Crop water demand judgment irrigation method and system based on image recognition

Abstract

The invention provides a crop water demand judgment irrigation method and system based on image recognition, and relates to the technical field of agricultural irrigation. An area to be irrigated is divided into a plurality of subareas, a plurality of image acquisition points are set in each subarea, and an optimal cruise path is calculated according to geographic coordinates of the image acquisition points; an unmanned aerial vehicle or a track robot carrying an image acquisition terminal is used for acquiring crop image data at the image acquisition points of all the subareas along the optimal cruise path; the method comprises the following steps: standardizing crop image data, extracting a spectral feature value for judging the crop moisture condition, inputting a pre-trained water demand judgment model to obtain a water demand index at each image acquisition point, and finally generating a corresponding control instruction based on the water demand index at each image acquisition point, and the irrigation execution unit receives the control instruction and completes irrigation operation. According to the invention, mechanical, automatic and intelligent precise irrigation based on image recognition can be realized, and efficient utilization of water resources is realized.

Description

Technical Field

[0001] This invention relates to the field of agricultural irrigation technology, specifically to an irrigation method and system for determining crop water requirements based on image recognition. Background Technology

[0002] Agricultural irrigation is a core link in ensuring food production. Traditional irrigation management relies heavily on farmers' experience and judgment, that is, deciding whether to irrigate by observing the dryness and wetness of the soil surface or the degree of wilting of crop leaves. This method is highly subjective and inaccurate, which can easily lead to untimely or excessive irrigation, resulting in serious waste of water resources, decline in crop yield and quality, and may also cause ecological problems such as soil salinization and nutrient loss.

[0003] To improve irrigation precision, modern agriculture has introduced sensor-based automated irrigation technology. This technology primarily uses soil moisture sensors deployed in the field to monitor the real-time volumetric water content of the soil in the crop root zone, automatically activating the irrigation system when the moisture content falls below a preset threshold. However, this method has several inherent design flaws: First, achieving accurate monitoring of large areas of farmland requires a densely deployed sensor network, resulting in high hardware costs. Sensor installation, wiring, and subsequent maintenance are cumbersome and easily damaged by agricultural machinery operations or animal gnawing. Furthermore, soil moisture sensors only reflect localized information at their installation points, failing to comprehensively represent the entire field, especially given the spatial variability of moisture across different soil types and terrain variations, leading to "point-to-area" decision-making errors. Moreover, this irrigation method has a lag effect; soil moisture changes lag behind crop water stress. By the time the sensor detects dry soil, the crop may have already suffered from water stress for some time, affecting its growth, making the monitoring results unpredictable.

[0004] In recent years, with the development of smart agriculture, irrigation decision-making systems based on meteorological station data have emerged. Although this method can estimate water use at the regional scale, it cannot perceive the specific crop water status at the field scale, nor can it cope with the spatial differences within the field.

[0005] With the advancement of image recognition and machine learning technologies, new approaches have been made possible for the non-destructive monitoring of crop physiological states. When crops are subjected to water stress, their canopy morphology, leaf tilt angle, and color (especially the spectral reflectance characteristics in the near-infrared band) undergo subtle but measurable changes. Existing technologies have attempted to use fixed cameras for continuous monitoring of crops or to take photos intermittently using handheld devices. However, fixed cameras have a fixed field of view, limited coverage, and are difficult to avoid analysis errors caused by image distortion and angle changes. Manual inspection methods are inefficient and cannot achieve large-scale, standardized data collection. Summary of the Invention

[0006] Purpose of the invention: The first purpose of this invention is to provide a crop water requirement determination irrigation method based on image recognition for precise zoning irrigation; the second purpose is to provide a crop water requirement determination irrigation system based on image recognition.

