Automatic detection method and system for rice diseases

Abstract

The invention discloses an automatic detection method and system for rice diseases, and relates to the field of crop detection.The automatic detection method comprises the steps that a biodegradable optical marking agent is prepared, the biodegradable optical marking agent is sprayed to rice leaves through a high-pressure micro-fluidic spraying device, and marked rice leaves are formed; scanning leaf vascular tissues in the marked rice leaves, capturing a GFP fluorescence signal and a CQD fluorescence signal, generating a three-dimensional fluorescence intensity graph, and processing the three-dimensional fluorescence intensity graph by using ImageJ image analysis software to obtain a disease positioning fluorescence graph of the rice leaves; and according to the disease positioning fluorogram of the rice leaves, scanning and marking the reflection spectrum of the rice leaves to obtain absorption characteristics of the reflection spectrum, and comparing the absorption characteristics with a pre-constructed disease fingerprint database to obtain rice disease diagnosis data. By preparing the double-labeled biodegradable optical labeling agent, the interference among fluorescence signals is reduced, and meanwhile, the accuracy of three-dimensional reconstruction of a rice disease area is also improved.

Description

Technical Field

[0001] The present invention relates to the field of crop detection, and in particular to an automatic detection method and system for rice diseases. Background Art

[0002] In recent years, automated rice disease detection and management technologies have made rapid progress in the agricultural sector. With the widespread application of near-infrared spectroscopy, fluorescence imaging, and biomarkers, a variety of disease detection methods based on spectral analysis and image processing have been developed. For example, traditional near-infrared spectrometers scan the reflectance spectrum of rice leaves and, combined with chemometric models, identify the absorption characteristics of common diseases such as rice blast. Fluorescent labeling technology captures the fluorescence signal of GFP or CQDs to accurately locate diseased areas.

[0003] Existing automated rice disease detection methods have areas that need to be addressed. Conventional spectral analysis methods, while capable of identifying disease types, are limited in their ability to precisely locate diseased areas, complicating operations within rice fields. Fluorescence imaging techniques, which rely on a single marker (such as GFP), limit signal resolution due to tissue penetration depth and signal crosstalk, resulting in inaccurate reconstruction of three-dimensional disease distribution. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an automatic detection method for rice diseases, which solves the problems of low disease detection and positioning accuracy and insufficient diagnostic comprehensiveness in existing automatic detection methods for rice diseases.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides an automatic detection method for rice diseases, comprising preparing a biodegradable optical marker, spraying the biodegradable optical marker onto rice leaves using a high-pressure microfluidic injection device to form marked rice leaves; The leaf vascular tissue in the labeled rice leaves was scanned to capture GFP fluorescence signals and CQD fluorescence signals, and a three-dimensional fluorescence intensity map was generated. The image was then processed using ImageJ image analysis software to obtain a fluorescence map of the disease location on the rice leaves. Based on the disease location fluorescence image of rice leaves, the reflectance spectrum of the marked rice leaves is scanned to obtain the absorption characteristics of the reflectance spectrum and compared with the pre-built disease fingerprint library to obtain rice disease diagnosis data; Use an intelligent sprayer to automatically dispense medicine and spray it onto the disease-infected area in the disease location fluorescence image of rice leaves. Track the degradation process of the biodegradable optical marker and output a rice disease management report.

[0007] As a preferred embodiment of the automatic detection method for rice diseases of the present invention, the preparation of the biodegradable optical marker comprises inserting GFP into Bacillus subtilis by electroporation, screening positive transformants, and culturing the screened positive transformants to obtain a Bacillus subtilis suspension; CQDs were synthesized by hydrothermal carbonization and encapsulated in chitosan to form a chitosan-CQD hydrogel microsphere solution; The Bacillus subtilis suspension and the chitosan-CQD hydrogel microsphere solution are mixed to obtain a biodegradable optical marker.

[0008] As a preferred embodiment of the automatic detection method for rice diseases of the present invention, the step of forming the marked rice leaves includes configuring a high-pressure microfluidic injection device, setting injection parameters, and spraying the biodegradable optical marker into the stomatal area on the back of the rice leaves.

[0009] As a preferred embodiment of the automatic detection method for rice diseases of the present invention, the capturing of GFP fluorescence signals and CQD fluorescence signals includes selecting a Thorlabs Bergamo II as a two-photon microscope and performing registration, and focusing the laser beam generated by the registered two-photon microscope on the marked rice leaves.

[0010] As a preferred embodiment of the automatic detection method for rice diseases of the present invention, obtaining a disease location fluorescence image of rice leaves includes separating GFP fluorescence signals and CQD fluorescence signals using an optical path unit of a registered two-photon microscope, and recording the fluorescence intensities of the separated GFP fluorescence signals and CQD fluorescence signals, respectively, to obtain a rice leaf fluorescence image dataset; The rice leaf fluorescence image dataset was pseudo-colored to obtain a coded fluorescence image dataset, and the coded fluorescence image dataset was multi-channel stacked using the Bio-Formats plug-in to form a three-dimensional dataset. The fluorescence intensity threshold was set, and the fluorescence intensity of the GFP fluorescence signal and the CQD fluorescence signal in the three-dimensional data set was counted using ImageJ software and compared with the fluorescence intensity threshold to determine the diseased area of the rice. The volume and centroid coordinates of the voxels in the rice disease area were counted to obtain the disease location fluorescence map of the rice leaf.

