Intelligent agricultural weeding track control system and method thereof

Abstract

The invention discloses an intelligent agricultural weeding track control system, which relates to the field of agricultural related technologies and comprises an identification and positioning module, a track planning module, a navigation optimization module, a collaborative operation module and a man-machine interaction module. The invention further discloses an intelligent agricultural weeding track control method which comprises the steps that the unmanned aerial vehicle or the ground robot cruises above a farmland or on the ground, the depth camera collects farmland image data in real time, the image processing unit conducts denoising, filtering and edge detection processing on the original image data collected by the depth camera, and the weeding track is obtained. The two-dimensional image and the three-dimensional point cloud data are registered, so that the two-dimensional image and the three-dimensional point cloud data are consistent in space, and weeds and crops are distinguished through an image recognition technology. Through the path planning algorithm, the optimal weeding path can be calculated according to the actual terrain, weed distribution and obstacle conditions of the farmland, blind walking and repeated operation of the robot in the operation process are avoided, and therefore the weeding efficiency is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of agriculture-related fields, and in particular to a smart agriculture weeding trajectory control system and method thereof. Background Art

[0002] With the rapid development of modern agriculture, the efficiency and quality requirements of agricultural production are constantly improving. However, weeds are a major problem in the growth process of crops, which not only affect the growth rate and yield of crops, but also may reduce the quality of agricultural products.

[0003] At present, with the advancement of agricultural modernization, smart agriculture has gradually become an important trend in agricultural production. However, traditional weeding methods mainly rely on manual or chemical spraying, which is not only inefficient, but also easy to pollute the environment and even affect the healthy growth of crops; and some mechanical weeding is also easy to accidentally damage crops, making it difficult to achieve efficient and accurate weeding.

[0004] Therefore, it is necessary to propose a smart agricultural weeding trajectory control system and method to solve the above problems. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the shortcomings of the prior art, the present invention provides a smart agricultural weeding trajectory control system and method, which solves the problem that the existing weeding methods are not only inefficient, but also easily pollute the environment and affect the healthy growth of crops; and mechanical weeding is prone to accidentally damage crops.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A smart agricultural weeding trajectory control system, the trajectory control system comprising:

[0010] Identification and positioning module: Use drones and ground robots to obtain information on the physiological status of plants, and use deep learning algorithms to continuously optimize the model through a large amount of training data to distinguish different types of weeds and crops;

[0011] Trajectory planning module: It uses sensors to collect environmental information of farmland to provide basic data for path planning. It uses the path planning algorithm to generate the optimal walking path according to the changes in field terrain, the distribution of obstacles and the information of weeded areas. It uses the reinforcement learning algorithm to allow the robot to continuously learn and optimize the path planning strategy in actual operation.

[0012] Navigation optimization module: According to the path generated by the data processing and decision-making unit, the robot's movement is controlled so that it moves along the predetermined path. It includes a motor drive control unit, a steering control unit, and a speed control unit. Through precise navigation control, it ensures that the robot can accurately reach the target position and perform weeding operations according to the planned path.

[0013] Collaborative operation module: For large-scale farmland, multiple weeding robots are used to work together. The robots share information and coordinate planning through wireless communication to avoid operation conflicts and improve overall operation efficiency.

[0014] Human-computer interaction module: provides users with an interface for interacting with the robot, allowing users to easily monitor the robot's working status, set parameters, and issue instructions.

[0015] Optionally, the identification and positioning module includes the following devices and subunits:

[0016] Drones and ground robots: Equipped with depth camera units. Depth cameras are one of the basic hardware of the recognition and positioning module. Depth cameras measure the distance from objects to the camera by emitting infrared light and receiving reflected light, thereby generating three-dimensional point cloud data. The data contains the depth information in the scene, which helps the system understand the spatial structure of the scene.

[0017] Image processing unit: The image processing unit is responsible for processing the raw image data collected by the depth camera to improve the image quality and recognition accuracy. At the same time, it is responsible for registering the two-dimensional image with the three-dimensional point cloud data to ensure the spatial consistency of the two.

