Greenhouse plant remote disease collaborative diagnosis method and system based on cloud platform

Abstract

The invention belongs to the technical field of intelligent agriculture, and relates to a greenhouse plant remote disease collaborative diagnosis method and system based on a cloud platform, and the method comprises the steps: obtaining real-time image data of greenhouse plants, and eliminating image shielding interference to form enhanced visual data; generating a preliminary disease feature judgment result; automatically triggering a plant surface sampling device to collect microtissue fluid to perform rapid biological detection, and obtaining pathological judgment evidence by identifying germ characteristic molecular identifiers in biological detection; performing dynamic weighting calculation by combining the illumination accumulation amount of the plant in the historical growth period and a moisture absorption curve to obtain a comprehensive diagnosis conclusion; a pesticide spraying path is monitored in real time through an action sensor arranged on spraying equipment, and the matching degree of an actual execution action and a guidance scheme is judged; and adaptively adjusting monitoring parameters of other plants in the same area according to the deviation type to form a dynamic protection strategy. The problem of pesticide waste caused by equipment calibration errors or manual operation errors in actual execution is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart agriculture and relates to a method and system for remote disease collaborative diagnosis of greenhouse plants based on a cloud platform. Background Art

[0002] Traditional greenhouse disease monitoring mainly relies on manual inspections combined with simple sensor technologies, suffering from problems such as low early disease recognition rates and lagged responses. Manual inspections are easily affected by subjective experience and difficult to capture hidden lesions in the lower layers of leaves. Temperature and humidity sensors can only provide environmental parameters and cannot correlate with disease characteristics, resulting in diseases being detected only when obvious symptoms appear, missing the best prevention and control window period. Leaf overlap and occlusion, equipment deployment costs, and complex lighting conditions increase the risk of misjudgment, and existing solutions are difficult to meet the technical requirements of precision prevention and control in modern agriculture.

[0003] Existing automated solutions use fixed camera devices and static threshold judgment mechanisms, and some systems introduce chemical sampling detection for auxiliary diagnosis. Although periodic data collection can be achieved, due to the lack of multispectral compensation technology and dynamic confidence algorithms, the biological detection link often relies on offline laboratory analysis, with high time consumption from sampling to result feedback and low correlation between historical growth data and environmental parameters, unable to form real-time comprehensive diagnosis.

[0004] Based on the above problems, traditional mechanical spraying equipment generally adopts a preset path planning mode and lacks the ability to verify the operation process. Actual execution results in pesticide waste due to equipment calibration errors or human operation mistakes. Summary of the Invention

[0005] To solve the above problems, the present invention provides a method and system for remote disease collaborative diagnosis of greenhouse plants based on a cloud platform.

[0006] In the first aspect, the present invention provides a method for remote disease collaborative diagnosis of greenhouse plants based on a cloud platform, adopting the following technical solutions: The method for remote disease collaborative diagnosis of greenhouse plants based on a cloud platform includes the following steps: S1. Obtain real-time image data of greenhouse plants. When it is detected that the overlapping part of the leaves exceeds a preset ratio, adjust the shooting angle through a mobile rail-mounted camera and activate the multispectral supplementary lighting device to eliminate image occlusion interference and form enhanced visual data; S2. Combine the enhanced visual data with the environmental monitoring data of the area where the plants are located, and generate a preliminary disease characteristic judgment result by analyzing the spatio-temporal distribution correlation between the temperature and humidity fluctuation trajectories and the abnormal areas in the images; S3. When the confidence level of the preliminary judgment result is lower than the set threshold, automatically trigger the plant surface sampling device to collect a small amount of tissue fluid for rapid biological detection, and the biological detection obtains pathological judgment evidence by identifying the pathogen characteristic molecular identifier; S4. Integrate the preliminary judgment result with the pathological judgment evidence, and perform dynamic weighted calculation by combining the light accumulation amount and water absorption curve of the historical growth cycle of the plant to obtain a comprehensive diagnosis conclusion; S5. Generate operation guidance information according to the comprehensive diagnosis conclusion, and simultaneously activate the user operation verification function. Real-time monitor the pesticide spraying path by setting an action sensor on the spraying device to judge the matching degree between the actual executed action and the guidance plan; S6. When there is an execution deviation in the operation verification result, freeze the associated data of the current case and activate the expert review channel. At the same time, adaptively adjust the monitoring parameters of other plants in the same area according to the deviation type to form a dynamic protection strategy.

[0007] A further solution of the present invention is to form enhanced visual data, including the following steps: Continuously collect plant leaf images through a mobile rail-mounted camera. When the image processing module detects that the pixel density in the intersection area of the leaf edge contours exceeds a preset ratio, calculate the optimal azimuth angle and moving speed of the camera; Start the combination of visible light and infrared band light sources in the preset multispectral supplementary lighting device to eliminate the shadow interference in the leaf overlapping area; Perform multispectral correction on the real-time image data according to the leaf coverage density determination formula to generate enhanced visual data containing clear vein textures.

[0008] A further solution of the present invention is to generate a preliminary disease characteristic judgment result, including the following steps: Extract the area expansion rate of the abnormal texture area in the enhanced visual data, and perform spatial registration with the temperature gradient distribution heat map in the environmental monitoring data; Use the sliding window method to calculate the weighted sum of the temperature change rate, humidity change rate and abnormal area growth rate to generate a mold growth index; When the mold growth index exceeds the preset threshold in the disease type characteristic library, match the disease code and mark the spatial coordinates of the associated area.

[0009] A further solution of the present invention is the dynamic weighted calculation, including the following steps: Assign a basic weight factor to the preliminary disease characteristic judgment result, and assign a verification weight factor to the pathological judgment evidence; Generate an environmental correction coefficient according to the light accumulation amount and historical mean value, and the change amount of the slope of the water absorption curve and the standard variation rate; Adjust the ratio of the basic weight factor and the verification weight factor through the dynamic weighted formula, and output the comprehensive diagnosis conclusion in combination with the stress resistance compensation factor.