[0011] As a preferred embodiment of the automatic rice disease detection method of the present invention, obtaining rice disease diagnostic data includes configuring an infrared handheld detection device, using the infrared handheld detection device to scan the rice disease area in the disease location fluorescence image of the rice leaf, capturing the reflected light intensity of the marked rice leaf, and generating reflected light intensity data points; Calculate the reflectance of the reflected light intensity data points and save them in TXT format to obtain the reflectance spectrum dataset of rice leaves; An ARM Cortex-M4 microcontroller was used to apply Savitzky-Golay filtering to the reflectance spectrum dataset of rice leaves to obtain a smoothed reflectance spectrum dataset. An absorption dip threshold is set and compared with the reflected light intensity in the smoothed reflectance spectrum dataset to obtain the absorption characteristics of the reflectance spectrum and perform normalization to generate an absorption spectrum feature dataset; Rice leaves with diseased samples were collected to construct a disease fingerprint library. The cosine similarity between the reflected light intensity in the disease fingerprint library and the normalized reflected light intensity in the absorption spectrum feature dataset was calculated to obtain the rice disease diagnosis results and evaluate the severity of the rice disease. The rice disease diagnosis results, the severity of rice diseases and the centroid coordinates were integrated into a CSV file to obtain the rice disease diagnosis data.

[0012] As a preferred embodiment of the automatic detection method for rice diseases of the present invention, the outputting of the rice disease management report includes grouping the centroid coordinates in the rice disease diagnosis data using a hierarchical clustering algorithm to generate a minimum covering rectangle; The S-shaped path algorithm is used to generate a preliminary spraying path in the minimum coverage rectangle, and the preliminary spraying path is optimized using RTK-GPS navigation to generate the final spraying path. Use a smart sprayer to spray pesticides along the final spray path to the disease-infected area to obtain pesticide application data, and use a soil sensor network to monitor the degradation process of biodegradable optical markers to obtain CQD fluorescence data; Integrate CQD fluorescence data and pesticide application data to generate rice disease management reports.

[0013] In a second aspect, the present invention provides an automatic detection system for rice diseases, comprising: a preparation module for preparing a biodegradable optical marker, and spraying the biodegradable optical marker onto rice leaves using a high-pressure microfluidic injection device to form marked rice leaves; The scanning module scans the leaf vascular tissue in the marked rice leaves, captures the GFP fluorescence signal and the CQD fluorescence signal, generates a three-dimensional fluorescence intensity map, and uses ImageJ image analysis software to process it to obtain a disease location fluorescence map of the rice leaves; The comparison module scans the reflectance spectrum of the marked rice leaves based on the disease location fluorescence image of the rice leaves, obtains the absorption characteristics of the reflectance spectrum and compares it with the pre-built disease fingerprint library to obtain rice disease diagnosis data; The report generation module uses an intelligent sprayer to automatically dispense pesticides, sprays them onto the disease-infected areas in the disease location fluorescence image of rice leaves, tracks the degradation process of biodegradable optical markers, and outputs a rice disease management report.

[0014] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, any step of the automatic detection method for rice diseases as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the automatic detection method for rice diseases as described in the first aspect of the present invention is implemented.

[0016] The present invention has the following beneficial effects: By mixing a Bacillus subtilis suspension with a chitosan-CQD hydrogel microsphere solution, a GFP- and CQD-dually labeled biodegradable optical marker is prepared. Through signal separation and degradation monitoring, this method addresses the issues of low positioning accuracy and insufficient diagnostic comprehensiveness. The synergistic effect of the dual markers reduces interference between fluorescent signals, improving the accuracy of three-dimensional reconstruction of diseased rice areas while ensuring uniform adhesion of the markers to rice leaves, thus overcoming the low signal resolution limitations of single markers. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] 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 description of 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.

[0018] Figure 1 Flowchart of the automatic rice disease detection method.

[0019] Figure 2 Schematic diagram of the preparation of biodegradable optical markers.

[0020] Figure 3 Schematic diagram of the generation of fluorescence intensity maps for localizing rice leaf diseases.

[0021] Figure 4 This is a flow chart for reflectance spectrum analysis and disease diagnosis. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0023] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0025] Reference Figures 1 to 4 , is an embodiment of the present invention, which provides an automatic detection method for rice diseases, comprising the following steps: S1. Preparation of biodegradable optical markers, using a high-pressure microfluidic injection device to spray the biodegradable optical markers onto rice leaves, forming a marked rice leaf comprising the following steps, S1.1. Insert the green fluorescent protein (GFP) gene into Bacillus subtilis using electroporation. The steps are as follows: First, construct an expression plasmid containing the GFP gene. Use an electroporator to transfer the GFP expression plasmid into the WB800N strain of Bacillus subtilis for electroporation. After electroporation, culture the strain in LB medium containing chloramphenicol under specific environmental conditions, such as 37°C and shaking at 200 rpm, and screen for positive transformants.

[0026] The positive transformants after screening were cultured again. After the culture was completed, the cells of the positive transformants after screening were collected by centrifugation and resuspended in sterile phosphate buffer at a concentration of, for example, 10 9 CFU / mL, and obtain Bacillus subtilis suspension.

[0027] S1.2. Carbon quantum dots (CQDs) were synthesized by hydrothermal carbonization and encapsulated in chitosan to form a chitosan-CQD hydrogel microsphere solution.

