Plant water potential real-time nondestructive testing system and method

Abstract

The invention discloses a plant water potential real-time nondestructive detection system and method, and relates to the technical field of plant water potential detection.The method comprises the steps that firstly, multispectral reflection data and surface deformation data of a target leaf are obtained; then calculating a leaf curl index, and extracting a spectral distortion amount; judging whether the curl index exceeds a preset threshold value or not, if yes, switching the detection wavelength to calculate a compensated water potential value, and if not, calculating a standard water potential value; and finally, adjusting mapping relation parameters according to the water potential value time sequence change. Real-time performance is achieved, and a water potential result can be rapidly obtained; non-destructive testing is adopted, and plants are not damaged; multispectral and deformation data are integrated, and the mapping relation between the two is combined, so that the detection accuracy is high; the wavelength can be adaptively switched according to the curling condition of the blade, and the adaptability is high; parameters can be dynamically adjusted, so that a detection model is continuously optimized, and a reliable means is provided for monitoring the plant moisture state.

Description

Technical Field

[0001] The present invention relates to the technical field of plant water potential detection, and particularly to a real-time non-destructive detection system and method for plant water potential. Background Art

[0002] Plant water potential is a key indicator for measuring the water status of plants. It reflects the energy level of water in plants and is of crucial significance for understanding plant physiological processes, growth and development, and the ability to adapt to the environment. Accurately detecting plant water potential helps agricultural producers irrigate rationally, avoid over-irrigation or under-irrigation, thereby improving water resource utilization efficiency and ensuring crop yield and quality. In ecological research, monitoring plant water potential can help scientists understand the water cycle of ecosystems and the dynamic changes of plant communities. Existing plant water potential detection technologies have many deficiencies in terms of real-time performance, non-destructiveness, and accuracy, and cannot meet the requirements of modern agricultural precision irrigation and ecological environment monitoring. Summary of the Invention

[0003] Aiming at the deficiencies of the existing technology, the present invention provides a real-time non-destructive detection system and method for plant water potential.

[0004] To achieve the above object, the technical solution of the present invention is as follows:

[0005] In the first aspect, the present invention discloses a real-time non-destructive detection method for plant water potential, including the following steps:

[0006] Obtain the multi-spectral reflection data and surface deformation data of the target leaf. The multi-spectral reflection data includes the reflectance time series data within a preset wavelength range, and the surface deformation data includes the three-dimensional point cloud data of the leaf surface;

[0007] Calculate the curling index of the leaf according to the surface deformation data, and extract the spectral distortion amount of a preset water absorption peak according to the multi-spectral reflection data;

[0008] Judge whether the curling index exceeds a preset deformation threshold;

[0009] If it exceeds, switch the detection wavelength to the preset secondary absorption peak wavelength according to the mapping relationship between the spectral distortion amount and the curling index, and calculate the compensated water potential value based on the reflectance data in the secondary absorption peak band;

[0010] If it does not exceed, calculate the standard water potential value based on the reflectance data in the preset water absorption peak band;

[0011] Dynamically adjust the model parameters of the mapping relationship according to the temporal variation characteristics of the compensated water potential value or the standard water potential value.

[0012] In a second aspect, the present invention discloses a real-time non-destructive detection system for plant water potential, which uses the above-mentioned real-time non-destructive detection method for plant water potential, and includes:

[0013] A data acquisition module, configured to obtain multi-spectral reflection data and surface deformation data of the leaf;

[0014] A data processing module, configured to calculate a curling index, extract a spectral distortion variable, and determine whether it exceeds a deformation threshold;

[0015] A water potential calculation module, configured to select a preset water absorption peak band or a secondary absorption peak band according to the curling index, and calculate a standard water potential or a compensated water potential;

[0016] A dynamic compensation and optimization module, configured to dynamically adjust the model parameters of the mapping relationship according to the temporal variation characteristics of the compensated water potential value or the standard water potential value.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0018] 1. It can obtain the multi-spectral reflection data and surface deformation data of plant leaves in real time, and calculate the plant water potential value in a timely manner. Compared with traditional detection methods, the detection cycle is greatly shortened, and it can timely reflect the change of the water state of plants, providing timely and accurate information for decision-making;

[0019] 2. It adopts a non-contact detection method, and the water potential detection can be completed without damaging plant tissues. This not only avoids interfering with plant growth, but also enables continuous monitoring of the same plant, more accurately reflecting the water dynamics of plants at different growth stages;

[0020] 3. It comprehensively considers the multi-spectral reflection data and surface deformation data of plant leaves. By establishing a mapping relationship between the spectral distortion variable and the curling index, it effectively compensates for the influence of physical morphological changes such as leaf curling on spectral reflection, improving the accuracy of detection results; dynamically adjusting the model parameters of the mapping relationship according to the temporal variation characteristics of the water potential value further optimizes the detection model, making the detection results more in line with the actual situation. Description of the Drawings

[0021] The disclosure of the present invention will be described with reference to the accompanying drawings. It should be understood that the drawings are only for illustrative purposes and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components. Among them:

[0022] Figure 1 is the step flow chart of the present invention;

[0023] Figure 2 is the acquisition process diagram of the surface deformation data of the present invention;

[0024] Figure 3 It is the calculation flow chart of the curl index of the present invention;

[0025] Figure 4 It is the calculation flow chart of the spectral distortion variable of the present invention;

[0026] Figure 5 It is the calculation flow chart of water potential compensation based on the secondary absorption peak of the present invention;

[0027] Figure 6 It is the flow chart of dynamic adjustment of model parameters of the present invention;

[0028] Figure 7 It is the flow chart of dynamic adjustment of the preset deformation threshold of the present invention;

[0029] Figure 8 It is the flow chart of dynamic correction of the curl index weight of the present invention;

[0030] Figure 9 It is the composition diagram of the system modules of the present invention. Detailed implementation manners

[0031] It is easy to understand that according to the technical solution of the present invention, without changing the essence of the present invention, those of ordinary skill in the art can propose various structural ways and implementation ways that can be mutually replaced. Therefore, the following detailed implementation manners and the accompanying drawings are only exemplary descriptions of the technical solution of the present invention, and should not be regarded as all of the present invention or regarded as a limitation or restriction on the technical solution of the present invention.

[0032] Application overview:

[0033] In the prior art, the detection of plant water potential mostly relies on in vitro sampling or the inversion of a single spectral parameter, and it is difficult to balance real-time performance and accuracy. When the leaf undergoes curling deformation in the traditional method, the spectral characteristics are easily interfered by surface wrinkles, resulting in an increase in the error of water potential inversion. Existing devices cannot synchronously perceive the influence of leaf morphological changes on optical detection. Especially when the leaf curling intensifies under drought stress, the single-wavelength detection model will show systematic deviations and it is difficult to meet the requirements of precise irrigation control.

[0034] To solve the above problems, the inventor found that there is an associated law between the degree of leaf curling and the distortion variable of a specific absorption peak, and an error compensation is achieved by establishing a morphology-spectroscopy coupling model. During the research process, it was found that the main absorption peak is highly sensitive to morphological changes, and although the secondary absorption peak has low sensitivity but good stability, thus the idea of dynamically switching the detection wavelength according to the deformation threshold was proposed. Further verified by experiments, the mapping relationship between the curl index and the distortion variable was introduced into the dynamic adjustment mechanism of model parameters to form a closed-loop feedback system.