[0010] A further solution of the present invention, the operation verification result, includes the following steps: Compare the target path vector coordinates in the operation guidance information with the actual spraying path collected by the motion sensor point by point; Calculate the intersection percentage of the coverage area of the actual spraying path and the deviation degree of the geometric center line, and quantify the execution deviation through the path matching degree formula; When the path matching degree is lower than the preset threshold, trigger an execution deviation alarm and record the three-dimensional coordinates of the deviation node.

[0011] A further solution of the present invention, the path matching degree formula:

[0012] Wherein, represents the path matching degree; represents the intersection area of the actual spraying area and the target area; represents the total area of the target area; represents the three-dimensional coordinates of the i-th sampling point of the actual path; represents the coordinates of the corresponding point of the target path; is the maximum allowable offset distance; n is the total number of sampling points; represents the total area of the actual spraying area; Set the path matching degree threshold according to the pesticide spraying path , When, trigger an execution deviation alarm.

[0013] A further solution of the present invention, forming a dynamic protection strategy, includes the following steps: Calculate the monitoring parameter adjustment amount of the target plant according to the deviation type coefficient and the sum of the mold growth indexes of the plants in the same area; Perform inverse weight allocation on the sampling frequency, multi-spectral supplementary light intensity and biological detection trigger threshold of the mobile orbital camera; Predict the disease diffusion path based on the spatio-temporal propagation model, and correct the image analysis parameters and environmental regulation thresholds of adjacent plants in advance.

[0014] A further solution of the present invention, freezing the current case-related data, includes the following steps: Package the enhanced visual data, biological detection spectral response value and operation guidance parameters of the entire diagnosis cycle into an unmodifiable data packet; Encrypt the data packet using the AES-256 encryption algorithm and store it in the isolated storage area; Push the encrypted data packet and the deviation analysis report to the expert terminal through the RESTful API interface.

[0015] A further solution of the present invention is to calculate the adjustment amount of the monitoring parameters of the target plant, including the following steps: When it is known to quantify the adjustment range of the protection strategy, a parameter correction formula is established:

[0016] Wherein, represents the adjustment amount of the monitoring parameters of the target plant; k is the deviation type coefficient. For example, in combination with expert experimental data, the pesticide residue deviation is taken as 1.2, and the path deviation is taken as 0.8; represents the sum of the mold growth indexes of other plants in the same area; represents the mold growth index of the plant before step S2; M is the calculation result of the path matching degree in step S5.

[0017] In a second aspect, the present invention provides a remote disease collaborative diagnosis system for greenhouse plants based on a cloud platform, adopting the following technical solutions: The remote disease collaborative diagnosis system for greenhouse plants based on a cloud platform includes the following modules: An image acquisition module for generating enhanced visual data through a mobile rail-mounted camera and a multi-spectral supplementary lighting device; An environmental data processing module for fusing the enhanced visual data with the temperature and humidity fluctuation trajectories to generate a preliminary disease characteristic judgment result; A biological detection module, when the confidence level of the preliminary judgment result is lower than a set threshold, automatically triggers a plant surface sampling device to collect a small amount of tissue fluid for rapid biological detection, and the biological detection obtains pathological judgment evidence by identifying the pathogen characteristic molecular identifier; A dynamic decision-making module for fusing the preliminary judgment result and the pathological judgment evidence, and performing dynamic weighted calculation in combination with the light accumulation amount and water absorption curve of the historical growth cycle of the plant to obtain a comprehensive diagnosis conclusion; An execution verification module for generating operation guidance information based on the comprehensive diagnosis conclusion, and synchronously starting the user operation verification function, and judging the matching degree between the actual execution action and the guidance plan by setting an action sensor on the spraying device to monitor the pesticide spraying path in real time; A safety response module for freezing the associated data of the current case and activating the expert review channel when there is an execution deviation in the operation verification result, and at the same time adaptively adjusting the monitoring parameters of other plants in the same area according to the deviation type to form a dynamic protection strategy.

[0018] In summary, the present invention includes the following beneficial technical effects: 1. Significantly improve the accuracy and efficiency of traditional agricultural disease control through an intelligent multi-modal data fusion and dynamic decision-making mechanism. Based on the collaborative operation of a mobile orbital camera and a multi-spectral supplementary lighting device, combined with the analysis of abnormal texture regions and the calculation of spatio-temporal phase differences, accurately identify the early disease characteristics and accurately locate the infected areas; 2. The dynamic resource allocation mechanism can intelligently optimize the utilization efficiency of regional prevention and control resources. When the operation verification module detects an execution deviation, synchronously reduce the detection threshold of high-threat plants in the same region and increase the monitoring parameter tolerance of low-risk plants through an adaptive adjustment algorithm; 3. The closed-loop verification system and data security guarantee mechanism provide full-process credibility guarantee for agricultural production. Eliminate the risk of liquid medicine waste or missed spraying caused by manual operation through the calculation of three-dimensional path matching degree and the real-time correction of pesticide spraying trajectory. Brief Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. The drawings are used to provide a further understanding of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 Disclose the flow schematic diagram of the remote disease collaborative diagnosis method for greenhouse plants based on the cloud platform.

[0021] Figure 2 Disclose the structural schematic diagram of the remote disease collaborative diagnosis system for greenhouse plants based on the cloud platform. Detailed Embodiments

[0022] In order to make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0023] The following combines the attached Figure 1 - Figure 2 Make a preferred and detailed description of the present invention.

[0024] Refer to the attached Figure 1 , the present invention proposes a remote disease collaborative diagnosis method for greenhouse plants based on the cloud platform, including the following steps: S1. Obtain the real-time image data of greenhouse plants. When it is detected that the overlapping part of the leaves exceeds the preset ratio, adjust the shooting angle through a mobile orbital camera and start the multi-spectral supplementary lighting device to eliminate the interference of image occlusion to form enhanced visual data; S2. Combine the enhanced visual data with the environmental monitoring data of the plant's location area, and generate a preliminary disease characteristic judgment result by analyzing the spatio-temporal distribution correlation between the temperature and humidity fluctuation trajectories and the abnormal image areas. S3. When the confidence level of the preliminary judgment result is lower than the set threshold, automatically trigger the plant surface sampling device to collect a small amount of tissue fluid for rapid biological detection, and obtain pathological judgment evidence through identifying the pathogen characteristic molecular identifiers in the biological detection. S4. Integrate the preliminary judgment result and the pathological judgment evidence, and perform dynamic weighted calculation by combining the light accumulation amount and the water absorption curve of the plant's historical growth cycle to obtain a comprehensive diagnosis conclusion. S5. Generate operation guidance information according to the comprehensive diagnosis conclusion, and synchronously activate the user operation verification function. Real-time monitor the pesticide spraying path by setting action sensors on the spraying equipment, and judge the matching degree between the actual executed actions and the guidance plan. S6. When there are execution deviations in the operation verification result, freeze the associated data of the current case and activate the expert review channel. At the same time, adaptively adjust the monitoring parameters of other plants in the same area according to the deviation type to form a dynamic protection strategy.