[0018] Target recognition unit: Use deep learning technology to analyze the processed images and point cloud data, and use target recognition algorithms to identify weeds and crops. The target recognition algorithms include convolutional neural networks (CNN) and regional convolutional neural networks (R-CNN). By training a large amount of labeled data, the feature representations of different categories are learned, thereby achieving accurate classification of weeds and crops.

[0019] Positioning unit: responsible for calculating the specific position of weeds and crops in space based on the results of target recognition, which involves coordinate transformation and geometric calculation mathematical operations. For example, the position and posture of the robot can be optimized by solving the least squares problem to make it closer to the target object.

[0020] Optionally, the mathematical model of the identification and positioning module is as follows:

[0021] Depth information calculation model: The depth camera calculates the distance from the object to the camera by measuring the flight time of infrared light. Assuming the speed of infrared light is c and the time difference between emission and reception is t, the distance d from the object to the camera can be expressed as: d = 0.5 × c × t;

[0022] Least squares model: In the positioning algorithm, the least squares method is often used to optimize the position and posture of the robot. Given a set of observation data (xi, yi) and the corresponding model prediction value (xi, yi), the optimal parameter θ is solved by minimizing the sum of squared errors: θ = argmin∑(xi-xi) 2 +(yi-yi) 2 .

[0023] Optionally, the trajectory planning module includes the following subunits and mathematical algorithms:

[0024] Environmental perception unit: uses a variety of sensors to collect environmental information of farmland and provide basic data for subsequent path planning. The sensors include GPS receivers, heading and angle sensors, cameras, and lidars. The GPS receiver is used to obtain the robot's location information, the heading and angle sensors are used to determine the robot's direction and posture, and the cameras and lidars are used to perceive surrounding obstacles (such as crops, trees, stones) and the distribution of weeds. The sensors transmit the collected data to the data processing and decision-making unit in real time.

[0025] Data processing and decision-making unit: Process and analyze the data collected by the environmental perception layer, filter and fuse the sensor data to improve the accuracy and reliability of the data, and use the path planning algorithm to generate the optimal weeding path based on the processed data and the specific positions of weeds and crops in space. The coordinate transformation formula in the rasterized path planning and the artificial potential field obstacle avoidance algorithm are used to calculate the terrain and obstacle distribution of the working area. At the same time, the path needs to be dynamically adjusted and optimized according to the actual situation to adapt to environmental changes.

[0026] Optionally, the algorithm model in the trajectory planning and navigation optimization module is as follows:

[0027] Path tracking algorithm: PID control is used to accurately track the predetermined path. The control formula is:

[0028]

[0029] Among them, u(t) is the output of the controller, that is, the expected turning angle of the robot's front wheels; e(t) is the current heading deviation, that is, the difference between the robot's current posture and the predetermined path; Kp, Ki, and Kd are the proportional, integral, and differential coefficients, respectively. The values ​​of these three parameters have an important influence on the performance of the controller and need to be adjusted and optimized according to the characteristics of the actual system;

[0030] Coordinate conversion formula in rasterized path planning: In the rasterized path planning method, the actual position coordinates of the robot are converted into coordinates in the grid coordinate system. Assuming the position coordinates of the robot in the working environment are (x, y), and the size of the grid is a×a, the grid coordinates (gx, gy) of the robot can be expressed as:

[0031]

[0032] in, represents the floor function, in this way, the position of the robot is mapped to the grid coordinate system for path planning and decision making;

[0033] Artificial potential field obstacle avoidance algorithm: In the artificial potential field obstacle avoidance algorithm, the potential field function is defined to calculate the virtual force exerted on the robot in the environment. The repulsive potential field function is usually defined as:

[0034]

[0035] Among them, Uo(q) is the repulsive potential field function, ρ(q,qobat) is the distance between the robot position q and the obstacle qobat, ρmax is the range of action of the repulsive potential field, and the gravitational potential field function can be defined as: Ug(q)=ρ(q,qgoal), where ρ(q,qgoal) is the distance between the robot position q and the target position qgoal. The virtual resultant force acting on the robot is calculated according to the gradients of the repulsive potential field and the gravitational potential field, thereby guiding the robot to avoid obstacles and move towards the target direction.

