A chlorophyll inversion method and system based on flat-curling blade spectral stability evaluation

Abstract

The application discloses a chlorophyll inversion method and system based on flat-curling leaf spectrum stability evaluation, and the chlorophyll inversion method first acquires multi-angle directional reflectance spectrum data of each leaf sample in a natural curling state and a flat state and corresponding measured values of chlorophyll content. Secondly, a plurality of candidate vegetation indexes are calculated according to the multi-angle directional reflectance spectrum data, and a chlorophyll inversion relationship is established; the chlorophyll inversion relationship is applied to the multi-angle directional reflectance spectrum data in the natural curling state and the flat state respectively, and inversion error indexes under different candidate vegetation indexes are obtained. Then, a configuration change index is constructed based on the inversion error indexes, a robust connection index is screened from the plurality of candidate vegetation indexes according to the configuration change index, and a chlorophyll inversion model is established based on the robust connection index. Finally, the chlorophyll inversion model is used to detect the chlorophyll content of the measured leaf in the natural curling state.

Description

Technical Field

[0001] This invention belongs to the field of chlorophyll content inversion technology, specifically relating to a chlorophyll inversion method and system based on the evaluation of spectral stability of flat-curled leaves. Background Technology

[0002] Chlorophyll content is an important physiological indicator characterizing a plant's photosynthetic capacity, nutritional status, and health level. Non-destructive chlorophyll content retrieval using leaf reflectance or hyperspectral data is a crucial technical approach in precision agriculture and plant phenotypic analysis.

[0003] Existing chlorophyll retrieval methods typically rely on empirical vegetation indices or radiative transfer models. To reduce measurement uncertainty, many experimental and model calibration processes assume that leaves are flat or nearly flat. However, leaves in their natural growth state often exhibit curled, bent, or tilted configurations, and there is a significant difference between the actual configuration and the ideal flat assumption.

[0004] Changes in leaf configuration alter the normal distribution on the leaf surface and the light-observation geometry, causing variations in the spectral response at the same chlorophyll level. For naturally curled leaves, configuration changes may also alter the intensity and spatial directionality of surface specular reflection; specular reflection does not directly carry information about chlorophyll absorption within the leaf, but it is superimposed on the total apparent reflectance spectrum, causing traditional vegetation indices to fluctuate under different configurations or observation geometries. Summary of the Invention

[0005] The purpose of this invention is to provide a chlorophyll inversion method and system based on the evaluation of the spectral stability of flat-curled leaves.

[0006] In a first aspect, the present invention provides a chlorophyll inversion method based on the evaluation of the spectral stability of flat-curled leaves, the method comprising:

[0007] The reflectance spectra of each leaf sample in multiple leaf samples under natural curled and flat states were obtained from multiple angle directions, as well as the measured values ​​of chlorophyll content of each leaf sample.

[0008] Multiple candidate vegetation indices were calculated based on multi-angle directional reflectance spectral data, and chlorophyll inversion models for detecting chlorophyll content were established based on each candidate vegetation index and the measured chlorophyll content. The chlorophyll inversion models corresponding to each candidate vegetation index were applied to multi-angle directional reflectance spectral data of natural curled state and flat state to obtain the chlorophyll inversion results of each candidate vegetation index under natural curled state and flat state, and the corresponding inversion error index was calculated.

[0009] Based on the difference in inversion error indices of the same candidate vegetation index under flat and naturally curled states, the configuration change index corresponding to the candidate vegetation index is determined; based on the inversion error index and configuration change index, robust connectivity indices are screened from multiple candidate vegetation indices, and the robust connectivity indices are prioritized.

[0010] Acquire spectral data of the naturally curled leaves, determine whether the spectral data meets the calculation conditions of the corresponding robust connectivity index according to the priority of the robust connectivity index, calculate the robust connectivity index with the highest priority that meets the calculation conditions, input the robust connectivity index into its corresponding chlorophyll inversion model, and obtain the chlorophyll content of the naturally curled leaves.

