Plant disease identification method based on chlorophyll fluorescence dynamic image

Abstract

The invention discloses a plant disease identification method based on a chlorophyll fluorescence dynamic image, and belongs to the technical field of plant disease identification, and the method comprises the following steps: constructing a laser-induced green fluorescence dynamic image acquisition system; the method comprises the following steps: placing a laser-induced green fluorescence dynamic image acquisition system in a black box, and acquiring characteristic values and characteristic images of plant leaves; constructing a feature image data set based on the feature values and the feature images of the plant leaves; according to the feature image data set, performing prediction training on the plant leaf disease degree by using an RF classification regression tree to obtain a trained RF classification regression tree model; and acquiring a new feature image of the plant leaf by using a laser-induced green fluorescence dynamic image acquisition system, inputting the feature image into the trained RF classification regression tree model, and identifying to obtain the plant leaf disease degree of the plant leaf. According to the invention, the problem of insufficient plant disease identification precision and accuracy is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of plant disease identification, and in particular relates to a plant disease identification method based on chlorophyll fluorescence dynamics images. Background Art

[0002] With the rapid development of remote sensing, computer vision and machine learning, optical remote sensing has been widely used in the field of crop growth and disease monitoring. The reflectance spectrum and chlorophyll fluorescence images of plant leaves have been widely used in remote sensing monitoring of plant diseases and have achieved remarkable results. Reflectance spectrum data can effectively reflect the changes in plant canopy structure, laying an important foundation for large-scale remote sensing crop disease detection. In recent years, many spectral disease indices have been introduced. However, the impact of disease on the spectrum mainly depends on the change rate of its physiological disease symptoms. Therefore, the plant disease stress response based on reflectance spectrum data has obvious lag. In the early stage after disease infection, plants mainly adapt to external stress quickly by adjusting physiological mechanisms, which is manifested as changes in photosynthetic physiological parameters. In addition, chlorophyll fluorescence data can sensitively reflect the changes in plant photosynthetic physiological characteristics and can be used to achieve early detection of diseases.

[0003] Due to the competitive relationship between chlorophyll fluorescence, energy dissipation and photochemical reactions of plants, any changes in photochemical reactions and energy dissipation will cause fluorescence effects. Studies have shown that when photochemical and thermal dissipation are low, the fluorescence yield is high. In contrast, when the photosynthetic rate or thermal dissipation increases, the fluorescence yield decreases. Therefore, changes in chlorophyll fluorescence parameters can be used to obtain information on photochemical efficiency and plant stress levels. When plants are stressed by diseases, the value of the fluorescence parameter Fv / Fm (variable fluorescence / maximum fluorescence yield) decreases because the dynamic balance of chlorophyll molecule synthesis and degradation is disrupted, reducing the PS II (photosystem II) activity of diseased plants. Chlorophyll fluorescence detection can be roughly divided into active chlorophyll fluorescence detection and SIF (solar induced chlorophyll fluorescence) detection. The SIF signal is very weak, usually less than 2% of the incident radiation, and is mixed with the reflected signal, which poses a great challenge to the extraction of the SIF signal. Active chlorophyll fluorescence detection includes two methods: CFI (chlorophyll fluorescence induction) technology and laser induced fluorescence technology. However, the tissue structure and chlorophyll content in different parts of the leaves are different, resulting in the heterogeneity of leaf photosynthesis. Fluorescence imaging can provide information about certain features, such as the color texture and fluorescence intensity of plants, and can also reveal the spatiotemporal heterogeneity of stressed leaves. Therefore, chlorophyll fluorescence imaging methods are used to distinguish diseased plants from healthy plants, achieve early diagnosis of infected crops, and detect crop diseases in a timely manner. However, the existing plant disease identification accuracy and precision still need to be improved. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a plant disease identification method based on chlorophyll fluorescence dynamic images, which solves the problem of insufficient precision and accuracy in plant disease identification.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0006] The present invention provides a plant disease identification method based on chlorophyll fluorescence dynamics images, comprising the following steps:

[0007] S1. Construct laser-induced green fluorescence dynamics image acquisition system;