[0035] Specifically, the detection system first synchronously acquires the hyperspectral image and three-dimensional morphological data of the leaf. By analyzing the longitudinal curvature and transverse fold density in the three-dimensional point cloud data, a quantitative index characterizing the degree of leaf curling is calculated. At the same time, the wavelength shift of the main water absorption peak band is extracted from the multispectral data as the spectral distortion variable. When the detected curling index exceeds the set threshold, the system automatically switches to the secondary absorption peak wavelength for detection. Utilizing the characteristic that this band is less affected by deformation, the water potential value compensation calculation is carried out in combination with the pre-established relationship model between the distortion variable and the curling index. During the continuous detection process, according to the temporal variation characteristics of the water potential value, the system real-time corrects the mapping model parameters through a feedback mechanism to form a dynamically optimized closed loop. For the case where the deformation threshold is not exceeded, the standard inversion model of the main absorption peak band is continued to output the water potential value.

[0036] Compared with the prior art, the traditional method relies on a single spectral feature and lacks a morphological compensation mechanism, which is prone to systematic errors when the leaf curls. This solution innovatively integrates multispectral and three-dimensional morphological data, and realizes the dynamic compensation of detection errors by establishing a deformation-spectral coupling model. Different from the existing static detection models, this solution can intelligently switch the detection mode according to the real-time deformation degree, and continuously optimize the model parameters through a closed-loop feedback, significantly improving the detection reliability under complex working conditions.

[0037] Through the above technical solution, this application effectively overcomes the problem of spectral feature distortion caused by leaf deformation, improves the water potential inversion accuracy while ensuring the detection real-time performance. The dynamic wavelength switching mechanism takes into account the high sensitivity of the main absorption peak and the stability advantage of the secondary absorption peak, and the model parameter adaptive adjustment function ensures the accuracy of long-term detection. This method provides a reliable technical means for the continuous non-destructive monitoring of plant water status, and is particularly suitable for the precise irrigation control scenario under drought stress.

[0038] After introducing the basic concept of the present invention, the embodiments of the present invention will be specifically introduced below with reference to the accompanying drawings.

[0039] Embodiment 1:

[0040] As Figure 1 shown, the real-time non-destructive detection method for plant water potential includes the following steps:

[0041] Step 1: Obtain the multispectral reflection data and surface deformation data of the target leaf. The multispectral reflection data includes the reflectance time series data within a preset wavelength range, and the surface deformation data includes the three-dimensional point cloud data of the leaf surface.

[0042] Multispectral reflection data refers to the time-series information of the reflectance of plant leaves at different wavelengths obtained through a spectroscopic device. Specifically, it can be achieved by continuously scanning in the near-infrared band using a hyperspectral imager. This data is used to capture the dynamic changes of water absorption characteristics over time. Surface deformation data refers to the quantitative parameters reflecting the three-dimensional morphological characteristics of the leaves. Specifically, structured light projection technology can be used in combination with a phase unwrapping algorithm to generate a point cloud model. This data is used to identify deformation states such as leaf curling and wrinkling.

[0043] The preset wavelength range is the range where the characteristic absorption peaks related to plant water potential are located, and this range covers at least two characteristic absorption peak intervals, with one of the intervals being the main detection interval.

[0044] The preset wavelength range refers to a specific spectral range covering the characteristic absorption peaks of plant water. For example, a region containing two water absorption peaks at 970 nm and 1450 nm in the near-infrared band can be selected, and the main detection interval can be set as the band near 970 nm.

[0045] The preset wavelength range needs to cover at least two different water absorption peaks. For example, it can cover the two bands of 970 nm and 1450 nm to ensure that when one band is interfered with, it can be switched to the other band.

[0046] By setting a detection interval containing multiple absorption peaks and establishing a primary and secondary absorption peak switching mechanism, it is possible to automatically select the band with less interference for compensation calculation when the leaf deforms. For example, in the prior art, the error of a detection system based on a single 970 nm band can reach 15% when the leaf curling index exceeds 0.5, while in this solution, the error can be controlled within 8% by switching to the secondary absorption peak.

[0047] As Figure 2 shown, it is the acquisition process of the surface deformation data; including:

[0048] Project a preset stripe pattern onto the leaf surface, collect the deformed stripe image, and generate the three-dimensional point cloud data based on the phase unwrapping algorithm; the three-dimensional point cloud data contains the quantitative parameters of the longitudinal curvature and transverse wrinkle density of the leaf.

[0049] The preset stripe pattern refers to an optical encoding pattern with specific spatial frequency and contrast, which can be projected onto the leaf surface by means of laser interference or digital projection specifically, and is used to reflect the microscopic deformation characteristics of the leaf surface through the degree of stripe deformation. The phase unwrapping algorithm refers to calculating the phase distribution change of the stripe image, and specifically, methods such as multi-frequency heterodyne or Gray code-assisted phase unwrapping can be adopted to achieve the precise reconstruction of three-dimensional deformation data. The three-dimensional point cloud data refers to the dataset of the leaf surface topological structure represented by three-dimensional coordinates, which can be generated by binocular vision or structured light scanning system specifically, and is used to quantify the geometric parameters of the longitudinal bending and transverse wrinkles of the leaf. The longitudinal curvature refers to the degree of bending along the main vein direction of the leaf, which can be specifically quantified by calculating the included angle of the normal vectors of adjacent points in the point cloud data, and reflects the overall curling state of the leaf. The transverse wrinkle density refers to the number of surface wrinkles perpendicular to the main vein direction, which can be specifically measured by counting the number of peaks and valleys within a unit length, and characterizes the local deformation characteristics of the leaf.

[0050] After projecting a regularly distributed stripe pattern on the leaf surface, the microscopic deformation of the leaf caused by water loss will cause geometric distortion of the projected stripes. By collecting the deformed stripe image with a high-resolution camera and combining the phase unwrapping algorithm to analyze the phase change of the stripes, the three-dimensional spatial coordinates of the leaf surface can be reconstructed. In this process, the longitudinal curvature is calculated by analyzing the height change curve of the main vein axis in the point cloud data, and the transverse wrinkle density is determined by detecting the surface undulation frequency perpendicular to the main vein direction. These two quantization parameters together constitute the characteristic description of the leaf deformation, providing reliable three-dimensional deformation data support for subsequent water potential detection.

[0051] Compared with the prior art, traditional methods rely on contact measurement or two-dimensional image analysis and cannot synchronously obtain the three-dimensional deformation characteristics of the leaf. For example, contact sensors will damage the leaf surface structure, and it is difficult to distinguish the coupled deformation of longitudinal bending and transverse wrinkles by two-dimensional image analysis. This solution can accurately quantify the three-dimensional deformation characteristics of the leaf during water loss through non-contact three-dimensional scanning technology combined with the stripe phase analysis algorithm, overcoming the problem of insufficient spatial resolution of traditional methods.

[0052] Through the above technical solution, this application realizes the non-destructive quantitative detection of the three-dimensional deformation characteristics of the leaf surface, providing a high-precision deformation data basis for accurately evaluating the plant water potential. By synchronously obtaining the longitudinal curvature and transverse wrinkle density parameters, it can comprehensively reflect the spatial deformation characteristics of the leaf at different water loss stages, effectively improving the anti-interference ability of the water potential detection model.

[0053] Step 2: Calculate the curling index of the leaf according to the surface deformation data, and extract the spectral distortion amount of the preset water absorption peak according to the multi-spectral reflection data.

