A system and method for distinguishing subtle differences in plant leaf color using multispectral and PLS-DA

Abstract

The present invention discloses a system and method for distinguishing subtle differences in plant leaf color using multispectral PLS-DA. The system includes: a processor, a light-shielding blackroom, a portable spectrometer, a light source, a liftable detection platform, and a movable transmission platform. The portable spectrometer is arranged at the top center of the light-shielding blackroom, the light source is arranged at the top of the light-shielding blackroom, the liftable detection platform and the movable transmission platform are both arranged inside the light-shielding blackroom cover, and the liftable detection platform is arranged on the movable transmission platform; a sample placement area to be tested is provided on the liftable detection platform, and the liftable detection platform is arranged below the portable spectrometer; and the processor is electrically connected to the portable spectrometer, the light source, the liftable detection platform, and the movable transmission platform. The present invention can achieve non-destructive, rapid, and accurate collection of spectral data of plant leaves, has a high level of automation, can effectively avoid errors caused by manual grading, and is conducive to the precise analysis of subtle differences in leaf color of the same plant species.

Description

Technical Field

[0001] The present invention relates to the technical field of plant leaf color trait detection, and in particular to a system and method for distinguishing subtle differences in plant leaf color by combining multispectral analysis (PLS-DA). Background Art

[0002] Among the methods for measuring and classifying plant leaf color phenotypes, traditional methods mainly rely on manual visual inspection, or use pigment measurement, colorimeter measurement, etc.

[0003] 1. Manual visual inspection: Manual visual inspection has always been the most common method for leaf color measurement. It directly uses the color difference of leaves to stimulate human vision for evaluation and grading. It is simple and easy to operate. However, this method is greatly affected by the environment and is easily limited by the sensory organs and the subjectivity of the measurer. It is difficult to distinguish subtle differences in color. It has poor expressiveness and reproducibility, and its accuracy is not high.

[0004] 2. Pigment determination: Plant leaf color is a dynamic, comprehensive, and apparent manifestation of pigments within the plant. Leaf color variation is often associated with changes in chloroplast pigment content, so pigment content can be used to assess plant leaf color. Spectrophotometry is currently the most commonly used method for determining chlorophyll content. However, preparing chlorophyll extracts using a spectrophotometer is cumbersome, time-consuming, and labor-intensive. Furthermore, pigment content cannot distinguish between darker shades of green, such as gray-green and dark green, based on leaf color. Therefore, this method is only suitable for measuring samples with significant leaf color variation in the laboratory.

[0005] 3. Colorimeter measurement: Due to the poor reproducibility of manual visual description, quantitative expression of leaf color, or digital representation, is crucial. This allows people to describe color to others at any time and in any place. This requires instrumental color measurement. However, colorimeter applications place stringent demands on the sample being measured, including surface flatness, consistent color, and size. These requirements are often not met in field plant sampling, limiting their scope of use.

[0006] Currently, there is no reliable method for accurately quantifying and distinguishing subtle differences in plant leaf color. Traditional manual measurement is subjective due to differences in human sensory perception, making it difficult to distinguish subtle differences in color. It also has poor expressiveness and reproducibility, and cannot match accurate genomic data. Pigment content determination not only takes a long time to measure, but is also unable to distinguish darker green leaf colors such as gray-green and dark green. Colorimeter measurements have many restrictions and a narrow scope of use. Therefore, against the backdrop of the rapid development of plant functional genomics, there is an urgent need for a precise quantitative plant leaf color phenotypic identification method to distinguish subtle differences in plant leaf color and ensure the scientific nature and accuracy of experimental data.

