based on the premise that when the number of target nutrients is three, the sample size is significantly larger than the number of nutrient variables to be solved, thus forming the percentile variation range of subsequent response components. For each calibrated leaf, the local absorption peak and its adjacent continuous wavelength intervals on both sides are searched in the calibrated optical density spectrum, and the candidate response band width is determined. Can be set to And each candidate response band contains at least One effective wavelength sampling point. This width is based on S1. The sampling interval and local spectral integration need to be set so that the response is not determined by a single wavelength noise, while avoiding the mixing of multiple independent spectral changes in an excessively wide range.

[0073] Furthermore, for candidate response bands, target nutrient-related response bands, shared overlapping bands, and non-overlapping shoulder bands, local baselines are constructed by connecting their two endpoints, and the response quantity is represented by the integral of the corrected optical density spectrum relative to this local baseline, thus ensuring that different types of response components use the same quantization caliber. For calibrated leaf numbers... Position group , side and response band The response quantity is obtained according to the following formula:

[0074]

[0075] in, Indicates response band The local baseline within the response band is formed by linearly connecting the corrected optical density values ​​at the two ends of the response band. Indicates the calibration blade In the position group , side The corrected optical density spectrum at the point is output by S2. Indicates response band The low-wavelength endpoint. Indicates response band The high wavelength endpoint. Indicates response band The set of continuous wavelength sampling points included. It can correspond to candidate response bands, target nutrient-related response bands, shared overlapping bands, or non-overlapping shoulder bands. Indicates the calibration blade In the position group , side For response band The integral response quantity, with dimensions of . Indicates the first The integration step size corresponding to each wavelength sampling point. For equally spaced spectra, its value is equal to the wavelength sampling interval of S1. . Indicates the sample number of the calibration blade. . This indicates the number of calibrated leaf samples; in this embodiment, it is taken as... . This represents the calibrated position integral response, calculated by taking the median of the responses on both sides of the main vein within the same position group. When there are only two sides, the median is equivalent to the average of the two values, used to establish the sorting direction of the position group.

[0076] Furthermore, after calibrating leaf samples by arranging them from low to high according to the independent nutrient content measurements of a specific target nutrient, the response bands of the same candidate nutrient were observed sequentially. The direction of change depends on the sorting order. This embodiment defines the effective unidirectional change ratio of adjacent sorted samples within a position group. Set to no less than Furthermore, it is required that the dominant increase or decrease direction is the same for all position groups before the candidate response band is retained as the corresponding target nutrient-related response band. The engineering boundary for calibrating the selection is based on allowing for a small number of adjacent order reversals caused by independent measurement errors and local spectral noise, while requiring that most samples still exhibit a consistent orientation. If opposite dominant orientations appear in different position groups, the candidate response band will not be included in the response relationship of the target nutrient, thus avoiding the misinterpretation of position-dependent random spectral changes as nutrient responses.

[0077] Furthermore, when two or more target nutrient correlation response bands share a common wavelength range, their intersection is recorded as a shared overlapping band. To avoid the boundary contact of a single sampling point being mistakenly considered as valid overlap, this embodiment requires the shared overlapping band to contain at least [missing information - likely referring to specific wavelength ranges]. A continuous effective wavelength sampling point. For each side of the shared overlapping band, a continuous interval that does not coincide with the nutrient correlation response peak of other targets is selected in the adjacent target nutrient correlation response interval as a non-overlapping shoulder band, and the width of each non-overlapping shoulder band is... Set as and at least includes One effective wavelength point. This range is determined based on the spectral sampling interval and the stability of the local shoulder response, which avoids using a single point response to distinguish nutrients and controls the shoulder band from extending into adjacent nutrient response peaks.