[0011] Preferably, the method for selecting the robust connectivity index is as follows:

[0012] Based on the inversion error index of each candidate vegetation index under the natural curl state, candidate vegetation indices that meet the preset accuracy requirements are selected from multiple candidate vegetation indices; among the candidate vegetation indices that meet the preset accuracy requirements, candidate vegetation indices with smaller configuration change indices are selected as robust connectivity indices.

[0013] Preferably, the candidate vegetation indices include one or more of the following: simple ratio indices, modified simple ratio indices, double difference classification indices, red-edge position indices, and wavelet red-edge position indices.

[0014] As a preferred method, the chlorophyll inversion relationship is as follows: the least squares method is used to establish a linear regression relationship between each candidate vegetation index and chlorophyll content, and the linear regression relationship is used as the chlorophyll inversion relationship of the corresponding candidate vegetation index.

[0015] Preferably, the inversion error index includes one or more of the following: root mean square error, bias, and coefficient of variation.

[0016] Preferably, the configuration change index includes one or more of the following: relative deviation change, deviation change, relative root mean square error change, root mean square error change, relative coefficient of variation change, and coefficient of variation change.

[0017] Preferably, the priority ranking method for the robust connectivity indices is as follows: sort the multiple robust connectivity indices in order of increasing relative root mean square error (RMSE); and determine the robust connectivity indices with smaller relative RMSE changes as higher priority robust connectivity indices.

[0018] Preferably, when the difference between the relative root mean square error changes of two adjacent robust connection indices is less than a first preset threshold, the relative coefficient of variation changes of the two indices are compared. If the difference between the relative coefficient of variation changes of the two indices is less than a second preset threshold, the relative deviation changes of the two indices are then compared.

[0019] As a preferred method, the method for obtaining multi-angle reflectance spectral data under the natural curled state and the flat state is as follows: first, collect multi-angle reflectance spectral data under the natural curled state when the leaf sample is in a natural curled state, and then collect multi-angle reflectance spectral data under the flat state after flattening the same leaf sample according to the same observation geometry scheme.

[0020] Secondly, the present invention provides a chlorophyll inversion system based on the spectral stability evaluation of flat-curled leaves, which is used to execute the above-mentioned chlorophyll inversion method; the chlorophyll inversion system includes a dual configuration data acquisition module, a candidate index modeling module, a dual configuration evaluation module, a configuration change evaluation module, a robust index screening module, a priority ranking module, and a chlorophyll inversion module; the dual configuration data acquisition module is used to receive multi-angle directional reflectance spectral data and measured chlorophyll content values ​​of the same leaf in natural curled and flat states; the candidate index modeling module is used to acquire multiple candidate vegetation indices and establish a chlorophyll inversion model between the candidate vegetation indices and chlorophyll content; The dual configuration evaluation module is used to obtain the chlorophyll inversion results of each candidate vegetation index under natural curled and flat states, and calculate the corresponding inversion error index; the configuration change evaluation module is used to determine the configuration change index based on the difference in the inversion error index of the same candidate vegetation index under flat and natural curled states; the robust index screening module is used to screen robust connectivity indices from multiple candidate vegetation indices based on the inversion error index and configuration change index; the priority ranking module is used to rank the priority of the screened robust connectivity indices based on the configuration change index; the chlorophyll inversion module is used to obtain the chlorophyll content of the tested leaves in natural curled state according to the chlorophyll inversion model.

[0021] Thirdly, the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described chlorophyll inversion method.

[0022] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described chlorophyll inversion method.

[0023] The beneficial effects of this invention are:

[0024] 1. This invention uses vegetation indices to invert and detect the chlorophyll content of naturally curled leaves, and selects vegetation indices that are more stable to configuration changes by comparing flat and curled states. Compared with most existing methods that flatten leaves and control geometric conditions indoors before detection, this invention is closer to the actual growth state of leaves, which can reduce the influence of leaf curling and specular reflection on the inversion results and improve the reliability in practical applications.