[0054] The curling index is a composite index that comprehensively reflects longitudinal bending and transverse wrinkling. Specifically, it can be obtained by the weighted sum of curvature calculation and density statistics, and is used to quantify the degree of leaf deformation. The spectral distortion amount refers to the offset of the wavelength position of the main absorption peak. Specifically, it can be determined by comparing the difference between the current detection wavelength and the reference wavelength, and reflects the interference degree of leaf deformation on spectral characteristics.

[0055] As Figure 3 shown, it is the calculation flow chart of the curling index; calculating the curling index of the leaf according to the surface deformation data includes:

[0056] Performing a weighted sum of the longitudinal curvature and the transverse wrinkling density, and the weight coefficients are predefined according to the leaf type;

[0057] The calculation formula of the curling index is:

[0058] Curling index = first preset coefficient × longitudinal curvature + second preset coefficient × transverse wrinkling density; where the sum of the first preset coefficient and the second preset coefficient is 1.

[0059] The longitudinal curvature refers to the degree of bending of the leaf along the main vein direction. Specifically, it can be quantified by the reciprocal of the radius of curvature extracted from the three-dimensional point cloud data, and is used to characterize the response degree of the longitudinal deformation of the leaf to the water state when calculating the curling index. The transverse wrinkling density refers to the density of the transverse ripples on the leaf surface. Specifically, it can be statistically calculated by the number of wave peaks per unit area in the point cloud data, and is used to reflect the optical detection error caused by the transverse deformation when calculating the curling index. The weight coefficient refers to the contribution ratio of the longitudinal curvature and the transverse wrinkling density in the curling index. Specifically, it can be preset according to the morphological characteristics of different plant leaves. For example, higher weight can be assigned to the transverse wrinkling density for gramineous plants, and the weight of the longitudinal curvature can be increased for woody plants, so as to optimize the deformation characteristics for different leaf types.

[0060] Specifically, after generating the three-dimensional point cloud data, the longitudinal curvature is obtained by extracting the curvature parameters of each point along the main vein of the leaf, and the transverse wrinkling density is calculated by statistically analyzing the distribution density of the transverse ripples on the leaf surface. According to the morphological characteristics of different plant leaves, such as the difference in the deformation patterns of coniferous tree leaves and broad-leaved plant leaves, the weight distribution ratio of the longitudinal curvature and the transverse wrinkling density is preset in advance. After linearly combining the two according to the preset weights, the curling index can reflect the comprehensive deformation state of the longitudinal curling and transverse wrinkling of the leaf at the same time. For example, in the detection of corn leaves, the influence of the transverse wrinkling density on spectral distortion is more significant. The second preset coefficient can be set to 0.6, and the first preset coefficient is correspondingly set to 0.4, so that the curling index can more accurately characterize the deformation characteristics of this type of leaf.

[0061] Complete calculation example of the curling index: corn leaf;

[0062] Input data:

[0063] 3D point cloud analysis results:

[0064] Average longitudinal curvature: 0.45 mm -1 ;

[0065] Transverse fold density: 7.2 per cm 2 ;

[0066] Preset weight (corn):

[0067] Longitudinal coefficient (first preset coefficient): 0.4;

[0068] Transverse coefficient (second preset coefficient): 0.6;

[0069] Calculation process:

[0070] Coiling index = 0.4 × 0.45 + 0.6 × 7.2 / 10 = 0.18 + 0.432 = 0.612;

[0071] The transverse fold density is normalized to the 0 - 1 range by dividing the fold density by 10.

[0072] As shown in the following table, it is a multi - plant type detection template:

[0073] Plant name First preset coefficient Second preset coefficient Recommended detection wavelength Wheat 0.35 0.65 980nm + 1440nm Rice 0.3 0.7 970nm + 1450nm Apple leaf 0.5 0.5 990nm + 1420nm

[0074] Compared with the prior art, the traditional method only judges the degree of leaf coiling through a single deformation parameter. For example, only using the longitudinal bending angle or the number of folds alone cannot accurately reflect the composite characteristics of leaf deformation of different plants. This solution combines and optimizes two types of deformation parameters with different physical meanings by introducing a weighted summation calculation method, which can be specifically adapted to the physiological structure characteristics of different plant leaves and solves the problem of poor type adaptability caused by single - parameter detection.

[0075] Through the above - mentioned technical solution, this application can dynamically adjust the weight distribution of deformation characteristics according to the plant leaf type, so that the coiling index simultaneously includes the comprehensive influence of longitudinal and transverse deformations, effectively improving the accuracy of water potential detection for different types of plants. By predefined weight coefficients, it can quickly adapt to the leaf characteristics of different crops such as wheat and rice, avoiding the problem of amplified detection errors caused by leaf morphological differences.

[0076] As Figure 4 shown, it is a calculation flow chart of spectral distortion amount; the spectral distortion amount of the preset water absorption peak is the wavelength offset of the preset water absorption peak band, and the wavelength offset is the difference between the current detection wavelength and the wavelength reference value under the standard water potential;

[0077] The preset water absorption peak includes at least one absorption peak within a preset wavelength range. The spectral distortion amount of the preset water absorption peak refers to the offset of the wavelength position of the main absorption peak band relative to the standard state. Specifically, a high-resolution spectrometer can be used to detect the central wavelength of the current absorption peak band and calculate the difference from the reference wavelength corresponding to the pre-stored standard water potential. This offset reflects the spectral feature deformation caused by leaf curling or changes in the water state. During the detection process, when the leaf curls, the reflectance of the main absorption peak band will be interfered by the surface deformation, resulting in an offset of the wavelength reference value. By calculating the difference between the currently detected wavelength and the standard reference value in real time, the spectral distortion amount can be quantified.

[0078] Complete calculation example of spectral distortion amount: Wheat leaf;

[0079] Detection data:

[0080] Current central wavelength of 970nm peak: 973.6nm;

[0081] Standard reference value: 970.2nm (under the standard water potential of this variety);

[0082] Detection value of the secondary absorption peak (1440nm): 1441.3nm.

[0083] Calculation process:

[0084] The spectral distortion amount of the main peak is: 973.6 - 970.2 = +3.4nm.

[0085] The spectral distortion amount of the secondary peak is: 1441.3 - 1439.8 = +1.5nm.

[0086] Step 3, determine whether the curling index exceeds a preset deformation threshold.

[0087] The preset deformation threshold, as the critical value for judging whether the leaf deformation significantly affects the detection accuracy, is usually calibrated through experiments and can be adjusted according to actual use.

[0088] Example: Real-time monitoring of corn fields;

[0089] Data input:

[0090] Curling index = 0.62 (longitudinal curvature 0.3 + transverse fold density 6.2 / 10);

[0091] Preset threshold = 0.5;

[0092] Judgment process:

[0093] 0.62 > 0.5 → Trigger the compensation mechanism;

[0094] System actions:

[0095] Proceed to Step 4 for subsequent actions.

[0096] Step 4: If it exceeds, according to the mapping relationship between the spectral distortion amount and the curling index, switch the detection wavelength to the preset secondary absorption peak wavelength, and calculate the compensated water potential value based on the reflectance data in the secondary absorption peak band.

