Rice seedling blast disease intelligent analysis system based on remote sensing data analysis

Abstract

The invention discloses a rice seedling blast disease intelligent analysis system based on remote sensing data analysis, and relates to the technical field of agricultural remote sensing and plant disease monitoring, the system comprises a physiological data acquisition module, a physiological feature extraction module and a disease severity calculation and severity grading module; wherein the physiological data acquisition module realizes in-situ non-invasive acquisition of the spectral reflectivity of rice seedling leaves, the maximum photochemical efficiency of a light system II and the water transport rate of stalks through various sensors such as a hyperspectral sensor and a chlorophyll luminoscope; the physiological feature extraction module fuses multiple parameters to calculate a leaf surface structure health index and a vascular function activity index; the disease grading module obtains an in-vivo disease load index through weighting, and outputs different disease grades in combination with a K-means clustering threshold value; the system solves the problems that traditional rice seedling blast diagnosis is high in subjectivity and hysteresis quality, and early-stage accurate recognition of diseases is achieved.

Description

Technical Field

[0001] This invention belongs to the field of agricultural remote sensing and plant disease monitoring technology, specifically relating to an intelligent analysis system for rice seedling blast disease based on remote sensing data analysis. Background Technology

[0002] Rice is a major food crop in my country. Seedling blast, a common fungal disease in rice seedlings, can cause seedlings to wither and die, and is a key obstacle to the safe production of rice.

[0003] Currently, the diagnosis of rice seedling blast disease mainly relies on manual visual inspection and traditional laboratory testing methods, both of which have significant technical shortcomings: Relying on agricultural technicians to diagnose diseases based on superficial characteristics such as leaf yellowing and lesion morphology is highly dependent on subjective experience and can only identify diseases in the later stages of their development, missing the optimal window for control. Furthermore, manual inspections require observation of each plant individually, which is extremely inefficient in large-scale rice fields, and each inspection is too time-consuming to meet real-time monitoring needs. Therefore, pathogen isolation and culture, along with PCR... While molecular identification and other methods for diagnosing diseases are accurate, they suffer from destructive and delayed results. They require collecting rice seedling samples and bringing them back to the laboratory, leading to plant death and hindering continuous in-situ monitoring. Furthermore, the testing process is lengthy, and by the time results are available, the disease may have already spread in the field, rendering control ineffective. Although recent research on crop disease monitoring based on remote sensing data has been conducted, it has largely focused on crops like wheat and corn. Technical solutions for rice seedlings have significant shortcomings: first, they rely solely on single spectral parameters, failing to consider the thin leaves and fragile vascular system of rice seedlings, thus failing to capture the early, latent damage to the photosynthetic and water transport systems caused by diseases; second, they do not consider the impact of environmental interference on the data, resulting in excessive monitoring errors; and third, they lack standardized disease grading models, failing to directly output mild, moderate, or severe grading results that can guide production, thus lacking practicality. In summary, current diagnostic technologies for rice seedling blast disease cannot meet the needs of early, in-situ, precise, and efficient production. There is an urgent need to build an intelligent analysis system that combines multi-source remote sensing data, adapts to the physiological characteristics of rice seedlings, and can directly output grading results. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art; to this end, this invention proposes a blueberry ripeness detection method and system based on computer vision, to solve the following technical problem: First, relying solely on a single spectral parameter without considering the physiological characteristics of thin leaves and fragile vascular bundles in rice seedlings makes it impossible to capture the latent damage to the photosynthetic and water transport systems caused by diseases in the early stages. Second, the impact of environmental interference on the data is not considered, resulting in excessive monitoring errors. Third, the lack of standardized disease grading models makes it impossible to directly output mild, moderate, or severe grading results that can guide production, thus lacking practicality.

[0005] To address the above problems, this invention provides an intelligent analysis system for rice seedling blast disease based on remote sensing data analysis, comprising the following modules: Physiological data acquisition module: used to acquire in situ physiological parameters of individual rice seedlings; the parameters include leaf spectral reflectance data acquired by a hyperspectral sensor, maximum photochemical efficiency data of photosystem II acquired by a chlorophyll fluorometer, and stem water transport rate data acquired by a plant stem micro-change sensor. Physiological feature extraction module: Based on the in-situ physiological parameters, it obtains the leaf structure health index by fusing the red edge position offset, leaf reflectance valley depth and leaf spectral slope, and obtains the vascular function activity index by fusing the maximum photochemical efficiency of the fusion photosystem II, stem water transport rate, environmental dynamic correction parameters, vascular function temporal change rate and vascular blockage early warning factor. Disease severity calculation and severity classification module: It is used to add the leaf structure health index and vascular function activity index with weights to obtain the in vivo disease load index of the degree of disease infection in rice seedlings. Based on the numerical range of the in vivo disease load index, it maps and outputs the disease severity level of individual rice seedlings.

[0006] Preferably, the physiological data acquisition module includes: In-situ, non-destructive parameter collection was performed on individual rice seedlings, specifically as follows: Leaf spectral reflectance data in the wavelength range of 350-1050nm were collected using a hyperspectral sensor with a wavelength interval of 2nm. The ambient light intensity and leaf temperature were recorded simultaneously during the data collection. The maximum photochemical efficiency data of photosystem II was collected using a chlorophyll fluorometer. A 30-minute dark adaptation treatment was set before the data collection. Each data collection was repeated 3 times and the average value was taken. The stem diameter micro-change data were collected by a plant stem micro-change sensor at a sampling frequency of 10Hz, and the stem water transport rate was calculated based on the data.

[0007] Preferably, in the physiological data acquisition module, when the hyperspectral sensor acquires leaf spectral reflectance data, it simultaneously acquires the data required for calculating the infrared position offset, including: The hyperspectral sensor simultaneously collects bidirectional reflectance data of rice seedling leaves, including spectral reflectance at three observation angles: 0°, 30°, and 45°. At the same time, the vertical distance between the sensor and the leaf surface is kept constant at 30 cm. The red edge position offset was calculated using a three-segment interpolation method. Specifically, spectral reflectance data in the 680-760nm band was extracted with a sampling interval of 1nm, and the band was divided into three sub-bands: 680-709nm, 710-730nm, and 731-760nm. The maximum and minimum reflectance values ​​for each sub-band are calculated to determine the red-edge inflection point for each sub-band. A line connecting the inflection points of the three bands is fitted using linear interpolation, and the wavelength value corresponding to the inflection point where the slope of the line changes from positive to negative is taken as the initial red-edge position offset. An angle correction coefficient is introduced to correct the initial red-edge position offset, specifically: in, This is the corrected offset of the red edge position. This is the initial red edge position offset. For the angle of observation, Angle correction coefficient.