[0078] Furthermore, for the leaves to be diagnosed, the response quantities of the target nutrient associated response band, shared overlapping band, and non-overlapping shoulder band are calculated using the same wavelength endpoints and local baseline rules as the calibrated leaves. The fixed order is first determined according to the target nutrient order, taking nitrogen, phosphorus, and potassium as an example, using the nitrogen-phosphorus-potassium order. For the same target nutrient, its positional nutrient response components are arranged sequentially as follows: the integrated response quantity of the target nutrient associated response band, the integrated response quantity of the shared overlapping band arranged from low to high wavelength, and the response quantities of the non-overlapping shoulder bands on both sides of the corresponding shared overlapping band. If a target nutrient does not have a shared overlapping band, only its actual target nutrient associated response quantity is saved. The fixed order is maintained consistently during the calibration and diagnosis process, ensuring that each response component has a unique wavelength range and a corresponding target nutrient.

[0079] Furthermore, the wavelength endpoints of each target nutrient's associated response band, its overlapping objects, shared overlapping bands, and left and right non-overlapping shoulder bands are saved together as response band overlap relationships. For example, when a shared overlapping band corresponds to two target nutrients, it is recorded that the shared overlapping band simultaneously belongs to the contribution calculation objects of two target nutrients, while the non-overlapping shoulder band of one of the target nutrients is only associated with that target nutrient. The response band overlap relationship does not include the unknown nutrient values ​​of the leaf to be diagnosed, but rather the wavelength structure relationship determined in the calibration stage. The first half of S3 ultimately obtains the nutrient response vector and response band overlap relationship saved by position group, and both are used together in the subsequent nutrient response contribution table establishment and multi-nutrient unsolved value solution process.

[0080] Using the nutrient response vectors and independent nutrient content measurements of calibrated leaf samples, the response contribution relationships of each target nutrient to the target nutrient-related response bands, shared overlapping bands, and non-overlapping shoulder bands were determined, and a nutrient response contribution table was formed according to the correspondence between target nutrients and response bands. Based on the overlap relationship of response bands, for each location group of tea leaves to be diagnosed, the location nutrient response components were represented as a combination of the response contributions of each target nutrient using the nutrient response contribution table. The reconstructed response of the shared overlapping band was composed of the response contributions of each target nutrient corresponding to the shared overlapping band, while the reconstructed response of the non-overlapping shoulder band was composed only of the response contributions of the target nutrient corresponding to the non-overlapping shoulder band, thus forming a response band overlap constraint. For the registered sampling locations on both sides of the main vein within the same location group, the same set of target nutrient values ​​were shared, and the positional residuals between the positional nutrient response components on both sides and the corresponding reconstructed responses were calculated, thus forming a positional consistency constraint. The reconstruction deviations of the shared overlapping zone, the reconstruction deviations of the non-overlapping shoulder zone, and the position residuals were normalized and summed according to the variation range of the corresponding response components in the calibrated leaf samples. The target nutrient value with the smallest total deviation was selected, provided that all target nutrient values ​​were non-negative. The median of the target nutrient values ​​obtained for all position groups was then taken for each target nutrient to form a multi-nutrient diagnostic value.

[0081] Furthermore, the nutrient response contribution table is established using the exact same response component order as the nutrient response vector. For a specific response component in the calibrated sample, the non-negative response contribution coefficient is determined using this response component as the explained variable and independent nutrient content measurements as input. If the response component belongs to a shared overlapping zone, only the target nutrient corresponding to that shared overlapping zone is allowed to participate in the contribution coefficient calculation. If the response component belongs to a non-overlapping shoulder zone of a target nutrient, the contribution coefficients of other target nutrients besides that target nutrient are fixed. The contribution coefficients are determined by all calibrated leaf samples using constrained least squares and are fixed for subsequent leaves to be diagnosed. They are not refitted for individual leaves to be diagnosed, thus ensuring that the diagnostic value is calculated based on independently measured calibrated samples.

[0082] Furthermore, to enable comparison of response amplitudes across different spectral bands and contributions from different nutrients within the same objective function, each response component is scaled according to the variation range of the calibrated leaf samples. This embodiment will... The range of change of each response component Defined as the first response component in all calibrated blade samples. Percentile and 1 The difference in percentile values. (Using...) The quantile range, rather than the maximum and minimum values, is chosen to reduce the dominance of individual outlier samples on the normalization scale. If a certain response component... Less than the largest quantile range among all response components Then, based on that The value is used as the minimum normalized denominator. This rule stems from the requirement of numerical stability and can prevent the almost unchanged response component from being abnormally amplified in the total deviation.