[0036] Optionally, the navigation optimization module utilizes the integrated high-precision GLONASS global navigation satellite system, as well as the inertial measurement unit (IMU) and visual odometer-assisted navigation technology to improve the positioning accuracy and stability of the robot, and utilizes SLAM (simultaneous localization and mapping) technology to build and update the three-dimensional map of the field environment in real time, providing a more accurate basis for path planning and navigation.

[0037] Optionally, the collaborative work module includes a communication and coordination unit, which is responsible for information exchange and coordinated control between multiple robots. It adopts wireless communication technology (such as Wi-Fi, Bluetooth, ZigBee) to establish a communication network between robots. Through the communication protocol, the robots send each other position information, work status, and task instruction data to perform collaborative work. At the same time, a master-slave control strategy is adopted, with a host robot responsible for global planning and scheduling, and other slave robots perform specific weeding tasks according to the instructions of the host.

[0038] Optionally, in the human-computer interaction module, the user enters the system of the human-computer interaction module through a mobile phone APP or computer software to view the robot's current position, weeded area, and remaining power information, and mark the areas that need to be weeded on the map or set the robot's working mode.

[0039] A smart agricultural weeding trajectory control method comprises the following steps:

[0040] S1: Image acquisition and processing: The drone or ground robot cruises over or on the farmland, and the depth camera collects farmland image data in real time. The image processing unit performs denoising, filtering and edge detection on the raw image data collected by the depth camera, and aligns the two-dimensional image with the three-dimensional point cloud data to make them consistent in space, and distinguishes weeds and crops through image recognition technology;

[0041] S2: Identification and positioning: Analyze the processed images and point cloud data through deep learning technology to identify weeds and crops. By training a large amount of labeled data, learn the feature representation of different categories, so as to achieve accurate classification of weeds and crops. It is responsible for calculating the specific location of weeds and crops in space based on the results of target recognition;

[0042] S3: Farmland environment perception: Collect farmland environment information through sensors to provide basic data for subsequent trajectory planning. The sensors transmit the collected data to the data processing and decision-making unit in real time;

[0043] S4: Path trajectory planning: Filter and fuse the data collected by the sensor to improve the accuracy and reliability of the data. Based on the processed data and the specific locations of weeds and crops in space, the path planning algorithm is used to generate the optimal weeding trajectory;

[0044] S5: Collaborative weeding: According to the weeding trajectory, the robot's movement is controlled so that it moves along the predetermined trajectory and performs weeding operations according to the planned trajectory. The collaborative operation mode of multiple weeding robots is adopted. The robots share information and coordinate planning through wireless communication. A host robot is responsible for global planning and scheduling, and other slave robots perform specific weeding tasks according to the host's instructions;

[0045] S6: Human-computer interaction: Users enter the interactive interface through a mobile phone APP or computer, monitor the robot's working status, and issue instructions based on the robot's current position, mowed area, and remaining power.

[0046] (III) Beneficial effects

[0047] The present invention provides a smart agricultural weeding trajectory control system and method thereof, which has the following beneficial effects:

[0048] 1. The present invention utilizes deep learning algorithms and image processing technology to quickly and accurately identify field weeds and crops. Compared with traditional manual identification methods, it is faster and can complete weed identification work in large areas of farmland in a shorter time, thereby improving the overall weed control efficiency.

[0049] 2. Through depth cameras, GPS and other technologies, the specific location and distribution of weeds can be determined, allowing the weeding robot to directly target the weeds and avoid accidental damage to crops.