[0008] S2, placing the laser-induced chlorophyll fluorescence dynamics image acquisition system in a black box, and acquiring characteristic values ​​and characteristic images of plant leaves;

[0009] S3, constructing a feature image dataset based on the feature values ​​and feature images of plant leaves;

[0010] S4, according to the feature image data set, using the RF classification regression tree to predict the degree of plant leaf disease and obtain a trained RF classification regression tree model;

[0011] S5. Use the laser-induced chlorophyll fluorescence dynamic image acquisition system to collect the characteristic image of the new plant leaf, and input the characteristic image into the trained RF classification regression tree model to identify the degree of plant leaf disease of the plant leaf.

[0012] Furthermore, the laser-induced green fluorescence dynamics image acquisition system comprises:

[0013] A laser source, used for emitting continuous laser light with a wavelength of 460 nm;

[0014] Optical fiber, used to transmit laser light to the beam expander;

[0015] A beam expander, used to diffuse the laser light so that the diffused light is irradiated onto the surface of the plant leaf on the sample stage;

[0016] A photoelectric detector is used to collect chlorophyll fluorescence dynamics images generated by light irradiating the surface of plant leaves through a bandpass filter;

[0017] The data acquisition unit is used to acquire characteristic images and corresponding characteristic values ​​in the chlorophyll fluorescence kinetics attenuation process according to the chlorophyll fluorescence kinetics image.

[0018] Furthermore, the method for acquiring characteristic images and corresponding characteristic values ​​in the chlorophyll fluorescence kinetics attenuation process according to the chlorophyll fluorescence kinetics image comprises the following steps:

[0019] using a photodetector and a data acquisition unit to collect a number of fluorescence images of the plant leaves when the laser source is not turned on, as uninduced fluorescence images;

[0020] Selecting pixels in the uninduced fluorescence image that are lower than a set pixel threshold to form a noise pixel set;

[0021] According to the noise pixel set, pixels in the chlorophyll fluorescence dynamics image that are opposite to the pixels in the noise pixel set are eliminated;

[0022] The pixel average value of the pixels in the chlorophyll fluorescence dynamics image and the non-noise pixel point set is calculated to obtain a characteristic image;

[0023] The descending curve area of ​​the characteristic image and the ratio of the chlorophyll fluorescence intensity peak value to the stable value are calculated as the characteristic values ​​corresponding to the characteristic image.

[0024] Furthermore, S2 comprises the following steps:

[0025] S21, placing the laser-induced green fluorescence dynamics image acquisition system in a black box;

[0026] S22, flattening the obtained plant leaves and placing them on a sample table;

[0027] S23, setting the photoelectric detector to a dynamic mode, setting the lens of the photoelectric detector vertically downward at a distance of 50 cm from the plant leaves, and setting the sampling period of the data acquisition unit to 0.8 s, the magnification to 20, and the number of continuously collected images to 600;

[0028] S24, start the laser-induced chlorophyll fluorescence dynamics image acquisition system, so that the laser emitted by the laser source is diffused by the beam expander and then irradiated onto the plant leaves, and obtains the characteristic image and its corresponding characteristic value in the chlorophyll fluorescence dynamics attenuation process through the photoelectric detector and the data acquisition unit.

[0029] Furthermore, there is a step A1 between S21 and S22, of pasting black anti-reflection paper on the bottom of the sample table.

[0030] Furthermore, the characteristic image is a chlorophyll fluorescence intensity kinetic curve diagram; the characteristic values ​​are the descending curve area of ​​the chlorophyll fluorescence intensity kinetic curve diagram and the ratio of the chlorophyll fluorescence intensity peak value to the stable value.

[0031] Furthermore, the S4 comprises the following steps:

[0032] S41, extracting several groups of feature images and their corresponding feature values ​​from the feature image data set as samples according to the bootstrap resampling method, and taking the plant leaf disease degree corresponding to the feature image as the true category of the sample;

[0033] S42, according to the RF classification and regression tree algorithm, based on each group of samples and their true categories, respectively construct a tree pair model;

[0034] S43, based on the decision tree algorithm, using the constructed tree pair model to repeatedly perform plant leaf disease degree prediction training on each sample until a preset number of training times is reached, and then proceed to S44;

[0035] S44. According to the majority voting method, based on the prediction accuracy of the leaf disease category of the object, the target tree pair model is selected from each tree pair model by voting as the trained RF classification regression tree model.

