A method and device for classifying drought stress degree of plants based on chlorophyll fluorescence image

Abstract

The application discloses a method and device for classifying drought stress degrees of plants based on chlorophyll fluorescence images, measures canopy gray images of each plant to be measured, carries out threshold segmentation on the canopy gray images excited by saturated light to obtain mask mask images, calculates chlorophyll fluorescence parameters based on the mask mask images, takes the chlorophyll fluorescence parameters of each plant to be measured and the proportion of abnormal points corresponding to the chlorophyll fluorescence parameters as characteristic parameters, inputs a plant drought stress degree classification model, and determines the drought stress degree of the plant to be measured. The application improves the accuracy of drought stress degree classification by introducing the proportion of abnormal points.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of plant detection, and particularly relates to a method and device for classifying drought stress degrees of plants based on chlorophyll fluorescence images. BACKGROUND

[0002] Chlorophyll fluorescence is known as a non-destructive detection probe of plants, and is closely related to photosynthesis of plants and participates in competitive distribution of energy of photosynthesis of plants. Since chlorophyll fluorescence is inversely proportional to energy participating in photosynthesis, the photosynthesis capacity of plants can be indirectly reflected by detecting the intensity of chlorophyll fluorescence of plants, so as to achieve the purpose of monitoring the health condition and physiological information of plants and taking measures in advance.

[0003] Chlorophyll fluorescence imaging (CFI) is a visual detection technology in chlorophyll fluorescence technology at present, and has great research potential in plant growth and environmental stress. The chlorophyll fluorescence imaging technology can obtain chlorophyll fluorescence parameters and spatial heterogeneity distribution images of chlorophyll fluorescence in a non-destructive, simple and rapid manner, and has become an important method for classifying photosynthesis effects of plants and response mechanisms of plant stress.

[0004] Three different intensity excitation lights, i.e. measuring light, actinic light and saturated pulse light, are used in the chlorophyll fluorescence imaging process, which causes a large difference in contrast between plant leaves and background after irradiation of different excitation lights. There is an obvious difference between plant leaves and background after irradiation of saturated pulse light, and the reflected light of non-plant leaf area is small. During the whole detection process, the position of the plant to be detected basically does not change, and the reflected light of the non-plant area is small.

[0005] A plant drought stress diagnosis method and device based on chlorophyll fluorescence imaging technology are disclosed in Chinese patent (CN106546567A), which comprises the following steps: dark adaptation is performed on a sample set of plants with known drought stress diagnosis results, crown layer chlorophyll fluorescence image data of the plants is collected, and feature parameters of the chlorophyll fluorescence image are extracted; a plant drought stress determination model is established by using a multi-classifier fusion method according to the collected chlorophyll fluorescence image data and feature parameters; chlorophyll fluorescence image data and feature parameters of a plant to be detected are collected, and are substituted into the plant drought stress determination model to perform drought stress diagnosis.

[0006] A crop physiological water deficiency diagnosis method and system are disclosed in Chinese patent (CN109239027A), which controls the water pump and electromagnetic valve through the embedded system to control the water content in the substrate in the cultivation tank, simultaneously uses the soil water content sensor and chlorophyll fluorescence sensor to collect the chlorophyll fluorescence parameters under different substrate water contents, and obtains the crop physiological water deficiency diagnosis characteristic value as the starting value of entering the logarithmic decline period through logistic data fitting analysis. SUMMARY

[0007] Therefore, the present application provides a method and device for classifying the drought stress degree of plants based on chlorophyll fluorescence images, which accurately realizes the segmentation of plant leaves and improves the accuracy of drought stress degree classification by introducing the proportion of abnormal points.

[0008] The present application realizes the above technical object through the following technical means.

[0009] A method for classifying the drought stress degree of plants based on chlorophyll fluorescence images:

[0010] Step (1), divide the plants to be tested into four groups, the first group is watered with 150ml every two days, the second group is watered with 100ml every two days, the third group is watered with 50ml every two days, and the fourth group is subjected to continuous drought treatment;

[0011] Step (2), measure the canopy gray image of the plant to be tested, and calculate the chlorophyll fluorescence parameters of the four groups, including the minimum fluorescence F0 under dark adaptation, the maximum fluorescence F m M under dark adaptation, the steady-state fluorescence F m ′ after light adaptation, and the maximum fluorescence F ′ after light adaptation;

[0012] Step (3), input the chlorophyll fluorescence parameters and the corresponding proportion of abnormal points as characteristic parameters into the plant drought stress degree classification model to determine the drought stress degree of the plant to be tested.