[0008] Preferably, the depth of the leaf reflectivity valley includes: The characteristic wavelengths of red valley and blue valley in the spectrum of rice seedling leaves were determined, with the red valley characteristic wavelength being 650-680nm and the blue valley characteristic wavelength being 450-480nm. The spectral sampling interval between the two bands was reduced from the basic 2nm to 1nm; The leaf reflectance valley depth is obtained by merging the reflectance valley depths of red and blue valleys using a weighted average method, as follows: Wherein, RGD is the depth of the leaf reflectance valley. This represents the maximum reflectivity within the characteristic band of the Red Valley. This represents the minimum reflectivity within the characteristic band of the Red Valley. This represents the maximum reflectivity within the characteristic band of the Blue Valley. This represents the minimum reflectivity within the characteristic band of the Blue Valley.

[0009] Preferably, the spectral slope of the leaf includes: The corresponding calculation bands were selected for different leaf ages of rice seedlings, specifically: 700-720nm band for 1-3 leaf age and 740-760nm band for 4-6 leaf age. The selected calculation band was divided into 11 data points according to wavelength intervals. When collecting reflectance data, the sensor integration time was adjusted according to the band. The 11 data points were divided into three feature intervals, A, B and C. Linear regression was performed on each of the three feature intervals to obtain the regression slope. Differential weights are assigned based on the biological significance of the three feature intervals, and the initial leaf spectral slope is obtained by weighting and adding them to the regression slope of each interval. Differential weighting coefficients were assigned to the 11 data points, and the weighted average wavelength and band arithmetic mean wavelength of the 11 data points were calculated. The initial leaf spectral slope was then corrected to obtain the corrected leaf spectral slope, specifically as follows: in, This is the corrected leaf spectral slope. The initial leaf spectral slope, For the weighted average wavelength, The arithmetic mean wavelength of the band; Three sets of 11 data points were collected continuously from the same leaf. The coefficient of variation of the leaf spectral slope of the three sets was calculated. Finally, the average value of the three sets of data with the coefficient of variation within the preset threshold range was taken as the final leaf spectral slope.

[0010] Preferably, the leaf structure health index includes: The leaf structure health index is calculated by fusing multiple parameters, including the red edge position offset, leaf reflectance valley depth, and leaf spectral slope, from the hyperspectral data. Specifically: Where A1 is the leaf structure health index, REP is the red edge position offset, RGD is the leaf reflectance valley depth, and RSS is the leaf spectral slope.

[0011] Preferably, the vascular function activity index includes: The vascular function activity index is calculated by coupling the maximum photochemical efficiency of photosystem II, stem water transport rate, environmental dynamic correction parameters, temporal change rate of vascular function, and vascular blockage early warning factor. Specifically: Where A2 is the vascular function activity index, α is the environmental dynamic correction coefficient, and β1 and β2 are dynamic weighting coefficients. For the variable fluorescence of optical system II, The maximum fluorescence of optical system II, The stem water transport rate. This is the varietal benchmark value for stem water transport rate. This is a reference value for the upper limit of varieties. The rate of change of vascular function over time. As an early warning factor for vascular blockage; The environmental dynamic correction coefficient is determined by coupling calculation of multiple environmental factor correction coefficients, which include temperature correction coefficient, humidity correction coefficient and illumination correction coefficient.

[0012] Preferably, the time-series rate of change of vascular function includes: The vascular function temporal change rate is combined with the dynamic adjustment cycle of leaf age stage. The dynamic adjustment cycle includes: the collection cycle of rice seedlings with 1-3 leaves is set to 12 hours / time; the collection cycle of rice seedlings with 4-6 leaves is set to 24 hours / time. According to the determined collection cycle, extract two types of core parameters from the system's parameter database: the current cycle and the parameters from the three consecutive cycles. The two core parameters are the maximum photochemical efficiency of the photosystem II after light adaptation correction and the standardized stem water transport rate. By weighting and adding the two types of core parameters for different periods, the comprehensive value of vascular function in the current period and the comprehensive value of vascular function three periods ago are obtained respectively. The difference between the two is divided by 3 to obtain the time series change rate of vascular function.

[0013] Preferably, the vascular blockage early warning factor includes: The data collection cycle for vascular blockage early warning factors should be consistent with the data collection cycle for the time-series change rate of vascular function. From all the stem water transport rate measurement data of this period, the stem water transport rate with the largest and smallest values ​​were selected, and the arithmetic mean was calculated. The fluctuation range was obtained by dividing the difference between the largest and smallest stem water transport rates by the arithmetic mean. The arithmetic mean multiplied by 10% is used as the threshold for whether a single fluctuation in stem water transport rate is significant. The difference between each stem water transport rate measurement data in this cycle and the previous stem water transport rate is checked sequentially. If the single fluctuation amplitude is greater than the threshold, it is determined to be a significant fluctuation. Finally, the number of all significant fluctuations in this cycle is counted as the fluctuation frequency. Based on the fluctuation amplitude and frequency, the risk level of vascular blockage is classified, and the value of the vascular blockage early warning factor is determined accordingly.

[0014] Preferably, the disease severity calculation and severity classification module includes: The weights of the leaf structure health index and the vascular function activity index were assigned based on leaf age. The two are weighted and added together to obtain the in vivo disease burden index, specifically: Where A is the disease load index, A1 is the leaf structure health index, A2 is the vascular function activity index, and γ1 and γ2 are the weights of the leaf structure health index and the vascular function activity index, respectively. Collect in vivo disease load index data of rice seedling samples with known health status, and determine their true disease severity through expert evaluation, including healthy, mild, moderate and severe; The in vivo disease burden index value of each sample is used as a one-dimensional data point to form a dataset; The K-means algorithm is applied to cluster one-dimensional data and calculate the threshold. The final threshold is then embedded into the disease severity calculation and severity classification module of the system. When the system diagnoses new rice seedlings, it calculates the disease load index value in the seedling and compares it directly with the threshold learned through clustering to output the disease severity level.