[0007] Technical solution: An irrigation method for determining crop water requirement based on image recognition, comprising the following steps:

[0008] S1. Divide the area to be irrigated into several zones and establish a mapping between the unique identifier of each zone and the corresponding geographic coordinate range;

[0009] S2. Set up several image acquisition points in each partition and calculate the optimal cruise route based on the geographical coordinates of the image acquisition points;

[0010] S3. Use a mobile carrier platform equipped with an image acquisition terminal to collect crop image data at each partition image acquisition point along the optimal cruise path;

[0011] S4. Standardize the crop image data;

[0012] S5. Extract spectral feature values ​​for judging crop water status from the standardized crop image data, input them into the pre-trained water demand judgment model, and output the water demand index at each image acquisition point.

[0013] S6. Based on the water demand index at each image acquisition point, generate corresponding control commands, and the irrigation execution unit receives the control commands and completes the irrigation operation.

[0014] Specifically, the area to be irrigated is divided into several zones, including:

[0015] The irrigated area is divided into several zones using either manual or automatic grid division. In the manual division method, the irrigated area is divided into several irregular polygonal zones based on crop planting blocks, soil moisture differences, and terrain conditions. In the automatic grid division method, after setting the area of ​​a single grid, the irrigated area is automatically divided into a regular grid array using an automatic gridding tool, with each grid serving as a zone. Each zone is assigned a unique identifier and mapped to its geographic coordinate range.

[0016] Specifically, the mobile carrier platform is either a track robot platform or a drone platform.

[0017] Specifically, the track robot platform includes a track system fixed to the area to be irrigated and a robot body set on the track system. The image acquisition terminal is connected to the robot body through a quick-release interface. The robot body moves along the optimal cruising path on the track system by motor drive, and simultaneously acquires crop image data.

[0018] Specifically, the drone platform includes a take-off and landing platform fixed to the area to be irrigated and a drone body set on the take-off and landing platform. The image acquisition terminal is connected to the drone body through a quick-release interface. The drone body realizes automatic navigation and crop image data acquisition by importing the optimal cruise path into the flight control system.

[0019] Specifically, the image acquisition terminal includes an image acquisition module, a positioning module, and a communication module. The image acquisition module uses a multispectral camera to acquire crop image data, the positioning module obtains the spatial coordinate information of the image acquisition terminal in real time, and the communication module completes the packaging and transmission of crop image data and corresponding spatial coordinate information.

[0020] Specifically, standardization processing of crop image data includes lens distortion correction, geometric correction, and image enhancement.

[0021] Specifically, the spectral characteristic values ​​include the normalized vegetation index and the normalized water index.

[0022] Specifically, the pre-trained water demand judgment model construction and training process includes:

[0023] A publicly available crop image dataset is obtained, and water requirement indicators are labeled on the crop image dataset. The dataset is then input into a machine learning model for training to obtain a water requirement judgment model that associates the spectral features of the images with water requirement indicators, wherein the water requirement indicators are normalized numerical indicators or discrete level indicators.

[0024] The present invention also provides an irrigation system for determining crop water requirement based on image recognition, comprising:

[0025] Irrigation Zoning Module: Used to divide the area to be irrigated into several zones and establish a mapping between the unique identifier of each zone and the corresponding geographic coordinate range;

[0026] Image acquisition point selection module: used to set several image acquisition points in each partition and calculate the optimal cruise path based on the geographical coordinates of the image acquisition points;

[0027] Image acquisition module: Utilizes a mobile carrier platform equipped with an image acquisition terminal to collect crop image data at image acquisition points in each zone along the optimal cruising path;

[0028] Image standardization module: used to standardize crop image data;

[0029] Water requirement assessment module: This module extracts spectral feature values ​​from standardized crop image data to assess crop moisture status, inputs them into a pre-trained water requirement assessment model, and outputs water requirement indicators for each image acquisition point.

[0030] Irrigation execution module: Based on the water demand index at each image acquisition point, it generates corresponding control commands. The irrigation execution unit receives the control commands and completes the irrigation operation.