[0013] Further, the process of calculating the chlorophyll fluorescence parameters of the four groups is as follows:

[0014] Collect the canopy gray image excited by saturated light, and perform threshold segmentation to obtain a mask mask image;

[0015] Perform pixel value matrix accumulation on the mask mask image to obtain the total number of leaf region pixel values num;

[0016] Multiply the mask mask image with the canopy gray images corresponding to F0, F m , F and F m ′ respectively to obtain the corresponding measurement pictures, and perform pixel value matrix accumulation on the measurement pictures to obtain the total value Data of the leaf region pixel values.

[0017] Data / (num*255) is calculated, i.e. a chlorophyll fluorescence parameter is obtained.

[0018] Further, the threshold segmentation adopts an OTSU threshold segmentation method.

[0019] Further, the abnormal point proportion comprises:

[0020] F v / F m Image 0.7-0.8 abnormal point proportion F v / F m _0 and F v / F m Image 0.6-0.7 abnormal point proportion F v / F m _1;

[0021] Y(II) Image 0.5-0.6 abnormal point proportion Y(II)_0, Y(II) Image 0.4-0.5 abnormal point proportion Y(II)_1, Y(II) Image 0.3-0.4 abnormal point proportion Y(II)_2, Y(II) Image 0.2-0.3 abnormal point proportion Y(II)_3, and Y(II) Image 0.1-0.2 abnormal point proportion Y(II)_4.

[0022] Further, the characteristic parameter further comprises PSII maximum photosynthetic efficiency F v / F m , PSII actual photosynthetic efficiency Y(II), photochemical quenching qP, non-photochemical quenching qN, photochemical quenching qL, non-photochemical quenching NPQ, and quantum yield of PSII regulated energy dissipation Y(NPQ).

[0023] Further, the calculation method of the abnormal point proportion is:

[0024] The pixel value matrix of the measurement image is traversed, data of all pixel values between 0.7-0.8, 0.6-0.7 is obtained and accumulated, as F v / F m Image abnormal point proportion;

[0025] The pixel value matrix of the measurement image is traversed, data of all pixel values between 0.5-0.6, 0.4-0.5, 0.3-0.4, 0.2-0.3 and 0.1-0.2 is obtained and accumulated, as the abnormal point proportion of the Y(II) image.

[0026] Further, the plant drought stress degree classification model is established by using a machine learning algorithm.

[0027] A device for classifying drought stress degree of plants based on chlorophyll fluorescence image, comprising: a CMOS camera, an STM32 single-chip microcomputer, an LED excitation light source, a CMOS camera trigger module and a notebook computer; the CMOS camera is arranged directly above the plant to be measured, used for collecting the canopy gray image of the plant to be measured and sending the image to the notebook computer, and the notebook computer processes and calculates the collected image; the LED excitation light source is arranged against the plant to be measured; the STM32 single-chip microcomputer is connected with the notebook computer and an LED switching power supply respectively, the notebook computer is connected with the CMOS camera trigger module and the CMOS camera in turn, the LED switching power supply is connected with the LED excitation light source, and the STM32 single-chip microcomputer controls the synchronous triggering of the CMOS camera and the LED excitation light source.

[0028] The present application has the following advantages:

[0029] (1) The present application takes the plant leaf image irradiated by saturated light as the target image of OTSU threshold segmentation to obtain a mask image, accumulates the pixel value matrix of the mask image to obtain the total number of leaf region pixel values num, multiplies the mask image with the F0, F m , F m and F v corresponding canopy gray images respectively to obtain the corresponding measurement pictures, and accumulates the pixel value matrix of the measurement pictures respectively to obtain the total value of leaf region pixel values Data; the Data / (num*255) is calculated, that is, the chlorophyll fluorescence parameter is obtained; the chlorophyll fluorescence parameter value generated by saturated light is larger, which can be well distinguished from the background, the edge extraction of the image is realized, and the chlorophyll fluorescence parameter is calculated more accurately.

