Intelligent prevention and control method, medium and equipment for crop diseases and insect pests

Abstract

According to the intelligent prevention and control method for crop diseases and insect pests, the medium and the equipment, the current terrain sensing data of the robot is obtained to adjust the pose of the robot, the current crop image is collected, the corresponding environment data is collected according to the current crop image, the image-environment data pair is obtained, and the early warning value of the image-environment data pair is obtained; and judging whether the result is greater than an early warning threshold, if not, rejecting the corresponding data pair, if so, obtaining an optimized data pair, and obtaining the disease cause type and severity of the corresponding crops according to the optimized data pair so as to adopt a corresponding prior prevention and control strategy. The closed-loop problem of'complex farmland all-terrain maneuvering inspection-local real-time fusion of multi-modal data-combined reasoning of plant diseases and insect pests and environmental inducements-automatic generation of accurate prevention and control decisions' cannot be simultaneously solved by using one set of system in the prior art is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural intelligent equipment, in particular to a crop disease and pest intelligent prevention and control method, medium and equipment. BACKGROUND

[0002] Crop disease and pest are the primary biological stress affecting grain quality and yield. According to the National Agricultural Technology Center, the occurrence area of major crop diseases and pests in China will reach 168 million hm2 per time in 2025, with a year-on-year growth of 6.2%. The prevention and control situation is severe. Traditional methods mainly rely on manual field inspection or fixed cameras, which have defects such as low efficiency, high false positives, and limited coverage. Although the statistical-meteorological coupling model that has appeared in recent years reduces subjective factors, it still requires a large amount of manpower and lacks automation. With the maturity of deep learning and multi-modal large models (such as Qwen, ERNIE-ViL, etc.), the use of image, text, and environmental sensor data for intelligent identification of diseases and pests has become a research hotspot. However, existing technologies still have obvious shortcomings in actual farmland scenarios: 1. Fixed monitoring nodes: limited by terrain, they cannot cover terraced fields and slopes, and static perspectives easily miss crop root diseases; 2. Unmanned aerial vehicle inspection: short endurance, inability to continuously monitor environmental parameters such as temperature, humidity, and soil EC value, insufficient image resolution due to high flight altitude, and high pest detection omission rate due to low flight altitude and noise interference; 3. Traditional agricultural robots: poor obstacle crossing ability due to wheel or track structure, difficulty in crossing ditches and trenches, and output of only disease and pest categories without considering environmental causes, leading to decision bias. Although existing patents (such as CN115062978A and CN117789344A) have made progress in unmanned aerial vehicle algorithm optimization or agricultural question and answer systems, they have not formed a complete closed loop of "full-terrain dynamic inspection - edge local dual-model collaboration - multi-modal data real-time fusion - automatic generation of prevention and control decisions", which is difficult to meet the high-precision, low-latency, and all-weather prevention and control needs in complex farmland scenarios. Wheeled-legged robots have the advantages of high speed and obstacle crossing of wheeled robots, and have been verified in industrial, emergency, and medical scenarios for their mobility and stability. However, they have not been systematically applied to crop disease and pest intelligent prevention and control. Therefore, there is an urgent need for a new method that integrates the adaptive motion ability of wheeled-legged robots and the deep analysis ability of multi-modal AI to break through the real-time and accurate decision-making bottleneck in complex farmland environments. Therefore, how to use a system to simultaneously solve the closed-loop problem of "complex farmland full-terrain mobile inspection - multi-modal data local real-time fusion - joint reasoning of diseases and pests and environmental causes - automatic generation of accurate prevention and control decisions" is a technical problem that needs to be solved in this field. SUMMARY

[0003] Based on this, the present application aims to provide a crop disease and pest intelligent prevention and control method, medium and equipment to solve at least one technical problem mentioned in the background art.

[0004] In a first aspect, the present application provides a crop disease and pest intelligent prevention and control method, comprising: Obtaining current terrain sensing data of the robot to adjust the pose of the robot and collect current crop images; According to the current crop image collection, the corresponding environment data is obtained, and the image-environment data pair is obtained; Obtaining the warning value of the image-environment data pair, judging whether it is greater than the warning threshold, if not, eliminating the corresponding data pair, if yes, obtaining the optimized data pair; According to the optimized data pair, the cause type and severity of the corresponding crop are obtained to take the corresponding prior prevention and control strategy.