[0031] Beneficial effects: Compared with the prior art, the significant effects of the present invention are:

[0032] 1. This invention directly monitors the physiological morphology and spectral characteristics of crop canopy through image recognition technology. It can detect water stress as early as possible before water molecule deficiency causes visible damage to crop growth, achieving direct and non-destructive diagnosis of crop water demand. It overcomes the inherent defects of traditional soil moisture sensors, such as lagging monitoring and partial measurement, making irrigation decisions more scientific and accurate, and effectively avoiding misjudgment from the source.

[0033] 2. By using data collected from high-precision zoning, this invention can generate a field water demand distribution map and drive the irrigation system to perform variable irrigation or zoning irrigation. This innovation means that water resources can be precisely applied to the areas that need them most, completely eliminating water waste under traditional uniform irrigation methods and effectively improving water resource utilization.

[0034] 3. This technical solution innovatively proposes an image acquisition terminal that can be adapted to multiple mobile platforms (rail robots / drones). Users can flexibly choose the most suitable deployment scheme according to the size of the field, terrain, crop type and budget. For greenhouses and contiguous flat land, a stable and reliable rail-based scheme can be selected; for complex terrain and large fields, a mobile and flexible drone scheme can be selected. This dual-mode flexible switching design greatly expands the application scenarios and market potential of this invention. Attached Figure Description

[0035] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention.

[0036] Figure 2 This is a flowchart of the image acquisition terminal operation in Embodiment 1 of the present invention.

[0037] Figure 3 This is a flowchart of the crop image data processing workflow of Embodiment 1 of the present invention.

[0038] Figure 4 This is a flowchart of the crop irrigation process in Embodiment 1 of the present invention. Detailed Implementation

[0039] A preferred embodiment of the present invention will be further described below with reference to the accompanying drawings.

[0040] Example 1

[0041] Please see Figure 1As shown, this embodiment provides an irrigation method for determining crop water requirements based on image recognition, including the following steps:

[0042] S1. Divide the area to be irrigated into several zones and establish a mapping between the unique identifier of each zone and the corresponding geographical coordinate range.

[0043] In the initial stage, users divide the electronic map or digital boundary map of the area to be irrigated into several zones using either manual or automatic grid division. In the manual division method, the area to be irrigated is divided into several irregular polygonal zones based on crop planting blocks, soil moisture differences, and terrain conditions. In the automatic grid division method, after setting the area of ​​a single grid, the automatic gridding tool automatically divides the area to be irrigated into a regular grid array, with each grid serving as a zone. A unique identifier (ID) is assigned to each zone, and a mapping is established between the ID and the geographic coordinate range of that zone.

[0044] S2. Set several image acquisition points in each partition and calculate the optimal cruise route based on the geographical coordinates of the image acquisition points.

[0045] Specifically, image acquisition points are usually selected from the geometric center of the partition or feature points that can represent the overall moisture status of the partition. Based on the characteristics of the mobile carrier platform, existing path planning algorithms such as A* algorithm and artificial potential field method are used to calculate an optimal cruise path based on the distribution of image acquisition points to ensure the highest acquisition efficiency and lowest energy consumption.

[0046] S3. Use a mobile carrier platform equipped with an image acquisition terminal to collect crop image data at each partition image acquisition point along the optimal cruise path.

[0047] The image acquisition terminal includes an image acquisition module, a positioning module, and a communication module. The image acquisition module uses a multispectral camera to acquire crop image data, the positioning module acquires the spatial coordinate information of the image acquisition terminal in real time, and the communication module packages and sends the crop image data and corresponding spatial coordinate information. The image acquisition terminal is mechanically and electrically connected to the mobile carrier platform through a standardized quick-release interface.

[0048] Please refer to Figure 2As shown, after setting the image acquisition points and planning the corresponding optimal cruise path, the mobile carrier platform receives the control command corresponding to the optimal cruise path, positions itself above the first image acquisition point, and the image acquisition terminal vertically downwards to capture images of the crop canopy, obtaining image data of the crop canopy within that zone. Simultaneously with image acquisition, the communication module records the current high-precision coordinates, timestamp, and corresponding zone ID. Subsequently, the module packages the image data, spatial coordinates, timestamp, and zone ID into a standard data packet. After completing the acquisition of one zone, the image acquisition terminal moves via the mobile carrier platform to the acquisition point of the next zone, until all zones have been traversed.