[0083] Furthermore, considering that the shared overlapping zone simultaneously contains contributions from multiple target nutrients, while the non-overlapping shoulder zone can only be explained by its corresponding target nutrient, and that the nutrient status of the same group of leaves should correspond to the same group of leaves on both sides of the main vein at the same location, this embodiment simultaneously retains the overlapping zone reconstruction bias, shoulder zone reconstruction bias, and positional residuals on both sides in the same objective function, and requires that both sides share the same solution vector:

[0084]

[0085] in, This indicates the target nutrient vector to be solved. Next, the Position groups, sides The Reconstructed response of each response component. Indicates the first The position groups are used together to solve the target nutrient vector on both sides of the main vein. Each component corresponds to a target nutrient, and the unit is consistent with the corresponding independent nutrient content measurement value. Indicates the first Target nutrients in each location group The unsolved value, and limited . The target nutrients in the nutrient response contribution table represent For response components The response contribution coefficient is determined by non-negative constraint least squares from the nutrient response vector of the calibrated leaf sample and the independent nutrient content measurement values. This indicates that permission is allowed for the first... Each response component contributes to the set of target nutrients. For shared overlapping zones, the set contains two or more target nutrients corresponding to that overlapping zone. For non-overlapping shoulder zones, it contains only one target nutrient corresponding to that shoulder zone. Indicates the tea leaf piece to be diagnosed. Position groups, sides The The nutrient response components at each measured location are formed by the aforementioned local baseline integration rule. Indicates the first The position group is in the target nutrient unsolved vector The total deviation formed is a dimensionless non-negative quantity. This represents the set of indices of all shared overlapping response components in the nutrient response vector. This indicates the number of response components sharing an overlap band. The corresponding term is calculated only if an overlap band exists; otherwise, this term is omitted. . This represents the set of indices for all non-overlapping shoulder strap response components. This indicates the number of non-overlapping shoulder strap response components. If no non-overlapping shoulder straps exist, the corresponding item is taken as... . It represents the total number of all response components in a location nutrient response vector. This represents the response component index, determined according to the aforementioned fixed response order. This indicates the target nutrient index. Indicates the first The first response component in the calibrated blade sample Percentile and 1 The difference in percentile values. This represents the normalized minimum response range; in this embodiment, it is set to all. The maximum value This is used to prevent the range of variation close to zero from becoming too small a denominator. This represents a vector where all target nutrient components are zero. (Constraint) This means that the unsolved values ​​of each target nutrient must not be negative. This means selecting, from all target nutrient vectors satisfying the non-negativity constraint, the one that minimizes the total deviation. The smallest vector.

[0086] Furthermore, in the aforementioned total deviation, the first term specifically strengthens the joint reconstruction relationship of shared overlapping zones; therefore, the reconstruction response of a shared overlapping zone can only involve the target nutrients specified by its response band overlap relationship. The second term strengthens the single nutrient constraint of non-overlapping shoulder bands; therefore, other target nutrients cannot offset the overlapping zone error by means of the shoulder band response. The third term combines the complete positional nutrient response components on both sides of the main vein with the same... The corresponding reconstruction responses are compared. Since different solution values ​​cannot be used to reduce the residuals on both sides, the spectra on both sides at the same longitudinal position of the leaf jointly constrain the same set of target nutrient values, forming a positional consistency constraint. All three parts are summed after being divided by the calibration range to avoid the spectral bands with larger absolute amplitudes naturally occupying higher weights.