[0030] (2) F v / F m image 0.7-0.8 abnormal point proportion F v / F m _0, F v / F m image 0.6-0.7 abnormal point proportion F v / F m _1, Y(II) image 0.5-0.6 abnormal point proportion Y(II)_0, Y(II) image 0.4-0.5 abnormal point proportion Y(II)_1, Y(II) image 0.3-0.4 abnormal point proportion Y(II)_2, Y(II) image 0.2-0.3 abnormal point proportion Y(II)_3, Y(II) image 0.1-0.2 abnormal point proportion Y(II)_4, and the accuracy of classification and identification is improved. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1This is a schematic diagram of the device structure for classifying plant drought stress based on chlorophyll fluorescence images according to the present invention, wherein: 1-CMOS camera, 2-STM32 microcontroller, 3-lens and filter, 4-LED excitation light source, 5-plant to be tested, 6-CMOS camera trigger module, 7-notebook computer. Detailed Implementation

[0032] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the scope of protection of the present invention is not limited thereto.

[0033] like Figure 1 As shown, this invention discloses a device for classifying plant drought stress based on chlorophyll fluorescence images, comprising a CMOS camera 1, an STM32 microcontroller 2, a lens and filter 3, an LED excitation light source 4, a CMOS camera trigger module 6, and a laptop computer 7. The CMOS camera 1 is positioned directly above the plant 5 under test and is used to acquire grayscale images of the plant's canopy, which are then sent to the laptop computer 7. The laptop computer 7 processes and calculates the acquired images, and diagnoses the plant's drought stress based on the results. The lens and filter 3 are mounted below the CMOS camera 1 and directly facing the plant under test. An 8mm focal length lens with an aperture F / value range of 1.8–16 is selected, and a 580nm long-pass filter is used. The LED excitation light source 4 is set towards the plant under test 5, and the LED excitation light source 4 consists of four LED light source boards emitting blue light. The STM32 microcontroller 2 is connected to the laptop 7 and the LED switching power supply respectively. The laptop 7 is connected to the CMOS camera trigger module 6 and the CMOS camera 1 in sequence. The LED switching power supply is connected to the LED excitation light source 4. The STM32 microcontroller 2 controls the synchronous triggering of the CMOS camera 1 and the LED excitation light source 4.

[0034] A method for classifying plant drought stress levels based on chlorophyll fluorescence images is proposed. This method involves setting different drought stress levels, measuring the canopy grayscale images of the plants under test, performing OTSU threshold segmentation to obtain mask images, calculating chlorophyll fluorescence parameters, and constructing a drought stress level classification model based on a deep learning algorithm. The feature parameters of the plants under test are then input into the drought stress level classification model to determine the degree of drought stress. This method not only accurately segments plant leaves but also improves the accuracy of drought stress level classification by incorporating the proportion of outliers in chlorophyll fluorescence parameters. The specific steps include:

[0035] Step (1): Divide the plants to be tested into four groups. The first group (HW) is watered with 150ml every two days, the second group (MW) is watered with 100ml every two days, the third group is watered with 50ml every two days, and the fourth group (NW) is subjected to continuous drought treatment.

[0036] Step (2), first, the plant to be tested is dark adapted for a period of time, then the plant sample to be tested is measured every night at 21:00, and four groups of chlorophyll fluorescence parameters are calculated; the specific process is as follows:

[0037] The plant to be tested is placed directly below the CMOS camera 1, and it is ensured that the leaves of the plant to be tested are all within the shooting range of the CMOS camera 1; in the dark adaptation stage: the 1Hz measurement light is turned on, the CMOS camera 1 triggers to shoot the canopy gray image at this time, and the minimum fluorescence F0 under dark adaptation is obtained after processing by the notebook computer 7; the 1Hz measurement light is turned off, and the saturation light is then turned on and then turned off; the 20Hz measurement light is turned on, the CMOS camera 1 triggers to shoot the canopy gray image at this time, and the maximum fluorescence F m ′ under dark adaptation is obtained after processing by the notebook computer 7; the 20Hz measurement light is turned off; in the light adaptation stage: the actinic light is turned on and then turned off, the 10Hz measurement light is turned on, the CMOS camera 1 triggers to shoot the canopy gray image at this time, and the steady-state fluorescence F after light adaptation is obtained after processing by the notebook computer 7; then the 10Hz measurement light is turned off, the saturation light is turned on and then turned off; the 20Hz measurement light is turned on, the CMOS camera 1 triggers to shoot the canopy gray image at this time, and the maximum fluorescence F m ′ after light adaptation is obtained after processing by the notebook computer 7; finally, the 20Hz measurement light is turned off, and the imaging process is completed.