[0028] Specifically, citric acid and ethylenediamine were dissolved in deionized water, stirred evenly, and then transferred to a stainless steel reactor lined with polytetrafluoroethylene for chemical reaction. After the reaction was completed, the mixture was cooled to room temperature and filtered through a filter membrane to obtain a carbon quantum dot solution.

[0029] Take part of the carbon quantum dot solution and the chitosan solution, mix them, add a small amount of glutaraldehyde, such as 0.5 mL, as a cross-linking agent, and stir to form a chitosan-CQD hydrogel microsphere solution.

[0030] S1.3. Mixing the Bacillus subtilis suspension and the chitosan-CQD hydrogel microsphere solution to obtain a biodegradable optical marker.

[0031] The specific operation is: take part of the Bacillus subtilis suspension and part of the chitosan-CQD hydrogel microsphere solution in a volume ratio of 1:1, add Tween-80 as a dispersant, stir for 30 to 40 minutes, add carboxymethyl cellulose (CMC) as an adhesive, and stir again for 10 to 15 minutes to form the final biodegradable optical marker.

[0032] S1.4. Use a high-pressure microfluidic injection device to spray the biodegradable optical marker onto the stomatal area on the back of the rice leaves to form labeled rice leaves.

[0033] Specifically, a customized high-pressure microfluidic injection device was selected, equipped with a precision flow controller and a GPS positioning system. The biodegradable optical marker was loaded into the liquid storage tank of the high-pressure microfluidic injection device, and the injection parameters, including spraying pressure and spraying flow, were set. The high-pressure microfluidic injection device moved along the field at a constant speed for spraying, with the spraying focus on the back of the rice leaves.

[0034] S2. Scanning the leaf vascular tissue in the labeled rice leaves, capturing the fluorescence signals of GFP and CQD, generating a three-dimensional fluorescence intensity map, and processing it using ImageJ image analysis software to obtain a disease localization fluorescence map of the rice leaves includes the following steps: S2.1. Select a Thorlabs Bergamo II equipped with a titanium sapphire laser as a two-photon microscope and align the two-photon microscope. The specific alignment process is as follows: Set the parameters of the titanium sapphire laser, including wavelength, pulse width, laser frequency, and repetition rate. The laser frequency is controlled at approximately 20 mW using an adjustable attenuator to avoid photodamage to rice leaf tissue.

[0035] The two-photon microscope was equipped with a water-immersion objective and focal length, covering the vascular tissue of rice leaves, including the xylem and phloem. A Hamamatsu H7422P-40 photomultiplier tube with a bandpass filter was used. The microscope was mounted on a portable track device, with the speed and scanning area set. The track was laid along the rice field and equipped with a servo motor to control the movement speed.

[0036] S2.2. Use the aligned two-photon microscope to scan the GFP and CQD in the biodegradable optical marker to obtain GFP fluorescence signals and CQD fluorescence signals.

[0037] The specific operation is as follows: the laser beam generated by the aligned two-photon microscope is focused on the xylem and phloem of rice leaves through a water immersion objective lens. Due to the photon absorption characteristics of GFP and CQD, GFP fluorescence signals and CQD fluorescence signals are generated under the action of the laser beam. The GFP and CQD fluorescence signals are separated by the optical path unit of the two-photon microscope. The separated GFP fluorescence signals and CQD fluorescence signals are guided to the photomultiplier tube through a bandpass filter. The photomultiplier tube records the fluorescence intensity of the separated GFP fluorescence signals and CQD fluorescence signals in real time at a sampling rate of about 1 MHz, in the range of 10 3 ~10 6 RFU. Each layer scan generates a two-dimensional image of 1000×1000 pixels, where the pixel value corresponds to the fluorescence intensity and is stored in grayscale format. The GFP fluorescence signal image and the CQD fluorescence signal image are recorded as independent channels, and the separation of each channel is achieved by switching the bandpass filter to avoid signal crosstalk. During the scanning process, the two-photon microscope will display the distribution of fluorescence intensity in real time. The high-intensity area, which corresponds to the diseased area, is observed on the screen. A single rice leaf generates 75MB of fluorescence image data, i.e., 75 layers × 1000 × 1000 pixels, forming a rice leaf fluorescence image dataset.

[0038] The rice leaf fluorescence image dataset is pseudo-colored and obtained as a coded fluorescence image dataset.

[0039] Using a two-photon microscope, the grayscale images of the GFP fluorescence signal and the CQD fluorescence signal are assigned pseudo-color images, for example, a green tone is assigned to the GFP channel and a blue tone is assigned to the CQD channel.

[0040] A lookup table (LUT) was used to map the grayscale images of the GFP fluorescence signal and the CQD fluorescence signal into the RGB color space.

[0041] The specific steps are as follows: Use ThorImage software to read grayscale images of the GFP and CQD channels, set a green LUT and a blue LUT, use the green LUT to map the grayscale image of the GFP channel to green brightness, and use the blue LUT to map the grayscale image of the CQD channel to blue brightness, forming single-channel color images: green and blue images. Then, use the MergeChannels function in ThorImage software to overlay the green and blue images to form a composite pseudo-color image, which is saved in TIFF format to obtain the encoded fluorescence image dataset.

[0042] The purpose of pseudo-color coding is to visually distinguish the spatial distribution of GFP fluorescence signals and CQD fluorescence signals and highlight the fluorescence characteristics of the diseased area, such as the signal overlap in the bacterial blight area.