[0087] Furthermore, the total deviation is in a non-negative quadratic form and carries a non-negative constraint on the target nutrient. In this embodiment, it can be obtained using an activity set solution method with non-negative constraints. The initial value is the median of the independent measurements of the corresponding target nutrient in the calibration sample. If any initial component is less than zero, then... Maximum number of iterations Set as The relative change in the total deviation between two adjacent iterations is less than Stop when the time comes. and The purpose of this numerical solution for engineering parameters is to ensure that low-dimensional, multi-nutrient solutions have clear termination conditions and avoid ineffective iterations. If... If the convergence boundary is still not met, the non-negative solution with the smallest total deviation is retained and the solution is marked as not fully converged for that position group, but not replaced with a negative nutrient value.

[0088] Furthermore, for each complete location group, the target nutrient vector to be solved is obtained according to the same contribution table, response band overlap relationship, and normalization rule. For the same target nutrient, all position groups corresponding to... The median value is used as the final diagnostic value for the target nutrient, sorted by numerical value. When the number of position groups is odd, the median value is used. When the number of position groups is even, the arithmetic mean of the two median values ​​is used. The final diagnostic value maintains the same units and dry matter basis as the independent nutrient content measurements during the calibration phase. Using the median of all position groups avoids representing the entire leaf from a single location, ensuring that local tissue residuals do not directly determine the final multi-nutrient diagnostic result.

[0089] Furthermore, the multi-nutrient diagnostic values ​​output in this step can form a set of diagnostic results according to a fixed order of the target nutrients. For example, when the target nutrients are nitrogen, phosphorus, and potassium, the corresponding three diagnostic values ​​are output sequentially. Each diagnostic value can be traced back to the corrected optical density spectrum, target nutrient associated response band, shared overlapping band, non-overlapping shoulder band, nutrient response contribution table, and independent nutrient calibration value for a specific location group, without introducing undefined comprehensive indices or black-box scoring. When a certain response component of the leaf to be diagnosed exceeds that of the calibration sample... When the range changes, the non-negative solution is still performed according to the same contribution relationship, but states exceeding the calibrated main range of change can be recorded simultaneously to distinguish between in-range diagnosis and extrapolation diagnosis. This state does not change the final median formation rule defined in the claims.

[0090] It should be noted that S3 does not directly calculate the value of each nutrient using an independent spectral response relationship. Instead, it explicitly preserves the shared overlapping bands when the target nutrient-related response bands overlap, and uses non-overlapping shoulder bands located on both sides of the overlapping bands, corresponding only to specific target nutrients, to limit the response contributions of different nutrients. Furthermore, the same set of nutrient values ​​can only be shared on both sides of the same main vein group. This overcomes the two types of solution degrees of freedom: an overlapping spectral response can be arbitrarily interpreted by multiple nutrient values, and a single-sided local spectrum can independently generate a seemingly optimal solution. This ensures that the final diagnostic result is constrained by both wavelength structural relationships and true spatial location relationships.

[0091] Example 2, an embodiment of the present invention, provides a multi-nutrient diagnostic system for tea trees based on transmission spectroscopy, including a vein-avoidance registration transmission module, a dual-reference spectrum correction module, and an overlapping constraint nutrient solution module.

[0092] The vein-avoidance registration and transmission module is used to set registration sampling positions in the vein-avoidance area of ​​tea leaves to obtain dark reference spectrum, empty light path reference spectrum and transmission spectrum of leaves at each position.

[0093] The dual-reference spectrum correction module is used to convert the leaf transmission spectrum into an optical density spectrum and correct structural interference based on the water absorption reference band and the scattering reference band.

[0094] The overlapping constraint nutrient solution module is used to construct nutrient response vectors from the calibrated optical density spectrum and solve multi-nutrient diagnostic values ​​according to the overlapping constraint and positional consistency constraint of the response band.

Claims

1. A method for diagnosing multiple nutrients in tea plants based on transmission spectroscopy, characterized in that, include: Registration sampling positions were set in the vein-avoiding area of ​​tea leaves to obtain dark reference spectrum, empty light path reference spectrum and transmission spectrum of leaves at each position. The leaf transmission spectrum was converted into an optical density spectrum, and structural interference was corrected based on the water absorption reference band and the scattering reference band. Nutrient response vectors are constructed from the calibrated optical density spectrum, and multi-nutrient diagnostic values ​​are calculated based on response band overlap constraints and position consistency constraints.