[0038] The calculation process of the chlorophyll fluorescence parameters is as follows:

[0039] Step (2.1), the canopy gray image excited by the saturation light is collected, and OTSU threshold segmentation is performed to obtain a mask mask image;

[0040] The CMOS camera 1 collects the canopy gray images of the four groups of plants to be tested excited by the saturation light, and adopts the threshold (OTSU) segmentation method to segment the canopy gray images excited by the saturation light to obtain a mask mask image. The OTSU threshold segmentation of the canopy gray image can effectively extract the chlorophyll fluorescence imaging image of only the plant leaf area. The mask mask image obtained by default threshold segmentation is a binary image, that is, the leaf area is white, and the corresponding pixel value is 1, and the other areas are black, and the corresponding pixel value is 0.

[0041] The specific steps of OTSU threshold segmentation are as follows: first, the number of pixels corresponding to each gray scale from 0 to 255 is calculated; second, each gray scale from 0 to 255 is iterated and set as a threshold T; third, the average gray scale and the proportion of the number of pixels of the background image are calculated; fourth, the average gray scale and the proportion of the number of pixels of the foreground image are calculated; when the gray scale iteration is completed, the maximum value of the average gray scale and the proportion of the number of pixels corresponding to the inter-class variance is saved; finally, the maximum value of the inter-class variance is found to obtain the threshold result.

[0042] Step (2.2), the calculation process of chlorophyll fluorescence parameters is as follows:

[0043] The pixel value matrix of the mask image is accumulated to obtain the total number of pixel values ​​num in the leaf region;

[0044] The mask image is compared with F0 and F1 respectively. m F and F m The corresponding canopy grayscale images are multiplied to obtain the corresponding measurement images. The pixel value matrix of each measurement image is accumulated to obtain the total pixel value Data of the leaf area.

[0045] Calculate Data / (num×255) to obtain the chlorophyll fluorescence parameters.

[0046] Step (3) Establish a classification model for the degree of drought stress in plants. Use chlorophyll fluorescence parameters to determine the characteristic parameters of the classification model for the degree of drought stress in plants, so as to classify the degree of drought stress in plants.

[0047] Considering that when plants are subjected to drought stress, the chlorophyll fluorescence image at the leaf margin changes first compared to the central area of ​​the leaf, the proportion of outliers is also selected as part of the feature parameters.

[0048] Therefore, the characteristic parameters of the plant drought stress classification model include: minimum fluorescence F0 under dark adaptation and maximum fluorescence F under dark adaptation. m PSⅡ maximum photosynthetic efficiency F v / F m Steady-state fluorescence F after light adaptation; Maximum fluorescence F after light adaptation m ′, PSⅡ actual photosynthetic efficiency Y(II), photochemical quenching (based on the photosynthetic unit "swamp" model) qP, non-photochemical quenching (F0′ needs to be calculated) qN, photochemical quenching (based on the photosynthetic unit "lake" model) qL, non-photochemical quenching (F0′ does not need to be calculated) NPQ, quantum yield of PSⅡ regulated energy dissipation Y(NPQ), F v / F m Image 0.7–0.8 outlier percentage F v / F m _0、F v / F m Image 0.6–0.7 outlier percentage F v / F m1, Y(II) image 0.5-0.6 abnormal point ratio Y(II)_0, Y(II) image 0.4-0.5 abnormal point ratio Y(II)_1, Y(II) image 0.3-0.4 abnormal point ratio Y(II)_2, Y(II) image 0.2-0.3 abnormal point ratio Y(II)_3, and Y(II) image 0.1-0.2 abnormal point ratio Y(II)_4.

[0049] The calculation method of the abnormal point ratio is: traversing the pixel value matrix of the measurement picture, obtaining all pixel values between 0.7-0.8, 0.6-0.7, and accumulating them as F v / F m The abnormal point ratio of the image; the same method is adopted to obtain the abnormal point ratios of the pixel values in the Y(II) picture between 0.5-0.6, 0.4-0.5, 0.3-0.4, 0.2-0.3, and 0.1-0.2.

[0050] Method for classifying drought stress degree of plants: using machine learning algorithms such as extreme gradient boosting algorithm (XGBoost), radial basis neural network (RBF), and random forest (Random Forest) to establish a plant drought stress degree classification model, inputting four groups of feature parameters into the plant drought stress degree classification model to determine the drought stress degree of the plant to be tested, including normal state, mild drought stress, moderate drought stress, severe drought stress, and complete drought stress.