[0025] 2. This invention determines the configuration change index based on the difference in inversion error index of the same candidate vegetation index under flat and naturally curled states. The configuration change index can quantify the impact of the leaf changing from a flat state to a naturally curled state on the chlorophyll inversion results, so that the sensitivity of candidate vegetation indices to configuration changes such as leaf curling, bending or tilting can be objectively evaluated, thereby providing a quantitative basis for the screening of robust connectivity indices.

[0026] 3. This invention selects robust connectivity indices based on both inversion error index and configuration change index, which can simultaneously consider the inversion accuracy and configuration stability of candidate vegetation indices, thereby improving the applicability of the selected robust connectivity indices in the detection of naturally curled leaves. Attached Figure Description

[0027] Figure 1 This is the overall flowchart of the present invention.

[0028] Figure 2 The image shows a comparison of the natural curled state, the flat state, and the normal vector of the same blade in this invention; where (a) is the naturally curled state blade; (b) is the flat state blade; and (c) is the normal vector of the naturally curled state blade and the flat state blade.

[0029] Figure 3 This is a diagram showing the directional reflection distribution of the flat blade under different zenith angles of light sources in this invention.

[0030] Figure 4 This is a diagram showing the directional reflection distribution of the naturally curled blades under different zenith angles of light sources in this invention. Detailed Implementation

[0031] The present invention will be further described below with reference to the accompanying drawings.

[0032] Example 1

[0033] A chlorophyll inversion method based on the evaluation of spectral stability of flat-curled leaves includes the following steps:

[0034] Step 1: Obtain spectral data of the same leaf with dual configuration.

[0035] like Figure 1 and Figure 2As shown, the same leaf sample was selected, and multi-angle BRF spectral data (multi-angle directional reflectance spectral data) were collected first under natural curled state. The multi-angle BRF spectral data includes spectral data at different observation zenith angles, observation azimuth angles, and direct downward viewing angles at each light source zenith angle. Among them, the light source zenith angle... Observing the zenith angle ; Observation azimuth Direct downward viewing angle That is, the multi-angle BRF spectral data is actually 508 sets of observation geometry: 4 source zenith angles, each source zenith angle contains 7 observation zenith angles × 18 observation azimuth angles, plus 1 direct downward angle, that is, 4 × (7 × 18 + 1) = 508.

[0036] The leaf was flattened, and multi-angle BRF spectral data were collected under the same observation geometry in a flat state. Simultaneously, the chlorophyll content of the leaf sample was measured using chemical methods. Figure 3 and Figure 4 As shown, the flat and naturally curled blades exhibit different directional reflectance distribution characteristics under different zenith angles of light sources, indicating that changes in blade configuration alter the BRF spectral response. Figure 3 The upper row corresponds to a wavelength of 780 nm, the lower row corresponds to a wavelength of 1700 nm, and the zenith angles of the light sources corresponding to each column are 15°, 30°, 40° and 55° respectively; Figure 4 The upper row corresponds to a wavelength of 780 nm, and the lower row corresponds to a wavelength of 1700 nm. The zenith angles of the light sources corresponding to each column are 15°, 30°, 40° and 55°, respectively.

[0037] In some embodiments, a subset of polarization observations is obtained, and existing polarization separation methods are used to evaluate the impact of specular reflection on chlorophyll inversion results.

[0038] Step 2: Candidate Index Modeling and Dual Configuration Inversion Evaluation

[0039] Multiple candidate vegetation indices were extracted from multi-angle BRF spectral data to construct a candidate vegetation index set. The candidate vegetation indices include simple ratio class index (SR), modified simple ratio class index (mSR), double difference class index (DD), and red edge position class index (REP). The specific candidate vegetation indices are shown in Table 1.

