A system for precisely spraying pesticides and controlling pests and diseases with a drone

Abstract

The present invention discloses a system for precisely spraying and controlling pests and diseases by an unmanned aerial vehicle, which includes an unmanned aerial vehicle, a spraying module, a sensor module, a data processing module and a control module. The spraying module, the sensor module, the data processing module and the control module are all installed on the unmanned aerial vehicle. The sensor module is electrically connected to the data processing module and transmits the collected data to the data processing module. The data processing module is electrically connected to the control module, and the data processing module issues a control strategy to the control module. The control module is electrically connected to the unmanned aerial vehicle and the spraying module to execute the control strategy. Through this integrated unmanned aerial vehicle system, precise identification and control of pests and diseases can be achieved, reducing the use of pesticides, lowering environmental pollution, and at the same time improving the yield and quality of crops.

Description

Technical Field

[0001] The present invention relates to the technical field of agriculture, and particularly to a system for precise spraying of drones to control pests and diseases. Background Art

[0002] With the rapid progress of drone technology, drones are increasingly widely used in the agricultural field, especially in pest and disease control. Precise spraying by drones can effectively improve the utilization rate of pesticides, reduce environmental pollution, and enhance agricultural production efficiency. However, existing drone spraying technologies still have some deficiencies in terms of spraying accuracy, intelligence level, and adaptability, which affect their actual application effects.

[0003] After retrieval, a spraying device for a pesticide spraying operation drone with a publication number of CN113443143B was disclosed on December 22, 2023. This patent can change the angle of the adjustment device by setting a second motor, a second rotating rod, and a rotating plate in cooperation, thereby changing the shape of the liquid medicine ejected. However, this technical solution mainly focuses on the adjustment of the nozzle angle, lacks real-time data collection and intelligent control during the spraying process, and cannot be dynamically adjusted according to the actual environment and crop conditions, which may lead to uneven spraying and resource waste.

[0004] After retrieval, a spraying device for a pesticide spraying operation drone with a publication number of CN113306723B was disclosed on January 10, 2023. This patent effectively avoids spraying the liquid medicine onto the drone body when the spraying assembly sprays the liquid medicine through the cooperation of a telescopic adjustment frame and support feet, avoids waste of the liquid medicine and pollution of the drone body, and has a good shock absorption effect. However, this technical solution mainly focuses on the improvement of the mechanical structure, lacks real-time data collection and intelligent control during the spraying process, cannot achieve precise spraying, and has poor adaptability in complex environments, making it difficult to meet the control requirements of different crops and different pests and diseases.

[0005] The above problems indicate that existing pesticide spraying drones still have certain deficiencies in terms of spraying accuracy, intelligence level, and adaptability. Summary of the Invention

[0006] The purpose of the present invention is to provide a system for precise spraying of drones to control pests and diseases to solve the above problems.

[0007] The present invention is achieved through the following technical solutions:

[0008] A system for precise spraying of drones to control pests and diseases includes a drone, a spraying module, a sensor module, a data processing module, and a control module. The spraying module, sensor module, data processing module, and control module are all installed on the drone;

[0009] The sensor module is electrically connected to the data processing module, and transmits the collected data to the data processing module. The data processing module is electrically connected to the control module, and the data processing module issues a control strategy to the control module. The control module is electrically connected to the drone and the spraying module to execute the control strategy.

[0010] Preferably, the sensor module includes the following types:

[0011] The environmental sensor module: installed at multiple positions of the drone, including a wind speed sensor module, a wind direction sensor module, a temperature sensor module, and a humidity sensor module, for real-time monitoring of the environmental data around the drone;

[0012] The crop sensor module: installed at the bottom of the drone, including a camera and an infrared sensor module, for monitoring the growth status and pest and disease conditions of the crops;

[0013] The position sensor module: includes a GPS module and an inertial measurement unit (IMU), for real-time monitoring of the position and attitude of the drone;

[0014] The liquid medicine sensor module: installed in the spraying module, including a flow sensor module and a pressure sensor module, for real-time monitoring of the flow rate and pressure of the liquid medicine.

[0015] Preferably, the data processing module includes an image processing and recognition unit, a sensor module data fusion unit, and a decision-making unit. Among them, the image processing and recognition unit is used to process the image data obtained from the crop sensor module and identify the growth status of the crops and the pest and disease areas. The sensor module data fusion unit is used to assist the decision-making unit in generating an accurate spraying strategy.

[0016] Preferably, the specific operation steps of the image processing and recognition unit are as follows:

[0017] S1: Obtain the farmland image I(x, y) from the crop sensor module;

[0018] S2: Generate a clear image I’(x, y) through preprocessing, including denoising and image enhancement;

[0019] S3: Use a convolutional neural network (CNN) to extract image features and output a feature map F k (x, y) ;

[0020] S4: Use a segmentation algorithm to locate the pest and disease area R(x, y).

[0021] Preferably, the convolution expression of the convolutional neural network is:

[0022] F k (x, y) = ∑ i,j W k (i, j)·I(x + i, y + j) + b k

[0023] Among them, i and j is the index of the convolutional kernel, representing the position offset of each element within the convolutional kernel. W k represents the (i, j) weight of the b k th convolutional kernel at position . F k (x, y) represents the bias term of the (x, y) th feature map of the convolutional kernel.

[0024] Activation function (ReLU):

[0025] A(x, y) = max(0, F(x, y)) ;

[0026] Parameter interpretation:

[0027] I(x, y) : The original input image;

[0028] W k (i, j) : Convolutional kernel parameters;

[0029] b k : Bias term;

[0030] F k (x, y) : Convolution output feature;

[0031] A(x, y) : Activated feature map.