[0025] In one embodiment of the present invention, step S1 includes the following steps: By setting a moving track camera on the top of the greenhouse, continuously collect the original images of the plant leaves. When the pixel density in the intersection area of the leaf edge contours is recognized by the image processing module to reach the preset ratio, trigger the track driving motor to control the camera to move along the horizontal guide rail to the best shooting position, and at the same time start the combination projection of the visible light and infrared band light sources in the surround multi-spectral lighting device to eliminate the shadow interference formed by the overlapping of the leaves, and form enhanced visual data with clear leaf textures and no occlusion. When establishing the mathematical representation of the leaf overlapping area, it is necessary to construct a leaf coverage density judgment formula:

[0026] Wherein, is the leaf overlapping ratio of the current frame image; is the actual pixel area of the th leaf mutual occlusion area, is the single leaf maximum circumscribed rectangle pixel area of all the leaves of the plant in the current camera field of view. The preset ratio is set to , if the leaf overlapping ratio of the current frame image, it is determined to trigger the displacement compensation mechanism of the moving track camera.

[0027] Among them, the multispectral supplementary lighting device includes a visible light array and an infrared light array that can be independently adjusted. For example, the visible light array selects an LED module with a color temperature of 5500K, and the infrared light array selects a tunable semiconductor light source with a wave peak at 760nm - 950nm.

[0028] The mobile rail-mounted camera refers to a visual acquisition unit installed along a straight rail on the top of the greenhouse.

[0029] The preset ratio refers to the ratio threshold of the effective leaf coverage area in the image to the area of the largest circumscribed circle of this area.

[0030] Exemplarily, when the image detection of the plants on the west side of the greenhouse shows that the overlapping area ratio of three adjacent leaves reaches 45%, exceeding the set preset ratio of 40%, the image processing module sends a position adjustment instruction to the mobile rail-mounted camera, calculates the target azimuth angle = π / 4 for the shooting position, the moving speed v of the camera is 0.8m / s, and the 565nm green light band and the 850nm near-infrared band are synchronously turned on for supplementary lighting, so that the lower leaf vein structure originally covered by the upper leaves can be resolved in the enhanced visual data.

[0031] In one embodiment of the present invention, step S2 includes the following steps: Collect the enhanced visual data generated in step S1, synchronously read the timing data recorded by the temperature sensor and the humidity sensor in the environmental monitoring data of the plant area, and create a data fusion matrix for the enhanced visual data and the environmental monitoring data through timestamp alignment. During spatial registration, based on the Z-axis height of the plant canopy coordinate system, calculate the coordinate deviation value between the thermal map of the temperature gradient distribution and the geometric center of the abnormal texture area in the enhanced visual data respectively; The temperature fluctuation trajectory refers to the temperature change curve of the temperature measurement points on the plant canopy recorded in minutes. The humidity fluctuation trajectory refers to the relative humidity timing data obtained by a capacitive humidity sensor suspended in the middle of the plant leaf layer, with a sampling rate of once per minute and data correction using cubic spline interpolation.

[0032] In the time dimension, the sliding window method is used to compare the phase difference between the humidity change rate curve and the area expansion rate of the image abnormal area within the past 72 hours. When the maximum temperature difference of the thermal map exceeds 5 and the absolute value of the phase difference is less than 15 minutes, it is determined that there is a spatio-temporal correlation between the environmental parameter fluctuation and the visual abnormality. According to the mold growth index threshold in the disease type feature library, a preliminary disease feature judgment result including the code of anthracnose or downy mildew is generated. When quantifying the interaction between the environment and visual data, a disease correlation determination formula needs to be established:

[0033] Among them,D represents the mold growth index; , , respectively represent the weight factor of the temperature change rate, the weight factor of the humidity change rate, and the weight factor of the change rate of the abnormal area; t is the derivative of temperature with respect to time (unit / h); is the reference value of the temperature change rate (unit / h); is the derivative of humidity with respect to time (unit / h); is the reference value of the humidity change rate (% / h); is the growth rate of the abnormal area diagnosed by the image (unit cm² / h); is the reference value of the area change rate (unit cm² / h).

[0034] Combined with the test data, the threshold of the mold growth index is set to , , and when it reaches this value, the judgment of the disease characteristics is triggered.

[0035] Among them, the data fusion matrix is a method of integrating enhanced visual data and environmental monitoring data into a unified framework. By aligning the timestamps, the synchronization of different data sources in time is ensured, so as to analyze the information of multiple data sources simultaneously.

[0036] The disease type feature library contains information such as the morphological, physiological, and ecological characteristics of historical diseases. By matching information such as the mold growth index threshold in the disease type feature library, a preliminary judgment of the disease is made.

[0037] Exemplarily, assuming that in the monitoring of zucchini in greenhouse No. GH-05, the enhanced visual data output in step S1 is identified by edge extraction, and the area of the leaf edge browning area expands from 8.3 cm² to 24.7 cm² during the period from 0900 to 1200. The environmental monitoring data in the same period shows that the temperature drops from 30.1 to 24.8 and the humidity rises from 69% to 93%. It is calculated that , , , and combined with the calibration of the test data, the reference value of the temperature change rate is set to , the reference value of the humidity change rate is set to , and the reference value of the area change rate is set to .

[0038] , , They represent that the weight factor of the temperature change rate is 0.35, the weight factor of the humidity change rate is 0.45, and the weight factor of the change rate of the abnormal area is 0.2. Substituting the above parameters into the formula, we get: Combined with the calibration of the test data, the threshold of the mold growth index is 0.19. Obviously, at this time , the system matches the high-humidity disease characteristics in the disease type feature library and generates a disease code of The preliminary judgment result of "suspected period of downy mildew", and the spatial coordinates of the marked associated area are (x = 1.2m, y = 0.8m, z = 0.6m).