[0043] S2.3. Use ImageJ software to stack the encoded fluorescence image dataset into a three-dimensional dataset.

[0044] Specifically, ImageJ software was launched, the Bio-Formats plug-in was loaded, and a coded fluorescence image dataset was selected. The Bio-Formats plug-in read all images from the coded fluorescence image dataset and sorted them according to the scanning order. In the Bio-Formats interface, the Group files with similar names and Openashyperstack options were checked. ImageJ software automatically stacked all images into a multichannel stack, which was displayed as a Hyperstack in the ImageJ window. To verify the correctness of the stack, the three-layer cross-section of the multichannel stack was checked using the Orthogonal Views function of ImageJ software to ensure that the alignment error between layers was less than 1 μm. The three-dimensional dataset consists of voxels, each of which contains the grayscale values of the GFP channel and the CQD channel.

[0045] The 3D dataset was rendered using the 3DViewer plug-in in ImageJ software.

[0046] The specific operation is as follows: In ImageJ software, open the three-dimensional dataset and start the 3DViewer plug-in through the Plugins>3DViewer function. After startup, the 3DViewer plug-in interface will display a three-dimensional rendering window. Select the Volume rendering mode in the three-dimensional rendering window and set the rendering parameters, that is, assign a green tone to the GFP channel and a blue tone to the CQD channel, and set the transparency to about 50%. Use the 3DViewer plug-in to map each voxel in the three-dimensional dataset to the color and brightness corresponding to the fluorescence intensity to generate a preliminary three-dimensional fluorescence intensity map.

[0047] Specifically, the grayscale value of each voxel is read one by one, and the voxels are projected along the line of sight, i.e., the z-axis, using a stereoscopic projection algorithm. The color, brightness, and transparency are superimposed one by one according to the set rendering parameters. The mapping process is displayed in real time in the 3DViewer window of ImageJ. The mapped voxels form a preliminary three-dimensional fluorescence intensity map, which is stored in OBJ format.

[0048] S2.4. Use ImageJ software to mark the preliminary three-dimensional fluorescence intensity image, find the diseased area of the rice leaves, and output the disease location fluorescence image of the rice leaves.

[0049] The specific steps are as follows: In ImageJ software, load the three-dimensional fluorescence intensity map and set the fluorescence intensity threshold, which is set according to the maximum value of the fluorescence intensity of the GFP fluorescence signal and the CQD fluorescence signal in the GFP and CQD channels, that is, the maximum fluorescence intensity. Use the Measure function in ImageJ software to count the fluorescence intensity of the GFP fluorescence signal and the CQD fluorescence signal in the three-dimensional data set. For example, the maximum intensity in the GFP channel is 60,000 grayscale values, and the CQD channel is 55,000 grayscale values. When the fluorescence intensity of the GFP fluorescence signal and the CQD fluorescence signal is greater than the fluorescence intensity threshold, the corresponding area of the rice leaf is marked as a diseased area, such as xylem blockage of bacterial blight, rice blast infection site, etc. Marking is achieved through the Analyze plug-in in ImageJ software. The Analyze plug-in automatically identifies fluorescence intensities above the fluorescence intensity threshold and generates the outline and intensity level of the diseased area. For example, red indicates a fluorescence intensity >90%. After marking, the volume and centroid coordinates of the voxels in the diseased area were counted using the 3DObjectCounter function and stored as a CSV file, including the diseased area ID, X coordinate, Y coordinate, Z depth, and voxel volume, to form a disease localization fluorescence map of the rice leaf.

[0050] S3. Based on the disease location fluorescence map of rice leaves, the reflectance spectrum of the marked rice leaves is scanned to obtain the absorption characteristics of the reflectance spectrum and compared with the pre-built disease fingerprint library to obtain rice disease diagnostic data. The process includes the following steps: S3.1. Configure an infrared handheld detection device with a diode laser and an InGaAs photodetector.

[0051] Specifically, the infrared handheld detection device is connected to a lithium battery and powered on to warm up for several minutes to stabilize the diode laser output. After warming up, the diode laser parameters—namely, laser wavelength, output power, and output spot diameter—need to be calibrated, and focused using the built-in collimating lens. The parameters of the InGaAs photodetector also require calibration, including the detection band, sampling rate, and cross-group gain.

[0052] The infrared handheld detection device is equipped with a touchscreen that displays the spectral curve and operation interface in real time. A fluorescence image of the rice leaf disease location is loaded through the touchscreen. The centroid coordinates from this image are then imported into the infrared handheld detection device's GPS navigation function to guide the scanning of the targeted diseased area. A reflectance standard plate is used as a calibration benchmark for reflected light intensity, and a reference reflected light intensity is measured. A reflectance standard plate is a highly reflective diffuse reflective material. Because its surface uniformly reflects near-infrared light, it is used as a calibration reference for spectrometers or detectors to provide known reflectance characteristics.

[0053] S3.2. Based on the disease location fluorescence image of the rice leaves, use the configured infrared handheld detection device to scan the reflection spectrum of the rice leaf surface and output the absorption characteristics of the reflection spectrum.

[0054] The reflection spectrum of the rice leaf surface refers to the continuous distribution curve of the intensity of the reflected light generated by the rice leaf surface in the near-infrared band in response to the incident laser as the wavelength changes. The reflection spectrum reflects the optical properties of the wax layer on the surface of the rice leaf. For healthy rice leaves, the reflection spectrum will appear smooth because the wax layer is intact and has weak absorption. For diseased rice leaves, the reflection spectrum will show absorption depressions because the pathogens will destroy the cell structure and wax layer of the rice leaves.