[0005] Further, the step of obtaining current terrain sensing data of the robot to adjust the pose of the robot and collect current crop images comprises: Obtaining current terrain sensing data of the robot; The terrain sensing data includes any one or more of IMU angular velocity, motor current and slope sensor data; According to the current terrain sensing data, the pitch angle compensation amount of the robot is calculated; According to the pitch angle compensation amount, the pose of the robot is adjusted to keep the pitch angle of the collection device, and the current crop image is collected.

[0006] Further, the step of obtaining the warning value of the image-environment data pair comprises: Alert = a * Conf + b * Env_Dev Env_Dev = Σ | (measured value - reference value) / s | Wherein, Alert is the warning value, Conf is the image confidence, Env_Dev is the environmental deviation, a and b are the set weight coefficients, and s is the standard deviation.

[0007] Further, the step of obtaining the cause type of the corresponding crop according to the optimized data pair comprises: Extracting the image features and environment data in the optimized data pair to obtain the image feature vector and the environment feature vector; Weighted fusion is performed on the image feature vector and the environment feature vector to obtain the initial probability of each cause; According to the initial probability of each cause, the prior knowledge base is matched to obtain a cause hypothesis set; According to the environmental constraint condition, the initial probability is corrected to obtain the final probability of each cause in the cause hypothesis set; The cause corresponding to the highest probability is the cause type of the current crop.

[0008] Further, the step of weighting and fusing the image feature vector and the environment feature vector to obtain the initial probability of each cause includes: P = α · F_img + β · F_env Wherein, α and β are set weight coefficients, F_img is the image feature vector, F_env is the environment feature vector, and P is the initial probability of each cause.

[0009] Further, the step of obtaining the cause type and severity of the current crop according to the optimization data to take the corresponding prior prevention and control strategy includes: Obtaining the severity level according to the image feature to obtain the lesion area ratio and the lesion area change rate; Matching the prior prevention and control strategy library according to the cause, the severity level and the environment data of the current crop to obtain the corresponding prevention and control strategy; Sending the corresponding prevention and control strategy to the robot to make the robot execute the prevention and control strategy.

[0010] In a second aspect, the present application further provides a computer storage medium, which stores executable program code; the executable program code is used for executing the crop disease and pest intelligent prevention and control method in any one of the first aspect.

[0011] In a third aspect, the present application further provides a terminal device, which comprises a memory and a processor; the memory stores program code executable by the processor; the program code is used for executing the crop disease and pest intelligent prevention and control method in any one of the first aspect.

[0012] The application provides a crop disease and pest intelligent prevention and control method, medium and equipment, which adjusts the pose of a robot by acquiring current terrain sensing data of the robot, collects current crop images, reduces image blur caused by bumping, ensures the accuracy of subsequent image analysis (such as leaf texture recognition), then collects corresponding environment data according to the current crop image, obtains an image-environment data pair, binds the image (visual feature) with the environment data (such as temperature and humidity, illumination, and soil pH), solves the problem of "different appearances of the same disease" (for example, yellow leaves caused by lack of nitrogen may be misjudged as diseases), and the data pair contains a space-time label, which is convenient for subsequent traceability analysis, then acquires a warning value of the image-environment data pair, judges whether the warning value is greater than a warning threshold, if not, the corresponding data pair is removed, if yes, an optimized data pair is obtained, low-risk data (such as healthy leaves) is removed through threshold filtering, meaningless model reasoning is avoided, edge computing delay is reduced, and finally the cause type and severity of the corresponding crop are obtained according to the optimized data pair, so that corresponding prior prevention and control strategies are taken, the cause type (such as fungal / bacterial / physiological diseases) and the severity (such as the area ratio of disease spots) are combined, low-toxicity pesticides or biological control can be selected, and chemical agent abuse is reduced. The application solves the closed-loop problem that the prior art cannot simultaneously solve the "complex farmland full-terrain mobile inspection - multi-modal data local real-time fusion - disease and pest and environment cause joint reasoning - automatic generation of precise prevention and control decision" closed loop problem. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 A flowchart of the crop disease and pest intelligent prevention and control method of the embodiments of the application. DETAILED DESCRIPTION