[0051] Embodiment:

[0052] I. Preparation of experimental materials. The experimental test sample is "Holland 8316" fruit cucumber. The cucumber seeds are cultivated in a seedling plug, and the substrate is a mixture of peat soil and vermiculite at a ratio of 3:1. When the seedlings grow to 3 leaves and 1 core, they are transplanted and cultivated in an artificial climate chamber. The daytime temperature in the artificial climate chamber is set to 28°C, and the nighttime temperature is set to 21°C. Multiple LED supplemental light lamps are used for supplemental lighting, with a PPFD (photosynthetic photon flux density) of about 400 μmol / m 2The light time is set to 12h per day (06:00-18:00). When the third true leaves of the cucumber seedlings grow, transplanting is carried out. On the first morning of the experiment, 200mL of water is poured into 4 groups of 24 cucumber samples, which is recorded as day1. Since the day of transplanting, the substrate water content is recorded every day using the weighing method. Among them, when the relative water content of the substrate is ≥60%, it is recorded as sufficient water; when the relative water content of the substrate is 60%-50%, it is recorded as light drought; when the relative water content of the substrate is 50%-40%, it is recorded as moderate drought; when the relative water content of the substrate is 40%-30%, it is recorded as severe drought; and when the relative water content of the substrate is 30%-0%, it is recorded as complete drought. In order to obtain the relative water content of the substrate of different experimental groups calculated by the weighing method, corresponding amounts of water are poured on the 1st, 3rd, 5th, 7th and 9th days for HW, MW and LW, and NW is continuously drought treated.

[0053] II. Obtain the three excitation lights required by the experiment by changing the size of the LED switching power supply voltage: measuring light, actinic light, and saturated light. The crown layer gray scale image is collected by controlling the CMOS camera 1 and the LED switching power supply by the STM32 single-chip microcomputer 2. Since the plant needs to undergo a period of dark adaptation before the measurement of some chlorophyll fluorescence parameters, the crown layer gray scale image of the plant sample is collected at 21:00 every night.

[0054] III. After the imaging process is completed, OTSU threshold segmentation is performed to obtain the mask graph of the mask, and the chlorophyll fluorescence parameters are calculated.

[0055] IV. In order to compare the influence of different machine learning algorithms on the classification performance and accuracy of different drought stress degrees of plants, different methods are used to classify the drought stress degrees of plants. Extreme gradient boosting algorithm, radial basis neural network and random forest machine learning algorithm are used to establish and test the plant drought stress degree classification model, so as to realize the purpose of accurately classifying five kinds of drought stress degrees of plants, i.e. normal state, mild drought stress, moderate drought stress, severe drought stress and complete drought stress. The experimental results show that the XGBoost algorithm has the highest classification accuracy for the drought stress degree of plants, followed by the random forest algorithm and the RBF neural network algorithm. However, when classifying the test set of plants with mild drought stress, only the XGBoost algorithm has high accuracy, and the RBF neural network and the random forest algorithm misclassify mild drought stress as normal state. When classifying the test set of plants with moderate drought stress, only about 50% of the accuracy is obtained, while in the severe classification, the classification accuracy of the three algorithms is high, close to 100%. However, in the classification of complete drought stress, only the XGBoost algorithm has the highest classification accuracy, close to 80%, which is much higher than the 25% classification accuracy of the RBF neural network and the random forest algorithm. Table 1 shows the evaluation index comparison of different algorithms for classifying the drought stress degree of plants. Among them, Accuracy is the accuracy index, which represents the correct rate of sample classification; Precision is the precision index, which represents the proportion of actual positive examples among the positive examples; Recall is the recall index, which represents the proportion of positive examples among the positive examples; F1-Measure is a statistical quantity, which is an evaluation standard for the weighted harmonic average of Precision and statistical quantity. According to the evaluation indexes shown in Table 1, it is found that the XGBoost algorithm has the best classification effect for different drought stress degrees of plants.

[0056] Table 1

[0057]

[0058]

[0059] The embodiments are preferred embodiments of the present application, but the present application is not limited to the above embodiments. Any obvious improvements, replacements or modifications made by those skilled in the art without departing from the essential content of the present application shall fall within the protection scope of the present application.