[0049] The following describes the specific implementation scheme of the mobile carrier platform equipped with an image acquisition terminal used in this embodiment.

[0050] The mobile carrier platform selected in this embodiment is either a track robot platform or a drone platform. The track robot platform is suitable for flat and regular plots of land, while the drone platform is suitable for large areas and complex terrain.

[0051] The track robot platform includes a track system fixed to the area to be irrigated and a robot body set on the track system. The track system can be any of the following: rigid slide rail, flexible cable track, or elevated truss structure. The image acquisition terminal is connected to the robot body through a quick-release interface. The robot body moves along the optimal cruising path on the track system by motor drive, and simultaneously acquires crop image data.

[0052] A platform drive module is provided in the robot body. The robot body moves and displaces along the track system by being driven. In this embodiment, the platform drive module includes a drive unit, a transmission mechanism and a walking mechanism.

[0053] The drive unit uses a waterproof and dustproof DC servo motor or stepper motor as the power source; preferably, a motor with an integrated encoder is used to accurately feedback the motor speed and angle, thereby realizing closed-loop control of the robot's displacement.

[0054] Depending on the track type and load requirements, different transmission schemes can be adopted for the transmission mechanism: One scheme is to use a rack and pinion drive. If the track system is a rigid metal slide rail, a rack can be fixedly installed on one side of the track. The output shaft of the motor is connected to a small gear through a reduction gearbox. The small gear meshes with the rack on the track. When the motor rotates, the rotational motion is converted into linear motion of the robot body along the track through the meshing of the gear and rack. The advantages of this scheme are high transmission accuracy, no slippage, and strong load-bearing capacity. Another scheme is to use friction wheel drive. The output shaft of the motor directly drives one or more rubber friction wheels. The friction wheels apply a certain clamping force to the top or side of the track, and the robot moves by relying on friction. This scheme has a simple structure and low noise. It only requires attention to the wear and slippage of the friction wheels.

[0055] The walking mechanism includes an active walking wheel set and a driven support wheel set. The active walking wheel set is connected to the aforementioned transmission mechanism and directly provides driving force. The driven support wheel set usually contains multiple V-shaped wheels or grooved wheels, which clamp the track from both sides or from above. They mainly serve to support the robot body and prevent tipping and derailment, ensuring that the robot can operate smoothly even in windy or uneven track conditions.

[0056] When implementing displacement control, a displacement command (such as "move forward 5.5 meters") is sent to the robot body through the communication module; after receiving the command from the robot's built-in controller (such as a PLC or microcontroller), the drive motor rotates at a preset speed; the integrated encoder monitors the actual number of rotations of the motor in real time, and the closed-loop control algorithm (such as PID control) ensures the accuracy of the displacement and the accuracy of the stopping position, so that the image acquisition terminal can accurately reach each preset acquisition point.

[0057] In this embodiment, the platform drive module can also integrate a current detection module; when an abnormal increase in motor current is detected (indicating a possible obstacle or jamming), the controller can immediately trigger an emergency stop and send a fault alarm message to the data center to prevent equipment damage.

[0058] The robot body can be equipped with a platform power module (such as a lithium battery pack), which can automatically charge at a specific position on the track via a sliding contact line or a contact charging pile, thereby enabling long-term unattended operation.

[0059] In this embodiment, the drone platform adopts a split design, including a take-off and landing platform fixed to the area to be irrigated and a drone body set on the take-off and landing platform. The image acquisition terminal is connected to the drone body through a quick-release interface. The drone body realizes automatic navigation and crop image data acquisition by importing the optimal cruise path into the flight control system.

[0060] The drone is equipped with a flight control system that provides flight power and controls the rotor for flight attitude and stability, a drone navigation and communication module that receives flight path instructions and transmits status information, and a drone power module that supplies power to the drone.