[0032] Preferably, the data fusion unit of the sensor module operates specifically as follows:

[0033] T1: Obtain data from the environmental sensor module, including wind speed, wind direction, temperature, and humidity;

[0034] T2: Use Kalman filtering to fuse the data from the environmental sensor module and estimate the environmental state x^(t) ;

[0035] Preferably, the Kalman filter specifically includes the following formulas:

[0036] Kalman filter prediction update equation:

[0037] x ^ (t|t - 1) = Ax ^ (t - 1) + Bu(t - 1) ;

[0038] P(t|t - 1) = AP(t - 1)A + Q ;

[0039] Kalman gain:

[0040] K(t) = P(t|t - 1)H(HP(t|t - 1)H + R)^(-1) ;

[0041] State update:

[0042] x^(t|t) = x^(t|t - 1) + K(t)(z(t) - Hx^(t|t - 1)) ;

[0043] P(t|t) = (I - K(t)H)P(t|t - 1) ;

[0044] Parameter interpretation:

[0045] x^(t|t - 1) : Predicted state at time t;

[0046] P(t|t - 1) : Covariance matrix of the predicted state;

[0047] A, B : State transition matrix, control input matrix;

[0048] Q : Process noise covariance;

[0049] K(t) : Kalman gain;

[0050] z(t) : Sensor module measurement value;

[0051] H : Represents the observation matrix, used to establish a relationship between the state vector and the observation value;

[0052] R : Represents the observation noise covariance matrix, used to represent the error or noise of the observation value.

[0053] Preferably, the decision-making unit operates specifically as follows:

[0054] F1: Input the pest and disease area R(x, y) and the environmental state x^(t) into the decision-making unit for pest and disease assessment, and generate a pest and disease severity score S d(x, y) ;

[0055] F2: Determine whether to spray according to the score. If it is determined to spray, dynamically adjust the spraying angle during the spraying process;

[0056] Preferably, in F2, the spraying angle is dynamically adjusted according to the wind speed, wind direction, UAV position and attitude. The calculation formula is: ;

[0057] Wherein, is the initial spraying angle, is the wind speed, is the wind direction angle, is the pressure difference, and are correction factors.

[0058] Preferably, the spraying module further includes an anti-clogging device. The anti-clogging device includes an ultrasonic sensor and a vibration motor. The ultrasonic sensor is installed at the liquid inlet of the nozzle to detect the clogging condition of the nozzle. The vibration motor is installed on the base of the nozzle to periodically vibrate the nozzle to prevent the liquid medicine from clogging. The control module includes a central processor and a signal transceiver. When the central processor detects nozzle clogging through the ultrasonic sensor, it sends an instruction to the vibration motor through the signal transceiver, and the vibration motor periodically vibrates the nozzle to remove the clogging;

[0059] The UAV further includes an autonomous obstacle avoidance module. The autonomous obstacle avoidance module includes a millimeter wave radar and an obstacle avoidance algorithm unit. The millimeter wave radar is installed in the front, back, left and right directions of the UAV to detect obstacles around the UAV in real time. The obstacle avoidance algorithm unit is installed in the central processor to process the data of the millimeter wave radar and generate an obstacle avoidance instruction. The control module is also connected to a power management module. When the millimeter wave radar detects an obstacle, the obstacle avoidance algorithm unit sends an instruction to the power management module through the signal transceiver to adjust the rotation speed and power of the motor, so that the UAV automatically bypasses the obstacle to ensure flight safety.

[0060] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0061] 1. By accurately identifying the pest and disease area and adjusting the spraying amount according to environmental conditions, the drug spraying in ineffective areas is avoided;

[0062] 2. Through the real-time data provided by the sensor module, the pest and disease area can be accurately identified and the spraying strategy can be adjusted according to environmental factors, avoiding the waste of traditional spraying methods;

[0063] 3. The entire data processing module realizes the precise spraying of pests and diseases by drones through an automated process. This intelligent adjustment can respond to environmental changes in real time and dynamically adjust the spraying strategy, greatly improving agricultural production efficiency and reducing labor costs.

[0064] 4. Through image segmentation technology, the specific location and scope of pests and diseases can be accurately identified and located, avoiding the problem of manual estimation errors in traditional methods, ensuring that the drone can perform precise spraying according to the distribution of pests and diseases, and improving the control efficiency.

[0065] 5. Convolution operations can reduce the complexity of manual feature extraction, automatically learn features through the training process, and enhance the adaptability and flexibility of the system. For different farmland environments and crop types, this convolutional neural network can be adjusted adaptively to provide efficient and precise pest and disease control solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, form a part of this application, and do not limit the embodiments of the present invention. In the drawings:

[0067] Figure 1 is the system block diagram of the present invention;

[0068] Figure 2 is the specific operation step flowchart of the image processing and recognition unit;

[0069] Figure 3 is the specific operation step flowchart of the decision-making unit. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0070] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with the embodiments and the drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and do not limit the present invention. It should be noted that the present invention is already in the actual R & D and use stage.

[0071] As Figures 1 - 3 shown, an embodiment of the present invention provides a system for precise spraying of pests and diseases by drones, including a drone, a spraying module, a sensor module, a data processing module, and a control module. The spraying module, sensor module, data processing module, and control module are all installed on the drone;

[0072] The sensor module is electrically connected to the data processing module and transmits the collected data to the data processing module. The data processing module is electrically connected to the control module. The data processing module issues a control strategy to the control module. The control module is electrically connected to the drone and the spraying module and executes the control strategy.

[0073] In a possible implementation, first, the sensor module is installed on the drone and is responsible for collecting environmental and crop health data in real time. The sensor module includes environmental sensors (such as temperature, humidity, wind speed, and wind direction sensors) and crop sensors (such as high-definition cameras, infrared sensors, etc.). These sensors transmit the collected data to the data processing module through electrical connections. The sensor module and the data processing module are connected through a data bus, ensuring real-time transmission and processing of information.