[0007] With the rapid development of machine vision and intelligent technologies, multispectral plant phenotyping has been applied to plant phenotypic trait analysis due to its non-destructive, efficient, and accurate characteristics. Partial least squares analysis (PLS-DA) is a commonly used supervised discriminant classification method based on partial least squares regression in spectral technology. It uses the independent variable matrix X and the categorical variable Y to establish a regression model and predict the class of unknown samples using the partial least squares method. A growing number of studies are combining these two techniques and are widely applying them to fields such as food, industry, agriculture, and medicine, laying the foundation for their application in plant phenotypic trait analysis. Summary of the Invention

[0008] The purpose of the present invention is to provide a system and method for distinguishing subtle differences in plant leaf color by combining multispectral analysis with PLS-DA. The system and method have a high level of automation, a simple structure, and are easy to use. They can effectively avoid errors caused by manual grading and obtain accurate grading data. The system has obvious effects on analyzing subtle differences in leaf color of the same plant species and can be effectively applied to scientific research on plant leaf color trait investigations to ensure the scientific nature and accuracy of the test data.

[0009] To achieve the above object, the present invention provides the following solutions:

[0010] A system for distinguishing subtle differences in plant leaf color by combining multispectral PLS-DA, the system comprising: a processor, a light-shielding blackroom, a portable ground object spectrometer, a light source, a liftable detection platform, and a movable transmission platform, wherein the portable ground object spectrometer is arranged at the top center of the light-shielding blackroom for collecting spectral data of samples to be detected, the light source is arranged at the top of the light-shielding blackroom for irradiating the samples to be detected, the liftable detection platform and the movable transmission platform are both covered by the light-shielding blackroom, and the movable transmission platform extends out from a side wall of the light-shielding blackroom; the liftable detection platform is arranged on the movable transmission platform, and the movable transmission platform is used to drive the liftable detection platform to move left and right; a sample placement area for placing plant leaves to be detected is provided on the liftable detection platform, and the liftable detection platform is arranged below the portable ground object spectrometer; the processor is electrically connected to the portable ground object spectrometer, the light source, the liftable detection platform, and the movable transmission platform, respectively.

[0011] Furthermore, the liftable testing platform is also provided with a calibration whiteboard, and the calibration whiteboard is arranged on the left side of the placement area of ​​the sample to be tested.

[0012] Furthermore, the model of the portable ground object spectrometer is PSR-1100.

[0013] Furthermore, the processor is connected to a display.

[0014] Furthermore, two groups of light sources are provided, which are respectively located on both sides of the top of the light-shielding black room and are arranged inwardly at a 45-degree angle with the side walls of the light-shielding black room.

[0015] Furthermore, the movable transmission platform is a conveyor belt mechanism, including a bracket and a conveyor belt, a driving wheel, a driven wheel and a motor arranged on the bracket, the conveyor belt is arranged between the driving wheel and the driven wheel, the motor is driven and connected to the driving wheel, and the motor is electrically connected to the processor.

[0016] The present invention also provides a method for distinguishing subtle differences in plant leaf color using multispectral combined with PLS-DA, which is applied to the above-mentioned system for distinguishing subtle differences in plant leaf color using multispectral combined with PLS-DA, comprising the following steps:

[0017] S1, select standard plant leaf samples and place them in the sample placement area to be tested, classify them according to the subtle characteristics of leaf color depth within the test range, and use a portable ground feature spectrometer to collect the spectral reflectance of the standard plant leaf samples according to the grade;

[0018] S2, SG-SNV preprocessing of the collected spectral reflectance data using SG smoothing filter and SNV standard normalization method;

[0019] S3, assign categorical variable values ​​to the samples based on the leaf color depth of the standard plant leaf samples; use the PLS regression method to perform regression analysis on the spectral reflectance data of the standard plant leaf samples and the categorical variable values ​​corresponding to the samples, determine the optimal number of main factors and establish a PLS-DA model based on multispectral features and categorical variables;

[0020] S4, permutation test of the PLS-DA model;

[0021] S5, placing the leaf sample to be tested in the sample placement area to be tested, collecting multispectral data of the leaf sample to be tested, and importing it into the PLS-DA model to determine the leaf color depth level.