[0039] In one embodiment of the present invention, matching the mold growth index threshold according to the disease type feature library and generating a preliminary disease feature judgment result including the code of anthracnose or downy mildew includes the following steps: 1. Construct a disease type feature library; Specifically, the disease type feature library is a database representing the detailed information of disease types including multiple historical records. The detailed information of disease types includes morphological characteristics such as the shape, color, and size of disease spots, physiological characteristics such as the disease development speed and environmental adaptability, ecological characteristics such as the disease occurrence season and geographical distribution, and the corresponding mold growth index threshold.

[0040] 2. Set the mold growth index threshold; Specifically, the mold growth index threshold is set based on experimental data, historical case analysis, or expert experience. These mold growth index thresholds reflect the growth conditions of different disease types under specific environmental conditions. For example, for a certain disease, when the mold growth index reaches or exceeds a certain threshold, it can be considered that the disease is actively developing.

[0041] 3. Extract and match disease characteristics; Specifically, in the enhanced visual data generated in step S1, the characteristic information of the disease area is extracted through image processing technology, and this characteristic information matches the characteristics of a certain disease type in the disease type feature library. Calculate the mold growth index based on the environmental monitoring data, and this index reflects the growth activity of the mold under the current environmental conditions.

[0042] 4. Generate a disease code; Specifically, when the extracted disease characteristics match a certain disease type in the disease type feature library and the mold growth index reaches or exceeds the threshold corresponding to the disease type, a preliminary disease feature judgment result including the code of the disease is generated. The disease code is the unique identifier of the disease type. For example, "anthracnose" corresponds to the code " ", and "downy mildew" corresponds to the code " ".

[0043] 5. Output the preliminary judgment result; Specifically, the output of the preliminary disease characteristic judgment result includes the disease type code and the spatial coordinates of the marked associated area. In the monitoring case of zucchini with greenhouse number GH-05 in step S2, by comparing the disease characteristics in the enhanced visual data with the information in the disease type characteristic library, the characteristics of "downy mildew" were successfully matched. According to the environmental monitoring data, the mold growth index calculated exceeded the threshold corresponding to downy mildew. Generate a preliminary disease characteristic judgment result with the code " ", indicating that the current plant may be infected with downy mildew.

[0044] In one embodiment of the present invention, step S3 includes the following steps: Obtain the preliminary disease characteristic judgment result output in step S2, and calculate the dynamic confidence of the preliminary judgment result according to the mold growth index D value and the case matching degree of the disease type characteristic library. When quantifying the relationship between the confidence and the trigger condition, establish a confidence calculation formula:

[0045] Wherein, C is the dynamic confidence; D is the mold growth index described in step S2; is the maximum mold growth index recorded for the current plant variety at this growth stage; is the number of confirmed cases in the disease type characteristic library with the same code as the preliminary judgment result; represents the total number of investigated cases corresponding to the code.

[0046] The determination threshold set according to different disease types , when the dynamic confidence C is lower than the preset determination threshold , the control unit sends an activation instruction to the plant surface sampling device, drives the microprobe to pierce into the upper epidermis of the leaf to a depth of 50 - 80 um and extracts 2 - 5 uL of tissue fluid, and transports the sample to the detection chamber of the bio-detection chip through capillary action.

[0047] The bio-detection chip integrates an immunochromatographic membrane and a surface plasmon resonance sensor, and simultaneously identifies the characteristic peaks of fungal cell wall -1, 3-glucan and the binding signal of the specific antibody of bacterial flagellin. When the spectral response value of the detection characteristic molecule identifier exceeds the reference value, generate a pathological determination evidence with a unique pathological code.

[0048] Among them, the confidence of the preliminary judgment result refers to the confidence based on the current mold growth index DThe weighted calculation result of the value and the confirmed proportion of similar cases of the plant variety in the disease type feature library, combined with the calibration of the test data, the weight distribution is D The value accounts for 60%, and the historical confirmed rate accounts for 40%; The set judgment threshold refers to the pre-set confidence critical value, which is divided into intervals according to different disease types. For example, the downy mildew type is set at 85% confidence, and the anthracnose type is set at 78% confidence; The plant surface sampling device includes an XYZ three-axis positioning platform and a microprobe with a pressure feedback function. The aperture at the end of the probe is 100um, and the sampling stroke accuracy reaches ±5um; Characteristic molecular identifier. A standard mass spectrometry library containing the mass spectrometry data of characteristic metabolites or structural molecules of various known pathogens is established. For the pathogen-specific metabolites or structural molecules confirmed by comparing with the standard mass spectrometry library, a mass spectrometer is used to perform mass spectrometry analysis on the unknown sample. The mass spectrometer can measure the mass-to-charge ratio (m / z) of the molecules in the sample. By comparing the peaks in the mass spectrum of the unknown sample with the peaks in the standard mass spectrometry library, matching characteristic molecular identifiers can be found. For example, it includes but is not limited to the characteristic mass-to-charge ratio m / z = 249.18 of the Botrydial toxin produced by Botrytis cinerea, and the molecular ion peak m / z = 503.28 of Ralstonin A of Ralstonia solanacearum.

[0049] Exemplarily, when monitoring the tomato plants in greenhouse number GH-07, the step S2 outputs a preliminary judgment result of "suspected gray mold" (code ), and it is calculated that D = 3.8. The current plant is in the fruiting stage and , retrieve the historical case records corresponding to the code in the disease type feature library cases, cases; substitute into the confidence formula to get: That is, 69.7%.

[0050] Combined with the test data, the judgment threshold of gray mold , the dynamic confidence is lower than the preset judgment threshold , it is determined that the dynamic confidence is insufficient, and then the sampling device is activated.

[0051] The XYZ positioning platform moves the probe to the center point coordinates of the lesion (x = 0.45m, y = 1.2m, z = 0.3m), and penetrates the leaf tissue with a contact pressure of 150mN to extract 3.2uL of tissue fluid. The bio-detection chip detects within 120 seconds The characteristic peak and the color development intensity of the immunochromatographic line reach 7.8 kAU, triggering the pathogen determination condition and generating the pathological determination evidence marked as "Botrytis cinerea confirm".