[0074] Furthermore, the data processing module is responsible for receiving data from the sensor module, processing, and analyzing it. The data processing module first preprocesses the collected images and environmental data, such as denoising, image enhancement, etc., to facilitate the extraction of more accurate crop pest and disease information. Using image processing algorithms and convolutional neural networks (CNNs), the data processing module can accurately identify the areas and types of pests and diseases on the crops. At the same time, the data processing module also fuses the environmental data, and through algorithms such as Kalman filtering, estimates the most suitable timing and area for spraying, generating a precise control strategy.

[0075] Once the data processing module completes the analysis of the pest and disease areas and environmental conditions, it transmits the generated control strategy to the control module through electrical connections. The control module decides the flight path, spraying method, and spraying amount of the drone according to the strategy information provided by the data processing module. The control module is also responsible for adjusting the working state of the spraying module, such as spraying angle, flow rate, pressure, etc., to ensure the maximization of the spraying effect.

[0076] Furthermore, the spraying module is the device that executes the specific spraying task. It accurately sprays the liquid medicine onto the pest and disease areas according to the instructions issued by the control module. The spraying module includes nozzles, liquid medicine storage containers, pumps, and flow sensors, and can dynamically adjust the spraying strategy according to real-time environmental data to adapt to changes in wind speed, humidity, etc.

[0077] In the embodiment of the present invention, the sensor module includes:

[0078] Environmental sensor module: Installed at multiple positions on the drone, including a wind speed sensor module, a wind direction sensor module, a temperature sensor module, and a humidity sensor module, for real-time monitoring of environmental data around the drone;

[0079] Crop sensor module: Installed at the bottom of the drone, including a camera and an infrared sensor module, for monitoring the growth status and pest and disease conditions of the crops;

[0080] Position sensor module: Includes a GPS module and an inertial measurement unit, for real-time monitoring of the position and attitude of the drone;

[0081] Liquid medicine sensor module: Installed in the spraying module, it includes a flow sensor module and a pressure sensor module, and is used to monitor the flow rate and pressure of the liquid medicine in real time.

[0082] In a possible implementation manner, first, the environmental sensor module includes a wind speed sensor, a wind direction sensor, a temperature sensor, and a humidity sensor. Multiple environmental sensors are installed at different positions of the unmanned aerial vehicle to comprehensively monitor the environmental changes around the unmanned aerial vehicle. The real-time data of these sensors is transmitted to the data processing module through wireless connection or electrical connection. The data processing module analyzes factors such as wind speed, wind direction, temperature, and humidity by fusing environmental data, and generates a suitable spraying strategy. For example, too high temperature and humidity or too high wind speed will affect the spraying effect, and the data processing module will adjust the spraying timing and spraying amount according to this information to ensure the effective transmission and effect of the medicine.

[0083] Furthermore, the crop sensor module is installed at the bottom of the unmanned aerial vehicle and mainly includes a high-definition camera and an infrared sensor. These sensors are used to monitor the growth status, pest and disease conditions, etc. of the crops in real time. The crop sensor module transmits the collected image data and infrared data to the data processing module. The data processing module processes these images, and detects whether there are pests and diseases on the crop surface through an image recognition processing algorithm, and judges the types, distribution areas, and severity of the pests and diseases. The data processing module uses these data to generate a precise spraying area and drug dosage allocation plan.

[0084] Furthermore, the position sensor module includes a GPS module and an inertial measurement unit (IMU), and is used to monitor the position and attitude of the unmanned aerial vehicle in real time. The GPS module provides the position information of the unmanned aerial vehicle, while the IMU provides the flight attitude data of the unmanned aerial vehicle (such as tilt angle and acceleration). This information is transmitted to the data processing module through electrical connection, helping the system to adjust the flight path and spraying angle in real time. Ensure that the spraying module always maintains the correct flight trajectory and altitude to achieve the purpose of precise spraying.

[0085] Furthermore, the liquid medicine sensor module is installed in the spraying module and mainly includes a flow sensor and a pressure sensor. The flow sensor monitors the flow rate of the liquid medicine in the spraying system, and the pressure sensor monitors the pressure of the spraying system. The liquid medicine sensor module transmits the real-time data to the control module, and the control module adjusts the spraying pressure and flow rate according to these data to ensure uniform spraying of the liquid medicine and avoid drug waste or uneven spraying caused by unstable pressure or too large flow rate.

[0086] In an embodiment of the present invention, the data processing module includes an image processing and recognition unit, a sensor module data fusion unit, and a decision-making unit. The image processing and recognition unit is used to process the image data obtained from the crop sensor module and identify the crop growth status and pest and disease areas. The sensor module data fusion unit is used to assist the decision-making unit in generating an accurate spraying strategy.

[0087] In a possible implementation manner, the image processing and recognition unit obtains monitoring data such as the crop growth status and pest and disease information collected by the crop sensor in real time. At the same time, it uses image processing algorithms combined with convolutional neural networks (CNNs) to process and analyze the obtained image data, so as to accurately identify the pest and disease areas and types on the crops.

[0088] The main task of the image processing and recognition unit is to identify the health status of the crops, determine whether there are pests and diseases and their severity. For example, by analyzing information such as leaf color and morphology, the system can identify pests, disease spots, or other growth problems. For the pest and disease areas, this unit will also mark the specific positions to provide positioning information for subsequent precise spraying.

[0089] The sensor module data fusion unit is responsible for collecting data from various environmental sensors, including temperature, humidity, wind speed, etc. These data reflect the environmental factors affecting crop growth and influence the spraying decision. It combines these sensor data with the pest and disease information provided by the image processing unit to generate a comprehensive crop and environmental state model, through which the spread of pests and diseases and the crop requirements in the current environment are reflected, assisting the decision-making unit in formulating a spraying plan.

[0090] Furthermore, based on the pest and disease recognition data provided by the image processing and recognition unit and the environmental data provided by the sensor module, the decision-making unit conducts data analysis and processing, and combines the growth cycle of the crops, the current pest and disease situation, and the environmental conditions to generate an accurate spraying strategy.