[0022] Furthermore, in step S1, the detection range is within the spectral reflectance band of 320nm to 1100nm.

[0023] Furthermore, in step S4, a permutation test is performed on the PLS-DA model, specifically including:

[0024] Statistical inference is performed using the random arrangement of sample data, the samples are permuted in order, the statistical test quantity is recalculated, and the empirical distribution is constructed. Then, based on this, the asymptotic P value of the sample distribution is calculated based on the permutation test principle. The model situation is inferred according to the significance level of the P value (P<0.05) to determine whether the model has failed to build or is overfitting.

[0025] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the system and method provided by the present invention for distinguishing subtle differences in plant leaf color by combining multispectral and PLS-DA, the system improves the quality of spectral data acquisition by setting up a light-shielding black room to block external environmental interference, performs spectral scanning through the PSR-1100 portable ground feature spectrometer, and adopts a processor connected to the portable ground feature spectrometer, so that the user can observe relevant data conveniently, with a high degree of automation, ensuring fast and efficient plant leaf spectral acquisition; through the cooperation of the liftable detection platform and the movable transmission platform, the sample can be moved left and right and up and down, which is conducive to the spectrometer aligning the sample to be tested, obtaining clearer and more accurate spectral data, thereby being able to non-destructively, quickly and accurately detect plant leaf spectral related data, and provide reliable data support for establishing a PLS-DA discrimination model;

[0026] The method assigns categorical variable values ​​to the samples according to their actual characteristics, uses the PLS regression method to perform regression analysis on the multispectral characteristics of the samples and the categorical variables corresponding to the samples, determines the optimal number of principal factors and establishes a PLS-DA model based on the multispectral characteristics and the categorical variables, obtains a plant leaf color classification discrimination model established by combining the multispectral characteristics with the PLS-DA method, and establishes an effective plant leaf color grade detection method that can distinguish different plant leaf color grades and conduct accurate and efficient plant leaf color trait investigations. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0028] Figure 1 This is a schematic diagram of the system structure for distinguishing subtle differences in plant leaf color using multispectral analysis combined with PLS-DA according to an embodiment of the present invention;

[0029] Figure 2 This is a flow chart of a method for distinguishing subtle differences in plant leaf color using multispectral analysis combined with PLS-DA according to an embodiment of the present invention;

[0030] Figure 3 This is an example of a sample of plant leaf color grade accurately segmented by the PLS-DA model according to an embodiment of the present invention;

[0031] Figure 4 Schematic diagram of the comparison between the original spectrum after SNV processing and the spectrum after SG smoothing;

[0032] Figure 5 This is a schematic diagram comparing the spectral characteristic curves of Chinese cabbage with four different leaf colors;

[0033] Figure 6 PLS-DA and principal component score plots for 64 modeling samples;

[0034] Figure 7 Model T1 principal component score map for 64 samples PLS-DA;

[0035] Figure 8 Model T2 principal component score plot for 64 samples PLS-DA;

[0036] Figure 9 Model T3 principal component score plot for 64 samples PLS-DA;

[0037] Figure 10 It is the permutation test diagram of the PLS-DA model;

[0038] Explanation of the accompanying symbols: 1. Light-shielding black room; 2. Light source; 3. Calibration whiteboard; 4. Liftable detection platform; 5. Motor; 6. Movable transmission platform; 7. Placement area for samples to be tested; 8. Portable ground feature spectrometer; 9. Processor; 10. Display. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] The purpose of the present invention is to provide a system and method for distinguishing subtle differences in plant leaf color by combining multispectral analysis with PLS-DA. The system and method have a high level of automation, a simple structure, and are easy to use. They can effectively avoid errors caused by manual grading and obtain accurate grading data. The system has obvious effects on analyzing subtle differences in leaf color of the same plant species and can be effectively applied to scientific research on plant leaf color trait investigations to ensure the scientific nature and accuracy of the test data.