[0015] The beneficial effects of this invention are: This invention designs dynamic data collection rules to address the characteristics of rice seedlings, such as thin leaves and rapid growth. It allocates different suitable wavelengths (e.g., 700-720nm for 1-3 leaf age and 740-760nm for 4-6 leaf age), different time periods, and different environmental adaptation controls according to leaf age, ensuring that data collection is adapted to the physiological characteristics of the seedling stage and avoiding errors caused by a one-size-fits-all approach to data collection. This invention addresses the angle dependence problem of hyperspectral data by combining the initial red edge position offset calculation with an angle correction coefficient using a three-segment interpolation method. The correction coefficients are adapted for the three observation angles of 0°, 30°, and 45°. At the same time, the weighted average method is used to fuse the valley depths of red and blue valley reflectance, improving the sensitivity of spectral features to leaf structure damage, which is different from the simple processing method of calculating valley depth in a single band in the existing technology. This invention calculates and obtains the leaf structure health index and vascular function activity index by integrating multiple parameters. It combines the two types of features, leaf structure and vascular function, to achieve a comprehensive disease characterization that combines phenotypic and physiological aspects, thereby improving the accuracy of early identification. Furthermore, the K-means clustering algorithm is applied to rice seedling blast classification. By calibrating thresholds using samples with known health status, the invention directly outputs classification results of healthy, mild, moderate, and severe diseases. Moreover, the weight allocation (the weights of the leaf structure health index and the vascular function activity index are adjusted with leaf age) is adapted to the disease development pattern during the seedling stage. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the module flow of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 As shown, this invention is an intelligent analysis system for rice seedling blast disease based on remote sensing data analysis, comprising the following modules: Physiological data acquisition module: used to acquire in situ physiological parameters of individual rice seedlings; the parameters include leaf spectral reflectance data acquired by a hyperspectral sensor, maximum photochemical efficiency data of photosystem II acquired by a chlorophyll fluorometer, and stem water transport rate data acquired by a plant stem micro-change sensor. Physiological feature extraction module: Based on the in-situ physiological parameters, it obtains the leaf structure health index by fusing the red edge position offset, leaf reflectance valley depth and leaf spectral slope, and obtains the vascular function activity index by fusing the maximum photochemical efficiency of the fusion photosystem II, stem water transport rate, environmental dynamic correction parameters, vascular function temporal change rate and vascular blockage early warning factor. Disease severity calculation and severity classification module: It is used to add the leaf structure health index and vascular function activity index with weights to obtain the in vivo disease load index of the degree of disease infection in rice seedlings. Based on the numerical range of the in vivo disease load index, it maps and outputs the disease severity level of individual rice seedlings.

[0019] Specifically, a hyperspectral sensor was used to collect the spectral reflectance of rice seedling leaves in the 350-1050 nm band (intervals of 2 nm), while simultaneously recording light intensity and leaf temperature; a chlorophyll fluorometer was used to collect the maximum photochemical efficiency of photosystem II after 30 minutes of dark adaptation (averaged after three repetitions); a plant stem micro-change sensor was used to collect data on micro-changes in stem diameter at a frequency of 10 Hz and to calculate the water transport rate; based on the spectral reflectance, a three-segment interpolation method combined with angle correction was used to obtain the red edge position offset, and a weighted fusion of the red valley (650-68 nm) was performed. The leaf spectral slope is calculated by combining the reflectance valley depth of 0nm and Blue Valley (450-480nm) with the leaf age-appropriate band. The three are then fused to obtain the leaf structure health index. The vascular function activity index is obtained by fusion of photosystem II efficiency, stem water transport rate, environmental dynamic correction parameters, vascular function temporal change rate, and vascular blockage early warning factor. The two indices are weighted according to leaf age and weighted and added to obtain the in vivo disease load index. The disease load index is then compared with the threshold determined by K-means clustering to output the healthy, mild, moderate, and severe levels.

[0020] In one embodiment of the present invention, the physiological data acquisition module includes: In-situ, non-destructive parameter collection was performed on individual rice seedlings, specifically as follows: Leaf spectral reflectance data in the wavelength range of 350-1050nm were collected using a hyperspectral sensor with a wavelength interval of 2nm. The ambient light intensity and leaf temperature were recorded simultaneously during the data collection. The maximum photochemical efficiency data of photosystem II was collected using a chlorophyll fluorometer. A 30-minute dark adaptation treatment was set before the data collection. Each data collection was repeated 3 times and the average value was taken. The stem diameter micro-change data were collected by a plant stem micro-change sensor at a sampling frequency of 10Hz, and the stem water transport rate was calculated based on the data.

[0021] Specifically, the sampling frequency of the plant stem micro-change sensor was set to 10Hz, and stem diameter micro-change data was continuously collected for 1 hour. Based on the collected diameter change curve, the stem water transport rate was calculated using a model (stem diameter change / time interval corresponding to diameter change × rice seedling stem cross-sectional area × stem cross-sectional area coefficient). The plant stem micro-change sensor collected stem diameter data at a sampling frequency of 10Hz, generating one diameter data point every 0.1 seconds. Ten consecutive data points were selected, and the diameter change within that time period (maximum diameter minus minimum diameter) was calculated. During sensor installation, the stem diameter at the sensor's fixed position was measured using digital calipers. Each rice seedling was measured three times at the same location, and the average value was used as the baseline diameter. The cross-sectional area was calculated using the formula for the area of ​​a circle. The stem cross-sectional area coefficient was determined by selecting healthy plants of the target rice variety, covering 1-6 leaf ages (30 plants per leaf age). The actual stem water transport rate of each plant was measured using a weighing method (standard water transport rate measurement method). Simultaneously, the system's sensors collected data on the stem diameter change, the corresponding time interval for the diameter change, and the rice seedling stem cross-sectional area. These data were then substituted into the stem cross-sectional area coefficient formula (actual stem water transport rate / (stem diameter change / corresponding time interval for diameter change × rice seedling stem cross-sectional area)). The average value of the stem cross-sectional area coefficient values ​​from the 30 plants of the same leaf age was taken as the cross-sectional area coefficient for that leaf age. The stem water transport rate was then obtained. To ensure data accuracy, three sets of data were continuously collected from each rice seedling, and the average value was taken as the valid data. All the above data collection processes were carried out in the original location of the rice seedlings. All sensors and plants were designed to be non-destructive (e.g., flexible probes, pressure-free fixation) to avoid interfering with the growth of the rice seedlings.