[0091] For example, the decision-making unit will determine the drug type, spraying amount, spraying time, and spraying range according to the types of different pests and diseases. At the same time, according to the environmental data, the decision-making unit will adjust the spraying amount to avoid drug drift or waste when the wind speed is too high.

[0092] In an embodiment of the present invention, the specific operation steps of the image processing and recognition unit are as follows:

[0093] S1: Obtain the farmland image I(x, y) from the crop sensor module;

[0094] S2: Generate a clear image I’(x, y) through preprocessing, including denoising and image enhancement;

[0095] S3: Use a convolutional neural network to extract image features and output a feature map F k (x, y) ;

[0096] S4: Use a segmentation algorithm to locate the pest and disease area R(x, y).

[0097] In a possible implementation, the crop sensor module (such as a high-definition camera, an infrared camera, etc.) collects image data of the farmland in real time through sensors installed on the drone. The images captured show the overall view or local areas of the farmland, which contain information such as the health status of the crops, environmental changes, possible pests and diseases, etc. The obtained original image I(x, y) will be used as the input data for subsequent image processing.

[0098] The original image is usually disturbed by noise (for example, light changes, weather effects, etc.), so it must be preprocessed to improve the image quality. Denoising algorithms (such as Gaussian filtering, median filtering, etc.) are used to remove the noise in the image, while image enhancement (such as histogram equalization, contrast enhancement, etc.) can enhance the image details, especially the tiny features of pests and diseases. The clear image I’(x, y) generated after preprocessing will provide higher-quality data input for subsequent feature extraction and pest and disease detection.

[0099] A convolutional neural network (CNN) is used to automatically extract useful image features from the clear image I’(x, y). The CNN can identify local and global patterns in the image, such as the color, shape, texture of the crops and the features of pests and diseases. The CNN extracts features at different levels through multiple convolutional layers and finally generates a feature map F k (x, y), and each point corresponds to the feature information at a certain position in the image. This process can effectively extract the key information of pests and diseases and provide a reliable basis for subsequent region segmentation.

[0100] The feature map of the image F k (x, y) will be input into an image segmentation algorithm (such as a segmentation model based on U-Net). This algorithm can automatically distinguish the normal area and the pest and disease area by analyzing the feature information in the image, thereby accurately locating the occurrence position and scope of pests and diseases. The finally generated pest and disease area R(x, y) will be transmitted to the control module to guide the drone for precise spraying.

[0101] In the convolutional neural network convolution expression described in the embodiment of the present invention:

[0102] F k (x, y) = ∑ i,j Wk (i, j)·I(x + i, y + j) + b k ;

[0103] Among them, i and j are the indices of the convolutional kernel, representing the position offsets of each element within the convolutional kernel. W k represents the th convolutional kernel's weight at position (i, j) . b k represents the bias term of the th feature map of the convolutional kernel. F k (x, y) represents the value of the output feature map after the action of the th convolutional kernel at position (x, y) ;

[0104] Activation function (ReLU):

[0105] A(x, y) = max(0, F(x, y)) ;

[0106] Parameter interpretation:

[0107] I(x, y) : The original input image;

[0108] W k (i, j) : Convolutional kernel parameters;

[0109] b k : Bias term;

[0110] F k (x, y) : Convolution output features;

[0111] A(x, y) : Activated feature map.

[0112] In a possible implementation, the convolution expression represents the operation mode of the convolutional layer in a convolutional neural network. Here, the input image I(x, y) will be convolved with the convolutional kernel W k (i, j). The convolutional kernel is a matrix of a fixed size, which contains the parameters learned by the network. These parameters can learn features (such as edges, textures, etc.) in the image through training. The result of the convolution operation is a feature map F k(x, y) is generated by the weighted sum of local regions of the input image, thereby extracting local features of the image to a certain extent. The bias term b_k is a constant adjustment after the convolution operation, which is used to improve the performance of the network.

[0113] Furthermore, the ReLU function activates the feature map output by the convolutional layer, outputs negative values as zero, and keeps positive values unchanged. This non-linear operation helps introduce non-linear transformations, enabling the neural network to fit more complex patterns. The feature map A(x, y) after ReLU activation can capture and express the significant features of the image more effectively than the feature map F k (x, y). Through this activation process, the system can identify key elements in the image, such as changes on the crop surface, traces of pests and diseases, etc.

[0114] In a convolutional neural network, the convolution operation between the input image I(x, y) and the convolution kernel W k (i, j) generates a convolutional feature map F k (x, y) by performing weighted summation on local regions of the image. This local perception mechanism allows the network to extract useful spatial features from the input image, providing a basis for pest and disease detection.

[0115] The feature map F k(x, y) after being processed by the convolutional layer is further transformed by the ReLU activation function A(x, y), thereby filtering out negative values in the features and retaining positive activation values. This process improves the network's sensitivity to image details and enables the neural network to better capture subtle changes in pests and diseases, avoiding the influence of certain unimportant features on system judgment.

[0116] This convolution process is often not a one-time operation, but a multi-layer stack. After each combination of the convolution kernel and the activation function, different levels of features are extracted, gradually converting complex image information into high-level semantic information. This hierarchical feature extraction mechanism ensures that the network can gradually identify and locate pest and disease areas in the farmland, improving the accuracy of detection.

[0117] In the embodiment of the present invention, the specific operation steps of the sensor module data fusion unit are as follows:

[0118] T1: Obtain data from the environmental sensor module, including wind speed, wind direction, temperature, and humidity;

[0119] T2: Use Kalman filtering to fuse the data from the environmental sensor module and estimate the environmental state x^(t).