[0041] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] like Figure 1As shown, the multi-spectral combined PLS-DA system provided by the present invention for distinguishing subtle differences in plant leaf color includes: a processor 9, a light-shielding blackroom 1, a portable ground object spectrometer 8, a light source 2, a liftable detection platform 4 and a movable transmission platform 6, wherein the portable ground object spectrometer 8 is arranged at the top center of the light-shielding blackroom 1 for collecting spectral data of the sample to be detected, the light source 2 is arranged at the top of the light-shielding blackroom 1 for irradiating the sample to be detected, the liftable detection platform 4 and the movable transmission platform 6 are both covered by the light-shielding blackroom 1 and arranged inside, and the movable transmission platform 6 extends out from a side wall of the light-shielding blackroom 1; the liftable detection platform 4 is arranged on the movable transmission platform 6, and the movable transmission platform 6 is used to drive the liftable detection platform 4 to move left and right; a sample placement area 7 to be detected is provided on the liftable detection platform 4 for placing plant leaves, and the liftable detection platform 4 is arranged below the portable ground object spectrometer 8; the processor 9 is electrically connected to the portable ground object spectrometer 8, the light source 2, the liftable detection platform 4 and the movable transmission platform 6 respectively. The liftable testing platform 4 may adopt a conventional electrically controlled telescopic rod structure, with a testing platform arranged above the electrically controlled telescopic rod to achieve the up and down lifting movement of the sample to be tested.

[0043] The liftable testing platform 4 is also provided with a calibration whiteboard 3, which is positioned to the left of the sample placement area 7. During use, the liftable testing platform 4 can be moved left and right by the movable transmission platform 6. When the calibration whiteboard 3 is moved below the portable ground feature spectrometer 8, whiteboard calibration can be performed or sample replacement can be performed, thereby improving detection efficiency.

[0044] The model of the portable surface spectrometer 8 is PSR-1100. The PSR-1100 has a spectral range of 320-1100 nm (ultraviolet-visible-near-infrared). Composed of a spectrometer and optical fiber, the PSR-1100 is a passive reflectivity spectrometer, containing a 512-line array detector, a fixed holographic grating as a dispersion element, and a 25° fiber optic probe.

[0045] The processor 9 is connected to a display 10 for visually monitoring data.

[0046] The light source 2 is a halogen lamp (35W 230V), and two groups are provided, which are respectively located on both sides of the top of the light-shielding black chamber 1 and are arranged at a 45-degree angle inward with the side wall of the light-shielding black chamber 1. The two light sources form a fixed angle to illuminate the sample to be tested without blind spots.

[0047] The movable transmission platform 6 is a conveyor belt mechanism, including a bracket and a conveyor belt, a driving wheel, a driven wheel and a motor 5 arranged on the bracket. The conveyor belt is arranged between the driving wheel and the driven wheel, the motor is driven and connected to the driving wheel, and the motor 5 is electrically connected to the processor 9.

[0048] The multi-spectral combined with PLS-DA system provided by the present invention distinguishes subtle differences in plant leaf color by setting up a light-shielding black room to block external environmental interference, thereby improving the quality of spectral data acquisition, performing spectral scanning through the PSR-1100 portable ground feature spectrometer, and adopting a processor to connect the portable ground feature spectrometer, so that users can observe relevant data conveniently, with a high degree of automation, ensuring fast and efficient plant leaf spectral acquisition; through the cooperation of the liftable detection platform and the movable transmission platform, the sample can be moved left and right and up and down, which is conducive to the spectrometer aligning the sample to be tested, obtaining clearer and more accurate spectral data, thereby enabling non-destructive, rapid and accurate detection of plant leaf spectral related data, and providing reliable data support for establishing the PLS-DA discrimination model.