[0052] In one embodiment of the present invention, step S4 includes the following steps: Specifically, the preliminary disease characteristic judgment result output by step S2 and the pathological determination evidence obtained in step S3 are used as inputs, and the light accumulation amount and the water absorption curve in the historical growth data storage of the plant are synchronously called. A basic weight factor is assigned to the preliminary judgment result, and a verification weight factor is assigned to the pathological determination evidence. An environmental correction coefficient is generated based on the daily cumulative photon flux data of the light sensor and the water absorption speed data of the soil humidity sensor.

[0053] The basic weight factor and the verification weight factor are adjusted according to the environmental correction coefficient through a dynamic weighting algorithm. When the daily change standard deviation of the water absorption curve exceeds the preset parameter, the stress resistance compensation factor is activated. Finally, a comprehensive diagnosis conclusion integrating visual diagnosis, biological detection, and growth environment adaptability is output, and a comprehensive diagnosis model is established, satisfying the following formula:

[0054] Wherein, S represents the comprehensive diagnosis score, is the visual diagnosis weight factor, D is the mold growth index generated in step S2, is the biological evidence weight factor, C is the dynamic confidence of the pathological determination evidence in step S3, is the current light accumulation amount, is the average light accumulation amount in the same historical period of this variety, is the slope change amount of the water absorption curve in the most recent 24 hours, is the standard water absorption change rate of this plant variety.

[0055] Set the comprehensive diagnosis score threshold according to different disease types , When it is, a positive diagnosis conclusion is generated.

[0056] Among them, the light accumulation amount refers to the integral value of the number of photons received per second by the plant during the entire growth cycle within the daily cycle, which is recorded by the quantum sensor installed at the top of the plant canopy in as the unit and stored daily by the piecewise trapezoidal integration method.

[0057] The water absorption curve refers to the functional relationship between the soil water potential in the main root region measured by an embedded soil tensiometer and time. It is a continuous curve formed by processing the measured values in kPa through Kalman filtering to reduce noise, reflecting the water absorption of the plant.

[0058] Dynamic weighted calculation means that according to the metabolic characteristics corresponding to the current growth stage of the plant (germination stage, flowering stage or fruiting stage), different variable weight factors are assigned to visual diagnosis, biological evidence, and environmental parameters. For example, the weight distribution in the flowering stage is 30% for visual diagnosis, 45% for biological evidence, and 25% for environmental correction, reflecting the different impacts of different growth stages on the diagnosis results.

[0059] Exemplarily, taking the diagnosis of the eggplant tree plant in greenhouse number GH-09 as an example, receiving the preliminary judgment result of step S2 D = 4.2 (the code is "suspected early blight") and the pathological evidence of step S3 C = 0.92 (the characteristic peak of alternariol toxin m / z = 273.15 is detected). Call the growth data of this plant in the past 30 days, calculate the current light accumulation , and compare it with the historical mean ; Detect the water absorption curve , the standard variability of this variety . When in the flowering stage, take , . Substitute into the formula to get: . Combining experimental data, the determination threshold for early blight is set to = 4.8, When, a positive diagnosis conclusion, "confirmed (moderate infection)" is generated.

[0060] In one embodiment of the present invention, step S5 includes the following steps: The operation guidance engine calls the disease countermeasure database to generate an operation instruction set adapted to the current disease type, infection degree, and plant growth stage. The instruction set includes pesticide mixing ratio parameters, spraying path vector coordinates, and time series control data. Synchronously activate the user operation verification function. Install an action sensor at the root of the spray boom of the backpack spray device. The action sensor combines an accelerometer, a gyroscope, and a magnetometer to simultaneously measure the linear acceleration, angular velocity, and magnetic field direction of an object in three-dimensional space, and real-time collect the three-dimensional space coordinates and attitude angles of the nozzle movement trajectory, and construct a three-dimensional model of the actual spraying path through a data fusion algorithm.

[0061] Among them, the user operation verification function refers to the real-time data interaction mechanism that connects the spray device sensor and the control terminal through the Internet of Things gateway, and uses the MQTT protocol to transmit sensing data at a frequency of 20 frames per second. The MQTT protocol supports the publish / subscribe mode, can achieve efficient and reliable message transmission, and is suitable for low-bandwidth, high-latency or unreliable network environments.

[0062] The data fusion algorithm refers to the algorithm for integrating, analyzing and processing the information of multiple sensors or data sources. In agricultural spray devices, the data fusion algorithm is used to integrate the data collected by the motion sensor to construct a more accurate and complete three-dimensional model of the actual spraying path.

[0063] The operation guidance information refers to the structured instructions including the chemical mixing ratio, action time window, and droplet size requirements. For example, the operation instructions for preventing and controlling gray mold include "40 mL of pyrimethanil suspension / 15 L of water, droplet , operation time 09:00 - 11:00". The droplet size requirement refers to the diameter requirement of the droplets generated by spraying, and the droplet size requirement is one of the important factors affecting the spraying effect.

[0064] The actual path nodes are compared with the target path in the guidance scheme point by point to calculate the path coverage efficiency and the cumulative value of the deviation angle. The path coverage efficiency refers to the efficiency of the actual spraying path covering the target area, which is evaluated by calculating the ratio of the intersection area of the actual spraying area and the target area to the total area of the target area. A higher path coverage efficiency means that the spraying operation covers the target area more accurately. The cumulative value of the deviation angle refers to the cumulative value of the deviation angle between the actual spraying path and the target path. A smaller cumulative value of the deviation angle indicates that the actual spraying path is closer to the target path. When the position deviation of three consecutive trajectory points exceeds the preset tolerance range, it is determined that the execution is abnormal. When quantifying the path execution accuracy, a path matching degree calculation formula is established:

[0065] Among them, M represents the path matching degree; represents the intersection area of the actual spraying area and the target area; represents the total area of the target area; represents the three-dimensional coordinates of the th sampling point of the actual path; represents the coordinates of the corresponding point on the target path; is the maximum allowable offset distance; n is the total number of sampling points; represents the total area of the actual spraying area; Set the path matching degree threshold according to the pesticide spraying path , When, trigger the execution deviation alarm.