[0050] 3. The present invention uses a path planning algorithm to calculate the optimal weeding path according to the actual terrain, weed distribution and obstacles of the farmland, thereby avoiding blind walking and repetitive operations of the robot during the operation process, reducing ineffective labor time, and thus greatly improving the weeding efficiency. When encountering emergencies or new obstacles, it ensures that the robot always moves along the optimal path, allowing the robot to flexibly respond to various situations in a complex farmland environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a module schematic diagram of the weeding trajectory control system of the present invention;

[0052] Figure 2 It is a schematic diagram of the flow chart of the weeding trajectory control method of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Example 1

[0055] See also Figure 1 The present invention discloses a smart agricultural weeding trajectory control system, the trajectory control system includes:

[0056] Identification and positioning module: Use drones and ground robots to obtain information on the physiological status of plants, and use deep learning algorithms to continuously optimize the model through a large amount of training data to distinguish different types of weeds and crops;

[0057] Trajectory planning module: It uses sensors to collect environmental information of farmland to provide basic data for path planning. It uses the path planning algorithm to generate the optimal walking path according to the changes in field terrain, the distribution of obstacles and the information of weeded areas. It uses the reinforcement learning algorithm to allow the robot to continuously learn and optimize the path planning strategy in actual operation.

[0058] Navigation optimization module: According to the path generated by the data processing and decision-making unit, the robot's movement is controlled to make it travel along the predetermined path. It includes a motor drive control unit, a steering control unit, and a speed control unit. Through precise navigation control, the robot is ensured to accurately reach the target position and perform weeding operations according to the planned path. The robot's positioning accuracy and stability are improved by integrating the high-precision GLONASS global navigation satellite system, as well as the inertial measurement unit (IMU) and visual odometer-assisted navigation technology. The SLAM (simultaneous localization and mapping) technology is used to build and update the three-dimensional map of the field environment in real time, providing a more accurate basis for path planning and navigation.

[0059] Collaborative operation module: For large-scale farmland, a collaborative operation mode of multiple weeding robots is adopted. The robots share information and coordinate planning through wireless communication to avoid operation conflicts and improve overall operation efficiency. The collaborative operation module includes a communication and coordination unit, which is the core part of collaborative operation. The communication and coordination unit is responsible for information exchange and coordinated control between multiple robots. Wireless communication technology (such as Wi-Fi, Bluetooth, ZigBee) is used to establish a communication network between robots. Through the communication protocol, the robots send each other location information, operation status, and task instruction data for collaborative operation. At the same time, a master-slave control strategy is adopted, with a host robot responsible for global planning and scheduling, and other slave robots perform specific weeding tasks according to the host's instructions;

[0060] Human-computer interaction module: provides users with an interface for interacting with the robot, allowing users to easily monitor the robot's working status, set parameters, and issue instructions. Users enter the human-computer interaction module system through a mobile phone APP or computer software to view the robot's current location, weeded area, and remaining power information, and mark areas that need to be weeded on the map or set the robot's working mode.

[0061] In this embodiment, the intelligent weeding system uses advanced sensor technology and the Internet of Things to accurately identify the difference between weeds and crops, thereby achieving precise weeding. The navigation optimization module can reasonably control the driving speed of the robot according to the robot's driving path and current working status. In areas where fine weeding is required, the speed can be appropriately reduced to ensure the quality of weeding; in open areas, the speed can be increased to speed up the progress of the operation. This speed control strategy not only ensures the weeding effect, but also reduces energy consumption. The high-precision GLONASS global navigation satellite system, as well as auxiliary navigation technologies such as inertial measurement units (IMUs) and visual odometers can be used to achieve precise positioning and navigation of the robot, which enables the robot to accurately reach the designated location for weeding operations, avoid accidental damage to crops, and improve the accuracy and quality of weeding operations.

[0062] Smart weed control systems can also reduce dependence on chemical herbicides, thereby reducing environmental pollution and pesticide residues in agricultural products, which is not only beneficial to protecting the ecological environment, but also helps to improve the quality and safety of agricultural products; at the same time, especially in large areas of farmland, smart weed control systems allow weeding work to be completed by robots or automated equipment, greatly reducing manpower requirements, not only reducing farmers' labor intensity, but also reducing agricultural production costs.

[0063] Example 2

[0064] This embodiment is a further optimization based on the embodiment 1. Specifically, the identification and positioning module includes the following devices and subunits:

[0065] Drones and ground robots: Equipped with depth camera units. Depth cameras are one of the basic hardware of the recognition and positioning module. Depth cameras measure the distance from objects to the camera by emitting infrared light and receiving reflected light, thereby generating three-dimensional point cloud data. The data contains the depth information in the scene, which helps the system understand the spatial structure of the scene.