[0061] The flight control system is the core controller of the UAV. It is essentially an embedded computer system responsible for processing sensor data and controlling flight attitude and stability in real time. It includes the main controller, which uses a high-performance microprocessor (such as the ARM Cortex series) and runs a real-time operating system (RTOS) to ensure timely response to control commands. The flight control system integrates an inertial measurement unit (IMU) with accelerometers and gyroscopes to monitor the UAV's three-axis acceleration and angular velocity in real time. It is the most critical sensor for sensing the UAV's attitude and motion.

[0062] The flight control system also integrates a barometer for measuring ambient air pressure, using pressure differences to estimate and maintain the drone's flight altitude. It also integrates an electronic speed controller (ESC), which receives commands from the main controller and precisely controls the speed of each brushless motor, thereby adjusting the rotor lift and ultimately enabling the drone's ascent, descent, pitch, roll, and yaw maneuvers. The flight control system can incorporate PID (proportional-integral-derivative) control algorithms or adaptive control algorithms. It continuously compares the actual flight status (current attitude, altitude, position) measured by sensors such as the IMU and barometer with the target status (target point coordinates, target altitude, target attitude) set in the flight mission plan, and adjusts the commands sent to each motor in real time based on the calculated error value. This achieves high flight stability and precise hovering (hovering accuracy down to the centimeter level), providing a stable shooting platform for image acquisition.

[0063] The UAV body may also include a navigation and positioning system integrated into the UAV navigation and communication module. The navigation and positioning system is used to receive real-time differential positioning signals, so that the positioning accuracy reaches the centimeter level. The navigation and positioning system may adopt an RTK real-time dynamic carrier phase differential positioning system and / or a PPK dynamic post-processing differential positioning system.

[0064] The drone itself can also be equipped with an obstacle avoidance system, which includes one or more of visual sensors, ultrasonic sensors, or lidar, used to detect obstacles on the flight path and achieve automatic detour or hovering; visual sensors are used for detecting and identifying obstacles in front; ultrasonic sensors are used for accurate distance measurement of ground altitude at low altitudes (especially during takeoff and landing); forward and backward or multi-directional lidar is used to construct point cloud maps of the surrounding environment to achieve three-dimensional obstacle avoidance; the flight control system processes the data from the above sensors in real time, and when obstacles are detected on the flight path, it will automatically trigger detour or hovering strategies to ensure operational safety.

[0065] The UAV body also includes a mission payload interface, which is used to connect mechanically, electrically and data with the image acquisition terminal mounted on the UAV body through a standardized quick-release interface. The mission payload interface includes at least a power output end and a data communication end, which are used to charge or supply power to the image acquisition terminal during flight, and to perform data interaction at the same time.

[0066] In this embodiment, the image acquisition terminal is also equipped with a gimbal. One end of the gimbal is used to mount the image acquisition module and adjust its shooting posture to maintain a standard vertical downward posture. The other end of the gimbal is connected to the mobile support platform through a standardized quick-release interface. In this embodiment, the gimbal is a three-axis stabilized gimbal, whose mechanical structure includes three parts, each providing precise control in three rotational dimensions. The pitch axis controls the forward and backward pitch movement of the image acquisition module, adjusting the tilt angle of the lens relative to the horizontal plane; the roll axis controls the left and right tilt movement of the image acquisition module, ensuring that it can return to horizontal even when tilted; the yaw axis controls the horizontal rotation of the image acquisition module, adjusting its shooting direction.

[0067] The base of the gimbal integrates a standardized quick-release female connector; this connector not only provides a robust mechanical connection, but its internal electrical connector also powers the gimbal itself and establishes a data channel; this allows the entire image acquisition terminal to function as a complete, plug-and-play functional module that can quickly switch between different mobile platforms without complicated disconnection and debugging.

[0068] S4. Standardize the crop image data.

[0069] Please refer to Figure 3 As shown, after receiving crop image data packets from all partitions, the system first verifies the integrity of the data and then automatically categorizes and stores the crop image data into the corresponding partition database based on the "partition ID" field in the data packet. This strong correlation between "partition" and "data" lays the data foundation for the subsequent generation of water requirement level maps based on partitions and the precise execution of partition-based variable irrigation.