[0066] Among them, the path matching degree refers to the matching degree between the actual spraying path and the target path, comprehensively considering the percentage of the coverage area of the actual spraying trajectory on the target grid area in three-dimensional space and the deviation degree of the path geometric center line. A higher path matching degree indicates that the actual spraying path is closer to the target path and the spraying operation quality is higher.

[0067] Exemplarily, for the confirmed case of Botrytis cinerea of strawberries in greenhouse No. GH-12, the operation guidance requirements are to perform spraying operations on the range of 1.2m - 2.4m of the ridge surface according to a zigzag path at an interval of 0.3m. The operator starts the spraying equipment, and the motion sensor continuously uploads spatial coordinate data, detecting that the actual path shows continuous deviation in the section of Y = 0.8m. After calculation, the proportion of the intersection of the actual coverage area and the target area is 89%, and the average offset of the path points is 0.12m. It is set to 0.15m.

[0068] Substituting into the formula gives , and calibrating the path matching degree threshold in combination with experimental data to be 0.85, it can be known that , an audible and visual alarm is immediately issued and the deviation node coordinates (x = 1.8m, y = 0.82m) are marked on the operation interface, and the verification result is recorded as "path matching failure - insufficient area coverage rate".

[0069] In one embodiment of the present invention, the operation guidance engine calls the disease countermeasure database to generate an operation instruction set adapted to the current disease type, infection degree and plant growth stage, including the following steps: 1. Construction of the disease countermeasure database; Specifically, the database needs to contain operation instructions for different disease types, infection degrees and different plant growth stages. These instructions are formulated based on the knowledge and practical experience of agricultural experts and a large amount of field test data. The database structure design should be reasonable, and queries and matches are made according to the disease type, infection degree and plant growth stage.

[0070] 2. Development of the operation guidance engine; Specifically, the operation guidance engine is the core software module, responsible for receiving the diagnostic conclusions output by the comprehensive diagnosis module. The diagnostic conclusions should contain key information such as disease type, infection degree and plant growth stage. The operation guidance engine has a database query function, and according to the received diagnostic conclusions, retrieves the matching operation instructions from the disease countermeasure database. Usually, it involves writing SQL query statements, using the disease type, infection degree and plant growth stage as query conditions to obtain the corresponding operation instruction set from the database.

[0071] 3. Generation of the operation instruction set; Specifically, the operation instructions retrieved from the database contain detailed information such as pesticide formulation parameters, spraying path vector coordinates, time series control data, etc. The operation guidance engine integrates these operation instructions to generate a complete set of operation instructions. The set of operation instructions needs to clearly and accurately guide the spraying equipment to perform operations, ensuring the accuracy and effectiveness of the spraying operations.

[0072] In one embodiment of the present invention, step S6 includes the following steps: When the operation verification result output in step S5 determines an execution deviation, lock the key data chain of the current disease case. This data chain covers all links from diagnosis to operation, including original enhanced visual data, biological detection evidence, operation guidance parameters, path matching degree data, and mark it as frozen case associated data and upload it to the expert review queue. The frozen case associated data contains an unmodifiable data packet of the original data, operation records, and timestamps of the entire diagnosis cycle of a specific plant, aiming to retain the complete diagnosis operation record, and is stored in the isolated storage area using the AES-256 encryption algorithm to prevent data from being tampered with or leaked.

[0073] The AES-256 encryption algorithm is a symmetric encryption algorithm with a 256-bit key length. The AES-256 encryption algorithm is used to encrypt the frozen case associated data to ensure the security of the data during storage and transmission, preventing data from being tampered with or leaked.

[0074] While freezing the case associated data, start the regional adaptive adjustment mode. This mode recalculates the environmental monitoring thresholds and image analysis parameters of other plants in the same region according to the deviation type coefficients defined in the deviation type dictionary. Dynamically adjust the sampling frequency of the mobile orbital camera, the multi-spectral supplementary light intensity, and the biological detection trigger confidence threshold to form a chained dynamic protection strategy for the current abnormal features. The weight of this strategy is inversely proportional to the mold growth index of each plant, that is, the higher the probability of disease occurrence of a plant, the greater the weight of its protection strategy, ensuring that resources can be preferentially allocated to the plants most in need of protection.

[0075] The dynamic protection strategy generation mechanism is based on the spatio-temporal propagation model to predict the disease spread path. By analyzing the spatio-temporal propagation law of the disease, the parameters in stages S1 - S4 are corrected in advance during the monitoring of adjacent plants, so as to formulate a more accurate protection strategy. For example, when the cumulative deviation angle of the spraying path exceeds 20 degrees, the humidity monitoring threshold in the same region is adjusted from 75% to 65% to adapt to the current abnormal features. The spatio-temporal propagation model is a mathematical model used to predict the spatio-temporal propagation law of the disease. By analyzing the historical spread data of the disease and considering the influence of time, space, and environmental factors on the disease spread, it predicts the future spread path of the disease.

[0076] Among them, the expert review queue refers to a workflow system that pushes encrypted frozen case data packets and deviation analysis reports to authorized plant protection expert terminals through RESTful API interfaces. The RESTful API interface is used to build the interface style of network applications, push encrypted frozen case data packets and deviation analysis reports to authorized plant protection expert terminals, and achieve remote data transmission and sharing.

[0077] In one embodiment of the present invention, the deviation type coefficients defined in the deviation type dictionary are used to recalculate the environmental monitoring thresholds and image analysis parameters of other plants in the same area, including: 1. Define the deviation type dictionary; Specifically, the deviation type dictionary is a mechanism or data structure for defining different deviation types and their corresponding deviation type coefficients. For different deviation types such as insufficient chemical coverage, insufficient light intensity, and abnormal temperature, a unique code and its corresponding deviation type coefficient are assigned to each deviation type. For example, the deviation type code Ty = 3 represents insufficient chemical coverage, and the deviation type coefficient k takes 1.0. The deviation type coefficient is defined based on the experience of experts.