[0014] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0015] It should be noted that if the embodiments of the present application involve directional indications, such as up, down, left, right, front, back, etc., the directional indications are only used to explain the relative position relationship, motion condition, etc. between the components in a certain posture, and if the certain posture changes, the directional indications also change accordingly. In addition, if the embodiments of the present application involve descriptions such as "first, second", "S1, S2", "step one, step two" and the like, such descriptions are only for description purposes and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of technical features indicated or indicating the execution order of the method, etc. Any person skilled in the art can understand that within the technical concept of the present application, as long as it does not deviate from the essential points of the present application, it should be included in the protection scope of the present application.

[0016] As shown in Figure 1 The present application provides an intelligent crop disease and pest control method: S1: acquiring current terrain sensing data of the robot to adjust the pose of the robot and collect current crop images; Specifically, the robot can optionally but not limited to calculate the attitude error of itself on the ridge, slope or wet and soft ground in real time through the on-board IMU, motor current sensor and slope meter, and actively adjust the suspension of the chassis and the angle of the gimbal, so that the camera always maintains the best imaging pose of "facing the crop canopy, pitch angle ± 2°, constant height from the ground", and then triggers the shutter to complete the non-shaking shooting to collect the current crop images.

[0017] Preferably, the step of acquiring current terrain sensing data of the robot to adjust the pose of the robot and collect current crop images can optionally include: S11: acquiring current terrain sensing data of the robot; the terrain sensing data includes any one or more of IMU angular velocity, motor current and slope sensor data; S12: calculating the pitch angle compensation amount of the robot according to the current terrain sensing data; S13: adjusting the pose of the robot according to the pitch angle compensation amount to maintain the pitch angle of the collection device and collecting the current crop images.

[0018] Specifically, the terrain sensing data such as MU angular velocity, motor current and slope sensor data can be optionally but not limited to acquired, and the pitch angle compensation amount of the robot is calculated according to the current terrain sensing data, and then the pose of the robot is adjusted according to the pitch angle compensation amount to maintain the pitch angle of the collection device and collect the current crop images. By quantifying the terrain (slope, bump, roll) where the robot is located in real time, multi-source fusion (IMU, motor current, slope sensor) is realized, the system robustness is improved, it is not sensitive to the failure of a single sensor, and the data frequency is high, which meets the dynamic compensation demand.

[0019] S2: Collect corresponding environmental data based on the current crop image to obtain image-environment data pairs; Specifically, it is optional but not limited to latching the latest values ​​of the current environmental sensors (temperature and humidity, light, soil moisture, CO2) through hardware interrupts at the same time as the camera shutter is triggered, and adding μS-level timestamps to generate a complete data pair of "one image + multi-dimensional environment" to avoid data misalignment due to network delays or clock drift.

[0020] S3: Obtain the warning value of the image-environment data pair and determine whether it is greater than the warning threshold. If not, eliminate the corresponding image-environment data pair. If so, obtain the optimized image-environment data pair. Specifically, it is optional but not limited to obtaining the warning value of the image-environment data pair according to formula 3-1 and 3-2: Alert=α*Conf +β*Env_Dev 3-1 Env_Dev = Σ|(measured value - reference value) / σ| 3-2 Among them, Alert is the warning value, Conf is the image confidence, Env_Dev is the environmental deviation, α and β are the set weight coefficients, and σ is the standard deviation.

[0021] Therefore, each data pair can be screened according to its warning value and redundant data can be eliminated to reduce the processing volume of subsequent steps and improve efficiency.

[0022] S4: Obtain the disease cause type and severity of the corresponding crops based on the optimized image-environmental data pair, so as to take corresponding prior measures.

[0023] Specifically, it is optional but not limited to performing lightweight edge inference on the optimized image-environmental data pairs to obtain the disease type and severity, and matching the "pesticide, dosage, and operating parameters" three-in-one prevention and control strategy from the prior knowledge base, and sending it to the actuator via the CAN bus to complete precise spraying or fertilizing.