[0022] In one embodiment of the present invention, the physiological data acquisition module simultaneously acquires the data required for calculating the infrared position offset when the hyperspectral sensor acquires the leaf spectral reflectance data, including: The hyperspectral sensor simultaneously collects bidirectional reflectance data of rice seedling leaves, including spectral reflectance at three observation angles: 0°, 30°, and 45°. At the same time, the vertical distance between the sensor and the leaf surface is kept constant at 30 cm. The red edge position offset was calculated using a three-segment interpolation method. Specifically, spectral reflectance data in the 680-760nm band was extracted with a sampling interval of 1nm, and the band was divided into three sub-bands: 680-709nm, 710-730nm, and 731-760nm. The maximum and minimum reflectance values ​​for each sub-band are calculated to determine the red-edge inflection point for each sub-band. A line connecting the inflection points of the three bands is fitted using linear interpolation, and the wavelength value corresponding to the inflection point where the slope of the line changes from positive to negative is taken as the initial red-edge position offset. An angle correction coefficient is introduced to correct the initial red-edge position offset, specifically: in, This is the corrected offset of the red edge position. This is the initial red edge position offset. For the angle of observation, Angle correction coefficient.

[0023] Specifically, the angle correction coefficient is determined through a standard plate calibration experiment: a diffuse reflection standard plate (with known reflectivity) is selected, and its 680-760nm spectral data are collected at three observation angles: 0°, 30°, and 45°. The initial red edge position offset of the standard plate at each angle is calculated and compared with the standard value to obtain the correction coefficient corresponding to each angle. Specifically, the angle correction coefficient is 1.0 at 0°, 0.9 at 30°, and 0.85 at 45°.

[0024] In one embodiment of the present invention, the depth of the blade reflectivity valley includes: The characteristic wavelengths of red valley and blue valley in the spectrum of rice seedling leaves were determined, with the red valley characteristic wavelength being 650-680nm and the blue valley characteristic wavelength being 450-480nm. The spectral sampling interval between the two bands was reduced from the basic 2nm to 1nm; The leaf reflectance valley depth is obtained by merging the reflectance valley depths of red and blue valleys using a weighted average method, as follows: Wherein, RGD is the depth of the leaf reflectance valley. This represents the maximum reflectivity within the characteristic band of the Red Valley. This represents the minimum reflectivity within the characteristic band of the Red Valley. This represents the maximum reflectivity within the characteristic band of the Blue Valley. This represents the minimum reflectivity within the characteristic band of the Blue Valley.

[0025] Specifically, select healthy functional leaves of rice seedlings and gently remove dust, dew, and other impurities from the leaf surface with a soft brush; activate the humidity sensor built into the hyperspectral sensor, place the probe close to the leaf surface, and monitor the leaf surface humidity value in real time. If the detected leaf surface humidity is >15%, immediately pause data acquisition and keep the sensor position unchanged; activate the sensor's matching hot air blowing device, adjust the hot air temperature to 35-40℃ (to avoid high temperature damage to the leaves), and set the wind speed to low, continuously blowing on the leaf sampling area for 10 seconds; after blowing, use the humidity sensor again to detect the leaf surface humidity until the humidity is ≤15%, then proceed to the subsequent data acquisition steps; Open the hyperspectral sensor data acquisition software. In the band setting interface, locate the red valley characteristic band and the blue valley characteristic band respectively. Set the range of the red valley characteristic band to 650-680nm and the range of the blue valley characteristic band to 450-480nm. encrypt the sampling interval and start the acquisition process to collect data. Filter out the maximum and minimum reflectance values ​​from the reflectance data of the red and blue valley characteristic bands respectively. Subtract the minimum value from the maximum value to obtain the corresponding red valley reflectance valley depth and blue valley reflectance valley depth. Use a weighted average method to fuse the reflectance valley depths of the red and blue valleys, determining the weight of the red valley as 0.7 and the weight of the blue valley as 0.3.

[0026] In one embodiment of the present invention, the leaf spectral slope includes: The corresponding calculation bands were selected for different leaf ages of rice seedlings, specifically: 700-720nm band for 1-3 leaf age and 740-760nm band for 4-6 leaf age. The selected calculation band was divided into 11 data points according to wavelength intervals. When collecting reflectance data, the sensor integration time was adjusted according to the band. The 11 data points were divided into three feature intervals, A, B and C. Linear regression was performed on each of the three feature intervals to obtain the regression slope. Differential weights are assigned based on the biological significance of the three feature intervals, and the initial leaf spectral slope is obtained by weighting and adding them to the regression slope of each interval. Differential weighting coefficients were assigned to the 11 data points, and the weighted average wavelength and band arithmetic mean wavelength of the 11 data points were calculated. The initial leaf spectral slope was then corrected to obtain the corrected leaf spectral slope, specifically as follows: in, This is the corrected leaf spectral slope. The initial leaf spectral slope, For the weighted average wavelength, The arithmetic mean wavelength of the band; Three sets of 11 data points were collected continuously from the same leaf. The coefficient of variation of the leaf spectral slope of the three sets was calculated. Finally, the average value of the three sets of data with the coefficient of variation within the preset threshold range was taken as the final leaf spectral slope.