[0036] As a preferred solution, the characteristic image is a chlorophyll fluorescence intensity kinetic curve corresponding to a preset area range block on the chlorophyll fluorescence kinetic image, wherein the preset area range block is not smaller than the size of the plant leaf spot.

[0037] The beneficial effects of the present invention are as follows: a plant disease identification method based on chlorophyll fluorescence kinetic images provided by the present invention realizes efficient and accurate collection of plant leaf features by constructing a laser-induced chlorophyll fluorescence kinetic image acquisition system, and provides reliable basic data; by placing the laser-induced chlorophyll fluorescence kinetic image acquisition system in a black box for operation, the interference of external environmental factors is effectively reduced, and the accuracy and stability of the collected characteristic values ​​and characteristic images are improved; based on the collected plant leaf characteristic values ​​and characteristic images, a characteristic image data set is constructed, which provides rich samples for the tree pair model to train the plant leaf disease degree prediction on samples, and helps to improve the generalization ability and accuracy of the model; the trained RF classification regression tree model can automatically identify new plant leaf characteristic images, and accurately judge the disease degree of plant leaves, realizing the intelligent and automatic disease identification; the present invention not only improves the efficiency of plant disease identification, but also improves the accuracy of plant disease degree identification by means of chlorophyll fluorescence kinetics and machine learning, and provides strong technical support for the prevention and control of plant diseases.

[0038] Other advantages of the present invention will be analyzed in more detail in subsequent embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 The present invention is a flowchart of a method for identifying plant diseases based on chlorophyll fluorescence dynamic images in an embodiment of the present invention.

[0041] FIG. 2( a ) is a chlorophyll fluorescence dynamics image of a plant leaf in an embodiment of the present invention.

[0042] FIG. 2( b ) is a schematic diagram of a leaf region of a plant leaf in an embodiment of the present invention.

[0043] FIG. 2( c ) is a graph showing the kinetics of chlorophyll fluorescence intensity of plant leaves in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0045] like Figure 1 As shown, in one embodiment of the present invention, the present invention provides a plant disease identification method based on chlorophyll fluorescence dynamics image, comprising the following steps:

[0046] S1. Construct laser-induced green fluorescence dynamics image acquisition system;

[0047] The laser-induced green fluorescence dynamics image acquisition system comprises:

[0048] A laser source, used for emitting continuous laser light with a wavelength of 460 nm;

[0049] Optical fiber, used to transmit laser light to the beam expander;

[0050] A beam expander is used to diffuse the laser so that the diffused light is irradiated onto the surface of the plant leaves on the sample stage; in this embodiment, the selected plant leaves are wolfberry leaves;

[0051] The photodetector is used to collect the chlorophyll fluorescence dynamics image generated by the light irradiating the surface of the plant leaves through the bandpass filter; the photodetector used in this embodiment is an EMCCD (electron multiplying charge coupled device), which has extremely high sensitivity and signal-to-noise ratio, and is particularly suitable for detecting weak fluorescence signals. In this embodiment, a 690FS10-50 bandpass filter is used to selectively pass the main wavelength component of chlorophyll fluorescence in the red light region, while filtering out light of other wavelengths, thereby improving the contrast and signal-to-noise ratio of the fluorescence image.

[0052] The data acquisition unit is used to acquire characteristic images and corresponding characteristic values ​​in the chlorophyll fluorescence kinetics attenuation process according to the chlorophyll fluorescence kinetics image.