[0120] In a possible implementation, the system first obtains multiple data of the current environment through the environmental sensor module, including wind speed, wind direction, temperature, humidity, etc. These data can reflect the environmental conditions and directly affect the accuracy of the spraying operation and the distribution effect of the medicament. For example, wind speed and wind direction may affect the propagation direction and diffusion range of the spraying droplets, while temperature and humidity may affect the evaporation speed and effect of the medicament. Therefore, obtaining these sensor data provides a basis for subsequent calculations and judgments.

[0121] After obtaining various sensor data, the Kalman filtering algorithm is then used to fuse these data. Kalman filtering is a recursive estimation algorithm that can predict and correct the state of a dynamic system and estimate relatively accurate state information in the case of incomplete or noisy data. In this system, Kalman filtering fuses data such as wind speed, wind direction, temperature, and humidity from different sensors to calculate an optimal estimated value (environmental state), thereby providing a more accurate environmental state model.

[0122] The specific steps of Kalman filtering include the following parts:

[0123] Prediction step: Predict the environmental state at the current moment based on the estimated state and dynamic model at the previous moment. At this time, the predicted state is not necessarily accurate and may contain certain errors.

[0124] Update step: Use the newly obtained sensor data to correct the predicted state and reduce the estimation error. By calculating the Kalman gain, the system can weigh the importance of the current sensor data and the predicted data, and finally update the estimated value of the environmental state.

[0125] In the embodiment of the present invention, the Kalman filtering specifically includes the following formulas:

[0126] Kalman filtering prediction update equation:

[0127] x ^ (t|t - 1) = Ax ^ (t - 1) + Bu(t - 1) ;

[0128] P(t|t - 1) = AP(t - 1)A + Q ;

[0129] Kalman gain:

[0130] K(t) = P(t|t - 1)H(HP(t|t - 1)H + R)^(-1) ;

[0131] State update:

[0132] x^(t|t) = x^(t|t - 1) + K(t)(z(t) - Hx^(t|t - 1)) ;

[0133] P(t|t) = (I - K(t)H)P(t|t - 1) ;

[0134] Parameter Interpretation:

[0135] x^(t|t - 1) : Predicted state at time t;

[0136] P(t|t - 1) : Covariance matrix of the predicted state;

[0137] A, B : State transition matrix, control input matrix;

[0138] Q : Process noise covariance;

[0139] K(t) : Kalman gain;

[0140] z(t) : Measurement value of the sensor module;

[0141] H : Represents the observation matrix, used to establish the relationship between the state vector and the observation value;

[0142] R : Represents the observation noise covariance matrix, used to represent the error or noise of the observation value.

[0143] In a possible implementation, the core of the Kalman filter is to estimate the state by combining the prediction and the measurement value. In the UAV spraying system, the information such as wind speed and humidity provided by the environmental sensor module is usually disturbed by noise. Therefore, there may be errors in directly using these measurement data for decision-making. The Kalman filter effectively reduces the influence of noise through the weighted fusion of the prediction and the actual measurement results, and obtains a more accurate estimation of the environmental state. In this way, the system can perform spraying control according to more accurate environmental conditions.

[0144] In practical applications, the sensor data may be delayed or unstable. The Kalman filter can dynamically adjust the confidence in the data according to the previous state prediction and the current measurement information, so that the system can cope with sudden sensor failures or data anomalies. This robustness is crucial for the continuous operation of the UAV under complex meteorological conditions.

[0145] The accurate estimation of the environmental state can help the UAV to adjust the spraying parameters in real time during the spraying process, such as flight speed, spraying angle and dosage, to ensure the accurate coverage of the medicament and reduce waste. For example, if the wind speed is too high, the system can increase the amount of medicament sprayed to make up for the drift of the medicament, or adjust the flight altitude to reduce the influence of the wind, thereby improving the spraying efficiency and operation accuracy.

[0146] In the embodiment of the present invention, the specific operation steps of the decision-making unit are as follows:

[0147] F1: Input the pest and disease area R(x,y) and the environmental status x^(t) into the decision-making unit for pest and disease assessment to generate a pest and disease severity score S d (x, y) ;

[0148] F2: Based on the score, decide whether to spray. If it is decided to spray, dynamically adjust the spraying angle during the spraying process.

[0149] In a possible implementation manner, in step F1, the decision-making unit receives two input signals:

[0150] Pest and disease area: This is a set of pest and disease information collected in real time by sensors carried by drones (such as image recognition devices, infrared sensors, etc.). Each coordinate point represents a part of the crop field, and the sensor data will indicate the type and severity of pests and diseases in this area. This information can help the decision-making unit understand the distribution of pests and diseases and provide a basis for spraying decisions.

[0151] Environmental status: Environmental information provided in real time by environmental sensors (such as temperature, humidity, wind speed, etc.). The environmental status has an important impact on the spread and diffusion speed of pests and diseases as well as the spraying effect of pesticides. Therefore, the decision-making unit needs to integrate this data to predict the diffusion effect of pesticides after spraying.

[0152] Combining the above input signals together, generate a pest and disease severity score through a built-in pest and disease assessment algorithm. This score reflects the severity of pests and diseases in different areas. The higher the score, the more severe the pests and diseases in this area, and the greater the amount of pesticides required for spraying. The generation of the score not only considers the quantity and type of pests and diseases but also combines the influence of environmental factors.

[0153] Specifically, first, the decision-making unit obtains input signals from two main sources: pest and disease area information and environmental status information. The pest and disease area information comes from the image processing and recognition unit, indicating whether there are pests and diseases on the crops, as well as the types and distributions of pests and diseases. Each position (x, y) contains the pest and disease information in this area, including the type of pests and diseases, the area, and the possible degree of influence. This data is usually processed by the image processing and recognition unit (such as a convolutional neural network CNN) to generate the pest and disease area R(x, y).