[0097] The preset secondary absorption peak is the wavelength range where the secondary absorption peak, which is lower than the preset main water absorption peak band affected by leaf curling and is related to plant water potential, is located. When it is detected that the offset of the main absorption peak band exceeds the threshold, it automatically switches to the secondary absorption peak band. The selection of this band is based on the observation results in the experimental data with a lower degree of influence by leaf curling. For example, at the same curling degree, the spectral distortion amount in the 1450 nm band is reduced by about 40% compared to the 970 nm band.

[0098] Through the above technical solutions, the problem of spectral data distortion caused by leaf curling can be effectively solved, and the detection accuracy can still be maintained when the plant undergoes moderate or more severe water stress. By presetting the multi-absorption peak detection interval and the dynamic switching mechanism, the systematic error caused by the interference of a single band is avoided, and at the same time, the degree of deformation influence is quantified by using the wavelength offset amount, providing a reliable data basis for the water potential compensation calculation.

[0099] Further explanations regarding the preset wavelength interval, preset water absorption peak, and preset secondary absorption peak:

[0100] Preset wavelength interval: It is the range where the characteristic absorption peaks related to plant water potential are located. This range includes at least two characteristic absorption peak intervals, and one of the intervals is selected as the main detection interval. The reason for covering at least two characteristic absorption peak intervals is to obtain information related to plant water potential from multiple perspectives and improve the accuracy and reliability of detection. The main detection interval is usually the interval with obvious absorption peak characteristics and strong correlation with plant water potential.

[0101] Preset water absorption peak: Plants have different light absorption characteristics at different wavelengths, and absorption peaks will appear at certain wavelengths related to plant water potential. The preset water absorption peak is the absorption peak selected in advance for detecting plant water potential.

[0102] Preset secondary absorption peak: It refers to the absorption peak in a certain wavelength interval. The absorption peak in this wavelength interval is less affected by leaf curling than the preset main water absorption peak band and is related to plant water potential. When the leaf curling degree is large and the detection result of the preset main water absorption peak band may be greatly interfered, the detection can be switched to the wavelength interval where the preset secondary absorption peak is located to obtain more accurate plant water potential data.

[0103] For example: When studying the water potential of a certain type of plant, through a large number of experiments and analyses, it is found that the characteristic absorption peaks related to the water potential of this plant mainly appear in the wavelength range of 900 - 2500 nm, and 900 - 2500 nm is the preset wavelength interval. Within this range, there are obvious characteristic absorption peaks in the two intervals of 1450 nm - 1470 nm and 1930 nm - 1950 nm. The interval of 1450 nm - 1470 nm is used as the main detection interval, that is, the interval where the preset water absorption peak is located.

[0104] When the leaf is not curled or the curling degree is small, the water potential of the plant is mainly calculated by detecting the absorption peak in the interval of 1450 nm - 1470 nm. However, when the curling degree of the leaf exceeds a certain threshold, the absorption peak in the interval of 1450 nm - 1470 nm may be distorted due to the curling of the leaf, resulting in inaccurate detection results. At this time, it is found that the absorption peak in the wavelength interval of 1640 nm - 1660 nm is relatively less affected by leaf curling and is also somewhat correlated with the plant water potential. Then, the interval of 1640 nm - 1660 nm can be used as the preset secondary absorption peak, and the detection can be switched to this interval to calculate the compensated water potential value.

[0105] As Figure 5 shown, it is the flowchart of water potential compensation calculation based on the secondary absorption peak; the process of calculating the compensated water potential value based on the reflectivity data of the secondary absorption peak band is as follows:

[0106] Extract the second derivative eigenvalue of the reflectivity of the secondary absorption peak band, and input the second derivative eigenvalue of the reflectivity into a pre-calibrated piecewise water potential inversion function;

[0107] The piecewise water potential inversion function includes at least two exponential function intervals, and each interval is divided according to the range of the curling index.

[0108] The second derivative eigenvalue of the reflectivity refers to the numerical feature obtained by performing second-order differential processing on the reflectivity curve of the secondary absorption peak band. Specifically, it can be realized by using the Savitzky-Golay filter combined with the numerical differential algorithm to eliminate the reflectivity noise caused by the microscopic deformation of the leaf surface. The pre-calibrated piecewise water potential inversion function refers to a mathematical model composed of multiple exponential function segments established based on experimental data. Specifically, it can be realized by piecewise fitting the mapping relationship curve between the second derivative of the reflectivity and the water potential value under different curling degrees to adapt to the influence of the change in the curling degree of the leaf on the inversion accuracy. The exponential function interval refers to multiple function segments divided based on the numerical range of the curling index. Specifically, it can be divided by clustering analysis or decision tree algorithm for the data distribution under different curling degrees to minimize the water potential inversion error within each interval.

[0109] Specifically, when it is detected that the degree of leaf curling exceeds the set threshold, the system automatically switches to the secondary absorption peak band for measurement. After the reflectivity data in the secondary absorption peak band is subjected to a second derivative transformation, it can effectively suppress the distortion of the reflectivity signal caused by leaf curling. According to the numerically calculated curling index in real time, the system selects the exponential function in the corresponding interval for water potential inversion. For example, when the curling index is in the interval of 0.3 - 0.5, the first set of exponential coefficients is used, and when it is in the interval of 0.5 - 0.7, the second set of exponential coefficients is used. For leaves that have not undergone significant curling, the first derivative feature of the main absorption peak is directly adopted, and the standard water potential value is quickly calculated through a pre-calibrated linear model. This measurement strategy for different scenarios not only ensures the detection efficiency under normal conditions but also maintains the measurement accuracy when the leaf deforms.

[0110] Example: Wheat leaf;

[0111] Input data:

[0112] Reflectivity data in the secondary absorption peak band:

[0113] Wavelength: 1445 - 1455 nm band;

[0114] Reflectivity values: [45.2, 43.8, 42.1, 40.5, 39.3, 38.8, 39.1, 40.2, 42.0]%;

[0115] Current curling index: 0.52 (belonging to the 0.3 - 0.7 interval).

[0116] Eigenvalue calculation process:

[0117] First derivative:

[0118] At 1447.5 nm: (43.8 - 45.2) / 5 = -0.28;

[0119] At 1452.5 nm: (40.2 - 39.1) / 5 = +0.22.

[0120] Second derivative:

[0121] At 1450 nm (0.22 - (-0.28)) / 10 = 0.05;

[0122] The final eigenvalue D2 = -0.08 after SG filtering.

[0123] Water potential inversion calculation:

[0124] Select the function in the corresponding interval:

[0125] Ψ = 3.1×exp(-2.8×(-0.08)) + 0.2 = 3.1×1.25 + 0.2 = 4.08 MPa;

[0126] Verification value of the pressure chamber: 4.12 MPa (error 1%).

[0127] Step Five: If not exceeded, calculate the standard water potential value based on the reflectance data in the preset water absorption peak band.

[0128] The process of calculating the standard water potential value based on the reflectance data in the preset water absorption peak band is as follows:

[0129] Extract the reflectance eigenvalue in the preset water absorption peak band, and the reflectance eigenvalue is the first derivative of the reflectance in this band;

[0130] Input the reflectance eigenvalue into a pre-calibrated linear water potential inversion model, and the linear water potential inversion model is obtained by fitting the reflectance data under the standard water potential.