[0055] The specific scanning process is as follows: using a configured infrared handheld detection device to move along the veins of rice leaves (main veins or primary lateral veins) at a constant speed, scanning the diseased areas in the disease location fluorescence image of the rice leaves one by one, irradiating the rice leaves with a laser beam generated by a diode laser, and an InGaAs photodetector capturing the reflected light intensity of the rice leaves, generating multiple reflected light intensity data points.

[0056] Calculate the reflectivity of the reflected light intensity data point. The expression is: ; in, Represents the reflectivity of the reflected light intensity data point, represents the intensity of reflected light from rice leaves, Indicates the reference reflected light intensity of the reflectivity standard plate. For example, assuming the reflected light intensity of a rice leaf is measured to be 8.5×10 5 μW / cm², while the reference reflected light intensity of the reflectivity standard plate is 10 6 μW / cm², the reflectivity of the reflected light intensity data point is calculated using this formula, resulting in a reflectivity of 85%. The reflectivity of healthy rice leaves is usually greater than 85%, while the reflectivity of diseased rice leaves is usually less than <65%.

[0057] The reflectance of all reflected light intensity data points is saved in TXT format to form a reflectance spectrum dataset of rice leaves, including wavelength points, reflection intensity and reflectance.

[0058] S3.3. Identify and extract absorption features of the reflectance spectrum of the rice leaf reflectance spectrum dataset.

[0059] The absorption characteristics of the reflectance spectrum refer to the reduced reflectivity and concave pattern in the spectral curve caused by rice diseases in specific spectral bands, such as 1050-1100nm. Specific absorption characteristics include rice blast, sheath blight, and bacterial blight. Rice blast presents a double-valley curve with two absorption peaks, sheath blight presents a single broad peak, and bacterial blight presents a single narrow peak. These three absorption characteristics reflect tissue changes caused by rice diseases. For example, sheath blight causes rice tissue necrosis, forming a broad peak. Each absorption characteristic includes wavelength, peak width, and absorption depth.

[0060] The reflectance spectrum dataset of rice leaves was imported into an ARM Cortex-M4 microcontroller, and Savitzky-Golay filtering was applied to the reflectance spectrum dataset of rice leaves using SpectralWorks software to remove noise.

[0061] Specifically, the filtering parameters, namely the window width and fitting polynomial, are set. For each wavelength point, SpectralWorks selects multiple preceding and succeeding wavelengths to construct a sliding window. This sliding window moves along the spectral band, repeatedly fitting the polynomial at each wavelength point to generate a smoothed spectrum that preserves the shape of the absorption dip in the spectral curve. For wavelength points at the edge, such as 1052.5nm, an asymmetric window is used, namely multiple preceding or multiple succeeding wavelength points, to output a smoothed reflectance spectral dataset.

[0062] SpectralWorks software was used to identify absorption features in the smoothed reflectance spectral dataset. An absorption dip threshold was set based on the reflected light intensity and signal-to-noise ratio measured in the field experiment. When the reflected light intensity in the smoothed reflectance spectral dataset fell below the absorption dip threshold, the reflected light intensity was marked as an absorption feature. SpectralWorks software also recorded the wavelength position, wavelength depth, and wavelength width of the absorption feature.

[0063] The absorption features are normalized to obtain the absorption spectrum feature data set.

[0064] Specifically, SpectralWorks software was used to find the highest and lowest reflected light intensities in the spectral band, and the difference between the highest and lowest reflected light intensities was calculated. The reflected light intensity of each absorption feature was subtracted from the lowest reflected light intensity and divided by the difference between the highest and lowest reflected light intensities to obtain the normalized reflected light intensity and store it in JSON format to form an absorption spectrum feature dataset.

[0065] S3.4. Construct a disease fingerprint library. This database stores the absorption characteristics of common rice diseases. The specific construction process is as follows: 1,000 sets of diseased leaf samples were collected from rice-growing areas, covering different rice varieties, such as indica and japonica, and at different infection stages of rice diseases. An Ocean Optics NIRQuest near-infrared spectrometer was used to scan the diseased leaf samples and record their reflectance spectra. This process was repeated multiple times for each sample group, and the average value was taken. The reflectance spectra of the diseased samples were subjected to Savitzky-Golay filtering, similarly identifying absorption characteristics. The absorption characteristics were organized into a JSON file containing disease type, reflected light intensity, spectral curve, peak position, and peak width, forming the disease fingerprint library.

[0066] S3.5. Compare the absorption spectrum feature dataset with the disease fingerprint library, use the ARMCortex-M4 microcontroller to run the cosine similarity algorithm, calculate the cosine similarity, and generate rice disease diagnosis data.

[0067] The normalized reflected light intensity in the absorption spectrum feature dataset is compared with the reflected light intensity in the disease fingerprint library one by one, and the cosine similarity can be calculated using the following expression: ; in, Indicates cosine similarity. The value range of cosine similarity is 0~1, 1 means a perfect match. represents the normalized reflected light intensity in the absorption spectrum feature dataset, Represents the reflected light intensity in the disease fingerprint library; The similarity threshold is set based on the field variability and ROC curve. Field variability refers to the fact that rice leaves are affected by the variety, growth stage, and environment, resulting in slight changes in the spectral curve. For example, setting the similarity threshold to 0.85 can tolerate ±5% variation and ensure field applicability.