[0027] Specifically, corresponding calculation bands were selected for different leaf ages of rice seedlings. For 1-3 leaf age, the 700-720nm band was selected (1-3 leaf age rice seedlings are in the seedling stage, with thin leaves and low chlorophyll content; the red edge effect is concentrated in the 700-720nm band, and the reflectance changes in this band can accurately reflect the early physiological abnormalities of seedling blight). For 4-6 leaf age, the 740-760nm band was selected (4-6 leaf age rice seedlings have thicker leaves and increased chlorophyll content; the infrared effect shifts to the long-wave direction to the 740-760nm band, and this band can effectively capture the spectral changes caused by vascular tissue lesions in the leaves). The selected band was divided into 11 data points (λ1-R1 to λ) according to a 2nm wavelength interval. 11 -R 11 The wavelengths of the 11 data points in the 700-720nm band are as follows: λ1=700nm, λ2=702nm, ..., λ 11 =720nm; Following the same method, the wavelengths of 11 data points corresponding to the 740-760nm band can be obtained; The sensor integration time is adjusted according to the band characteristics: 10ms for leaf ages 1-3 (700-720nm, short-wave red edge), and 15ms for leaf ages 4-6 (740-760nm, long-wave red edge) due to weaker light energy, ensuring that the signal-to-noise ratio (SNR) of reflectance at all data points is ≥35dB. Specifically, the 718-720nm wavelength for leaf ages 1-3 and 4-6... The SNR of data points at 758-760nm (red edge tail) for leaf age is ≥38dB; based on the biological significance of different leaf ages, the characteristic interval division is adjusted to ensure that each interval can accurately capture the disease-sensitive spectral signals of the corresponding leaf age: for leaf ages 1-3: interval A (λ1-λ3, edge transition zone, 700-704nm) corresponds to the reflectance characteristics of leaf surface cells. Early damage to the cuticle of the epidermis in the early stage of blast disease will lead to an increase in reflectance in this interval; interval B (λ4-λ8, infrared core region, 706-714nm): the region with the most significant red edge effect, directly related to chlorophyll content and cell structure integrity. After blast disease infection, the reflectance slope in this region will be significantly reduced; interval C (λ9-λ 11 The tail-end transition zone (716-720 nm) corresponds to the reflectance characteristics of the leaf spongy tissue. Necrosis of spongy tissue cells caused by blast disease will lead to abnormal fluctuations in reflectance in this zone. Linear regression fitting was performed on each of the three zones to obtain the regression slope. Among them, the goodness-of-fit R² for intervals A and C is ≥0.92, and the goodness-of-fit R² for interval B is ≥0.98 (the core area is more sensitive to disease and requires higher fitting accuracy); for 4-6 leaf age: interval A (λ1-λ3, edge transition zone, 740-744nm): corresponds to the reflectance characteristics of the palisade tissue of the leaf. Disordered arrangement of the palisade tissue caused by wilt disease will cause changes in reflectance in this interval; interval B (λ4-λ8, red-edge core zone, 746-754nm): long-wave red-edge core zone, related to the vascular tissue function of the leaf and the chlorophyll a / b ratio. Vascular blockage caused by wilt disease will cause abnormal reflectance slope in this interval; interval C (λ9-λ 11 The tail-end transition zone (756-760nm) corresponds to the reflectance characteristics of the area where the lower epidermis of the leaf connects to the petiole. The reflectance in this zone increases significantly when the blight spreads to the petiole. Similarly, linear regression fitting is performed, with the goodness-of-fit requirement consistent with that for leaf ages 1-3. If the goodness-of-fit for any leaf age interval B is <0.98, the sensor is triggered to re-acquire data for that interval. The number of re-acquisitions should not exceed 3. If the goodness-of-fit is still <0.98 after 3 acquisitions, the data is marked as invalid, and a different leaf is collected to ensure the reliability of the spectral data in the core area. Differential weights are assigned based on the biological significance of the three characteristic intervals (i.e., different segment intervals B are most sensitive to blight in seedlings, each assigned a weight of 70%, and the other two intervals assigned 15%). These weights are then added to the regression slope of each interval to obtain the initial leaf spectral slope, specifically: ,in, The initial leaf spectral slope, These represent the regression slopes for intervals A, B, and C, respectively. Based on the disease sensitivity of data points at each leaf age, differential weighting coefficients are assigned to enhance the contribution of sensitive data points in the core area and at the tail end: 1-3 / 4-6 leaf ages: λ1=λ 11 =0.05、λ2=λ 10 =0.08, λ3=λ9=0.12, λ4=λ8=0.15, λ5=λ7=0.18, and λ6=0.2; calculate the weighted average wavelength and band arithmetic mean wavelength of each leaf age data point, and obtain the corrected leaf spectral slope by correcting it with the initial leaf spectral slope; formulate matching stability verification rules for the band characteristics of different leaf ages: continuously collect 3 sets of complete data for the same leaf (3 sets of 11 data points for leaf ages 1-3, and 3 sets of 11 data points for leaf ages 4-6), and calculate the coefficient of variation of the 3 sets of corrected leaf spectral slopes, specifically: Where C is the coefficient of variation. The corrected leaf spectral slope corresponds to the i-th group. The average of the three sets of data is required to be within a preset threshold range. The preset threshold is based on the statistical distribution characteristics of the coefficient of variation of the spectral slope data of rice seedling leaves. By setting the percentile corresponding to a reasonable and stable data range, the coefficient of variation value corresponding to the percentile is used as the threshold. Only the dataset with a coefficient of variation ≤ the threshold (i.e., data within the "reasonable and stable range" of the statistical distribution) is retained, and unstable data at the tail of the distribution is removed.

[0028] In one embodiment of the present invention, the leaf structure health index includes: The leaf structure health index is calculated by fusing multiple parameters, including the red edge position offset, leaf reflectance valley depth, and leaf spectral slope, from the hyperspectral data. Specifically: Where A1 is the leaf structure health index, REP is the red edge position offset, RGD is the leaf reflectance valley depth, and RSS is the leaf spectral slope.

[0029] In one embodiment of the present invention, the vascular function activity index includes: The vascular function activity index is calculated by coupling the maximum photochemical efficiency of photosystem II, stem water transport rate, environmental dynamic correction parameters, temporal change rate of vascular function, and vascular blockage early warning factor. Specifically: Where A2 is the vascular function activity index, α is the environmental dynamic correction coefficient, and β1 and β2 are dynamic weighting coefficients. For the variable fluorescence of optical system II, The maximum fluorescence of optical system II, The stem water transport rate. This is the varietal benchmark value for stem water transport rate. This is a reference value for the upper limit of varieties. The rate of change of vascular function over time. As an early warning factor for vascular blockage; The environmental dynamic correction coefficient is determined by coupling calculation of multiple environmental factor correction coefficients, which include temperature correction coefficient, humidity correction coefficient and illumination correction coefficient.