[0049] like Figure 2 As shown, the present invention also provides a method for distinguishing subtle differences in plant leaf color by combining multispectral and PLS-DA, which is applied to the above-mentioned system for distinguishing subtle differences in plant leaf color by combining multispectral and PLS-DA, including the following steps:

[0050] S1: Select standard plant leaf samples and place them in the sample placement area to be tested. Grades are assigned based on subtle features of leaf color depth within the test range, and the spectral reflectance of the standard plant leaf samples is collected using a portable surface feature spectrometer based on the grade. After collecting the plant leaves, flatten them, wipe off dust and stains from the surface, and use an improved portable surface feature spectrometer to collect spectral data, collecting spectral reflectance in the 320nm to 1100nm band.

[0051] S2, using the SG smoothing filter and SNV standard normalization method to perform SG-SNV preprocessing on the collected spectral reflectance data; this can reduce instrument and environmental noise in the data, weakening and eliminating the effects of sample inhomogeneity, baseline drift, high-frequency noise, stray light, etc. on the spectral signal; the SG smoothing filter is a method based on polynomials and moving windows, using the least squares method to achieve optimal fitting in the time domain. It can not only improve the signal-to-noise ratio of the signal, but also better preserve the useful information in the signal (spectrum), and is widely used in spectral analysis;

[0052] S3, assign categorical variable values ​​to the samples based on the leaf color depth of the standard plant leaf samples; use the PLS regression method to perform regression analysis on the spectral reflectance data of the standard plant leaf samples and the categorical variable values ​​corresponding to the samples, determine the optimal number of main factors and establish a PLS-DA model based on multispectral features and categorical variables;

[0053] S4, permutation test of the PLS-DA model;

[0054] S5, placing the leaf sample to be tested in the sample placement area to be tested, collecting multispectral data of the leaf sample to be tested, and importing it into the PLS-DA model to determine the leaf color depth level.

[0055] Wherein, in the step S1, the detection range is within the spectral reflectance band of 320nm to 1100nm.

[0056] In step S4, a permutation test is performed on the PLS-DA model, specifically including:

[0057] Statistical inference is performed using the random arrangement of sample data, the samples are permuted in order, the statistical test quantity is recalculated, and the empirical distribution is constructed. Then, on this basis, the P value is calculated to infer the model situation. (P<0.05) is used as the judgment basis to determine whether the model has failed to build or is overfitting.

[0058] When investigating plant leaf color traits, leaf colors are often similar with small differences, making it impossible to manually quantify and difficult to meet the needs of accurate investigation. Figure 3 As shown, eight Chinese cabbage samples have minimal leaf color differences and the same chlorophyll content, but are visually classified into different leaf color categories: leaves 1 and 2 are dark green; leaves 3 and 4 are oily green; leaves 5, 6, and 7 are dark green; and leaf 8 is grayish green. The following uses the detection of subtle differences in Chinese cabbage leaf color as an example, with reference to the accompanying drawings and specific embodiments, to further illustrate the present invention:

[0059] 1. The leaf color grade for this test was determined based on the test range, and 64 standard grade leaf samples were collected according to four leaf color grades. Chinese cabbage leaves were flattened and cleaned of dust and stains. Spectral data was collected using an improved plant leaf spectrum acquisition system, collecting spectral reflectance from 320nm to 1100nm.

[0060] 2. In order to reduce the instrument and environmental noise in the data, weaken and eliminate the influence of sample inhomogeneity, baseline drift, high-frequency noise, stray light, etc. on the spectral signal, it is necessary to consider preprocessing the spectral data. The present invention uses 15-point 5-time SG smoothing for spectral preprocessing and performs data normalization. SG smoothing can not only improve the signal-to-noise ratio of the signal, but also better maintain the useful information in the signal (spectrum). It is widely used in spectral analysis, such as Figure 5 shown.