[0053] The method for acquiring characteristic images and corresponding characteristic values ​​in the chlorophyll fluorescence dynamics decay process according to the chlorophyll fluorescence dynamics image comprises the following steps:

[0054] using a photodetector and a data acquisition unit to collect a number of fluorescence images of the plant leaves when the laser source is not turned on, as uninduced fluorescence images;

[0055] Selecting pixels in the uninduced fluorescence image that are lower than a set pixel threshold to form a noise pixel set;

[0056] According to the noise pixel set, pixels in the chlorophyll fluorescence dynamics image that are opposite to the pixels in the noise pixel set are eliminated;

[0057] The pixel average value of the pixels in the chlorophyll fluorescence dynamics image and the non-noise pixel point set is calculated to obtain a characteristic image;

[0058] The descending curve area of ​​the characteristic image and the ratio of the chlorophyll fluorescence intensity peak value to the stable value are calculated as the characteristic values ​​corresponding to the characteristic image.

[0059] S2, placing the laser-induced chlorophyll fluorescence dynamics image acquisition system in a black box, and acquiring characteristic values ​​and characteristic images of plant leaves;

[0060] The S2 comprises the following steps:

[0061] S21, placing the laser-induced green fluorescence dynamics image acquisition system in a black box;

[0062] As a preferred solution, there is a step A1 between S21 and S22, that is, pasting black anti-reflection paper on the bottom of the sample stage. In this embodiment, by pasting black anti-reflection paper on the bottom of the sample stage, the plant leaves are imaged on a black light-absorbing background, which can absorb the light of the surrounding environment to the maximum extent, thereby avoiding the interference of external light on the chlorophyll fluorescence signal, ensuring the accuracy of the collection result, and secondly, on the black background, the chlorophyll fluorescence signal emitted by the plant leaves will be more prominent, making the fluorescence image clearer, which is convenient for subsequent data processing and analysis.

[0063] S22. Flatten the obtained plant leaves and place them on a sample table; when collecting fluorescent images on a black background, the plant leaves are in an extended state, which can more realistically reflect the photosynthetic efficiency and physiological state of the leaves.

[0064] S23, setting the photoelectric detector to a dynamic mode, setting the lens of the photoelectric detector vertically downward at a distance of 50 cm from the plant leaves, and setting the sampling period of the data acquisition unit to 0.8 s, the magnification to 20, and the number of continuously collected images to 600;

[0065] S24, start the laser-induced chlorophyll fluorescence dynamics image acquisition system, so that the laser emitted by the laser source is diffused by the beam expander and then irradiated onto the plant leaves, and obtains the characteristic image and its corresponding characteristic value in the chlorophyll fluorescence dynamics attenuation process through the photoelectric detector and the data acquisition unit.

[0066] The characteristic image is a chlorophyll fluorescence intensity kinetic curve graph; the characteristic value is the descending curve area of ​​the chlorophyll fluorescence intensity kinetic curve graph, and the ratio of the chlorophyll fluorescence intensity peak value to the stable value. In this embodiment, the chlorophyll fluorescence intensity stable value is the chlorophyll fluorescence intensity when the rate of decrease of the chlorophyll fluorescence intensity value over time is less than the preset change rate threshold for the first time, or the average value of the chlorophyll fluorescence intensity when the chlorophyll fluorescence intensity kinetic curve is flat.

[0067] In this embodiment, wolfberry is taken as an example, as shown in FIG2(a), which is a partial chlorophyll fluorescence dynamics image during the acquisition process from the start of the laser-induced chlorophyll fluorescence dynamics image acquisition system to 480 seconds after the start, specifically, the chlorophyll fluorescence images of three wolfberry leaves at different times, wherein the three wolfberry leaves are normal leaves, Pre-diseased leaves with early leaf blight disease, and Leaf blight diseased leaves;

[0068] As shown in Figure 2(b), this scheme recorded the change of chlorophyll fluorescence intensity of one leaf area of ​​each wolfberry leaf within 0 to 480 seconds after the laser-induced chlorophyll fluorescence dynamic image acquisition system was started. The leaf areas of the three leaves included the leaf area in box 1 of the normal leaf, the leaf area in box 2 of the Pre-diseased leaf, and the leaf area in box 3 of the Leaf blight diseased leaf.