[0070] Specifically, the system continuously listens for and receives data packets from one or more image acquisition terminals via wireless networks (such as 4G / 5G, LoRaWAN, and Wi-Fi). Each data packet contains a crop canopy image, corresponding centimeter-level precision spatial coordinates, a timestamp, and a terminal / partition identifier. The system verifies the data packets' validity (e.g., CRC check). After successful verification, the image data, coordinate data, and other metadata are separated and stored in a temporary database or message queue for further processing.

[0071] After data separation, lens distortion correction, geometric correction, and image enhancement are performed on the crop images.

[0072] Distortion correction: Based on the camera's intrinsic parameters and lens distortion parameters, the image is corrected to eliminate barrel or pincushion distortion caused by the lens's optical characteristics.

[0073] Geometric correction: Based on the posture information at the time of shooting, the image is rotated and affine transformed to ensure that the image perspective is standard vertically downward, so that images taken at different times and from different angles are comparable.

[0074] Image enhancement: Perform operations such as contrast stretching and histogram equalization on the image as needed to highlight the characteristic information of the crop canopy.

[0075] S5. Extract spectral feature values ​​for judging crop moisture status from the standardized crop image data, input them into the pre-trained water demand judgment model, and output the water demand index at each image acquisition point.

[0076] Please refer to Figure 3 As shown, before training the model, feature extraction is performed first. The spectral feature values ​​mainly include the normalized vegetation index and the normalized water index, and the calculation formula is as follows:

[0077]

[0078]

[0079] In the formula: The Normalized Difference Vegetation Index (NDVI) is primarily used to assess plant greenness and biomass. The normalized water index is more sensitive to the water content inside the leaves. This refers to the reflectance value in the near-infrared band. The reflectance value is for the red band. This represents the reflectance value for the green band.

[0080] The extracted spectral feature values ​​are input into a pre-trained machine learning model (such as CNN). The training process is as follows: obtain a publicly available crop image dataset, label the crop image dataset with water requirement indicators, and then input it into the machine learning model for training. This model can learn the complex nonlinear relationship between the canopy phenotypic features (such as leaf color, texture, curl, etc.) and spectral features of crops under different water stresses. Then, a water requirement judgment model that associates the image spectral features with water requirement indicators is obtained. The water requirement indicators are normalized numerical indicators or discrete level indicators, such as continuous water requirement index (ranging from 0 to 1, with larger values ​​indicating more urgent water requirements) or discrete water requirement levels (divided into four levels: "severe water shortage", "mild water shortage", "adequate water", and "excessive water").

[0081] Machine learning models can be deployed on cloud platforms or servers to receive and collect data across multiple scenarios, issue control commands, and receive and send user interaction information.

[0082] S6. Based on the water demand index at each image acquisition point, generate corresponding control commands, and the irrigation execution unit receives the control commands and completes the irrigation operation.

[0083] Please refer to Figure 3 As shown in this embodiment, by combining the spatial coordinates of each image acquisition point with its corresponding water demand index or water demand level, and using spatial interpolation algorithms (such as Kriging interpolation or inverse distance weighted interpolation IDW), a continuous and visualized water demand level distribution map covering the entire irrigated area can be generated based on the data of these discrete points. This distribution map can intuitively show the spatial differences in field water conditions.

[0084] In this embodiment, the irrigation execution unit is a zone-controlled irrigation network, including multiple solenoid valves corresponding to each zone, a water distribution network, and a central controller. Control commands can independently control the opening and closing of specific solenoid valves. The central controller is a programmable logic controller (PLC) or an industrial-grade IoT gateway, responsible for receiving irrigation control commands from the cloud or local terminal. Each zone corresponds to at least one solenoid valve, preferably a normally closed diaphragm valve powered by 24V DC. The solenoid valve is electrically connected to the central controller via a cable, receiving switching control signals from it. The water distribution network consists of a main water supply pipe, branch pipes, and capillary pipes (drip irrigation tape / micro-sprinkler tape), responsible for delivering water to the field. The inlet end of the solenoid valve is connected to the main water supply pipe, and the outlet end is connected to the branch pipe and capillary pipes (such as drip irrigation tape, sprinklers, etc.) leading to the corresponding zone. The solenoid valves and the mobile support platform can be powered by the same power supply unit or by multiple distributed power supply units.