[0066] Image processing unit: The image processing unit is responsible for processing the raw image data collected by the depth camera to improve the image quality and recognition accuracy. At the same time, it is responsible for registering the two-dimensional image with the three-dimensional point cloud data to ensure the spatial consistency of the two.

[0067] Target recognition unit: Use deep learning technology to analyze the processed images and point cloud data, and use target recognition algorithms to identify weeds and crops. The target recognition algorithms include convolutional neural networks (CNN) and regional convolutional neural networks (R-CNN). By training a large amount of labeled data, the feature representations of different categories are learned, thereby achieving accurate classification of weeds and crops.

[0068] Positioning unit: responsible for calculating the specific position of weeds and crops in space based on the results of target recognition, which involves coordinate transformation and geometric calculation mathematical operations. For example, the position and posture of the robot can be optimized by solving the least squares problem to make it closer to the target object.

[0069] The mathematical model of the recognition and positioning module is as follows:

[0070] Depth information calculation model: The depth camera calculates the distance from the object to the camera by measuring the flight time of infrared light. Assuming the speed of infrared light is c and the time difference between emission and reception is t, the distance d from the object to the camera can be expressed as: d = 0.5 × c × t;

[0071] Coordinate transformation model: In order to convert the two-dimensional image coordinates into three-dimensional space coordinates, coordinate transformation is required. Suppose the image coordinates are (u, v) and the corresponding three-dimensional space coordinates are (X, Y, Z), then the conversion can be performed using the following formula: [X, Y, Z] T =R×[x,y,z] T +T, where R and T are rotation matrix and translation vector, respectively, which are used to describe the transformation relationship from image coordinate system to world coordinate system;

[0072] Least squares model: In the positioning algorithm, the least squares method is often used to optimize the position and posture of the robot. Given a set of observation data (xi, yi) and the corresponding model prediction value (xi, yi), the optimal parameter θ is solved by minimizing the sum of squared errors: θ = argmin∑(xi-xi) 2 +(yi-yi) 2 .

[0073] In this embodiment, deep learning algorithms and image processing technology are used to quickly and accurately identify field weeds and crops. Compared with traditional manual identification methods, it is faster and can complete the weed identification work of a large area of ​​farmland in a shorter time, thereby improving the overall weeding efficiency. Through depth cameras, GPS and other technologies, the specific location and distribution of weeds can be determined, so that the weeding robot can directly operate on the weeds, avoiding accidental damage to crops.

[0074] Example 3

[0075] This embodiment is a further optimization based on the embodiment 1. Specifically, the trajectory planning module includes the following sub-units and mathematical algorithms:

[0076] Environmental perception unit: uses a variety of sensors to collect environmental information of farmland and provide basic data for subsequent path planning. The sensors include GPS receivers, heading and angle sensors, cameras, and lidars. The GPS receiver is used to obtain the robot's location information, the heading and angle sensors are used to determine the robot's direction and posture, and the cameras and lidars are used to perceive surrounding obstacles (such as crops, trees, stones) and the distribution of weeds. The sensors transmit the collected data to the data processing and decision-making unit in real time.

[0077] Data processing and decision-making unit: Process and analyze the data collected by the environmental perception layer, filter and fuse the sensor data to improve the accuracy and reliability of the data, and use the path planning algorithm to generate the optimal weeding path based on the processed data and the specific locations of weeds and crops in space. The coordinate transformation formula in the rasterized path planning and the artificial potential field method obstacle avoidance algorithm are used to calculate the terrain and obstacle distribution in the working area. At the same time, the path needs to be dynamically adjusted and optimized according to the actual situation to adapt to environmental changes;

[0078] The algorithm model in the trajectory planning and navigation optimization module is as follows:

[0079] Path tracking algorithm: Use PID control. The PID controller is a classic control algorithm used to accurately track the predetermined path. Its control formula is:

[0080]