[0154] The environmental status information is the environmental sensor data fused by the data fusion unit of the sensor module. These environmental factors have important impacts on the spread of pests and diseases and the spraying effect of pesticides. For example, a higher temperature may promote the spread of pests and diseases, a higher humidity may increase the instability of the spraying effect, and too high a wind speed may cause the drift of pesticides. All this information is processed by algorithms such as Kalman filtering to provide an accurate estimate of the environmental status x^(t).

[0155] Then, a pest and disease severity score is generated. After obtaining the pest and disease area information and the environmental status information, the decision-making unit performs a fusion analysis on these two signals through a built-in pest and disease assessment algorithm to generate a pest and disease severity score (Sd(x,y)). This score not only considers the types, quantities, and distributions of pests and diseases but also combines the environmental status to evaluate the potential threat of pests and diseases. The specific process is as follows: For the assessment of pest and disease severity, the algorithm first calculates the pest and disease index of the area based on the characteristics of the pest and disease area R(x,y). For example, within a specific area, different weights are assigned based on the type (such as insect pests, fungi, etc.) and quantity of pests and diseases, and the area size and the spread range of pests and diseases also affect the final score. The weight values of pest and disease types are determined through prior learning and database accumulation.

[0156] In addition, according to the regulatory role played by the crop and environmental state model based on the environmental state \(x^{(t)}\) during the evaluation process. For example, when the temperature is high and the humidity is high, the spread rate of pests and diseases may be faster. For the case of high wind speed, the system will consider the adjustment of the spread range of the pesticide and the spraying angle, which in turn affects the adjustment of the spraying amount. Environmental factors such as wind speed and humidity have a significant impact on the spread of pests and diseases and the spraying effect of pesticides. Therefore, it is necessary to comprehensively consider the impact of the environment on spraying accuracy and drug efficacy. The crop and environmental state model plays a core role in the evaluation process of pest and disease severity. Specifically, this model combines crop state information such as pest and disease types, quantities, spread ranges, etc., and environmental state information such as wind speed, humidity, temperature, etc., to form a comprehensive pest and disease severity score. For a specific area, different weights are assigned to the types (such as insect pests, fungi, etc.) and quantities of pests and diseases, and these weights are determined based on historical data and learning algorithms. The spread range and affected area size of pests and diseases will further affect the score. In addition, the environmental state \(x^{(t)}\) plays a regulatory role during the evaluation process. For example, when the temperature is high and the humidity is high, the spread rate of pests and diseases may be faster, and the score will be correspondingly increased; when the wind speed is high, the system will comprehensively evaluate the spread range and spraying angle of the pesticide and adjust the spraying amount to ensure the drug efficacy. Through the crop and environmental state model, the system can comprehensively evaluate the threat level of pests and diseases and formulate a precise spraying plan accordingly. Finally, based on all the above information, through weighted processing, the pest and disease severity score \(S_d(x, y)\) of each area is generated. The higher the score, the more severe the pests and diseases in the area, and the more urgent the spraying need in this area is considered by the system.

[0157] In addition, the flight path is used for the drone to determine the flight route during the spraying process to ensure that the pesticide can evenly cover the pest and disease area to be treated. The process of calculating the optimal flight path is as follows:

[0158] Input data: Based on the pest and disease severity score, environmental state information (such as wind speed, temperature, humidity), and geographical data of the flight area.

[0159] Path planning method: Use the shortest path algorithm (such as A* algorithm, Dijkstra algorithm, etc.) or area-based coverage algorithms (such as grid coverage algorithm, heuristic algorithm, etc.) to generate the flight path of the drone. By calculating the distribution density of the pest and disease area, flight distance, flight height, obstacles, and the impact of environmental factors on the flight path, ensure that the path can effectively cover the pest and disease area and avoid obstacles and dangerous areas.

[0160] Output: The flight path trajectory, usually including the starting point, ending point, path points, and flight height of the flight, etc.

[0161] The spraying decision is used to dynamically adjust the spraying parameters during the flight of the drone to ensure that the pesticide can be accurately sprayed onto the pest and disease areas under different environmental conditions. The spraying control algorithms for the spraying volume, spraying angle, and spraying timing are as follows:

[0162] Input data: severity score of the pest and disease area, environmental status information, drone flight path, technical parameters of the pesticide sprayer, etc.

[0163] Control strategy: Adjust the spraying volume according to the severity score of the pests and diseases. Increase the spraying volume in areas with more severe pests and diseases, and decrease the spraying volume in areas with less severe pests and diseases.

[0164] Spraying angle adjustment: Dynamically adjust the spraying angle according to information such as wind speed, wind direction, and flight altitude to reduce the drift of the pesticide and ensure accurate spraying of the pesticide.

[0165] Spraying timing: Dynamically determine the spraying timing based on real-time environmental data (such as wind speed changes, temperature and humidity, etc.) to avoid poor pesticide effectiveness or waste caused by excessive wind speed or high humidity.

[0166] Output: Spraying instructions, including spraying volume, spraying angle, spraying time, and the coverage area of the spraying area, etc.

[0167] In step F2, based on the severity score of the pests and diseases, the decision-making unit decides whether to initiate spraying. If the severity score of a certain area exceeds the preset threshold, the decision-making unit will issue a spraying instruction to direct the drone to spray the pesticide on that area. Otherwise, the spraying will be skipped to avoid unnecessary pesticide waste.

[0168] During the spraying process, the decision-making unit will also dynamically adjust the spraying angle according to the real-time environmental status. This process is crucial for improving the spraying accuracy:

[0169] Influence of environmental factors: For example, changes in wind speed and wind direction will affect the spraying effect of the pesticide. If the wind speed is high, the pesticide may drift with the wind, causing waste. The decision-making unit will adjust the spraying angle according to the real-time wind speed and wind direction data to ensure that the pesticide is accurately sprayed onto the area with the most severe pests and diseases.

[0170] Morphology of the pest and disease area: The morphologies and distributions of different pest and disease areas vary greatly. The decision-making unit dynamically adjusts the spraying angle according to the geometric shape and severity score of the pest and disease area to ensure uniform coverage of the pesticide in the pest-dense area and avoid unnecessary waste.