[0131] The first derivative of the reflectance refers to the numerical feature obtained by performing a first-order differential process on the reflectance curve in the preset water absorption peak band. Specifically, the slope feature at the edge of the absorption peak can be extracted by the central difference method to characterize the linear relationship between the water potential and the spectral feature under the standard state. The pre-calibrated linear water potential inversion model refers to a linear regression model established through the reflectance data under the standard water potential condition. Specifically, the corresponding relationship between the first derivative of the reflectance and the water potential value can be fitted by the least squares method to provide a reference water potential measurement when the leaf does not undergo significant deformation.

[0132] Compared with the prior art, the traditional method only uses the single-order derivative feature of a single absorption peak for water potential inversion and cannot effectively distinguish the deformation interference from the real water potential change when the leaf curls. In this solution, through the second-order derivative processing combined with the design of a piecewise function, it can adaptively select the optimal inversion model according to different curling degrees. At the same time, by using the complementary characteristics of the primary and secondary absorption peaks, the applicability of the detection scenario is extended while ensuring the measurement accuracy.

[0133] Through the above technical solution, this application can automatically select the secondary absorption peak band and eliminate the deformation interference through the extraction of the second-order derivative feature when the leaf undergoes curling deformation, and realize the accurate compensation calculation of the water potential value in combination with the pre-calibrated piecewise inversion model. For leaves that do not undergo significant deformation, the first derivative feature of the primary absorption peak is used for fast linear inversion, taking into account both the detection efficiency and accuracy. This measurement strategy for different scenarios effectively solves the technical problem of inaccurate measurement of the traditional method under the leaf deformation condition, and at the same time avoids the cumbersome operation of manual intervention to adjust the measurement parameters.

[0134] Example: Wheat leaf;

[0135] Reflectivity data:

[0136] 968 nm: 45%;

[0137] 970 nm: 40%;

[0138] 972 nm: 46%;

[0139] Derivative calculation:

[0140] (40 - 45) / 2 = -2.5 (left);

[0141] (46 - 40) / 2 = +3.0 (right);

[0142] Eigenvalue: (-2.5 + 3.0) / 2 = +0.25% / nm;

[0143] Water potential inversion:

[0144] Ψ = 2.3×0.25 - 0.5 = -0.075 MPa (normal range).

[0145] Step Six: Dynamically adjust the model parameters of the mapping relationship according to the temporal variation characteristics of the compensated water potential value or the standard water potential value.

[0146] As Figure 6 shown, it is the flowchart for dynamic adjustment of model parameters; dynamically adjusting the model parameters of the mapping relationship includes:

[0147] During continuous detection, record the real-time corresponding relationship between the curl index and the spectral distortion variable;

[0148] When the deviation of the corresponding relationship exceeds the preset feedback threshold in N consecutive detections, update the linear coefficient of the mapping relationship based on the least squares method;

[0149] where N is an integer not less than 2.

[0150] Recording the real-time corresponding relationship during continuous detection means synchronously storing the values of the curl index and the spectral distortion variable changing with time, and specifically, data matching can be achieved through timestamp alignment. The deviation exceeding the preset feedback threshold means that the combined calculation result of the difference in curl index and the difference in spectral distortion variable between two adjacent detections exceeds the allowable range, and specifically, it can be quantitatively judged by using the deviation factor calculation formula. Updating the linear coefficient by the least squares method means fitting the optimal mapping relationship parameters according to historical data, and specifically, coefficient optimization can be achieved by constructing a model that minimizes the sum of squared errors. N being an integer not less than 2 means that the determination times for triggering parameter update include at least two consecutive detections, for example, it can be set to 3 or 5 times to ensure adjustment stability.

[0151] Specifically, during the continuous detection process, the system automatically records the correspondence between the curling index and the spectral distortion to form a time series data set. When it is detected that the deviation factor corresponding to N consecutive data exceeds the threshold, the system automatically calls the least squares algorithm to refit the linear coefficient of the mapping relationship. For example, when the deviation factor exceeds the threshold three times in a row, the system uses the most recent 10 sets of data as input, calculates the new linear coefficient through the least squares method, and replaces the original parameters. This process does not require human intervention and can dynamically adjust the model parameters according to the real-time changes in plant water potential to ensure the accuracy of the detection results.

[0152] Compared with existing technologies, existing methods usually use fixed model parameters for water potential calculation, which cannot adapt to the dynamic changes in the relationship between leaf deformation and spectral distortion. However, this solution realizes adaptive adjustment of the model through real-time data recording and deviation trigger mechanism, combined with least squares parameter optimization, and significantly improves the detection stability in complex environments.

[0153] Through the above technical solution, this application effectively solves the problem of detection error accumulation caused by the solidification of model parameters in traditional methods, and can automatically correct the mapping relationship between curl index and spectral distortion during plant growth, ensuring the calculation accuracy of water potential values during long-term continuous detection. For example, when plants experience diurnal water changes or environmental temperature and humidity fluctuations, the system can automatically identify the dynamic association between leaf deformation characteristics and spectral response, and adjust model parameters in time to match the current physiological state.

[0154] The deviation calculation process of the corresponding relationship is:

[0155] Extract the curl index difference ΔCI and spectral distortion difference Δλ between two adjacent tests;

[0156] Calculate the deviation factor: Deviation factor = |Δλ-(a×ΔCI+b)|, where a and b are the linear coefficients of the current mapping relationship respectively;

[0157] When the deviation factor exceeds a preset feedback threshold, the linear coefficient of the current mapping relationship is updated.

[0158] The difference in curl index refers to the difference in the degree of leaf surface deformation between two consecutive detections. Specifically, it can be achieved through the algebraic difference between the curl index at the current moment and the previous detection value, and is used to characterize the dynamic change trend of the leaf curling state. The difference in spectral distortion amount refers to the difference in the wavelength offset amount of the water absorption peak between two consecutive detections. Specifically, it can be achieved through the difference between the wavelength offset amount at the current moment and the historical value, and is used to quantify the fluctuation amplitude of spectral characteristics with leaf deformation. The deviation factor is an index that reflects the stability of the correlation between the curl index and the spectral distortion amount. Specifically, it adopts the combined operation form of the product of the linear coefficient and the difference, and is used to determine whether the current mapping relationship deviates from the actual correlation law. The linear coefficient is a parameter that reflects the strength of the correlation between the curl index and the spectral distortion amount. Specifically, the initial value can be obtained through regression analysis of historical data and dynamically corrected during operation.

[0159] Specifically, during continuous detection, the curl index and the spectral distortion amount obtained in two consecutive times are respectively extracted and the difference is calculated. When the deviation factor exceeds the preset threshold, it indicates that the current mapping relationship can no longer accurately reflect the actual correlation between the curling deformation and the spectral characteristics. At this time, the linear coefficient update process is started. The update process uses the least squares method to fit the latest data sequence and generates a new linear coefficient to replace the original parameter. This deviation-triggered mechanism can ensure that the model parameters are always synchronized with the real-time detection data.

[0160] Example: Wheat field monitoring;

[0161] Background conditions:

[0162] Original mapping relationship: Δλ = 5.0 × CI;

[0163] Preset threshold: Deviation factor > 0.25;

[0164] Trigger times: N = 3;

[0165] Abnormal data sequence:

[0166] Time CI Δλ / nm Deviation factor Overlimit count 09:00 0.40 +2.0 - - 09:05 0.42 +3.1 1.1 1 09:10 0.45 +3.8 1.3 2 09:15 0.47 +4.5 1.4 3 (trigger)

[0167] Parameter update process:

[0168] Select the nearest 10 groups of data to fit the new relationship:

[0169] Calculate the new coefficient: Δλ = 7.2 × CI;

[0170] Verification after update:

[0171] Data at 09:20: CI = 0.49, actual Δλ = 3.9 nm;

[0172] New model prediction: 7.2 × 0.49 = 3.53 nm;

[0173] The error has decreased from 42% to 9% in the original model.