[0081] Among them, u(t) is the output of the controller, that is, the expected turning angle of the robot's front wheels; e(t) is the current heading deviation, that is, the difference between the robot's current posture and the predetermined path; Kp, Ki, and Kd are the proportional, integral, and differential coefficients, respectively. The values ​​of these three parameters have an important influence on the performance of the controller and need to be adjusted and optimized according to the characteristics of the actual system;

[0082] Coordinate conversion formula in rasterized path planning: In the rasterized path planning method, the actual position coordinates of the robot are converted into coordinates in the grid coordinate system. Assuming the position coordinates of the robot in the working environment are (x, y), and the size of the grid is a×a, the grid coordinates (gx, gy) of the robot can be expressed as:

[0083]

[0084] in, represents the floor function, in this way, the position of the robot is mapped to the grid coordinate system for path planning and decision making;

[0085] Artificial potential field obstacle avoidance algorithm: In the artificial potential field obstacle avoidance algorithm, the potential field function is defined to calculate the virtual force exerted on the robot in the environment. The repulsive potential field function is usually defined as:

[0086]

[0087] Among them, Uo(q) is the repulsive potential field function, ρ(q,qobat) is the distance between the robot position q and the obstacle qobat, ρmax is the range of action of the repulsive potential field, and the gravitational potential field function can be defined as: Ug(q)=ρ(q,qgoal), where ρ(q,qgoal) is the distance between the robot position q and the target position qgoal. The virtual resultant force acting on the robot is calculated according to the gradients of the repulsive potential field and the gravitational potential field, thereby guiding the robot to avoid obstacles and move towards the target direction.

[0088] In this embodiment, the path planning algorithm can calculate the optimal weeding path according to the actual terrain, weed distribution and obstacles of the farmland, thereby avoiding blind walking and repeated operations of the robot during the operation, reducing invalid labor time, and greatly improving the weeding efficiency. In large-scale farmland, traditional manual weeding methods require a lot of time and manpower, while the intelligent weeding robot based on path planning can complete large-scale weeding tasks in a short time; and can monitor the robot's position and surrounding environment changes in real time. When encountering emergencies or new obstacles, the path planning can be adjusted in time to ensure that the robot always moves along the optimal path. This dynamic adjustment capability enables the robot to flexibly respond to various situations in complex farmland environments, further improving the weeding efficiency;

[0089] Reasonable path planning can prevent the robot from making too many round trips and detours in the field, thereby reducing energy consumption. For electric-driven weeding robots, saving energy means extending working time and driving range, reducing charging times and maintenance costs.

[0090] See also Figure 2 The present invention also discloses a smart agricultural weeding trajectory control method, comprising the following steps:

[0091] S1: Image acquisition and processing: The drone or ground robot cruises over or on the farmland, and the depth camera collects farmland image data in real time. The image processing unit performs denoising, filtering and edge detection on the raw image data collected by the depth camera, and aligns the two-dimensional image with the three-dimensional point cloud data to make them consistent in space, and distinguishes weeds and crops through image recognition technology;

[0092] S2: Identification and positioning: Analyze the processed images and point cloud data through deep learning technology to identify weeds and crops. By training a large amount of labeled data, learn the feature representation of different categories, so as to achieve accurate classification of weeds and crops. It is responsible for calculating the specific location of weeds and crops in space based on the results of target recognition;

[0093] S3: Farmland environment perception: Collect farmland environment information through sensors to provide basic data for subsequent trajectory planning. The sensors transmit the collected data to the data processing and decision-making unit in real time;

[0094] S4: Path trajectory planning: Filter and fuse the data collected by the sensor to improve the accuracy and reliability of the data. Based on the processed data and the specific locations of weeds and crops in space, the path planning algorithm is used to generate the optimal weeding trajectory;

[0095] S5: Collaborative weeding: According to the weeding trajectory, the robot's movement is controlled so that it moves along the predetermined trajectory and performs weeding operations according to the planned trajectory. The collaborative operation mode of multiple weeding robots is adopted. The robots share information and coordinate planning through wireless communication. A host robot is responsible for global planning and scheduling, and other slave robots perform specific weeding tasks according to the host's instructions;

[0096] S6: Human-computer interaction: Users enter the interactive interface through a mobile phone APP or computer, monitor the robot's working status, and issue instructions based on the robot's current position, mowed area, and remaining power.