[0171] In the embodiment of the present invention, in F2, the spraying angle is dynamically adjusted according to the wind speed, wind direction, drone position, and attitude, and the calculation formula is: ;

[0172] Where, is the initial spraying angle, is the wind speed, is the wind direction angle, is the pressure difference, and are correction factors.

[0173] In a possible implementation, the initial spraying angle is a preset value of the UAV spraying system, usually determined by the design parameters of the UAV or the requirements of the spraying task. This angle represents the ideal working angle of the spraying device without the influence of external factors (such as wind speed and wind direction changes).

[0174] The wind speed and wind direction angle are obtained in real time through the wind speed sensor and wind direction sensor on the UAV. The wind speed represents the intensity of air flow, and the wind direction angle represents the direction of the wind, usually expressed as the angle with the spraying direction. The wind speed and wind direction are important factors affecting the spraying effect. Excessive wind speed may cause the drift of the chemical agent, while the wind direction determines whether the chemical agent is blown to the areas that do not need to be sprayed.

[0175] By calculating , the system can dynamically adjust the spraying angle according to the magnitude and direction of the wind speed. Specifically, if the wind speed is large or the wind direction deviates from the ideal spraying direction, the spraying angle will be adjusted accordingly to ensure the accurate spraying of the chemical agent.

[0176] The pressure difference represents the change in the nozzle pressure of the chemical agent spraying device. The change in the pressure difference directly affects the spraying speed and flow rate of the chemical agent, and thus affects the spraying range of the chemical agent. By adjusting the spraying angle, the influence of the pressure difference on the spraying effect is compensated to ensure that the chemical agent can be evenly covered in the predetermined area.

[0177] in the calculation formula is used to correct the spraying angle according to the change in the pressure difference to compensate for the influence of pressure fluctuations on the spraying accuracy.

[0178] The correction factors are constants obtained through experimental or simulation data debugging. Their role is to quantify the relationship between factors such as wind speed, wind direction, and pressure difference and the spraying angle. Due to different environmental conditions and different spraying tasks, the correction factors will be different, so they need to be adjusted according to the actual situation.

[0179] In an embodiment of the present invention, the spraying module further includes a clogging prevention device, which includes an ultrasonic sensor and a vibration motor. The ultrasonic sensor is installed at the liquid inlet of the nozzle for detecting the clogging condition of the nozzle. The vibration motor is installed on the base of the nozzle for periodically vibrating the nozzle to prevent the liquid medicine from clogging. The control module includes a central processing unit and a signal transceiver. When the central processing unit detects nozzle clogging through the ultrasonic sensor, it sends an instruction to the vibration motor through the signal transceiver, and the vibration motor periodically vibrates the nozzle to remove the clogging object;

[0180] The unmanned aerial vehicle further includes an autonomous obstacle avoidance module, which includes a millimeter-wave radar and an obstacle avoidance algorithm unit. The millimeter-wave radar is installed in the front, rear, left, and right directions of the unmanned aerial vehicle for real-time detection of obstacles around the unmanned aerial vehicle. The obstacle avoidance algorithm unit is used to process the data of the millimeter-wave radar and generate an obstacle avoidance instruction. The control module is also connected to a power management module. When the millimeter-wave radar detects an obstacle, the obstacle avoidance algorithm unit sends an instruction to the power management module through the signal transceiver to adjust the rotation speed and power of the motor, so that the unmanned aerial vehicle automatically bypasses the obstacle.

[0181] In a possible implementation manner, the ultrasonic sensor is installed at the liquid inlet of the nozzle, and by transmitting and receiving ultrasonic signals, it monitors the flow state of the liquid medicine in real time. When the liquid medicine flow rate is abnormal, the ultrasonic sensor can detect whether the nozzle is clogged. The sensor transmits the detected clogging information to the central processing unit.

[0182] The vibration motor is installed on the base of the nozzle and is responsible for periodically vibrating the nozzle. Its operation is controlled by the central processing unit. When the ultrasonic sensor detects nozzle clogging, the central processing unit sends an instruction to the vibration motor through the signal transceiver to start the vibration function. The vibration motor effectively removes the clogging object in the nozzle through periodic vibration and restores the normal operation of the nozzle.

[0183] During the spraying operation, the ultrasonic sensor continuously monitors the liquid flow of the nozzle. If a clogging condition is detected, the signal is transmitted to the central processing unit. The central processing unit immediately sends an instruction to the vibration motor through the signal transceiver to start the vibration motor for cleaning, ensuring that the nozzle is unblocked and the spraying operation is restored. In this way, the nozzle can continuously work efficiently, reducing downtime or operation interruption caused by clogging.

[0184] The millimeter-wave radar is installed in the front, rear, left, and right directions of the unmanned aerial vehicle for real-time detection of obstacles in the surrounding environment. The radar can accurately measure the position and distance of the obstacle by transmitting millimeter-wave signals and receiving the reflected signals. When an obstacle enters the detection range, the radar transmits the data to the central processing unit.

[0185] The obstacle avoidance algorithm unit is installed in the central processing unit and is responsible for processing data from the millimeter-wave radar. By analyzing the obstacle data, the obstacle avoidance algorithm unit can generate the optimal obstacle avoidance path and strategy. The generated obstacle avoidance instructions are sent to the power management module of the UAV through the signal transceiver.

[0186] After receiving the obstacle avoidance instructions, the power management module adjusts the motor speed and power of the UAV to ensure that the UAV bypasses the obstacles. During this process, the power management module adjusts the operating state of the motor in real time, enabling the UAV to autonomously avoid obstacles and maintain its heading and flight stability.