2. The method for diagnosing multiple nutrients in tea plants based on transmission spectroscopy as described in claim 1, characterized in that: The step of setting registration sampling positions in the vein-avoiding area of ​​tea leaves includes... Establish the leaf length reference axis with the extension direction of the main vein of the tea leaf, and determine the paired candidate positions in the leaf mesophyll area on both sides of the main vein according to the same projection position along the leaf length reference axis. Candidate locations where the measurement spot covers the first-order side veins are eliminated, and the remaining paired candidate locations are determined as registration sampling locations; Assign a position group identifier to the same position group based on the same projection position along the leaf length reference axis.

3. The method for diagnosing multiple nutrients in tea plants based on transmission spectroscopy as described in claim 2, characterized in that: The acquisition of the dark reference spectrum, the empty optical path reference spectrum, and the transmission spectrum of the blades at each position includes... Keeping the light source spectrum, incident light path, receiving light path and detection conditions unchanged, the incident light is blocked to obtain the dark reference spectrum, and the tea leaves are removed to obtain the empty light path reference spectrum. Sequentially cover the measurement spot with each registration sampling position and acquire the leaf transmission spectrum; Align the dark reference spectrum, the empty optical path reference spectrum, and the leaf transmission spectrum with the same wavelength sampling points and associate them with the position group identifier to form a registered transmission spectrum group.

4. The method for diagnosing multiple nutrients in tea plants based on transmission spectroscopy as described in claim 3, characterized in that: The process of converting the leaf transmission spectrum into an optical density spectrum includes... For each wavelength sampling point of the registered transmission spectrum group, the net transmission intensity is obtained by subtracting the dark reference intensity from the leaf transmission intensity, and the net reference intensity is obtained by subtracting the dark reference intensity from the empty optical path reference intensity. The ratio of net transmitted intensity to net reference intensity is determined as the corrected transmittance, and the optical density value is obtained by performing a negative logarithmic transformation on the corrected transmittance. The optical density values ​​are arranged according to the wavelength sampling points to form an optical density spectrum corresponding to the registration sampling position and position group identifier.

5. The method for diagnosing multiple nutrients in tea plants based on transmission spectroscopy as described in claim 4, characterized in that: The correction of structural interference based on water absorption and scattering reference bands includes, The water absorption reference band is separated from the target nutrient correlation response band used to construct the nutrient response vector, and the scattering reference band avoids the target nutrient correlation response peak. For the optical density spectrum at each registration sampling position, a local baseline is formed within the water absorption reference band by connecting the two ends of the band. The band area of ​​the spectral line relative to the local baseline is calculated, and the slope of the spectral line baseline is calculated within the scattering reference band. A moisture reference value was formed by the median area of ​​all registered sampling locations of the same tea leaf, and a scattering reference value was formed by the median slope of the spectral baseline. Based on the deviation of the band area relative to the moisture reference value and the deviation of the spectral baseline slope relative to the scattering reference value at each registration sampling position, an absorption amplitude correction factor and a baseline morphology correction factor are formed respectively. The spectral band amplitude of the corresponding optical density spectrum is scaled using the absorption amplitude correction factor, and then the tilted baseline is compensated using the baseline shape correction factor to obtain the corrected optical density spectrum with the position group identifier preserved. The water absorption reference band characterizes the changes in absorption amplitude caused by leaf water content, while the scattering reference band characterizes the changes in transmission baseline caused by leaf tissue structure.