[0061] 3. Establishment of the PLS-DA discriminant model of Chinese cabbage leaf color and the permutation test process: According to the characteristics of the light and dark color of Chinese cabbage leaves, Chinese cabbage is divided into four types (grades): gray-green, dark green, oily green, and dark green, as shown in Table 1, and the sample is assigned classification variable values ​​H, S, Y, and M; the PLS regression method is used to perform regression analysis on the multi-spectral and classification variables corresponding to the sample, and the optimal number of principal factors is determined and a PLS-DA model is established combining the multi-spectral characteristics with the classification variables to obtain the Chinese cabbage leaf color classification discriminant model established by the multi-spectral combined with the PLS-DA method. In the present invention, different numbers of principal factors are first selected. It can be seen from experiments that when the number of principal factors is 3, the model discrimination performance is relatively stable and has a higher discrimination accuracy. Figures 5 to 9 shown.

[0062] Table 1 Categorical variables of leaf samples

[0063]

[0064] like Figure 6-9 As shown, the PLS-DA model of the present invention contains three principal components, and the fitting parameter is R 2 X=0.0.996、R 2 Y=0.0.836、Q 2 =0.597. R 2 X and R 2 The closer Y is to 1, the more stable the model is. 2 A score greater than 0.5 indicates a high prediction accuracy, and the analysis results can be displayed on a score plot. The projected scores of each sample onto the plane formed by the first and second principal components are spatial coordinates, which intuitively reflect the similarities or differences between samples. If two samples are significantly different, the two coordinate points will be relatively far apart on the score plot, and vice versa.

[0065] like Figure 10 As shown, the permutation test is used to validate the PLS-DA model. Statistical inference is performed using random permutations of sample data. The samples are permuted in order, the statistical test is recalculated, an empirical distribution is constructed, and then the P value is calculated based on this for inference. The order of y is randomly repeated 200 times, and the separation model is fitted to all permuted y, extracting as many components as the original matrix y. 2 The larger the value, the better the predictive ability of the model; R 2 is the cumulative variance value, which indicates how much original data is used to establish a new PLS-DA discriminant model. A larger value indicates a stronger explanatory power of the model. Figure 10It can be seen that within the 95% confidence interval (P<0.05), the R 2 and Q 2 The values ​​are all lower than the rightmost R 2 and Q 2 value, and Q 2 The intercepts of the regression lines are all negative, indicating that the constructed PLS-DA discriminant models have no overfitting phenomenon, have good predictive ability, and can be effectively used for discriminant analysis of various categories.

[0066] 4. Input the spectrum of the sample to be identified into the PLS-DA model to obtain the leaf color type of the Chinese cabbage to be identified for technical verification.

[0067] As shown in Table 2, the PLS-DA model established through PLS discriminant analysis performed well in classifying the 64 Chinese cabbage leaf samples. With the exception of one misclassified sample, the recognition rate was 95.83%, while all other classifications achieved 100% recognition. The overall recognition rate was 98.44%, demonstrating that the multispectral combined with PLS-DA discriminant analysis method can effectively detect and discriminate the leaf color of Chinese cabbage of varying shades, providing an effective method for rapidly identifying different grades of Chinese cabbage leaf color.

[0068] Table 2 Classification results of 60 Chinese cabbage leaf samples based on PLS-DA model

[0069]

[0070] The present invention provides a method for distinguishing subtle differences in plant leaf color by combining multispectral analysis (PLS-DA). Through physical and chemical tests, a classification and discrimination model for plant leaf color grades and spectral data is established. This method can more simply and quickly identify subtle differences in plant leaf color and obtain plant leaf color grades. Spectra are sensitive tools for non-destructively measuring leaf color grades. This method is applied to scientific research on plant leaf color trait investigations, serves for the location of leaf color genes in plant genome data, and assists in molecular breeding. In addition, in order to reduce instrument and environmental noise in the data, weaken and eliminate the effects of sample unevenness, baseline drift, high-frequency noise, stray light, etc. on spectral signals, SG smoothing is used to improve the signal-to-noise ratio of the signal, and SNV is used to standardize the data. According to the actual characteristics of the sample, a classification variable value is assigned to the sample, and the PLS regression method is used to perform regression analysis on the multispectral and classification variables corresponding to the sample. The optimal number of principal factors is determined and a PLS-DA model is established based on the multispectral characteristics and classification variables. A plant leaf color classification discrimination model established by the multispectral combined with PLS-DA method is obtained, and an effective method for detecting plant leaf color grades is established.