[0085] The strategy for generating irrigation control commands is as follows: based on the obtained visualized water demand level distribution map, the average water demand level is calculated for each logical partition, and the average water demand level of the partition is compared with the preset irrigation threshold. For partitions whose water demand level exceeds the threshold, the required irrigation duration or irrigation amount for the partition is calculated linearly or nonlinearly based on the value of its water demand level, and commands to control the opening of the corresponding solenoid valve are generated.

[0086] Please refer to Figure 4As shown, during actual operation, the central controller issues a command to its digital output (DO) channel according to the control instructions, outputting a 24V DC signal to the designated solenoid valve (such as the valve with ID A05). This signal drives the solenoid coil inside the solenoid valve to generate a magnetic field, overcoming the spring force and fluid pressure, lifting the valve core, opening the valve passage, and initiating irrigation. After a preset time (e.g., 300 seconds), the central controller cuts off the output signal, the solenoid valve closes under the action of the spring, and irrigation stops.

[0087] Because each solenoid valve is independently connected to a different output channel of the central controller, the system can control irrigation in different zones concurrently and independently, achieving true "on-demand irrigation" and avoiding waste of water resources.

[0088] Solenoid valves can integrate Hall effect sensors or microswitches to detect the actual open or closed state of the valve (whether it is open or closed). This status signal can be fed back to the central controller to verify whether the commands are executed correctly, thus achieving closed-loop control.

[0089] In this embodiment, flow sensors can also be installed on the main water supply pipe or each branch pipe. The system can compare the opening command of the solenoid valve with the real-time monitored flow data. If a valve is commanded to open but the flow is zero, it may indicate that there is a pipe blockage or rupture in that section. Conversely, if the valve is closed but there is still flow, it indicates that the valve is faulty and leaking. The system can then issue an alarm immediately.

[0090] Based on the above description, the final report will be pushed to the user terminal, which includes a visualized water demand level distribution map, irrigation decision details (set of irrigation control instructions) for the area to be irrigated, and relevant system log information.

[0091] Example 2

[0092] This embodiment provides an image recognition-based crop water requirement determination irrigation system, including:

[0093] Irrigation Zoning Module: Used to divide the area to be irrigated into several zones and establish a mapping between the unique identifier of each zone and the corresponding geographic coordinate range;

[0094] Image acquisition point selection module: used to set several image acquisition points in each partition and calculate the optimal cruise path based on the geographical coordinates of the image acquisition points;

[0095] Image acquisition module: Utilizes a mobile carrier platform equipped with an image acquisition terminal to collect crop image data at image acquisition points in each zone along the optimal cruising path;

[0096] Image standardization module: used to standardize crop image data;

[0097] Water requirement assessment module: This module extracts spectral feature values ​​from standardized crop image data to assess crop moisture status, inputs them into a pre-trained water requirement assessment model, and outputs water requirement indicators for each image acquisition point.

[0098] Irrigation execution module: Based on the water demand index at each image acquisition point, it generates corresponding control commands. The irrigation execution unit receives the control commands and completes the irrigation operation.

Claims

1. A method for determining crop water requirement and irrigation based on image recognition, characterized in that, Includes the following steps: S1. Divide the area to be irrigated into several zones and establish a mapping between the unique identifier of each zone and the corresponding geographic coordinate range; S2. Set up several image acquisition points in each partition and calculate the optimal cruise route based on the geographical coordinates of the image acquisition points; S3. Use a mobile carrier platform equipped with an image acquisition terminal to collect crop image data at each partition image acquisition point along the optimal cruise path; S4. Standardize the crop image data; S5. Extract spectral feature values ​​for judging crop water status from the standardized crop image data, input them into the pre-trained water demand judgment model, and output the water demand index at each image acquisition point. S6. Based on the water demand index at each image acquisition point, generate corresponding control commands, and the irrigation execution unit receives the control commands and completes the irrigation operation.