[0174] The threshold selection principles for some plants are shown in the following table:

[0175] Plant type Recommended threshold Scientific basis Corn 0.30 The leaf deformation response is relatively fast Wheat 0.25 The waxy layer results in a gentle change Fruit tree 0.35 The lignified tissue buffers the deformation

[0176] Influence of the value of N:

[0177] N = 2: High sensitivity but prone to misadjustment;

[0178] N = 5: Strong stability may cause delay;

[0179] Compromise solution: N = 3 (recommended).

[0180] Typical application scenarios:

[0181] Adaptation to seasonal conversion:

[0182] Spring → Summer: The increase in temperature causes changes in the leaf elasticity, and the system automatically adjusts the coefficient from 5.0 to 6.3.

[0183] Handling of variety differences:

[0184] When a new variety of leaf is detected: If the deviation exceeds the threshold for 5 consecutive times, a new mapping relationship file will be automatically created.

[0185] Compensation for equipment aging:

[0186] When the performance of the spectrometer decays: The accuracy is maintained through gradual adjustment of the coefficient, and at the same time, a maintenance reminder is triggered.

[0187] Verification of implementation effect:

[0188] Measured data in the vineyard:

[0189] Average error in the 3 days before adjustment: 14%;

[0190] After enabling dynamic adjustment:

[0191] Error in the first week: 9%;

[0192] Error in the fourth week: 5% (3 automatic adjustments completed).

[0193] Compared with the existing technology, traditional methods usually adopt fixed time intervals or manually set parameter update rules, and cannot effectively identify abnormal offsets in the correlation relationship. There is a lack of quantitative evaluation indicators for the dynamic relationship between the curl index and the spectral distortion amount in the existing technology, resulting in the lag of model parameter updates behind the actual state changes. This solution can capture abnormal fluctuations in the correlation relationship in real time by constructing a deviation factor calculation model, and automatically trigger parameter optimization through mathematical operations, significantly improving the adaptive ability of the model.

[0194] Through the above technical solution, this application solves the problem of model misalignment caused by fixed parameters in traditional detection methods. By using a deviation factor to evaluate the correlation stability between the curling deformation and spectral features in real time, parameter updates are automatically triggered when abnormal deviations are detected, ensuring that the water potential calculation model always matches the current plant physiological state. This solution effectively reduces the impact of cumulative errors caused by leaf deformation on the detection results and improves the long-term stability of water potential detection.

[0195] As Figure 7 shown, it is a flow chart for dynamically adjusting the preset deformation threshold; dynamically adjusting the model parameters of the mapping relationship further includes: adjusting the value of the preset deformation threshold according to the standard deviation of the compensated water potential values within a preset time period. The specific process is as follows:

[0196] When the standard deviation is less than the preset stability threshold, reduce the preset deformation threshold by a first preset ratio;

[0197] When the standard deviation is greater than the preset stability threshold, increase the preset deformation threshold by a second preset ratio.

[0198] The preset time period refers to a continuous time window used to statistically analyze the data fluctuations of the compensated water potential values, which can be specifically implemented using a fixed time length or a variable time interval. For example, it can be set to 30 minutes or 5 consecutive detection cycles. The preset stability threshold refers to the critical value for determining the stability of the water potential value data, which can be specifically determined through statistical analysis of historical data. For example, it is set based on the mean standard deviation in the normal state of the same plant. The first preset ratio and the second preset ratio refer to the amplitude parameters for adjusting the preset deformation threshold, which can be specifically implemented using an equal ratio or a gradient ratio method. For example, the first preset ratio is 5% and the second preset ratio is 3%. The calculation process of the standard deviation uses statistical methods to quantify the degree of dispersion of the compensated water potential values and is used to characterize the volatility of the detection results.

[0199] Specifically, during continuous detection, the standard deviation of the compensated water potential values reflects the stability of the detection system. When the standard deviation is low, it indicates that the water potential value fluctuates less in the current detection environment. At this time, reducing the preset deformation threshold can improve the sensitivity to slight leaf deformation and avoid missing early water stress signals. On the contrary, when the standard deviation is high, it indicates the existence of external interference or system noise. At this time, increasing the preset deformation threshold can reduce false positives and preferentially use the main absorption peak data for calculation. The adjustment amplitude of the preset ratio needs to be associated with the degree of deviation of the standard deviation from the stability threshold. For example, when the standard deviation is 20% lower than the threshold, the first preset ratio adjustment is triggered. This dynamic adjustment mechanism is realized through feedback control logic, making the setting of the deformation threshold match the stability of the real-time detection environment.

[0200] Example: Maize field monitoring;

[0201] Initial state:

[0202] Preset deformation threshold: 0.50;

[0203] Stability threshold: 0.15 MPa;

[0204] Water potential data for the last 30 minutes: [-1.2, -1.3, -1.25, -1.18, -1.22] MPa.

[0205] Standard deviation calculation result:

[0206] Mean: -1.23 MPa;

[0207] Standard deviation: 0.048 MPa.

[0208] Threshold adjustment:

[0209] 0.048 < 0.15 → Trigger the reduction rule;

[0210] New threshold = 0.50 × (1 - 5%) = 0.475.

[0211] Effect verification:

[0212] Before adjustment: Ignore slight curling between 0.45 - 0.50;

[0213] After adjustment: Can identify curling above 0.45;

[0214] Subsequent detection shows:

[0215] Curling index 0.48 (previously not alarming), actual water potential -1.6 MPa (already water - deficient).

[0216] Compared with the prior art, in the traditional method, the deformation threshold is usually a fixed value and cannot adapt to the dynamic changes of different environmental conditions or plant states. For example, in a high - temperature or high - humidity environment, the natural deformation of leaves may intensify, resulting in frequent triggering of the fixed threshold and switching of secondary absorption peaks, increasing calculation errors. In this solution, through the linked adjustment of the standard deviation and the deformation threshold, it can actively reduce the detection sensitivity when the data stability is high to capture subtle changes; when the data fluctuates greatly, it can improve the detection fault tolerance to ensure the effective utilization of the main absorption peak data, thus maintaining the detection accuracy under complex environmental conditions.

[0217] Through the above - mentioned technical solution, this application solves the problem of cumulative detection errors caused by a fixed deformation threshold in the prior art and realizes the dynamic optimization of the threshold based on environmental adaptability. This solution can automatically adjust the detection strategy according to the real - time data stability. While ensuring the reliability of the main absorption peak data, it can effectively capture the early deformation characteristics of leaves and improve the robustness and adaptability of the water potential detection system.

[0218] Such asFigure 8 As shown, it is the flow chart of dynamic correction of the curl index weight;

[0219] After the wavelength is switched, the fluctuation range of the second derivative eigenvalue of the reflectance in the secondary absorption peak band is monitored in real time;

[0220] When the fluctuation range exceeds the preset fluctuation threshold, a feedback instruction is generated to re-collect the surface deformation data;

[0221] Based on the re-collected data, the calculation weight coefficient of the curl index is corrected;

[0222] The fluctuation range of the second derivative eigenvalue of the reflectance refers to the amplitude change interval of the second differential curve of the reflectance in the secondary absorption peak band. Specifically, it can be realized by calculating the variance of the sliding window, which is used to characterize the influence degree of the leaf surface deformation on the spectral data. Among them, the preset fluctuation threshold refers to the allowable upper limit value of the change of the second derivative of the reflectance set in advance. Specifically, it can be determined by historical data statistics or experimental calibration, which is used to judge whether it is necessary to trigger the data re-collection mechanism.