[0040] Table 1 List of candidate vegetation indices

[0041]

[0042] Based on independently calibrated data (including candidate vegetation index values ​​and measured chlorophyll content values), a linear regression relationship between each candidate vegetation index and chlorophyll content was established using ordinary least squares (OLS) to obtain the chlorophyll inversion model corresponding to each candidate vegetation index. The inversion relationship was applied to multi-angle BRF spectral data of both naturally curled and flat vegetation configurations to obtain chlorophyll inversion results for each candidate vegetation index under both configurations (flat and naturally curled). Based on the chlorophyll inversion results and measured chlorophyll content values, chlorophyll inversion error indices for flat and naturally curled vegetation configurations were calculated, including RMSE (root mean square error), Bias, and CV (coefficient of variation). Error indices for flat vegetation are shown in Table 2, and those for curled vegetation are shown in Table 3.

[0043] Table 2. Average error index of chlorophyll inversion under 508 angle combinations for flat leaves

[0044]

[0045] Table 3. Average error index of chlorophyll inversion for naturally curled leaves under the same angle combination.

[0046]

[0047] Step 3: Screening for Stable Indices

[0048] The configurational change indices of each candidate vegetation index between the flat and naturally curled states were calculated. The configurational change indices include ΔBias_Rel (relative deviation change), ΔBIAS (deviation change), ΔRMSE_Rel (relative root mean square error change), ΔRMSE (root mean square error change), ΔCV_Rel (relative coefficient of variation change), and ΔCV (coefficient of variation change). The specific results are shown in Table 4.

[0049] Table 4. Configurational variation indices between flat and naturally curled states.

[0050]

[0051] Based on the inversion average error index of each candidate vegetation index (Table 3), candidate vegetation indices that meet the accuracy requirements are selected from all candidate vegetation indices, and those with smaller configuration change indices are selected as robust connectivity indices. Table 4 shows that mDATT has a ΔRMSE of 0.5074, a ΔRMSE_Rel of 0.135, and a ΔCV of 1.2424; WREP-S4 has a ΔRMSE of 0.6465, a ΔRMSE_Rel of 0.184, and a ΔCV of 1.6096; and REPPF has a ΔRMSE of 0.8291, a ΔRMSE_Rel of 0.231, and a ΔCV of 2.0886. Therefore, WREP-S4, mDATT, and REPPF are identified as robust connectivity indices for the inversion of chlorophyll from flat to naturally curled states. Note that WREP-S4, mDATT, and REPPF are selections from the candidate vegetation index set and are not considered as the newly proposed vegetation indices of this invention.

[0052] The robust connectivity indices are prioritized based on the configuration change index; the prioritization rules are as follows:

[0053] First, the robust connectivity indices are sorted in ascending order of the relative root mean square error change ΔRMSE_Rel or the root mean square error change ΔRMSE. If the difference between ΔRMSE_Rel or ΔRMSE of two adjacent robust connectivity indices is less than a first preset threshold, then the relative coefficient of variation change ΔCV_Rel or the coefficient of variation change ΔCV is compared. If the difference between ΔCV_Rel or ΔCV is less than a second preset threshold, then the relative deviation change ΔBias_Rel or the deviation change ΔBIAS is compared, and the smaller robust connectivity index is selected as the higher priority.

[0054] Step 4: Chlorophyll content detection

[0055] Spectral data of naturally curled leaves were collected. Based on the priority of robust connectivity indices, it was determined whether the spectral data of naturally curled leaves met the calculation conditions of robust connectivity indices. Chlorophyll content was detected according to the chlorophyll inversion model corresponding to the robust connectivity indices that met the calculation conditions.

[0056] Example 2

[0057] A chlorophyll inversion system based on the spectral stability evaluation of flat-curled leaves is provided for performing the chlorophyll inversion method of Example 1. This chlorophyll inversion system can be deployed in computer equipment, spectral data processing terminals, servers, or plant phenotypic analysis platforms. It can also be used in conjunction with multi-angle spectral acquisition devices to achieve leaf dual-configuration spectral data processing, robust connectivity index screening, and chlorophyll content inversion of naturally curled leaves.

[0058] The chlorophyll inversion system includes a biconfiguration data acquisition module, a candidate index modeling module, a biconfiguration evaluation module, a configuration change evaluation module, a robust index screening module, a priority ranking module, and a chlorophyll inversion module.