[0097] In summary: the system integrates advanced sensors, image recognition, data analysis and other technologies to achieve intelligent identification and precise weeding of farmland weeds. Specifically, the system uses drones or ground robots equipped with high-definition cameras to monitor farmland in real time and accurately distinguish between crops and weeds through image recognition technology. The system automatically plans weeding trajectories based on the distribution of weeds and controls weeding equipment for precise operations, effectively avoiding damage to crops.

[0098] Compared with traditional manual identification methods, it is faster and can complete the weed identification work of large areas of farmland in a shorter time, thereby improving the overall weeding efficiency. Especially in large areas of farmland, traditional manual weeding methods require a lot of time and manpower, while intelligent weeding robots based on path planning can complete large-scale weeding tasks in a short time.

[0099] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A smart agricultural weeding trajectory control system, characterized in that: The trajectory control system comprises: Identification and positioning module: Use drones and ground robots to obtain information on the physiological status of plants, and use deep learning algorithms to continuously optimize the model through training data; Trajectory planning module: It uses sensors to collect environmental information of farmland to provide basic data for path planning. It uses the path planning algorithm to generate walking paths based on changes in field terrain, the distribution of obstacles, and information about weeded areas. It uses a reinforcement learning algorithm to allow the robot to continuously learn and optimize path planning strategies during actual operations. Navigation optimization module: controls the movement of the robot according to the path generated by the data processing and decision-making unit, so that it can travel along the predetermined path, including the motor drive control unit, steering control unit, and speed control unit, and performs weeding operations according to the planned path; Collaborative operation module: For large-scale farmland, multiple weeding robots are used to work together, and the robots share information and coordinate planning through wireless communication; Human-computer interaction module: provides users with an interface for interacting with the robot, allowing users to easily monitor the robot's working status, set parameters, and issue instructions.

2. According to claim 1, a smart agricultural weeding trajectory control system is characterized by: The identification and positioning module includes the following devices and subunits: Drones and ground robots: Equipped with depth camera units, the depth camera emits infrared light and receives reflected light to measure the distance from the object to the camera, thereby generating 3D point cloud data that contains the depth information of the scene; Image processing unit: The image processing unit is responsible for processing the raw image data collected by the depth camera to improve the image quality and recognition accuracy. At the same time, it is responsible for registering the two-dimensional image with the three-dimensional point cloud data; Target recognition unit: Use deep learning technology to analyze the processed images and point cloud data, use the target recognition algorithm to identify the weeds and crops, and learn the feature representation of different categories by training a large amount of labeled data to classify weeds and crops; Positioning unit: responsible for calculating the position of weeds and crops in space based on the results of target recognition.

3. The intelligent agricultural weeding trajectory control system according to claim 2 is characterized in that: The mathematical model of the identification and positioning module is as follows: Depth information calculation model: The depth camera calculates the distance from the object to the camera by measuring the flight time of infrared light. Assuming the speed of infrared light is c, and the time difference between emission and reception is t, the distance d from the object to the camera is expressed as: d = 0.5 × c × t; Least squares model: In the positioning algorithm, the least squares method is used to optimize the position and posture of the robot. Given a set of observation data (xi, yi) and the corresponding model prediction value (xi, yi), the optimal parameter θ is solved by minimizing the sum of squared errors: θ = argmin∑(xi-xi) 2 +(yi-yi) 2 .

4. The smart agricultural weeding trajectory control system according to claim 1 is characterized by: The trajectory planning module includes the following subunits and mathematical algorithms: Environmental perception unit: uses a variety of sensors to collect environmental information of farmland and provide basic data for subsequent path planning. The sensors include GPS receivers, heading and angle sensors, cameras, and lidars. The GPS receiver is used to obtain the robot's location information, the heading and angle sensors are used to determine the robot's direction and posture, and the cameras and lidars are used to perceive surrounding obstacles and the distribution of weeds. The sensors transmit the collected data to the data processing and decision-making unit in real time. Data processing and decision-making unit: Process and analyze the data collected by the environmental perception layer, filter and fuse the sensor data, and use the path planning algorithm to generate a weeding path based on the processed data and the position of weeds and crops in space. The coordinate transformation formula in the rasterized path planning and the artificial potential field obstacle avoidance algorithm are used to calculate the terrain and obstacle distribution in the working area.