[0078] 2. Execute the deviation determination condition; Specifically, executing the deviation determination condition refers to an event trigger mechanism in which the output value of the path matching degree calculation formula in step S5 is lower than the set threshold three times in a row. This mechanism can ensure that execution deviations are captured in a timely manner and corresponding countermeasures are taken.

[0079] 3. Dynamic protection strategy generation mechanism; Specifically, based on the spatio-temporal propagation model to predict the disease spread path, by analyzing the spatio-temporal propagation law of the disease, the areas and plants that the disease may spread to can be predicted, so as to correct the parameters in advance and formulate a more accurate protection strategy. The spatio-temporal propagation model analyzes the historical spread data of the disease, and considers time, space, environmental factors, and the disease spread path as the training set. The spatio-temporal propagation model is trained with the training set, and time, space, and environmental factors are used as inputs to predict the future spread path of the disease.

[0080] 4. Quantify the adjustment range of the protection strategy; It can be seen that when quantifying the adjustment range of the protection strategy, a parameter correction formula is established:

[0081] Among them, represents the adjustment amount of the monitoring parameters of the target plant; k is the deviation type coefficient. For example, combined with expert experimental data, the pesticide residue deviation takes 1.2, and the path deviation takes 0.8; represents the sum of the mold growth indices of other plants in the same area; It represents the mold growth index of the plant before step S2; M is the calculation result of the path matching degree in step S5.

[0082] Exemplarily, for the prevention and control of downy mildew in greenhouse No. GH-15, when the spraying path matching degree M = 0.76 (set threshold 0.85) is detected in step S5, the execution deviation code = 2 is triggered, and the case data packet (including 12 groups of enhanced images, 3 frames of biological detection spectra, and 217 points of path coordinates) is frozen. At the same time, the disease probability indices of 6 tomato plants in the same area are obtained as . Calculate the total disease probability sum , and select k = 1.0.

[0083] When adjusting the parameters of the plant, . Automatically adjust the overlapping determination threshold in step S1 from 40% to , and lower the biological detection confidence threshold in S3 from 75% to . Synchronously increase the camera sampling frequency in this area from 5 seconds / time to 3 seconds / time. After receiving the data packet through the review channel, the plant pathology expert confirms that the actual deviation reason is that the operator did not calibrate the equipment according to the coordinate spacing. Based on this, the system updates the learning parameters in the deviation type dictionary .

[0084] See the appendix Figure 2 . The present invention also proposes a remote disease collaborative diagnosis system for greenhouse plants based on a cloud platform, including the following modules: An image acquisition module for generating enhanced visual data through a mobile orbital camera and a multispectral supplementary lighting device; An environmental data processing module for fusing the enhanced visual data with the temperature and humidity fluctuation trajectories to generate a preliminary disease feature judgment result; A biological detection module, when the confidence level of the preliminary judgment result is lower than the set threshold, automatically triggers a plant surface sampling device to collect trace tissue fluid for rapid biological detection, and the biological detection obtains pathological judgment evidence by identifying the pathogen characteristic molecular identifier; A dynamic decision-making module for fusing the preliminary judgment result and the pathological judgment evidence, and performing dynamic weighted calculation in combination with the light accumulation amount and water absorption curve of the historical growth cycle of the plant to obtain a comprehensive diagnosis conclusion; An execution verification module for generating operation guidance information based on the comprehensive diagnosis conclusion, and synchronously starting the user operation verification function, and judging the matching degree of the actual execution action and the guidance plan by setting an action sensor on the spraying device to monitor the pesticide spraying path in real time; A safety response module, which is used to freeze the associated data of the current case and activate the expert review channel when there is an execution deviation in the operation verification result, and at the same time, adaptively adjust the monitoring parameters of other plants in the same region according to the deviation type to form a dynamic protection strategy.

[0085] It should be noted that for the formulas mentioned above, through the principle of dimensional consistency and mathematical standardization means (such as normalization processing, dimensionless parameter conversion, or unit system unification), physical quantities with different attributes can be translated into dimensionless standard values or superposable parameters with the same dimension, so as to eliminate the interference of different dimensions on the operation logic, and make the formulas have mathematical operation rationality and objective law adaptability while retaining the original data distribution characteristics. The above is only an exemplary embodiment of the present invention, and the scope of the present invention cannot be limited thereby.

[0086] Each of the above-mentioned modules can be implemented in whole or in part by software, hardware, and their combination, support being embedded in the processor of the computer device in hardware form or being independent of it, and at the same time, also support being stored in the memory of the computer device in software form, so as to facilitate the processor to call and execute the operations corresponding to each of the above-mentioned modules.

[0087] It should be noted that the human body information (including but not limited to human body device information and personal information, etc.) and data (including but not limited to data for analysis, stored data, and displayed data, etc.) involved in the present invention are all information and data authorized by the human body or fully authorized by all parties. The collection, use, and processing of relevant data require relevant legal standards.

[0088] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A cloud platform-based collaborative remote disease diagnosis method for greenhouse plants, characterized in that: The following steps are involved: S1. Acquire real-time image data of greenhouse plants. When it is detected that the overlap of leaves exceeds a preset ratio, adjust the shooting angle by moving the track camera and start the multi-spectral fill light device to eliminate image occlusion interference to form enhanced visual data; S2. Combine the enhanced visual data with the environmental monitoring data of the area where the plant is located, and generate preliminary disease feature judgment results by analyzing the correlation between the temperature and humidity fluctuation trajectory and the spatiotemporal distribution of the abnormal image area; S3. When the confidence level of the preliminary judgment result is lower than the set threshold, the plant surface sampling device is automatically triggered to collect a trace amount of tissue fluid for rapid biological detection, and the biological detection obtains pathological judgment evidence by identifying the characteristic molecular identifier of the pathogen; S4. Integrate the preliminary diagnosis results with the pathological evidence, combine the light accumulation and water absorption curve of the plant's historical growth cycle to perform dynamic weighted calculations, and draw a comprehensive diagnosis conclusion; S5. Generate operation guidance information based on the comprehensive diagnosis conclusion, and simultaneously start the user operation verification function, by setting the action sensor on the spray equipment to monitor the pesticide spraying path in real time, and judge the matching degree between the actual execution action and the guidance plan; S6. When there is an execution deviation in the operation verification result, freeze the current case-related data and activate the expert review channel. At the same time, adaptively adjust the monitoring parameters of other plants in the same area according to the deviation type to form a dynamic protection strategy.