[0059] The dual-configuration data acquisition module is used to acquire multi-angle reflectance spectral data of each leaf sample in natural curled and flat states, as well as the measured chlorophyll content of each leaf sample.

[0060] The candidate index modeling module calculates multiple candidate vegetation indices based on multi-angle directional reflectance spectral data provided by the dual-configuration data acquisition module, and establishes chlorophyll inversion models between each candidate vegetation index and chlorophyll content. The module can obtain the correspondence between each candidate vegetation index value and the measured chlorophyll content value based on independently calibrated data, and establish a linear regression relationship between each candidate vegetation index and chlorophyll content using the least squares method. This linear regression relationship is then used as the chlorophyll inversion model for the corresponding candidate vegetation index.

[0061] The dual-configuration evaluation module is used to apply the chlorophyll inversion models established by the candidate index modeling module to the multi-angle reflectance spectral data of the natural curled state and the flat state, respectively, to obtain the chlorophyll inversion results of each candidate vegetation index under the natural curled state and the flat state, and to calculate the corresponding inversion error index based on the chlorophyll inversion results and the measured chlorophyll content.

[0062] The configuration change evaluation module is used to determine the configuration change index corresponding to the candidate vegetation index based on the difference in inversion error indices between the flat and naturally curled states. Through this module, the degree of change in the inversion performance of the same candidate vegetation index after the leaves change from a flat to a naturally curled state can be quantified, thus characterizing the sensitivity of the candidate vegetation index to changes in leaf configuration.

[0063] The robustness index screening module is used to screen robust connectivity indices from multiple candidate vegetation indices based on inversion error and configuration change indices. The inversion error index determines whether a candidate vegetation index meets a preset inversion accuracy requirement, while the configuration change index evaluates the configuration stability of candidate vegetation indices that meet the preset inversion accuracy requirement. The priority ranking module is used to prioritize the screened robust connectivity indices based on the configuration change index. The chlorophyll inversion module is used to obtain the chlorophyll content of naturally curled leaves based on a chlorophyll inversion model.

[0064] The above system can be used to process multi-angle reflectance spectral data of naturally curled and flat leaves in a unified manner, and select robust connectivity indices suitable for detecting chlorophyll content in naturally curled leaves from multiple candidate vegetation indices, thereby achieving stable inversion of chlorophyll content in naturally curled leaves.

Claims

1. A chlorophyll inversion method based on the evaluation of spectral stability of flat-curled leaves, characterized in that: The method includes: The reflectance spectra of each leaf sample in multiple leaf samples under natural curled and flat states were obtained from multiple angle directions, as well as the measured values ​​of chlorophyll content of each leaf sample. Multiple candidate vegetation indices were calculated based on multi-angle directional reflectance spectral data, and chlorophyll inversion models for detecting chlorophyll content were established based on each candidate vegetation index and the measured chlorophyll content. The chlorophyll inversion models corresponding to each candidate vegetation index were applied to multi-angle directional reflectance spectral data of natural curled state and flat state to obtain the chlorophyll inversion results of each candidate vegetation index under natural curled state and flat state, and the corresponding inversion error index was calculated. Based on the difference in inversion error indices of the same candidate vegetation index under flat and naturally curled states, the configuration change index corresponding to the candidate vegetation index is determined; based on the inversion error index and configuration change index, robust connectivity indices are screened from multiple candidate vegetation indices, and the robust connectivity indices are prioritized. Acquire spectral data of the naturally curled leaves, determine whether the spectral data meets the calculation conditions of the corresponding robust connectivity index according to the priority of the robust connectivity index, calculate the robust connectivity index with the highest priority that meets the calculation conditions, input the robust connectivity index into its corresponding chlorophyll inversion model, and obtain the chlorophyll content of the naturally curled leaves.