[0071] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for distinguishing subtle differences in plant leaf color using multispectral and PLS-DA systems, characterized in that: The multi-spectral combined with PLS-DA system for distinguishing subtle differences in plant leaf color includes: a processor, a light-shielding black room, a portable ground object spectrometer, a light source, a liftable detection platform and a movable transmission platform, wherein the portable ground object spectrometer is arranged at the top center of the light-shielding black room for collecting spectral data of samples to be detected, the light source is arranged at the top of the light-shielding black room for irradiating samples to be detected, the liftable detection platform and the movable transmission platform are both arranged inside the light-shielding black room cover, and the movable transmission platform extends out from a side wall of the light-shielding black room; the liftable detection platform is arranged on the movable transmission platform, and the movable transmission platform is used to drive the liftable detection platform to move left and right; a sample placement area for to be detected is provided on the liftable detection platform for placing plant leaves, and the liftable detection platform is arranged below the portable ground object spectrometer; the processor is electrically connected to the portable ground object spectrometer, the light source, the liftable detection platform and the movable transmission platform respectively; The method for distinguishing subtle differences in plant leaf color using multispectral and PLS-DA includes the following steps: S1, select standard plant leaf samples and place them in the sample placement area to be tested, classify them according to the subtle characteristics of leaf color depth within the test range, and use a portable ground feature spectrometer to collect the spectral reflectance of the standard plant leaf samples according to the grade; S2, SG-SNV preprocessing of the collected spectral reflectance data using SG smoothing filter and SNV standard normalization method; S3, assign categorical variable values ​​to the samples based on the leaf color depth of the standard plant leaf samples; use the PLS regression method to perform regression analysis on the spectral reflectance data of the standard plant leaf samples and the categorical variable values ​​corresponding to the samples, determine the optimal number of main factors and establish a PLS-DA model based on multispectral features and categorical variables; S4, permutation test of the PLS-DA model; S5, placing the leaf sample to be tested in the sample placement area to be tested, collecting multispectral data of the leaf sample to be tested, and importing the data into the PLS-DA model to determine the leaf color depth level; In step S4, a permutation test is performed on the PLS-DA model, specifically including: Statistical inference is performed using random permutations of sample data. The samples are permuted sequentially, the statistical test quantity is recalculated, and an empirical distribution is constructed. Based on this, the asymptotic P value of the sample distribution is calculated based on the permutation test principle. The model is inferred based on the significance level of the P value to determine whether the model has failed or is overfitting. The light source is provided in two groups, which are respectively located on both sides of the top of the black chamber and are arranged inwardly at a 45-degree angle with the side wall of the black chamber; The movable transmission platform is a conveyor belt mechanism, comprising a bracket and a conveyor belt provided on the bracket, a driving wheel, a driven wheel, and a motor, wherein the conveyor belt is sleeved between the driving wheel and the driven wheel, the motor is drivingly connected to the driving wheel, and the motor is electrically connected to the processor; In step S1, the detection range is within the spectral reflectance band of 320nm to 1100nm.

2. The method of the system for distinguishing subtle differences in plant leaf color by combining multispectral and PLS-DA according to claim 1, characterized in that: The liftable testing platform is further provided with a calibration whiteboard, and the calibration whiteboard is arranged on the left side of the placement area of ​​the sample to be tested.

3. The method of the multispectral combined with PLS-DA system for distinguishing subtle differences in plant leaf color according to claim 1, characterized in that: The model of the portable ground object spectrometer is PSR-1100.

4. The method of the system for distinguishing subtle differences in plant leaf color by combining multispectral and PLS-DA according to claim 1, characterized in that: The processor is connected to a display.

Images