Claims

1. A method for classifying the degree of drought stress of plants based on chlorophyll fluorescence images, characterized by: step (1), dividing the plants to be tested into four groups, the first group being watered with 150 ml every two days, the second group being watered with 100 ml every two days, the third group being watered with 50 ml every two days, and the fourth group being subjected to continuous drought treatment; step (2), obtaining the chlorophyll fluorescence parameters of the plants to be tested; step (3), inputting the chlorophyll fluorescence parameters and the proportion of abnormal points corresponding thereto as characteristic parameters into a plant drought stress degree classification model to determine the degree of drought stress of the plants to be tested; the threshold segmentation uses the OTSU threshold segmentation method; the method for calculating the proportion of abnormal points is: traversing the pixel value matrix of the measurement picture to obtain all pixel values between 0.5-0.6, 0.4-0.5, 0.3-0.4, 0.2-0.3 and 0.1-0.2 and accumulating them as the proportion of abnormal points of the Y(II) image; the plant drought stress degree classification model is established using a machine learning algorithm. The system comprises: a CMOS camera (1), an STM32 single-chip microcomputer (2), an LED excitation light source (4), a CMOS camera triggering module (6) and a notebook computer (7); the CMOS camera (1) is arranged directly above the plants to be tested, is used to collect the crown layer gray-scale images of the plants to be tested and send them to the notebook computer (7), and the notebook computer (7) processes and calculates the collected images; the LED excitation light source (4) is arranged opposite to the plants to be tested; the STM32 single-chip microcomputer (2) is connected with the notebook computer (7) and an LED switching power supply, the notebook computer (7) is connected with the CMOS camera triggering module (6) and the CMOS camera (1) in sequence, the LED switching power supply is connected with the LED excitation light source (4), and the STM32 single-chip microcomputer (2) controls the synchronous triggering of the CMOS camera (1) and the LED excitation light source (4). Step (2), measure the canopy gray image of the plant to be tested, and calculate four groups of chlorophyll fluorescence parameters, including minimum fluorescence under dark adaptation , maximum fluorescence under dark adaptation , steady-state fluorescence F after light adaptation, and maximum fluorescence after light adaptation ; ​ The process of calculating four groups of chlorophyll fluorescence parameters is as follows: collecting a crown layer gray image excited by saturated light, and performing threshold segmentation to obtain a mask mask image; performing pixel value matrix accumulation on the mask mask image to obtain the total number of leaf area pixel values num; multiplying the mask mask image and the corresponding crown layer gray image F and 、 、F and respectively to obtain the corresponding measurement pictures, and performing pixel value matrix accumulation on the measurement pictures to obtain the total value of the leaf area pixel values Data; Calculations i.e. the chlorophyll fluorescence parameters are obtained; The abnormal point proportion includes: Image 0.7~0.8 abnormal point proportion _0 and Image 0.6~0.7 abnormal point proportion _1; Y (II) image 0.5~0.6 abnormal point proportion _0, Y (II) image 0.4~0.5 abnormal point proportion _1, Y (II) image 0.3~0.4 abnormal point proportion _2, Y (II) image 0.2~0.3 abnormal point proportion _3 and Y (II) image 0.1~0.2 abnormal point proportion _4.

2. The method of classifying the degree of drought stress of plants based on chlorophyll fluorescence images according to claim 1, characterized in that, ​ 3.The method of classifying the degree of drought stress of plants based on chlorophyll fluorescence images according to claim 1, characterized in that, The characteristic parameters further include PSII maximum photosynthetic efficiency PSII actual photosynthetic efficiency Y(II), photochemical quenching qP, non-photochemical quenching qN, photochemical quenching qL, non-photochemical quenching NPQ and quantum yield of PSII regulated energy dissipation Y(NPQ). 4.The method of classifying a degree of drought stress of a plant based on a chlorophyll fluorescence image according to claim 1, wherein, ​ Traverse the pixel value matrix of the measurement picture, obtain the data of all pixel values between 0.7-0.8, 0.6-0.7 and accumulate, as The proportion of abnormal points of the image; ​ 5. The method of classifying the degree of drought stress of plants based on chlorophyll fluorescence images according to claim 1, characterized in that, ​ 6. An apparatus for implementing the method of classifying the degree of drought stress of plants based on chlorophyll fluorescence images according to any one of claims 1 to 5, characterized in that, ​ ​

Images