[0024] Preferably, the step of obtaining the pathogenic type of the corresponding crop according to the optimized image-environment data pair may optionally include: S41: extracting image features and environmental data from the optimized data pair to obtain an image feature vector and an environmental feature vector; S42: Perform weighted fusion of the image feature vector and the environmental feature vector to obtain the initial probability of each cause; Specifically, the image feature vector and the environment feature vector may be weightedly fused according to Formula 4-1: P = α·F_img + β·F_env 4-1 Wherein, a, β are set weight coefficients, F img is an image feature vector, F env is an environment feature vector, and P is an initial probability of each cause.

[0025] By converting image pixels and environment sensor data into computable vectors, texture, color, shape features in the image, temperature, humidity, light, soil EC value and other features in the environment are extracted, and the two are weighted and fused to obtain the initial probability of each cause.

[0026] S43: According to the initial probability of each cause, the prior knowledge base is matched to obtain a cause hypothesis set; S44: According to the environmental constraint condition, the initial probability is corrected to obtain the final probability of each cause in the cause hypothesis set; S45: The cause corresponding to the highest probability is the cause type of the current crop.

[0027] Specifically, but not limited to, according to the initial probability of each cause, the prior knowledge base is matched to obtain a number of suspected causes, which is a cause hypothesis set, and then further screening is performed to obtain the final determined cause type of the current crop. While improving efficiency, the classification accuracy is also improved.

[0028] Further preferably, according to the optimization of image-environment data pairs, the cause type and severity of the corresponding crop are obtained to take corresponding prior measures, which can include: S46: According to the image feature, the lesion area ratio and lesion area change rate are obtained to obtain the severity level; S47: According to the cause, severity level and environment data of the current crop, the prior prevention and control strategy library is matched to obtain the corresponding prevention and control strategy; S48: The corresponding prevention and control strategy is sent to the robot, so that the robot executes the prevention and control strategy.

[0029] Specifically, but not limited to, the strategy library is configured freely according to regional regulations (banned or limited use of pesticides), organic / conventional planting mode, and then the lesion area ratio and change rate are used to distinguish acute spread vs. chronic local, guide pesticide dosage, and then "cause + severity + environment" is translated into an executable prescription to find a matching corresponding prevention and control strategy in the strategy library. Finally, the corresponding prevention and control strategy is sent to the robot, so that the robot executes the prevention and control strategy, and completes the intelligent prevention and control of crop diseases and pests.

[0030] In this embodiment, an intelligent crop disease and pest prevention and control method of the present application is given. The current terrain sensing data of the robot is acquired to adjust the pose of the robot, collect the current crop image, reduce the image blur caused by bumping, and ensure the accuracy of subsequent image analysis (such as leaf texture recognition). Then, the corresponding environment data is collected according to the current crop image, the image-environment data pair is obtained, the image (visual feature) is bound with the environment data (such as temperature and humidity, light, soil pH), the problem of "different images for the same disease" (such as yellow leaves caused by lack of nitrogen which may be misjudged as disease) is solved, and the data pair contains space-time tags for subsequent traceability analysis. Then, the warning value of the image-environment data pair is obtained, and it is judged whether it is greater than the warning threshold. If not, the corresponding data pair is removed. If yes, the optimized data pair is obtained. The low-risk data (such as healthy leaves) is removed through threshold filtering to avoid meaningless model reasoning and reduce edge computing delay. Finally, the cause type and severity of the corresponding crop are obtained according to the optimized data pair, so as to take corresponding prior prevention and control strategies. According to the cause type (such as fungus / bacteria / physiological disease) and severity (such as disease spot area ratio), low-toxicity pesticides or biological control can be selected, and the abuse of chemical agents is reduced. The closed-loop problem that the prior art cannot simultaneously solve "complex farmland full-terrain mobile inspection - multi-modal data local real-time fusion - disease and pest and environmental cause joint reasoning - automatic generation of precise prevention and control decision" is solved.

[0031] In another aspect, the present application also provides a computer storage medium storing executable program codes; the executable program codes are used to execute any of the above crop disease and pest intelligent prevention and control methods.

[0032] In another aspect, the present application also provides a terminal device comprising a memory and a processor; the memory stores program codes executable by the processor; the program codes are used to execute any of the above crop disease and pest intelligent prevention and control methods.