[0068] SpectralWorks software was used to identify the highest cosine similarity. When the highest cosine similarity exceeded the similarity threshold, the rice disease diagnosis result was obtained. This was the disease type in the disease fingerprint library corresponding to the highest cosine similarity output by the cosine similarity algorithm. For example, if the cosine similarity between the normalized reflected light intensity of the absorption spectral feature dataset and the reflected light intensity of the disease fingerprint library was 0.95, and higher than other types, such as sheath blight (0.6), the diagnosis result was rice blast.

[0069] The severity of rice diseases is assessed based on the reflectance in the reflectance spectrum dataset of rice leaves, which is divided into severe infection, moderate infection, and mild infection.

[0070] Specifically, based on multiple groups of disease samples collected previously, the reflectivity was measured using a handheld detection device, and the necrotic area on the rice leaf tissue was observed under a microscope to obtain the measurement results. For example, a certain disease caused severe damage to the wax layer of the rice leaves and the collapse of the cell walls. The average reflectivity was between 60%±5%, and the corresponding tissue necrotic area was greater than 50%. At this time, the disease can be classified as severe infection. The same applies to moderate and mild infections. For example, in the early stage of bacterial blight, the necrotic area of the leaf tissue was less than 20%, and the reflectivity was higher than 75%, indicating that only the wax layer of the rice leaves was slightly damaged, and the reflectivity was close to that of healthy leaves. The bacterial blight can be classified as a mild infection.

[0071] The rice disease diagnosis results, the severity of rice diseases, and the centroid coordinates in the disease location fluorescence image of rice leaves are integrated into a CSV file to form rice disease diagnosis data and transmitted to the smart sprayer via LoRa wireless communication.

[0072] S4. Using an intelligent sprayer to automatically dispense the drug, spraying the disease-infected area on the rice leaves as indicated by the disease location fluorescence image, tracking the degradation process of the biodegradable optical marker, and outputting a rice disease management report includes the following steps: S4.1. Calculate the required pesticide concentration for the smart sprayer. Specifically, the smart sprayer controller receives rice disease diagnostic data via a LoRa receiver. After receiving the data, it sets a pesticide formulation based on the diagnostic results and severity of the rice disease, as well as the standards of the Ministry of Agriculture and Rural Affairs. For example, for severe rice blast, use 0.3g / L tricyclazole; for moderate sheath blight, use 0.5g / L hexaconazole; and for mild bacterial blight, use 0.2g / L thiophanate-methyl. The mass of the pesticide to be dispensed is calculated based on the formulation parameters. For example, assuming a spraying volume of 2 liters and rice blast requires 0.3g / L tricyclazole, the mass of the pesticide to be dispensed is 0.3 × 2 = 0.6g.

[0073] According to the calculated drug mass, the drug concentration is calculated as follows: ; in, Indicates the drug concentration, Indicates the quality of the dispensed medicine, Indicates the volume of water; A hierarchical clustering algorithm is used to group the centroid coordinates in the rice disease diagnosis data and generate the minimum covering rectangle. Specifically: Calculate the Euclidean distance between each pair of centroid coordinates, the expression is: ; in, represents the Euclidean distance between each pair of centroid coordinates, Indicates the first The coordinates of the diseased points, Indicates the first The coordinates of the diseased points, Indicates the first The coordinates of the diseased points, Indicates the first The coordinates of the diseased point.

[0074] When the Euclidean distance between each pair of centroid coordinates is less than a fixed value, such as 2 cm, the pair of centroid coordinates are clustered into the same group.

[0075] Using the S-shaped path algorithm, a spraying path is generated within the minimum coverage rectangle to cover the disease-infected area.

[0076] Specifically, the spray path parameters, including spray width and step length, are set. The length of the spray path is calculated by dividing the area of the minimum coverage rectangle by the spray width, and the length of the spray path is used as the preliminary spray path.

[0077] The preliminary spraying path is optimized to generate the final spraying path.

[0078] Specifically, the smart sprayer's position is corrected using RTK-GPS navigation, ensuring that the starting point of the preliminary spraying path aligns with the lower left corner of the minimum coverage rectangle. The smart sprayer's controller compares reflectivity with rice disease diagnostic data to identify healthy rice leaf areas and remove non-disease-affected areas, typically with a reflectivity greater than 75%. For example, if the reflectivity within the minimum coverage rectangle is 80% within a 12-12.5 cm segment, the preliminary spraying path skips that segment. Cubic spline interpolation is applied to each turning point (approximately every 5 cm) to generate a smooth trajectory and reduce jitter in the preliminary spraying path.

[0079] S4.2. Use the smart sprayer to spray pesticides on the disease-infected area and obtain the pesticide application data.

[0080] Specifically, place the smart sprayer in the field, and use the touch screen to confirm the final spraying path and the status of the pesticide. At the same time, the controller activates the nozzle and preheats it for about 20 to 30 seconds.

[0081] After the spraying preparations are completed, the smart sprayer is moved at a constant speed along the final spraying path. GPS navigation tracks the coordinates of the center of mass. The nozzle dynamically adjusts the solenoid valve switch based on the coordinates to spray only the diseased areas of the rice. During the spraying process, the spraying amount per acre of field is recorded and the real-time spraying amount is controlled. The expression for the real-time spraying amount is: ; in, Indicates the real-time spraying amount. represents the area of the diseased area, Indicates the spraying amount per acre of field, Indicates the total area of the field.