[0030] Specifically, the dynamic weighting coefficients are adjusted according to the leaf age stage of rice seedlings. Specifically: for 1-3 leaf age (vascular development incomplete), β1 and β2 values ​​are set to 0.3 and 0.7 respectively, representing the contribution of enhanced water transport capacity; for 4-6 leaf age (vascular development mature), β1 and β2 values ​​are set to 0.6 and 0.4 respectively, representing the proportion of photosynthetic function supporting vascular metabolism. The environmental dynamic correction coefficients are first calibrated using sub-coefficients, including the temperature correction sub-coefficient (based on leaf temperature collected by environmental sensors; at 25℃, the temperature correction sub-coefficient is set to 1.0, increasing or decreasing by 0.05 for every 5℃ deviation; for example, at 30℃, the temperature correction sub-coefficient is set to 0.95), the humidity correction sub-coefficient (based on relative humidity collected; at 60%, the humidity correction sub-coefficient is set to 1.0, increasing or decreasing by 0.05 for every 20% deviation), and the light correction sub-coefficient (based on ambient light intensity collected; light intensity is 1000 μmol / m²). 2 At ⋅s, the illumination correction factor is assigned a value of 1.0, and for every deviation of 500 μmol / m 2 The weights of the three sub-coefficients were determined to be 0.4, 0.3, and 0.3 respectively by calibrating experimental data under 10 different environmental conditions (the coefficient of s increases or decreases by 0.03). The three sub-coefficients were then weighted and added together to obtain the environmental dynamic correction coefficient.

[0031] In one embodiment of the present invention, the time-series rate of change of vascular function includes: The vascular function temporal change rate is combined with the dynamic adjustment cycle of leaf age stage. The dynamic adjustment cycle includes: the collection cycle of rice seedlings with 1-3 leaves is set to 12 hours / time; the collection cycle of rice seedlings with 4-6 leaves is set to 24 hours / time. According to the determined collection cycle, extract two types of core parameters from the system's parameter database: the current cycle and the parameters from the three consecutive cycles. The two core parameters are the maximum photochemical efficiency of the photosystem II after light adaptation correction and the standardized stem water transport rate. By weighting and adding the two types of core parameters for different periods, the comprehensive value of vascular function in the current period and the comprehensive value of vascular function three periods ago are obtained respectively. The difference between the two is divided by 3 to obtain the time series change rate of vascular function.

[0032] Specifically, continuous data on target rice seedlings are selected from the system parameter database. Using the current collection period as a baseline, data from three preceding periods are extracted, resulting in four periods of core data. The maximum photochemical efficiency of Photosystem II is easily affected by ambient light intensity. When light is insufficient, the leaf photosynthetic system is not fully activated, and the measured raw value will be lower than the true level. The optimal light intensity for activating the photosynthetic system of rice seedlings during natural growth is 1000 μmol / m². 2 ⋅s, when the collected light intensity is not equal to 1000 μmol / m 2 When the light intensity is ≥1000 μmol / m², the original value under non-suitable lighting conditions needs to be converted into the corrected value under equivalent suitable lighting conditions using a correction formula; when the light intensity is ≥1000 μmol / m², 2 At ⋅s, with sufficient light, the photosynthetic system is fully activated, and the correction factor is directly taken as 1.0; when the light intensity is <1000μmol / m 2 At ⋅s, insufficient light means the photosynthetic system is not fully activated, requiring logarithmic compensation, specifically: in, The maximum photochemical efficiency of the photosystem II after light adaptation correction. 0.02 is the original value of the maximum photochemical efficiency of the optical system II, 1000 is the baseline value of the optimal activation light intensity of the optical system II, and L is the actual ambient light intensity for collecting the original value of the maximum photochemical efficiency of the optical system II. The standardized stem water transport rate is obtained by calling the baseline value of stem water transport rate of healthy rice plants of this variety stored in the database, and dividing the original value of stem water transport rate by the baseline value of stem water transport rate. Based on the degree of influence of the two types of core parameters on vascular function, fixed weights are assigned. The maximum photochemical efficiency of the photosystem II after light adaptation correction corresponds to a weight of 0.4, and the standardized stem water transport rate corresponds to a weight of 0.6. The two types of parameters for each period are weighted and summed to obtain the comprehensive value of vascular function for the corresponding period. The difference between the current comprehensive value and the comprehensive value three periods ago is divided by 3 to obtain the time series change rate. If it is greater than 0, it indicates that the vascular function is improving; if it is equal to 0, it indicates that the function is stable; if it is less than 0, it indicates that the function is declining (possibly due to disease).

[0033] In one embodiment of the present invention, the vascular blockage early warning factor includes: The data collection cycle for vascular blockage early warning factors should be consistent with the data collection cycle for the time-series change rate of vascular function. From all the stem water transport rate measurement data of this period, the stem water transport rate with the largest and smallest values ​​were selected, and the arithmetic mean was calculated. The fluctuation range was obtained by dividing the difference between the largest and smallest stem water transport rates by the arithmetic mean. The arithmetic mean multiplied by 10% is used as the threshold for whether a single fluctuation in stem water transport rate is significant. The difference between each stem water transport rate measurement data in this cycle and the previous stem water transport rate is checked sequentially. If the single fluctuation amplitude is greater than the threshold, it is determined to be a significant fluctuation. Finally, the number of all significant fluctuations in this cycle is counted as the fluctuation frequency. Based on the fluctuation amplitude and frequency, the risk level of vascular blockage is classified, and the value of the vascular blockage early warning factor is determined accordingly.

[0034] Specifically, based on the logic that fluctuation amplitude reflects the severity of congestion and fluctuation frequency reflects the frequency of congestion occurrence, a three-level risk standard is set (determined through calibrated data of 100 sets of known congestion degree-rate fluctuations). Among them, low risk: fluctuation amplitude ≤ 20% and fluctuation frequency ≤ 5 times / cycle; medium risk: 20% < fluctuation amplitude ≤ 50% and 5 times / cycle < fluctuation frequency ≤ 15 times / cycle; high risk: fluctuation amplitude > 50% and fluctuation frequency > 15 times / cycle. The vascular blockage warning factor is determined according to the risk level. The higher the value, the higher the risk of blockage and the stronger the negative impact on the vascular function activity index. Low risk: the corresponding vascular blockage warning factor is 0.1; medium risk: the corresponding vascular blockage warning factor is 0.5; high risk: the corresponding vascular blockage warning factor is 0.8.