[0069] As shown in Figure 2(c), in the chlorophyll fluorescence intensity kinetic curve, the fluorescence intensity of the three leaf types first increased rapidly and then decreased slowly with time, and the rate of decrease and steady-state value continued to change; compared with the other two leaf types, the fluorescence intensity of normal leaves Normal decreased faster from the maximum value, and the slope of the curve was larger, and the steady-state fluorescence value was relatively small; the fluorescence decrease rate of Pre-diseased leaves was slower than that of normal leaves, and the steady-state value was higher; compared with the other two leaf types, the fluorescence intensity value of Leaf blight diseased leaves decreased more slowly, and the steady-state fluorescence value increased significantly.

[0070] S3, constructing a feature image dataset based on the feature values ​​and feature images of plant leaves;

[0071] S4, according to the feature image data set, using the RF classification regression tree to predict the degree of plant leaf disease and obtain a trained RF classification regression tree model;

[0072] The S4 comprises the following steps:

[0073] S41, according to the bootstrap resampling method, extracting several groups of feature images and their corresponding feature values ​​from the feature image data set as samples, and taking the plant leaf disease degree corresponding to the feature image as the true category of the sample; in this scheme, the plant leaf disease degree includes normal leaves, early leaf blight leaves and leaf blight leaves;

[0074] S42. According to the RF classification and regression tree algorithm, tree pair models are constructed based on each group of samples and their true categories; the RF classification and regression tree algorithm is suitable for situations where the amount of data is limited and high precision is required. The advantage of the RF algorithm is that it can run large data sets, and due to the introduction of randomness, the possibility of overfitting is small, which means that the classifier will not produce an overfitting model. Its main advantage is that a large number of tree pair models reduce the prediction speed of the algorithm and facilitate real-time application;

[0075] S43, based on the decision tree algorithm, using the constructed tree pair model to repeatedly perform plant leaf disease degree prediction training on each sample until a preset number of training times is reached, and then proceed to S44;

[0076] S44. According to the majority voting method, based on the prediction accuracy of the leaf disease category of the object, the target tree pair model is selected from each tree pair model by voting as the trained RF classification regression tree model.

[0077] S5. Use the laser-induced chlorophyll fluorescence dynamic image acquisition system to collect the characteristic image of the new plant leaf, and input the characteristic image into the trained RF classification regression tree model to identify the degree of plant leaf disease of the plant leaf.

[0078] On the basis of the above embodiments, in a preferred implementation example of the present invention, the characteristic image is a chlorophyll fluorescence intensity kinetics curve graph corresponding to a preset area range block on the chlorophyll fluorescence kinetics image, wherein the preset area range block is not smaller than the size of the plant leaf spot.

[0079] In this scheme, as a preferred scheme, when the characteristic image is related to the preset area range block on the chlorophyll fluorescence kinetic image, the blocks with peak values ​​higher than the average peak values ​​in the preset area range blocks in the chlorophyll fluorescence kinetic image of the plant leaves are recalculated to obtain the corresponding chlorophyll fluorescence intensity kinetic curve graph, and then according to the RF classification regression tree algorithm, the curve graph is used as the characteristic image to perform plant leaf disease degree prediction training, thereby effectively improving the accuracy and efficiency of identifying the degree of plant leaf disease.

[0080] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A plant disease identification method based on chlorophyll fluorescence dynamics image, characterized in that: The steps include: S1. Construct laser-induced green fluorescence dynamics image acquisition system; S2, placing the laser-induced chlorophyll fluorescence dynamics image acquisition system in a black box, and acquiring characteristic values ​​and characteristic images of plant leaves; S3, constructing a feature image dataset based on the feature values ​​and feature images of plant leaves; S4, according to the feature image data set, using the RF classification regression tree to predict the degree of plant leaf disease and obtain a trained RF classification regression tree model; S5. Use the laser-induced chlorophyll fluorescence dynamic image acquisition system to collect the characteristic image of the new plant leaf, and input the characteristic image into the trained RF classification regression tree model to identify the degree of plant leaf disease of the plant leaf.