[0082] After spraying is completed, the controller of the smart sprayer records the application data, including rice leaf ID, disease type, pesticide type, dosage concentration, center of mass coordinates, and application amount.

[0083] S4.3. Monitoring the biodegradation process of the biodegradable optical marker through a soil sensor network to obtain CQD fluorescence data includes the following steps: Since the chitosan-CQD hydrogel microsphere solution in the biodegradable optical marker is hydrolyzed during its half-life due to the action of pesticide irrigation or rainwater after spraying pesticides, releasing CQDs, the residual concentration of chitosan in the hydrogel microspheres is calculated as follows: ; in, represents the residual concentration of chitosan in the hydrogel microspheres, represents the initial chitosan concentration, which is measured by the previously prepared chitosan-CQD hydrogel microsphere solution. It represents the multiple of half-life, which is defined as the hydrolysis time of chitosan from the start of pesticide spraying to the present moment, divided by the half-life.

[0084] It is further explained that this formula is used to estimate when chitosan has degraded to a safe level and to adjust the subsequent sensor sampling frequency according to the remaining concentration, thereby reducing blind monitoring.

[0085] A soil sensor network was selected as the monitoring device and deployed in the field, covering the disease-infected areas of rice leaves. The nodes of the soil sensor network measured the CQD fluorescence intensity in the runoff water four times a day and converted it into CQD concentration using the following conversion formula: ; in, represents the concentration of CQDs, represents the fluorescence intensity of CQDs in runoff water, The calibration slope represents the relationship between the fluorescence intensity of CQD in runoff water and the concentration of CQD. It is used to reflect the linear relationship between the fluorescence intensity of CQD in runoff water and the concentration of CQD. It is obtained by using a CQD solution of known concentration to measure the fluorescence intensity with a YSIEXO2 fluorescence probe and linearly fitting the fluorescence intensity with the laboratory standard curve (R²=0.99). The intercept of the linear relationship between the fluorescence intensity of CQDs in runoff water and the concentration of CQDs is determined by laboratory measurements of pure water without CQDs. is 0.

[0086] When the CQD concentration drops to a stable value, such as below 0.01 μg / L, it indicates 100% degradation. The CQD concentration is stored in CSV format to obtain CQD fluorescence data.

[0087] S4.4. Integrate the CQD fluorescence data and pesticide application data and upload them to the cloud database using the 4G network to generate a rice disease management report. The report includes the pesticide application record, namely the leaf ID, disease type, pesticide type, dosage concentration, dosage, degradation progress, and verification results. For example, 20 days after application, the CQD concentration is <0.01μg / L, indicating 100% clearance.

[0088] This embodiment also provides an automatic detection system for rice diseases, comprising: a preparation module for preparing a biodegradable optical marker, and spraying the biodegradable optical marker onto rice leaves using a high-pressure microfluidic injection device to form marked rice leaves; The scanning module scans the leaf vascular tissue in the marked rice leaves, captures the GFP fluorescence signal and the CQD fluorescence signal, generates a three-dimensional fluorescence intensity map, and uses ImageJ image analysis software to process it to obtain a disease location fluorescence map of the rice leaves; The comparison module scans the reflectance spectrum of the marked rice leaves based on the disease location fluorescence image of the rice leaves, obtains the absorption characteristics of the reflectance spectrum and compares it with the pre-built disease fingerprint library to obtain rice disease diagnosis data; The report generation module uses an intelligent sprayer to automatically dispense pesticides, sprays them onto the disease-infected areas in the disease location fluorescence image of rice leaves, tracks the degradation process of biodegradable optical markers, and outputs a rice disease management report.

[0089] This embodiment also provides a computer device suitable for the automatic detection method of rice diseases, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the automatic detection method of rice diseases proposed in the above embodiment.

[0090] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0091] This embodiment also provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the program implements the method for automatically detecting rice diseases as proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0092] In summary, the present invention prepares a biodegradable optical marker dually labeled with GFP and CQDs by mixing a Bacillus subtilis suspension with a chitosan-CQD hydrogel microsphere solution. Through signal separation and degradation monitoring, this approach addresses the issues of low positioning accuracy and insufficient diagnostic comprehensiveness. The synergistic effect of the dual markers reduces interference between fluorescent signals, improving the accuracy of 3D reconstruction of diseased rice areas while ensuring uniform adhesion of the markers to rice leaves, thus overcoming the low signal resolution limitations of single markers.

[0093] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for automatically detecting rice diseases, characterized in that: include, preparing a biodegradable optical marker, and spraying the biodegradable optical marker onto rice leaves using a high-pressure microfluidic injection device to form marked rice leaves; The leaf vascular tissue in the labeled rice leaves was scanned to capture GFP fluorescence signals and CQD fluorescence signals, and a three-dimensional fluorescence intensity map was generated. The image was then processed using ImageJ image analysis software to obtain a fluorescence map of the disease location on the rice leaves. Based on the disease location fluorescence image of rice leaves, the reflectance spectrum of the marked rice leaves is scanned to obtain the absorption characteristics of the reflectance spectrum and compared with the pre-built disease fingerprint library to obtain rice disease diagnosis data; Use an intelligent sprayer to automatically dispense medicine and spray it onto the disease-infected area in the disease location fluorescence image of rice leaves. Track the degradation process of the biodegradable optical marker and output a rice disease management report.