2. The crop water requirement determination and irrigation method based on image recognition according to claim 1, characterized in that, The process of dividing the area to be irrigated into several zones includes: The irrigated area is divided into several zones using either manual or automatic grid division. In the manual division method, the irrigated area is divided into several irregular polygonal zones based on crop planting blocks, soil moisture differences, and terrain conditions. In the automatic grid division method, after setting the area of ​​a single grid, the irrigated area is automatically divided into a regular grid array using an automatic gridding tool, with each grid serving as a zone. Each zone is assigned a unique identifier and mapped to its geographic coordinate range.

3. The crop water requirement determination and irrigation method based on image recognition according to claim 1, characterized in that, The mobile carrier platform is either a track robot platform or an unmanned aerial vehicle (UAV) platform.

4. The crop water requirement determination and irrigation method based on image recognition according to claim 3, characterized in that, The track robot platform includes a track system fixed to the area to be irrigated and a robot body set on the track system. The image acquisition terminal is connected to the robot body through a quick-release interface. The robot body moves along the optimal cruising path on the track system by motor drive, and simultaneously acquires crop image data.

5. The crop water requirement determination and irrigation method based on image recognition according to claim 3, characterized in that, The drone platform includes a take-off and landing platform fixed to the area to be irrigated and a drone body set on the take-off and landing platform. The image acquisition terminal is connected to the drone body through a quick-release interface. The drone body realizes automatic navigation and crop image data acquisition by importing the optimal cruise path into the flight control system.

6. The crop water requirement determination and irrigation method based on image recognition according to claim 1, characterized in that, The image acquisition terminal includes an image acquisition module, a positioning module, and a communication module. The image acquisition module uses a multispectral camera to acquire crop image data, the positioning module obtains the spatial coordinate information of the image acquisition terminal in real time, and the communication module completes the packaging and transmission of crop image data and corresponding spatial coordinate information.

7. The crop water requirement determination and irrigation method based on image recognition according to claim 1, characterized in that, The standardization processing of crop image data includes: lens distortion correction, geometric correction, and image enhancement of crop images.

8. The crop water requirement determination irrigation method based on image recognition according to claim 1, characterized in that, The spectral characteristic values ​​include the normalized vegetation index and the normalized water index.

9. The crop water requirement determination and irrigation method based on image recognition according to claim 1, characterized in that, The pre-trained water demand judgment model construction and training process includes: A publicly available crop image dataset is obtained, and water requirement indicators are labeled on the crop image dataset. The dataset is then input into a machine learning model for training to obtain a water requirement judgment model that associates the spectral features of the images with water requirement indicators, wherein the water requirement indicators are normalized numerical indicators or discrete level indicators.

10. An irrigation system for determining crop water requirement based on image recognition, characterized in that, include: Irrigation Zoning Module: Used to divide the area to be irrigated into several zones and establish a mapping between the unique identifier of each zone and the corresponding geographic coordinate range; Image acquisition point selection module: used to set several image acquisition points in each partition and calculate the optimal cruise path based on the geographical coordinates of the image acquisition points; Image acquisition module: Utilizes a mobile carrier platform equipped with an image acquisition terminal to collect crop image data at image acquisition points in each zone along the optimal cruising path; Image standardization module: used to standardize crop image data; Water requirement assessment module: This module extracts spectral feature values ​​from standardized crop image data to assess crop moisture status, inputs them into a pre-trained water requirement assessment model, and outputs water requirement indicators for each image acquisition point. Irrigation execution module: Based on the water demand index at each image acquisition point, it generates corresponding control commands. The irrigation execution unit receives the control commands and completes the irrigation operation.

Images