5. The intelligent agricultural weeding trajectory control system according to claim 4 is characterized in that: The algorithm model in the trajectory planning and navigation optimization module is as follows: Path tracking algorithm: PID control is used to accurately track the predetermined path. The control formula is: Among them, u(t) is the output of the controller, that is, the expected turning angle of the robot's front wheel; e(t) is the current heading deviation, that is, the difference between the robot's current posture and the predetermined path; Kp, Ki, and Kd are the proportional, integral, and differential coefficients, respectively; Coordinate conversion formula in rasterized path planning: In the rasterized path planning method, the actual position coordinates of the robot are converted into coordinates in the grid coordinate system. Assuming the position coordinates of the robot in the working environment are (x, y), and the size of the grid is a×a, the grid coordinates (gx, gy) of the robot can be expressed as: in, Represents the floor function, in this way, the position of the robot is mapped to the grid coordinate system for path planning and decision-making.

6. The smart agricultural weeding trajectory control system according to claim 1 is characterized by: The navigation optimization module uses the integrated high-precision GLONASS global navigation satellite system, as well as the inertial measurement unit (IMU) and visual odometer-assisted navigation technology, and uses SLAM (simultaneous localization and mapping) technology to build and update the three-dimensional map of the field environment in real time, providing a basis for path planning and navigation.

7. The smart agricultural weeding trajectory control system according to claim 1 is characterized by: The collaborative operation module includes a communication and coordination unit, which is responsible for information exchange and coordinated control between multiple robots. Wireless communication technology (such as Wi-Fi, Bluetooth, ZigBee) is used to establish a communication network between robots. Through the communication protocol, the robots send each other location information, operation status, and task instruction data to perform collaborative operations. At the same time, a master-slave control strategy is adopted, with a host robot responsible for global planning and scheduling, and other slave robots perform specific weeding tasks according to the instructions of the host.

8. The smart agricultural weeding trajectory control system according to claim 1 is characterized by: In the human-computer interaction module, the user enters the system of the human-computer interaction module through a mobile phone APP or computer software to view the robot's current position, mowed area, and remaining power information, and mark the area that needs to be mowed on the map or set the robot's working mode.

9. A smart agricultural weeding trajectory control method, characterized in that: The following steps are involved: S1: Image acquisition and processing: The drone or ground robot cruises over or on the farmland, and the depth camera collects farmland image data in real time. The image processing unit performs denoising, filtering and edge detection on the raw image data collected by the depth camera, and aligns the two-dimensional image with the three-dimensional point cloud data to make them consistent in space, and distinguishes weeds and crops through image recognition technology; S2: Identification and positioning: Analyze the processed images and point cloud data through deep learning technology to identify weeds and crops. Through training and labeling data, learn the feature representation of different categories to accurately classify weeds and crops. It is responsible for calculating the position of weeds and crops in space based on the results of target recognition. S3: Farmland environment perception: Collect farmland environment information through sensors to provide basic data for subsequent trajectory planning. The sensors transmit the collected data to the data processing and decision-making unit in real time; S4: Path trajectory planning: Filter and fuse the data collected by the sensor, and use the path planning algorithm to generate a weeding trajectory based on the processed data and the positions of weeds and crops in space; S5: Collaborative weeding: According to the weeding trajectory, the robot's movement is controlled so that it moves along the predetermined trajectory and performs weeding operations according to the planned trajectory. The collaborative operation mode of multiple weeding robots is adopted. The robots share information and coordinate planning through wireless communication. A host robot is responsible for global planning and scheduling, and other slave robots perform specific weeding tasks according to the host's instructions; S6: Human-computer interaction: Users enter the interactive interface through a mobile phone APP or computer, monitor the robot's working status, and issue instructions based on the robot's current position, mowed area, and remaining power.

Images