2. The chlorophyll inversion method based on the evaluation of spectral stability of flat-curled leaves according to claim 1, characterized in that: The selection method for the robust connectivity index is as follows: Based on the inversion error index of each candidate vegetation index under the natural curl state, candidate vegetation indices that meet the preset accuracy requirements are selected from multiple candidate vegetation indices; among the candidate vegetation indices that meet the preset accuracy requirements, candidate vegetation indices with smaller configuration change indices are selected as robust connectivity indices.

3. The chlorophyll inversion method based on the evaluation of spectral stability of flat-curled leaves according to claim 1, characterized in that: The candidate vegetation indices include one or more of the following: simple ratio indices, modified simple ratio indices, double difference classification indices, red-edge position indices, and wavelet red-edge position indices.

4. The chlorophyll inversion method based on the evaluation of spectral stability of flat-curled leaves according to claim 1, characterized in that: The chlorophyll inversion model is constructed by using the least squares method to establish a linear regression relationship between each candidate vegetation index and chlorophyll content, and using the linear regression relationship as the chlorophyll inversion relationship of the corresponding candidate vegetation index.

5. The chlorophyll inversion method based on the evaluation of spectral stability of flat-curled leaves according to claim 1, characterized in that: The inversion error index includes one or more of the following: root mean square error, bias, and coefficient of variation.

6. The chlorophyll inversion method based on the evaluation of spectral stability of flat-curled leaves according to claim 5, characterized in that: The configuration change index includes one or more of the following: relative deviation change, deviation change, relative root mean square error change, root mean square error change, relative coefficient of variation change, and coefficient of variation change.

7. The chlorophyll inversion method based on the evaluation of spectral stability of flat-curled leaves according to claim 6, characterized in that: The priority ranking method for the robust connectivity indices is as follows: multiple robust connectivity indices are ranked in ascending order of relative root mean square error change or root mean square error change; robust connectivity indices with smaller relative root mean square error change or root mean square error change are determined as higher priority robust connectivity indices.

8. The chlorophyll inversion method based on the evaluation of spectral stability of flat-curled leaves according to claim 7, characterized in that: When the difference between the relative root mean square error changes of two adjacent robust connection indices is less than a first preset threshold, the relative coefficient of variation changes of the two indices are compared. If the difference between the relative coefficient of variation changes of the two indices is less than a second preset threshold, the relative deviation changes of the two indices are then compared.

9. The chlorophyll inversion method based on the evaluation of spectral stability of flat-curled leaves according to claim 1, characterized in that: The method for obtaining multi-angle reflectance spectral data under the natural curled state and the flat state is as follows: first, collect multi-angle reflectance spectral data under the natural curled state when the leaf sample is in a natural curled state, and then collect multi-angle reflectance spectral data under the flat state after flattening the same leaf sample according to the same observation geometry scheme.

10. A chlorophyll inversion system based on the evaluation of spectral stability of flat-curled leaves, characterized in that: This method is used to perform the chlorophyll inversion method based on the spectral stability evaluation of flat-curled leaves as described in claim 1. The chlorophyll inversion system includes a dual-configuration data acquisition module, a candidate index modeling module, a dual-configuration evaluation module, a configuration change evaluation module, a robustness index screening module, a priority ranking module, and a chlorophyll inversion module. The dual-configuration data acquisition module receives multi-angle reflectance spectral data and measured chlorophyll content of the same leaf in both naturally curled and flat states. The candidate index modeling module acquires multiple candidate vegetation indices and establishes a chlorophyll inversion model between the candidate vegetation indices and chlorophyll content. The dual-configuration evaluation module... The module is used to obtain the chlorophyll inversion results of each candidate vegetation index under natural curled and flat states, and to calculate the corresponding inversion error index; the configuration change evaluation module is used to determine the configuration change index based on the difference in the inversion error index of the same candidate vegetation index under flat and natural curled states; the robust index screening module is used to screen robust connectivity indices from multiple candidate vegetation indices based on the inversion error index and configuration change index; the priority ranking module is used to rank the priority of the screened robust connectivity indices based on the configuration change index; the chlorophyll inversion module is used to obtain the chlorophyll content of the tested leaves in natural curled state according to the chlorophyll inversion model.

Images