[0223] The generation logic of the feedback instruction is as follows:

[0224] Statistical range of the second derivative eigenvalue of the reflectance within the current detection period;

[0225] When the ratio of the range to the historical mean exceeds the preset tolerance coefficient, it is determined as abnormal fluctuation;

[0226] The calculation of the historical mean is based on the past detection data of the same plant.

[0227] The feedback instruction refers to the control signal for triggering the re-collection of the surface deformation data. Specifically, it can be realized by interrupting the detection process and starting the three-dimensional point cloud data acquisition module, which is used to ensure the reliability of the calculation basic data of the curl index. Among them, the calculation weight coefficient of the curl index refers to the distribution ratio of the longitudinal curvature and the transverse fold density in the curl index formula. Specifically, it can be dynamically adjusted by an adaptive optimization algorithm, which is used to compensate for the measurement deviation caused by the leaf deformation. Among them, the ratio of the range to the historical mean refers to the proportional relationship between the difference between the maximum and minimum values of the second derivative of the reflectance within the current detection period and the average value of the past data of the same plant. Specifically, it can be realized by retrieving the time series database and real-time calculation, which is used to quantify the severity of the abnormal fluctuation.

[0228] Specifically, after completing the wavelength switching to the secondary absorption peak band, the variance or standard deviation of the second derivative of the reflectivity is calculated in real time to determine whether the data fluctuation within the current detection period exceeds a preset threshold. If the fluctuation is abnormal, a feedback instruction is triggered to interrupt the current detection process, and the three-dimensional point cloud data of the leaf surface is recollected. Based on the newly collected surface deformation data, the weight coefficients of the longitudinal curvature and the transverse wrinkle density are iteratively optimized. For example, the gradient descent method is used to adjust the ratio of the first preset coefficient and the second preset coefficient, thereby correcting the calculation result of the curling index. The generation of the feedback instruction depends on the dynamic evaluation of the range of the second derivative of the reflectivity. By comparing with the mean value of the historical detection data of the same plant, false judgments caused by environmental noise or equipment drift are avoided.

[0229] Example: Vineyard monitoring;

[0230] Initial state:

[0231] Weight coefficient: Longitudinal 0.5 / Transverse 0.5;

[0232] Historical mean: -0.07 (for the past 10 detections);

[0233] Preset tolerance: 30%;

[0234] Abnormal detection:

[0235] The data for the current period is shown in the following table:

[0236] Detection times Second derivative value 1 -0.12 2 +0.03 3 -0.18

[0237] Range calculation: 0.03 - (-0.18) = 0.21;

[0238] Ratio: 0.21 / 0.07 = 300% > 30%;

[0239] Execution process:

[0240] Trigger the recollect instruction:

[0241] The newly collected 3D point cloud shows:

[0242] Longitudinal curvature: 0.25 / mm, Transverse wrinkle: 5.2 / cm 2 .

[0243] Weight correction:

[0244] Analysis shows that the fluctuation mainly comes from the longitudinal change. New weight: Longitudinal 0.4 / Transverse 0.6.

[0245] Effect verification:

[0246] The corrected curling index: 0.25×0.4 + 5.2×0.6 = 3.22 → 0.322 (normalized).

[0247] The water potential error after compensation is reduced from 18% to 6%.

[0248] Compared with the prior art, the traditional method usually does not establish a dynamic monitoring and feedback mechanism after wavelength switching, resulting in the inability to timely correct the measurement error caused by leaf deformation. This solution realizes a rapid response to abnormal deformation by real-time monitoring of spectral data fluctuations and correlating historical detection data. At the same time, the accuracy of the curl index calculation is improved by optimizing the weight coefficient, overcoming the problem of environmental interference accumulation in a single detection mode.

[0249] Through the above technical solution, the present application can effectively identify abnormal spectral data caused by leaf deformation or external interference after wavelength switching, and avoid systematic errors caused by local deformation through a data re-acquisition and parameter correction mechanism, thereby ensuring the stability of the water potential detection result. At the same time, based on the dynamic threshold judgment of the plant historical data, the false trigger probability caused by instantaneous environmental changes is further reduced, enhancing the robustness of the detection system.

[0250] Embodiment 2:

[0251] As Figure 9 shown, a real-time non-destructive plant water potential detection system using the above real-time non-destructive plant water potential detection method includes:

[0252] A data acquisition module for obtaining multi-spectral reflection data and surface deformation data of the leaf;

[0253] A data processing module for calculating the curl index, extracting the spectral distortion amount, and determining whether it exceeds the deformation threshold;

[0254] A water potential calculation module for selecting a preset water absorption peak band or a secondary absorption peak band according to the curl index and calculating the standard water potential or the compensated water potential;

[0255] A dynamic compensation and optimization module for dynamically adjusting the model parameters of the mapping relationship according to the temporal variation characteristics of the compensated water potential value or the standard water potential value.

[0256] The data acquisition module refers to a device for synchronously obtaining leaf spectral information and morphological changes, which can be specifically implemented by combining a multispectral camera and a 3D scanner. The multispectral camera is used to collect the reflectance time-series data within a preset wavelength range, and the 3D scanner projects a fringe pattern and collects the deformed image to generate 3D point cloud data. The data processing module refers to an arithmetic unit for feature extraction of morphological and spectral data. Specifically, it can calculate the weighted values of the longitudinal curvature and the transverse fold density through an embedded processor, and judge the degree of leaf curling according to a preset deformation threshold. The water potential calculation module refers to a calculation unit for water potential inversion based on the reflectance data of different bands. Specifically, it can use a preset piecewise water potential inversion function and a linear water potential inversion model to process the data of the secondary absorption peak and the main absorption peak respectively. The dynamic compensation and optimization module refers to a control unit for real-time adjustment of model parameters. Specifically, it can record the deviation relationship between the curling index and the spectral distortion amount, and update the linear coefficient using the least squares method.

[0257] Specifically, the data acquisition module synchronously obtains the spectral reflectance signal and the surface morphological data of the leaf through the multispectral camera and the 3D scanner. The data processing module performs a weighted summation of the longitudinal curvature and the transverse fold density of the 3D point cloud data to obtain the curling index, and simultaneously extracts the spectral distortion amount of the preset water absorption peak. When the curling index exceeds the threshold, the water potential calculation module switches to the secondary absorption peak band and calculates the compensation water potential through the second derivative eigenvalue of the reflectance; when it does not exceed the threshold, it calculates the standard water potential based on the first derivative feature of the main absorption peak. The dynamic compensation and optimization module judges the deviation degree of the model parameters through the deviation factor according to the time-series change of the water potential value, triggers the update of the linear coefficient or the adjustment of the deformation threshold, and simultaneously monitors the fluctuation range of the reflectance in the secondary absorption peak band, and re-collects the deformation data and corrects the weight coefficient in case of abnormal fluctuation.