2. The automatic detection method for rice diseases according to claim 1, wherein: The preparation of the biodegradable optical marker comprises inserting GFP into Bacillus subtilis by electroporation, screening positive transformants, and culturing the screened positive transformants to obtain a Bacillus subtilis suspension; CQDs were synthesized by hydrothermal carbonization and encapsulated in chitosan to form a chitosan-CQD hydrogel microsphere solution; The Bacillus subtilis suspension and the chitosan-CQD hydrogel microsphere solution are mixed to obtain a biodegradable optical marker.

3. The automatic detection method for rice diseases according to claim 2, wherein: The method of forming the marked rice leaves comprises configuring a high-pressure microfluidic injection device, setting injection parameters, and spraying the biodegradable optical marker into the stomatal area on the back of the rice leaves.

4. The automatic detection method for rice diseases according to claim 3, wherein: The capturing of GFP fluorescence signals and CQD fluorescence signals includes selecting a Thorlabs Bergamo II as a two-photon microscope and performing alignment, and focusing the laser beam generated by the aligned two-photon microscope on the marked rice leaves.

5. The automatic detection method for rice diseases according to claim 4, characterized in that: Obtaining the disease location fluorescence image of the rice leaf includes separating the GFP fluorescence signal and the CQD fluorescence signal using the optical path unit of the registered two-photon microscope, and recording the fluorescence intensities of the separated GFP fluorescence signal and the CQD fluorescence signal respectively to obtain a rice leaf fluorescence image dataset; The rice leaf fluorescence image dataset was pseudo-colored to obtain a coded fluorescence image dataset, and the coded fluorescence image dataset was multi-channel stacked using the Bio-Formats plug-in to form a three-dimensional dataset. The fluorescence intensity threshold was set, and the fluorescence intensity of the GFP fluorescence signal and the CQD fluorescence signal in the three-dimensional data set was counted using ImageJ software and compared with the fluorescence intensity threshold to determine the diseased area of the rice. The volume and centroid coordinates of the voxels in the rice disease area were counted to obtain the disease location fluorescence map of the rice leaf.

6. The automatic detection method for rice diseases according to claim 5, characterized in that: The obtaining of rice disease diagnosis data includes configuring an infrared handheld detection device, using the infrared handheld detection device to scan the rice disease area in the disease location fluorescence image of the rice leaves, capturing the reflected light intensity of the marked rice leaves, and generating reflected light intensity data points; Calculate the reflectance of the reflected light intensity data points to obtain the reflectance spectrum data set of the rice leaves; An ARM Cortex-M4 microcontroller was used to apply Savitzky-Golay filtering to the reflectance spectrum dataset of rice leaves to obtain a smoothed reflectance spectrum dataset. An absorption dip threshold is set and compared with the reflected light intensity in the smoothed reflectance spectrum dataset to obtain the absorption characteristics of the reflectance spectrum and perform normalization to generate an absorption spectrum feature dataset; Rice leaves with diseased samples were collected to construct a disease fingerprint library. The cosine similarity between the reflected light intensity in the disease fingerprint library and the normalized reflected light intensity in the absorption spectrum feature dataset was calculated to obtain the rice disease diagnosis results and evaluate the severity of the rice disease. The rice disease diagnosis results, the severity of rice diseases and the centroid coordinates were integrated into a CSV file to obtain the rice disease diagnosis data.

7. The automatic detection method for rice diseases according to claim 6, characterized in that: Outputting the rice disease management report includes configuring a hierarchical clustering algorithm to group centroid coordinates in the rice disease diagnosis data to generate a minimum covering rectangle; The S-shaped path algorithm is used to generate a preliminary spraying path in the minimum coverage rectangle, and the preliminary spraying path is optimized using RTK-GPS navigation to generate the final spraying path. Use a smart sprayer to spray pesticides along the final spray path to the disease-infected area to obtain pesticide application data, and use a soil sensor network to monitor the degradation process of biodegradable optical markers to obtain CQD fluorescence data; Integrate CQD fluorescence data and pesticide application data to generate rice disease management reports.

8. An automatic detection system for rice diseases, based on the automatic detection method for rice diseases according to any one of claims 1 to 7, characterized in that: include, A preparation module is used to prepare a biodegradable optical marker, and a high-pressure microfluidic injection device is used to spray the biodegradable optical marker onto rice leaves to form marked rice leaves; The scanning module scans the leaf vascular tissue in the marked rice leaves, captures the GFP fluorescence signal and the CQD fluorescence signal, generates a three-dimensional fluorescence intensity map, and uses ImageJ image analysis software to process it to obtain a disease location fluorescence map of the rice leaves; The comparison module scans the reflectance spectrum of the marked rice leaves based on the disease location fluorescence image of the rice leaves, obtains the absorption characteristics of the reflectance spectrum and compares it with the pre-built disease fingerprint library to obtain rice disease diagnosis data; The report generation module uses an intelligent sprayer to automatically dispense pesticides, sprays them onto the disease-infected areas in the disease location fluorescence image of rice leaves, tracks the degradation process of biodegradable optical markers, and outputs a rice disease management report.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the automatic detection method for rice diseases according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the automatic detection method for rice diseases according to any one of claims 1 to 7 are implemented.

Images