2. The plant disease identification method based on chlorophyll fluorescence dynamics image according to claim 1, characterized in that: The laser-induced green fluorescence dynamics image acquisition system comprises: A laser source, used for emitting continuous laser light with a wavelength of 460 nm; Optical fiber, used to transmit laser light to the beam expander; A beam expander, used to diffuse the laser light so that the diffused light is irradiated onto the surface of the plant leaf on the sample stage; A photoelectric detector is used to collect chlorophyll fluorescence dynamics images generated by light irradiating the surface of plant leaves through a bandpass filter; The data acquisition unit is used to acquire characteristic images and corresponding characteristic values ​​in the chlorophyll fluorescence kinetics attenuation process according to the chlorophyll fluorescence kinetics image.

3. The plant disease identification method based on chlorophyll fluorescence dynamics image according to claim 2, characterized in that: The method for acquiring characteristic images and corresponding characteristic values ​​in the chlorophyll fluorescence dynamics decay process according to the chlorophyll fluorescence dynamics image comprises the following steps: using a photodetector and a data acquisition unit to collect a number of fluorescence images of the plant leaves when the laser source is not turned on, as uninduced fluorescence images; Selecting pixels in the uninduced fluorescence image that are lower than a set pixel threshold to form a noise pixel set; According to the noise pixel set, pixels in the chlorophyll fluorescence dynamics image that are opposite to the pixels in the noise pixel set are eliminated; The pixel average value of the pixels in the chlorophyll fluorescence dynamics image and the non-noise pixel point set is calculated to obtain a characteristic image; The descending curve area of ​​the characteristic image and the ratio of the chlorophyll fluorescence intensity peak value to the stable value are calculated as the characteristic values ​​corresponding to the characteristic image.

4. The plant disease identification method based on chlorophyll fluorescence dynamics image according to claim 3, characterized in that: The S2 comprises the following steps: S21, placing the laser-induced green fluorescence dynamics image acquisition system in a black box; S22, flattening the obtained plant leaves and placing them on a sample table; S23, setting the photoelectric detector to a dynamic mode, setting the lens of the photoelectric detector vertically downward at a distance of 50 cm from the plant leaves, and setting the sampling period of the data acquisition unit to 0.8 s, the magnification to 20, and the number of continuously collected images to 600; S24, start the laser-induced chlorophyll fluorescence dynamics image acquisition system, so that the laser emitted by the laser source is diffused by the beam expander and then irradiated onto the plant leaves, and obtains the characteristic image and its corresponding characteristic value in the chlorophyll fluorescence dynamics attenuation process through the photoelectric detector and the data acquisition unit.

5. The method for identifying plant diseases based on chlorophyll fluorescence dynamics images according to claim 4, characterized in that: There is also a step A1 between S21 and S22, of pasting black anti-reflection paper on the bottom of the sample stage.

6. The method for identifying plant diseases based on chlorophyll fluorescence dynamics images according to claim 4, characterized in that: The characteristic image is a chlorophyll fluorescence intensity kinetic curve diagram; the characteristic values ​​are the descending curve area of ​​the chlorophyll fluorescence intensity kinetic curve diagram and the ratio of the chlorophyll fluorescence intensity peak value to the stable value.

7. The method for identifying plant diseases based on chlorophyll fluorescence dynamics images according to claim 6, characterized in that: The S4 comprises the following steps: S41, extracting several groups of feature images and their corresponding feature values ​​from the feature image data set as samples according to the bootstrap resampling method, and taking the plant leaf disease degree corresponding to the feature image as the true category of the sample; S42, according to the RF classification and regression tree algorithm, based on each group of samples and their true categories, respectively construct a tree pair model; S43, based on the decision tree algorithm, using the constructed tree pair model to repeatedly perform plant leaf disease degree prediction training on each sample until a preset number of training times is reached, and then proceed to S44; S44. According to the majority voting method, based on the prediction accuracy of the leaf disease category of the object, the target tree pair model is selected from each tree pair model by voting as the trained RF classification regression tree model.

8. The method for identifying plant diseases based on chlorophyll fluorescence dynamics images according to claim 5, characterized in that: The characteristic image is a chlorophyll fluorescence intensity kinetic curve corresponding to a preset area range block on the chlorophyll fluorescence kinetic image, wherein the preset area range block is not smaller than the size of the plant leaf spot.

Images