2. The cloud platform-based greenhouse plant remote disease collaborative diagnosis method according to claim 1, characterized in that: Forming enhanced visual data includes the following steps: The images of plant leaves are continuously collected by a mobile track camera. When the image processing module detects that the pixel density of the intersection area of ​​the leaf edge contour exceeds a preset ratio, the optimal azimuth angle and moving speed of the camera are calculated; Start the combination of visible light and infrared light sources in the preset multi-spectral supplementary lighting device to eliminate the shadow interference in the overlapping area of ​​the leaves; The real-time image data is subjected to multi-spectral correction according to a leaf cover density determination formula to generate enhanced visual data containing clear leaf vein textures.

3. The cloud platform-based greenhouse plant remote disease collaborative diagnosis method according to claim 2, characterized in that: Generating preliminary disease characteristic judgment results includes the following steps: Extracting the area expansion rate of the abnormal texture region in the enhanced visual data, and performing spatial registration with the temperature gradient distribution thermodynamic map in the environmental monitoring data; The sliding window method is used to calculate the weighted sum of the temperature change rate, humidity change rate and abnormal area growth rate to generate the mold growth index; When the mold growth index exceeds a preset threshold in the disease type feature library, the disease code is matched and the spatial coordinates of the associated area are marked.

4. The cloud platform-based greenhouse plant remote disease collaborative diagnosis method according to claim 3, characterized in that: Dynamic weighted calculation includes the following steps: Allocating a basic weight factor to the preliminary disease feature judgment result, and allocating a verification weight factor to the pathological judgment evidence; Generate an environmental correction coefficient according to the light accumulation amount and the historical average value and the change amount and the standard change rate of the slope of the water absorption curve; The ratio of the basic weight factor to the verification weight factor is adjusted through a dynamic weighted formula, and a comprehensive diagnostic conclusion is output in combination with the stress resistance compensation factor.

5. The cloud platform-based greenhouse plant remote disease collaborative diagnosis method according to claim 1, characterized in that: The operation verification result includes the following steps: Comparing the target path vector coordinates in the operation guidance information with the actual spraying path acquired by the motion sensor point by point; Calculate the intersection percentage of the actual spray path coverage area and the deviation of the geometric center line, and quantify the execution deviation through the path matching formula; When the path matching degree is lower than a preset threshold, a deviation alarm is triggered and the three-dimensional coordinates of the deviation node are recorded.

6. The cloud platform-based greenhouse plant remote disease collaborative diagnosis method according to claim 5, characterized in that: The path matching formula is: in, Indicates the path matching degree; Indicates the intersection area of ​​the actual spraying area and the target area; Indicates the total area of ​​the target area; Represents the three-dimensional coordinates of the i-th sampling point on the actual path; Indicates the coordinates of the corresponding points on the target path; is the maximum allowable offset distance; n is the total number of sampling points; Indicates the total area of ​​the actual spraying area; Set path matching threshold according to pesticide spraying path , , an execution deviation alarm is triggered.

7. The cloud platform-based greenhouse plant remote disease collaborative diagnosis method according to claim 1, characterized in that: The formation of a dynamic protection strategy includes the following steps: Calculate the monitoring parameter adjustment amount of the target plant based on the deviation type coefficient and the sum of the mold growth index of the plants in the same area; Reverse weight distribution is performed on the sampling frequency, multi-spectral fill light intensity and biological detection trigger threshold of the mobile track camera; The disease spread path is predicted based on the spatiotemporal propagation model, and the image analysis parameters and environmental control thresholds of adjacent plants are corrected in advance.

8. The cloud platform-based greenhouse plant remote disease collaborative diagnosis method according to claim 7, characterized in that: Freeze the current case-related data, including the following steps: Encapsulate the enhanced visual data of the entire diagnostic cycle, the biological detection spectral response value and the operation guidance parameters into an unmodifiable data package; The data packet is encrypted using the AES-256 encryption algorithm and then stored in an isolated storage area; Push encrypted data packets and deviation analysis reports to expert terminals through the RESTful API interface.

9. The cloud platform-based greenhouse plant remote disease collaborative diagnosis method according to claim 8, characterized in that: Calculating the monitoring parameter adjustment amount of the target plant includes the following steps: It can be known that when the adjustment range of the quantitative protection strategy is determined, the parameter correction formula is established: in, Indicates the monitoring parameter adjustment amount of the target plant; k is the deviation type coefficient. For example, based on the expert experimental data, the drug residue deviation is 1.2 and the path deviation is 0.8; It represents the sum of the fungal growth index of other plants in the same area; represents the fungal growth index of the plant before step S2 is generated; M is the path matching degree calculation result of step S5.

10. A cloud-based greenhouse plant remote disease collaborative diagnosis system, characterized in that: Includes the following modules: An image acquisition module for generating enhanced visual data through a mobile track-mounted camera and a multi-spectral fill light device; An environmental data processing module, used to fuse the enhanced visual data with temperature and humidity fluctuation trajectories to generate preliminary disease feature judgment results; The biological detection module automatically triggers the plant surface sampling device to collect trace tissue fluid for rapid biological detection when the confidence level of the preliminary judgment result is lower than the set threshold. The biological detection obtains pathological judgment evidence by identifying the characteristic molecular identifier of the pathogen; The dynamic decision-making module is used to integrate the preliminary judgment results with the pathological judgment evidence, and to perform dynamic weighted calculations based on the light accumulation and water absorption curve of the plant's historical growth cycle to draw a comprehensive diagnosis conclusion; The execution verification module is used to generate operation guidance information based on the comprehensive diagnosis conclusions and simultaneously start the user operation verification function. By setting motion sensors on the spray equipment to monitor the pesticide spraying path in real time, it determines the matching degree between the actual execution action and the guidance plan; The safety response module is used to freeze the current case-related data and activate the expert review channel when there is an execution deviation in the operation verification results. At the same time, it adaptively adjusts the monitoring parameters of other plants in the same area according to the deviation type to form a dynamic protection strategy.

Images