[0035] In one embodiment of the present invention, the disease severity calculation and severity classification module includes: The weights of the leaf structure health index and the vascular function activity index were assigned based on leaf age. The two are weighted and added together to obtain the in vivo disease burden index, specifically: Where A is the disease load index, A1 is the leaf structure health index, A2 is the vascular function activity index, and γ1 and γ2 are the weights of the leaf structure health index and the vascular function activity index, respectively. Collect in vivo disease load index data of rice seedling samples with known health status, and determine their true disease severity through expert evaluation, including healthy, mild, moderate and severe; The in vivo disease burden index value of each sample is used as a one-dimensional data point to form a dataset; The K-means algorithm is applied to cluster one-dimensional data and calculate the threshold. The final threshold is then embedded into the disease severity calculation and severity classification module of the system. When the system diagnoses new rice seedlings, it calculates the disease load index value in the seedling and compares it directly with the threshold learned through clustering to output the disease severity level.

[0036] Specifically, based on the characteristics of disease impact at different leaf age stages (leaf structures are fragile in the 1-3 leaf age, and diseases first damage the leaf surface; the vascular system is fully developed in the 4-6 leaf age, and diseases easily block the vascular system), differentiated weights were set: 1-3 leaf age: leaf structure health index weight is 0.6, vascular function activity index weight is 0.4; 4-6 leaf age: leaf structure health index weight is 0.4, vascular function activity index weight is 0.6. Healthy seedlings, mildly diseased seedlings, moderately diseased seedlings, and severely diseased seedlings of the same rice variety were selected as calibration samples. The categories were defined as follows: healthy seedlings (no lesions, normal photosynthetic and vascular functions); mildly diseased seedlings (1-2 lesions <2mm in diameter on leaves, no significant functional decline); moderately diseased seedlings (lesions covering 10%-30% of leaf area, functional decline of 15%-30%); and severely diseased seedlings (lesions covering >30% of leaf area, functional decline >30%). The health status of all samples was confirmed on-site by plant pathology experts. The in vivo disease load index of each calibrated sample was calculated, with each sample's in vivo disease load index value serving as a one-dimensional data point to form a dataset. The K-means algorithm was then launched in the system's data analysis module, with the number of clusters set to K=4 (corresponding to the four disease levels: healthy, mild, moderate, and severe). The initial cluster centers were determined using the elbow rule, and the convergence condition was proximity. If the distance between cluster centers in two iterations is less than 0.01, the dataset is input into the algorithm. Through iterative calculation, the data points are divided into four clusters, each corresponding to a disease level—Cluster 1 (healthy), Cluster 2 (mild), Cluster 3 (moderate), and Cluster 4 (severe). The maximum and minimum values ​​of the disease load index in the four clusters are calculated to determine the boundary between adjacent levels. For example, the threshold between healthy and mild is obtained by averaging the minimum value of Cluster 1 and the maximum value of Cluster 2. The same method is used to obtain the thresholds between mild and moderate, and between moderate and severe, for a total of three thresholds: TD1, TD2, and TD3. If the disease load index value is ≤ TD1, it is judged as healthy; if TD1 < TD2, it is judged as mild; if TD2 < TD3, it is judged as moderate; if the disease load index value is > TD3, it is judged as severe.

[0037] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A smart analysis system for rice seedling blast disease based on remote sensing data analysis, characterized in that, Includes the following modules: Physiological data acquisition module: used to acquire in situ physiological parameters of individual rice seedlings; the parameters include leaf spectral reflectance data acquired by a hyperspectral sensor, maximum photochemical efficiency data of photosystem II acquired by a chlorophyll fluorometer, and stem water transport rate data acquired by a plant stem micro-change sensor. Physiological feature extraction module: Based on the in-situ physiological parameters, it obtains the leaf structure health index by fusing the red edge position offset, leaf reflectance valley depth and leaf spectral slope, and obtains the vascular function activity index by fusing the maximum photochemical efficiency of the fusion photosystem II, stem water transport rate, environmental dynamic correction parameters, vascular function temporal change rate and vascular blockage early warning factor. Disease severity calculation and severity classification module: It is used to add the leaf structure health index and vascular function activity index with weights to obtain the in vivo disease load index of the degree of disease infection in rice seedlings. Based on the numerical range of the in vivo disease load index, it maps and outputs the disease severity level of individual rice seedlings.

2. The intelligent analysis system for rice seedling blast disease based on remote sensing data analysis according to claim 1, characterized in that, The physiological data acquisition module includes: In-situ, non-destructive parameter collection was performed on individual rice seedlings, specifically as follows: Leaf spectral reflectance data in the wavelength range of 350-1050nm were collected using a hyperspectral sensor with a wavelength interval of 2nm. The ambient light intensity and leaf temperature were recorded simultaneously during the data collection. The maximum photochemical efficiency data of photosystem II was collected using a chlorophyll fluorometer. A 30-minute dark adaptation treatment was set before the data collection. Each data collection was repeated 3 times and the average value was taken. The stem diameter micro-change data were collected by a plant stem micro-change sensor at a sampling frequency of 10Hz, and the stem water transport rate was calculated based on the data.

3. The intelligent analysis system for rice seedling blast disease based on remote sensing data analysis according to claim 1, characterized in that, In the physiological data acquisition module, when the hyperspectral sensor acquires leaf spectral reflectance data, it simultaneously obtains the data required for calculating the infrared position offset, including: The hyperspectral sensor simultaneously collects bidirectional reflectance data of rice seedling leaves, including spectral reflectance at three observation angles: 0°, 30°, and 45°. At the same time, the vertical distance between the sensor and the leaf surface is kept constant at 30 cm. The red edge position offset was calculated using a three-segment interpolation method. Specifically, spectral reflectance data in the 680-760nm band was extracted with a sampling interval of 1nm, and the band was divided into three sub-bands: 680-709nm, 710-730nm, and 731-760nm. The maximum and minimum reflectance values ​​for each sub-band are calculated to determine the red-edge inflection point for each sub-band. A line connecting the inflection points of the three bands is fitted using linear interpolation, and the wavelength value corresponding to the inflection point where the slope of the line changes from positive to negative is taken as the initial red-edge position offset. An angle correction coefficient is introduced to correct the initial red-edge position offset, specifically: in, This is the corrected offset of the red edge position. This is the initial red edge position offset. For the angle of observation, Angle correction coefficient.