[0033] For example, the program codes can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the program codes in the terminal device.

[0034] The terminal device can be a desktop computer, a notebook computer, a palm computer, a cloud server and other computing devices. The terminal device can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the terminal device can also include input / output devices, network access devices, buses, etc.

[0035] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0036] The memory can be an internal storage unit of the terminal device, such as a hard disk or a memory. The memory can also be an external storage device of the terminal device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the memory can also include both the internal storage unit and the external storage device of the terminal device. The memory is used to store the program code and other programs and data required by the terminal device. The memory can also be used to temporarily store data that has been output or will be output.

[0037] The computer storage medium and the terminal device described above are created based on the crop pest and disease intelligent prevention and control method, and the technical effects and advantages thereof will not be repeated here. The technical features of the above-described embodiments can be combined in any manner. To make the description concise, all possible combinations of the technical features in the above-described embodiments are not described, however, as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the present disclosure.

[0038] The above-described embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be noted that for ordinary skilled in the art, without departing from the concept of the present application, a number of variations and improvements can be made, which are within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. A method for intelligent prevention and control of crop diseases and insect pests, characterized in that: include: Obtain the robot's current terrain sensor data to adjust the robot's posture and collect current crop images; Collect corresponding environmental data based on the current crop image to obtain image-environment data pairs; Obtain the warning value of the image-environment data pair and determine whether it is greater than the warning threshold. If not, the corresponding data pair is eliminated. If so, the optimized data pair is obtained. Based on the optimized data, the disease cause type and severity of the corresponding crops are obtained to adopt corresponding prior prevention and control strategies.

2. The method according to claim 1, characterized in that The steps for obtaining the robot's current terrain sensor data, adjusting the robot's posture, and collecting the current crop image include: Obtain the robot's current terrain sensor data; terrain sensor data includes any one or more of IMU angular velocity, motor current, and slope sensor data; Calculate the robot's pitch angle compensation based on the current terrain sensor data; The robot's posture is adjusted according to the pitch angle compensation to maintain the pitch angle of the acquisition device and collect the current crop image.

3. The method according to claim 1, characterized in that The steps of obtaining the warning value of the image-environment data pair include: Alert=α*Conf +β*Env_Dev Env_Dev = Σ|(measured value - reference value) / σ| Among them, Alert is the warning value, Conf is the image confidence, Env_Dev is the environmental deviation, α and β are the set weight coefficients, and σ is the standard deviation.

4. The method according to claim 1, wherein The steps of obtaining the pathogen type of the corresponding crop according to the optimized data include: Extracting image features and environmental data from the optimized data pair to obtain image feature vectors and environmental feature vectors; Perform weighted fusion on the image feature vector and the environmental feature vector to obtain the initial probability of each cause of disease; Match the prior knowledge base according to the initial probability of each cause to obtain a set of cause hypotheses; Modify the initial probability according to the environmental constraints to obtain the final probability of each cause in the cause hypothesis set; The disease cause with the highest probability is the disease cause type of the current crop.

5. The method according to claim 4, characterized in that The steps of weighted fusion of the image feature vector and the environment feature vector to obtain the initial probability of each cause of disease include: P = α·F_img + β·F_env Among them, α and β are the set weight coefficients, F_img is the image feature vector, F_env is the environment feature vector, and P is the initial probability of each cause.

6. The method according to claim 5, characterized in that The steps to obtain the type and severity of the disease cause of the corresponding crop based on the optimized data and to adopt the corresponding a priori prevention and control strategy include: The lesion area ratio and lesion area change rate are obtained based on image features to obtain the severity level; Match the a priori control strategy library according to the current crop disease causes, severity levels, and environmental data to obtain the corresponding control strategy; Send the corresponding prevention and control strategy to the robot so that the robot can execute the prevention and control strategy.

7. A computer storage medium, characterized in that Executable program code is stored; the executable program code is used to execute the intelligent control method for crop diseases and insect pests according to any one of claims 1 to 6.

8. A terminal device, characterized in that: It comprises a memory and a processor; the memory stores program code that can be executed by the processor; the program code is used to execute the intelligent control method for crop diseases and insect pests according to any one of claims 1 to 6.

Images