[0258] Compared with the prior art, the existing plant water potential detection systems usually adopt a single spectral band and do not consider the influence of leaf deformation on the detection accuracy, resulting in a significant increase in the water potential inversion error in the leaf curling state. This system integrates morphological detection and multi-band spectral analysis, dynamically selects the optimal detection band and establishes a deformation compensation mechanism, and solves the problem of spectral distortion interference caused by leaf curling. In addition, the prior art lacks the ability to update model parameters online, while this system adaptively adjusts the deformation threshold and the inversion model parameters by real-time monitoring the stability of the water potential value, improving the reliability of long-term detection.

[0259] Through the above technical solution, the present application can automatically switch to the secondary absorption peak band with less interference when the leaf undergoes curling deformation, dynamically compensate the water potential value in combination with the morphological data, and effectively reduce the detection error caused by the change of leaf posture. At the same time, by continuously tracking the temporal variation characteristics of the water potential value, the deformation threshold and the inversion model parameters are continuously optimized to ensure the stable detection accuracy of the system under different environmental conditions. In addition, the abnormal fluctuation monitoring mechanism can timely detect sensor drift or external interference, and maintain the robustness of the system through data re-acquisition and weight correction.

[0260] The technical scope of the present invention is not limited to the content described above. Those skilled in the art can make various deformations and modifications to the above embodiments without departing from the technical idea of the present invention, and these deformations and modifications should all fall within the protection scope of the present invention.

Claims

1. A real-time non-destructive detection method for plant water potential, characterized in that: It includes the following steps: Obtain the multi-spectral reflection data and surface deformation data of the target leaf. The multi-spectral reflection data includes the reflectance time-series data within a preset wavelength range, and the surface deformation data includes the three-dimensional point cloud data of the leaf surface; Calculate the curling index of the leaf according to the surface deformation data, and extract the spectral distortion amount of a preset water absorption peak according to the multi-spectral reflection data; Judge whether the curling index exceeds a preset deformation threshold; If it exceeds, switch the detection wavelength to the preset secondary absorption peak wavelength according to the mapping relationship between the spectral distortion amount and the curling index, and calculate the compensated water potential value based on the reflectance data of the secondary absorption peak band; If it does not exceed, calculate the standard water potential value based on the reflectance data of the preset water absorption peak band; Dynamically adjust the model parameters of the mapping relationship according to the temporal variation characteristics of the compensated water potential value or the standard water potential value.

2. The real-time non-destructive detection method for plant water potential according to claim 1, characterized in that: The acquisition process of the surface deformation data includes: Project a preset stripe pattern onto the leaf surface, collect the deformed stripe image, and generate the three-dimensional point cloud data based on the phase unwrapping algorithm; the three-dimensional point cloud data contains the quantization parameters of the longitudinal curvature and transverse fold density of the leaf.

3. The real-time non-destructive detection method for plant water potential according to claim 2, characterized in that: The calculation of the curling index of the leaf according to the surface deformation data includes: Perform weighted summation on the longitudinal curvature and transverse fold density, and the weight coefficients are predefined according to the leaf type; The calculation formula of the curling index is: Curling index = first preset coefficient × longitudinal curvature + second preset coefficient × transverse fold density; where the sum of the first preset coefficient and the second preset coefficient is 1.

4. The real-time non-destructive detection method for plant water potential according to claim 1, characterized in that: The spectral distortion amount of the preset water absorption peak is the wavelength offset of the preset water absorption peak band, and the wavelength offset is the difference between the current detection wavelength and the wavelength reference value under the standard water potential; the preset water absorption peak includes at least one absorption peak within a preset wavelength range, and the preset wavelength range is the range where the characteristic absorption peaks related to the plant water potential are located, and this range covers at least two characteristic absorption peak intervals, and one of the intervals is used as the main detection interval; the preset secondary absorption peak is the wavelength range where the secondary absorption peak related to the plant water potential is located and is lower than the preset main water absorption peak band affected by leaf curling.

5. The real-time non-destructive detection method for plant water potential according to claim 1, characterized in that: The process of calculating the compensated water potential value based on the reflectance data of the secondary absorption peak band is: Extract the second derivative eigenvalue of the reflectance of the secondary absorption peak band, and input the second derivative eigenvalue of the reflectance into a pre-calibrated piecewise water potential inversion function; The piecewise water potential inversion function includes at least two exponential function intervals, and each interval is divided according to the curling index range; The process of calculating the standard water potential value based on the reflectance data of the preset water absorption peak band is: Extract the reflectance eigenvalue of the preset water absorption peak band, and the reflectance eigenvalue is the first derivative of the reflectance of this band; Input the reflectance eigenvalue into a pre-calibrated linear water potential inversion model, and the linear water potential inversion model is fitted based on the reflectance data under the standard water potential.

6. The real-time non-destructive detection method for plant water potential according to claim 1, wherein: The dynamic adjustment of the model parameters of the mapping relationship includes: During the continuous detection process, record the real-time correspondence between the curling index and the spectral distortion amount; When the deviation of the corresponding relationship exceeds a preset feedback threshold in N consecutive detections, update the linear coefficient of the mapping relationship based on the least squares method; where N is an integer not less than 2.

7. The real-time non-destructive detection method for plant water potential according to claim 6, characterized in that: The deviation calculation process of the corresponding relationship is as follows: Extract the difference in curl index ΔCI and the difference in spectral distortion amount Δλ between two adjacent detections; Calculate the deviation factor: deviation factor = |Δλ - (a×ΔCI + b)|, where a and b are the linear coefficients of the current mapping relationship respectively; When the deviation factor exceeds the preset feedback threshold, trigger the update of the linear coefficient of the current mapping relationship.

8. The real-time non-destructive detection method for plant water potential according to claim 1, characterized in that: The dynamic adjustment of the model parameters of the mapping relationship further includes: adjusting the value of the preset deformation threshold according to the standard deviation of the compensated water potential value within a preset time period. The specific process is as follows: When the standard deviation is less than the preset stability threshold, reduce the preset deformation threshold by a first preset ratio; When the standard deviation is greater than the preset stability threshold, increase the preset deformation threshold by a second preset ratio.

9. The real-time non-destructive detection method for plant water potential according to claim 1, characterized in that: It further includes: After the wavelength is switched, continuously monitor the fluctuation range of the second derivative eigenvalue of the reflectance in the secondary absorption peak band; When the fluctuation range exceeds the preset fluctuation threshold, generate a feedback instruction to re-collect the surface deformation data; Correct the calculation weight coefficient of the curl index based on the re-collected data; The generation logic of the feedback instruction is as follows: Statistically calculate the range of the second derivative eigenvalue of the reflectance in the current detection period; When the ratio of the range to the historical mean value exceeds the preset tolerance coefficient, it is determined as abnormal fluctuation; The calculation of the historical mean value is based on the previous detection data of the same plant.

10. A real-time non-destructive detection system for plant water potential, characterized in that: Using the real-time non-destructive detection method for plant water potential as described in any one of claims 1 to 9, including: A data acquisition module for acquiring multi-spectral reflectance data and surface deformation data of the leaf; A data processing module for calculating the curl index, extracting the spectral distortion amount, and determining whether it exceeds the deformation threshold; A water potential calculation module for selecting a preset water absorption peak band or a secondary absorption peak band according to the curl index and calculating the standard water potential or the compensated water potential; A dynamic compensation and optimization module for dynamically adjusting the model parameters of the mapping relationship according to the temporal variation characteristics of the compensated water potential value or the standard water potential value.

Images