4. The intelligent analysis system for rice seedling blast disease based on remote sensing data analysis according to claim 1, characterized in that, The depth of the leaf reflectivity valley includes: The characteristic wavelengths of red valley and blue valley in the spectrum of rice seedling leaves were determined, with the red valley characteristic wavelength being 650-680nm and the blue valley characteristic wavelength being 450-480nm. The spectral sampling interval between the two bands was reduced from the basic 2nm to 1nm; The leaf reflectance valley depth is obtained by merging the reflectance valley depths of red and blue valleys using a weighted average method, as follows: Wherein, RGD is the depth of the leaf reflectance valley. This represents the maximum reflectivity within the characteristic band of the Red Valley. This represents the minimum reflectivity within the characteristic band of the Red Valley. This represents the maximum reflectivity within the characteristic band of the Blue Valley. This represents the minimum reflectivity within the characteristic band of the Blue Valley.

5. The intelligent analysis system for rice seedling blast disease based on remote sensing data analysis according to claim 1, characterized in that, The spectral slope of the leaf includes: The corresponding calculation bands were selected for different leaf ages of rice seedlings, specifically: 700-720nm band for 1-3 leaf age and 740-760nm band for 4-6 leaf age. The selected calculation band was divided into 11 data points according to wavelength intervals. When collecting reflectance data, the sensor integration time was adjusted according to the band. The 11 data points were divided into three feature intervals, A, B and C. Linear regression was performed on each of the three feature intervals to obtain the regression slope. Differential weights are assigned based on the biological significance of the three feature intervals, and the initial leaf spectral slope is obtained by weighting and adding them to the regression slope of each interval. Differential weighting coefficients were assigned to the 11 data points, and the weighted average wavelength and band arithmetic mean wavelength of the 11 data points were calculated. The initial leaf spectral slope was then corrected to obtain the corrected leaf spectral slope, specifically as follows: in, This is the corrected leaf spectral slope. The initial leaf spectral slope, For the weighted average wavelength, The arithmetic mean wavelength of the band; Three sets of 11 data points were collected continuously from the same leaf. The coefficient of variation of the leaf spectral slope of the three sets was calculated. Finally, the average value of the three sets of data with the coefficient of variation within the preset threshold range was taken as the final leaf spectral slope.

6. The intelligent analysis system for rice seedling blast disease based on remote sensing data analysis according to claim 1, characterized in that, The leaf structure health index includes: The leaf structure health index is calculated by fusing multiple parameters, including the red edge position offset, leaf reflectance valley depth, and leaf spectral slope, from the hyperspectral data. Specifically: Where A1 is the leaf structure health index, REP is the red edge position offset, RGD is the leaf reflectance valley depth, and RSS is the leaf spectral slope.

7. The intelligent analysis system for rice seedling blast disease based on remote sensing data analysis according to claim 1, characterized in that, The vascular function activity index includes: The vascular function activity index is calculated by coupling the maximum photochemical efficiency of photosystem II, stem water transport rate, environmental dynamic correction parameters, temporal change rate of vascular function, and vascular blockage early warning factor. Specifically: Where A2 is the vascular function activity index, α is the environmental dynamic correction coefficient, and β1 and β2 are dynamic weighting coefficients. For the variable fluorescence of optical system II, The maximum fluorescence of optical system II, The stem water transport rate. This is the varietal benchmark value for stem water transport rate. This is a reference value for the upper limit of varieties. The rate of change of vascular function over time. As an early warning factor for vascular blockage; The environmental dynamic correction coefficient is determined by coupling calculation of multiple environmental factor correction coefficients, which include temperature correction coefficient, humidity correction coefficient and illumination correction coefficient.

8. The intelligent analysis system for rice seedling blast disease based on remote sensing data analysis according to claim 7, characterized in that, The time-series rate of change of vascular function includes: The vascular function temporal change rate is combined with the dynamic adjustment cycle of leaf age stage. The dynamic adjustment cycle includes: the collection cycle of rice seedlings with 1-3 leaves is set to 12 hours / time; the collection cycle of rice seedlings with 4-6 leaves is set to 24 hours / time. According to the determined collection cycle, extract two types of core parameters from the system's parameter database: the current cycle and the parameters from the three consecutive cycles. The two core parameters are the maximum photochemical efficiency of the photosystem II after light adaptation correction and the standardized stem water transport rate. By weighting and adding the two types of core parameters for different periods, the comprehensive value of vascular function in the current period and the comprehensive value of vascular function three periods ago are obtained respectively. The difference between the two is divided by 3 to obtain the time series change rate of vascular function.

9. The intelligent analysis system for rice seedling blast disease based on remote sensing data analysis according to claim 7, characterized in that, The vascular blockage early warning factors include: The data collection cycle for vascular blockage early warning factors should be consistent with the data collection cycle for the time-series change rate of vascular function. From all the stem water transport rate measurement data of this period, the stem water transport rate with the largest and smallest values ​​were selected, and the arithmetic mean was calculated. The fluctuation range was obtained by dividing the difference between the largest and smallest stem water transport rates by the arithmetic mean. The arithmetic mean multiplied by 10% is used as the threshold for whether a single fluctuation in stem water transport rate is significant. The difference between each stem water transport rate measurement data in this cycle and the previous stem water transport rate is checked sequentially. If the single fluctuation amplitude is greater than the threshold, it is determined to be a significant fluctuation. Finally, the number of all significant fluctuations in this cycle is counted as the fluctuation frequency. Based on the fluctuation amplitude and frequency, the risk level of vascular blockage is classified, and the value of the vascular blockage early warning factor is determined accordingly.

10. The intelligent analysis system for rice seedling blast disease based on remote sensing data analysis according to claim 1, characterized in that, The disease severity calculation and severity classification module includes: The weights of the leaf structure health index and the vascular function activity index were assigned based on leaf age. The two are weighted and added together to obtain the in vivo disease burden index, specifically: Where A is the disease load index, A1 is the leaf structure health index, A2 is the vascular function activity index, and γ1 and γ2 are the weights of the leaf structure health index and the vascular function activity index, respectively. Collect in vivo disease load index data of rice seedling samples with known health status, and determine their true disease severity through expert evaluation, including healthy, mild, moderate and severe; The in vivo disease burden index value of each sample is used as a one-dimensional data point to form a dataset; The K-means algorithm is applied to cluster one-dimensional data and calculate the threshold. The final threshold is then embedded into the disease severity calculation and severity classification module of the system. When the system diagnoses new rice seedlings, it calculates the disease load index value in the seedling and compares it directly with the threshold learned through clustering to output the disease severity level.

Images