[0187] The specific implementation manners described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific implementation manners of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A system for precisely spraying and controlling pests and diseases by an unmanned aerial vehicle, characterized in that: It includes a drone, a spraying module, a sensor module, a data processing module, and a control module. The spraying module, sensor module, data processing module, and control module are all installed on the drone. The sensor module is electrically connected to the data processing module and transmits the collected data to the data processing module. The data processing module is electrically connected to the control module, and the data processing module issues control strategies to the control module. The control module is electrically connected to the drone and the spraying module to execute the control strategies. The data processing module includes an image processing and recognition unit, a sensor module data fusion unit, and a decision-making unit. Among them, the image processing and recognition unit is used to process the image data obtained from the crop sensor module and identify the crop growth status and pest and disease areas. The sensor module data fusion unit is used to assist the decision-making unit in generating accurate spraying strategies. The specific operation steps of the image processing and recognition unit are as follows: S1: Obtain the farmland image I(x, y) from the crop sensor module; S2: Generate a clear image I’(x, y) through preprocessing, including denoising and image enhancement; S3: Extract image features using a convolutional neural network and output a feature map F k (x,y) ; S4: Use a segmentation algorithm to locate the pest and disease area R(x, y); The specific operation steps of the sensor module data fusion unit are as follows: T1: Obtain the data of the environmental sensor module, including wind speed, wind direction, temperature, and humidity; T2: Use Kalman filtering to fuse the data of the environmental sensor module and estimate the environmental state x^(t); The specific operation steps of the decision-making unit are as follows: F1: Input the pest and disease area R(x,y) and the environmental status x^(t) into the decision-making unit for pest and disease assessment to generate a pest and disease severity score S d (x, y) ; F2: Decide whether to spray according to the score. If it is decided to spray, the spraying angle is dynamically adjusted during the spraying process; In F2, the spraying angle is dynamically adjusted according to the wind speed, wind direction, position, and attitude of the drone. The calculation formula is: ； Among them, is the initial spraying angle, is the wind speed, is the wind direction angle, is the pressure difference, and are correction factors.

2. The system for precisely spraying to control pests and diseases of an unmanned aerial vehicle according to claim 1, characterized in that: The sensor module includes: Environmental sensor module: Installed at multiple positions on the drone, including a wind speed sensor module, a wind direction sensor module, a temperature sensor module, and a humidity sensor module, used to monitor the environmental data around the drone in real time; Crop sensor module: Installed at the bottom of the drone, including a camera and an infrared sensor module, used to monitor the growth status and pest and disease conditions of the crops; Position sensor module: Includes a GPS module and an inertial measurement unit, used to monitor the position and attitude of the drone in real time; Liquid medicine sensor module: Installed in the spraying module, including a flow sensor module and a pressure sensor module, used to monitor the flow and pressure of the liquid medicine in real time.

3. The system for precisely spraying to control pests and diseases of an unmanned aerial vehicle according to claim 2, characterized in that: The convolution expression of the convolutional neural network is: F k (x,y)=∑ i,j W k (i,j)·I(x+i,y+j)+b k Among them, i and j are the indices of the convolution kernel, representing the position offsets of each element within the convolution kernel. W k represents the -th convolution kernel's weight at position (i,j) . b k represents the bias term of the -th feature map of the convolution kernel. F k (x,y) represents the value of the output feature map after the action of the -th convolution kernel at position (x,y) . Activation function (ReLU): A(x, y)= max(0, F(x, y)) Parameter interpretation: I(x,y) : Original input image; W k (i,j) : Convolution kernel parameters; b k : Bias term; F k (x,y) : Convolution output feature; A(x,y) : Feature map after activation.

4. The system for precisely spraying pesticides and controlling pests and diseases of an unmanned aerial vehicle according to claim 3, wherein: The specific Kalman filter includes the following formulas: Kalman filter prediction update equation: x ^ (t|t-1)=Ax ^ (t-1)+Bu(t-1) P(t|t-1)= AP(t-1)A+Q Kalman gain: K(t)= P(t|t-1)H(HP(t|t-1)H+R)-1 State update: x^(t|t)=x^(t|t-1)+K(t)(z(t)-Hx^(t|t-1)) P(t|t)=(I-K(t)H)P(t|t-1) Parameter interpretation: x^(t|t-1) : Predicted state at time t; P(t|t-1) : Covariance matrix of the prediction state; A,B : State transition matrix, control input matrix; Q : Process noise covariance; K(t) : Kalman gain; z(t) : Measured value of the sensor module; H : Represents the observation matrix, which is used to establish the relationship between the state vector and the observed value; R : Represents the observation noise covariance matrix, which is used to represent the error or noise of the observed values.

5. The system for precisely spraying to control plant diseases and insect pests by an unmanned aerial vehicle according to claim 4, characterized in that: The spraying module further includes a clogging prevention device. The clogging prevention device includes an ultrasonic sensor and a vibration motor. The ultrasonic sensor is installed at the liquid inlet of the nozzle for detecting the clogging condition of the nozzle. The vibration motor is installed on the base of the nozzle for periodically vibrating the nozzle to prevent the liquid medicine from clogging. The control module includes a central processing unit and a signal transceiver. When the central processing unit detects nozzle clogging through the ultrasonic sensor, it sends an instruction to the vibration motor through the signal transceiver, and the vibration motor periodically vibrates the nozzle to remove the clogging material; The drone further includes an autonomous obstacle avoidance module. The autonomous obstacle avoidance module includes a millimeter-wave radar and an obstacle avoidance algorithm unit. The millimeter-wave radar is installed in the front, back, left, and right directions of the drone for real-time detection of obstacles around the drone. The obstacle avoidance algorithm unit is used for processing the data of the millimeter-wave radar to generate obstacle avoidance instructions. The control module is also connected to a power management module. When the millimeter-wave radar detects an obstacle, the obstacle avoidance algorithm unit sends an instruction to the power management module through the signal transceiver to adjust the rotation speed and power of the motor, so that the drone automatically bypasses the obstacle.

Images