6. The method for diagnosing multiple nutrients in tea plants based on transmission spectroscopy as described in claim 5, characterized in that: The nutrient response vector constructed from the corrected optical density spectrum includes, The nutrient content of the calibrated leaf samples was determined independently. The local absorption peaks and their adjacent continuous wavelength ranges in the calibrated optical density spectrum were divided into candidate response bands, and the integral response of the calibrated leaf samples was calculated for each candidate response band. The calibrated leaf samples were sorted according to the independent nutrient content measurements of each target nutrient, and the candidate response bands with consistent sorting directions of the integrated response in each position group were retained as the corresponding target nutrient associated response bands. When there are overlapping bands in the target nutrient correlation response bands of two or more target nutrients, the overlapping bands are recorded as shared overlapping bands, and non-overlapping shoulder bands corresponding to each target nutrient and not overlapping with the correlation response peaks of other target nutrients are determined on both sides of the shared overlapping band. For each location group of tea leaves to be diagnosed, calculate the integral response of each target nutrient-related response band relative to the local baselines at both ends, and calculate the integral response of the shared overlapping band and the shoulder girdle response of the corresponding non-overlapping shoulder girdle relative to the local baseline. The integral response quantities corresponding to the same target nutrient, the integral response quantities of the shared overlapping zone, and the shoulder band response quantities are combined in a fixed order to form the location nutrient response components. Then, all location nutrient response components are combined in a fixed order according to the target nutrient to form a nutrient response vector that retains the location group identifier. The correspondence between the target nutrient-related response bands, shared overlapping bands, and non-overlapping shoulder bands is recorded as the response band overlap relationship.

7. The method for diagnosing multiple nutrients in tea plants based on transmission spectroscopy as described in claim 6, characterized in that: The calculation of multi-nutrient diagnostic values ​​based on response band overlap constraints and positional consistency constraints includes... Using the nutrient response vectors and independent nutrient content measurements of calibrated leaf samples, the response contribution relationship of each target nutrient to the target nutrient associated response band, shared overlapping band, and non-overlapping shoulder band was determined, and a nutrient response contribution table was formed according to the corresponding order of target nutrients and response bands. Based on the overlapping relationship of response bands, for each position group of tea leaves to be diagnosed, the position nutrient response component is represented as a combination of the target nutrient response contributions using a nutrient response contribution table. The reconstructed response of the shared overlapping band is composed of the target nutrient response contributions corresponding to the shared overlapping band, while the reconstructed response of the non-overlapping shoulder band is composed only of the target nutrient response contributions corresponding to the non-overlapping shoulder band, thus forming a response band overlapping constraint. For the same set of registered sampling positions located on both sides of the main vein, the same set of target nutrient unsolved values ​​are shared. The positional residuals between the nutrient response components on both sides and the corresponding reconstructed responses are calculated to form positional consistency constraints. The reconstruction deviation of the common overlapping zone, the reconstruction deviation of the non-overlapping shoulder zone, and the position residual were normalized and summed according to the variation range of the corresponding response components in the calibrated leaf samples. Under the condition that all target nutrient unsolved values ​​are non-negative, the target nutrient unsolved value with the smallest total deviation was selected. The median of the target nutrient values ​​obtained from all location groups is taken to form multi-nutrient diagnostic values.

8. A multi-nutrient diagnostic system for tea plants based on transmission spectroscopy, employing the multi-nutrient diagnostic method for tea plants based on transmission spectroscopy as described in any one of claims 1 to 7, characterized in that: Includes a pulse avoidance registration and transmission module, a dual-reference spectrum correction module, and an overlap constraint nutrient solution module; The vein-avoidance registration and transmission module is used to set registration sampling positions in the vein-avoidance area of ​​tea leaves to obtain dark reference spectrum, empty light path reference spectrum and transmission spectrum of leaves at each position. The dual-reference spectrum correction module is used to convert the leaf transmission spectrum into an optical density spectrum and correct structural interference based on the water absorption reference band and the scattering reference band. The overlapping constraint nutrient solution module is used to construct a nutrient response vector from the corrected optical density spectrum and to solve multiple nutrient diagnostic values ​​according to the overlapping constraint and position consistency constraint of the response band.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the tea tree multinutrient diagnosis method based on transmission spectroscopy as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the tea tree multinutrient diagnosis method based on transmission spectroscopy as described in any one of claims 1 to 7.