Method, system and equipment for extracting geometric parameters of wheat canopy and storage medium

Abstract

The invention discloses a wheat canopy geometric parameter extraction method, system and device and a storage medium. The method comprises the following steps: constructing a multi-modal sensor integrated acquisition system; a pitch axis anti-interference control strategy based on ESO-LQR is adopted for the acquisition system; the method comprises the following steps: acquiring point cloud data and image data of wheat by using an acquisition system, performing two-dimensional character extraction based on a color index by using the image data, and performing three-dimensional canopy structure distribution character extraction by using the point cloud data and the image data. The pitch axis anti-interference control strategy based on the ESO-LQR comprises the following steps: establishing a pitch axis dynamical model and carrying out disturbance analysis; lumped disturbance is expanded into a system state, a third-order expansion state observer ESO is designed, and a disturbance estimation value is generated; constructing a state vector for an optimal controller LQR, designing a cost function, solving an optimal feedback matrix through a Riccati equation, and generating a control quantity; and compensating the disturbance estimation value into the control quantity to generate final control input.

Description

Technical Field

[0001] The present invention relates to the field of agricultural detection technology, in particular to a method, system, equipment and storage medium for extracting geometric parameters of a wheat canopy. Background Art

[0002] Wheat is one of the world's major food crops, and its cultivated area and yield are crucial to farmers' economic benefits and national food security. Therefore, rapidly and accurately phenotyping wheat growth traits has become a hot topic in modern agricultural research. Geometric parameters such as wheat canopy cover, flatness, and yellowing index are key indicators for assessing crop growth status.

[0003] Traditional measurement methods rely on manual vision or single sensors (such as handheld laser rangefinders). These methods suffer from low efficiency, limited data dimensions, and poor dynamic adaptability, making them unable to meet the demands of modern agriculture for efficient and accurate data. Furthermore, complex field environments, such as gravel and vehicle vibration, often affect the stable operation of sensors, making efficient data collection difficult. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a method, system, equipment and storage medium for extracting wheat canopy geometric parameters. It utilizes an acquisition system integrated with multimodal sensors and adopts an ESO-LQR-based pitch axis anti-disturbance control strategy to overcome the defects of the existing technology such as the use of a single sensor resulting in a single data dimension and inaccurate collected data caused by field vibration.

[0005] To achieve the above-mentioned objectives, the present invention provides, on the one hand, a method for extracting geometric parameters of wheat canopy, comprising the following steps: constructing an acquisition system integrating multimodal sensors, the acquisition system comprising a stabilized acquisition device and multiple sensors, the multiple sensors comprising at least a camera, a tightly coupled lidar and an inertial measurement unit (IMU), the stabilized acquisition device comprising a three-axis imaging gimbal and a shock-absorbing fixing device, and the multiple sensors being mounted on the three-axis imaging gimbal; adopting an ESO-LQR-based pitch axis anti-disturbance control strategy for the acquisition system; utilizing the acquisition system to collect point cloud data and image data of wheat, utilizing the image data to perform two-dimensional trait extraction based on color index, and utilizing the point cloud data and the image data to perform three-dimensional canopy structure distribution trait extraction.

[0006] The pitch axis anti-disturbance control strategy based on ESO-LQR further includes the following steps: establishing a pitch axis dynamic model and performing disturbance analysis, wherein the pitch axis dynamic model includes lumped disturbances, and the lumped disturbances include low-frequency disturbances, high-frequency disturbances and coupled disturbances; expanding the lumped disturbances into system states, designing a third-order extended state observer (ESO), and generating disturbance estimates; constructing a state vector for the optimal controller LQR, designing a cost function, and solving the optimal feedback matrix through the Riccati equation to generate a control variable; and compensating the disturbance estimate into the control variable to generate a final control input.

[0007] The pitch axis anti-disturbance control strategy based on ESO-LQR for the acquisition system further includes the following steps: angle initialization, including manual angle initialization and automatic angle initialization, wherein:

[0008] The manual angle initialization includes: controlling the pitch axis angle by sending a PWM signal or an S.Bus signal through the remote control, and setting and locking the angles of the pitch axis and the yaw axis according to the angle value fed back by the inertial measurement unit IMU;

[0009] The automatic angle initialization includes: setting a target angle of the pitch axis, obtaining the current angle of the pitch axis, and automatically adjusting the pitch angle of the acquisition system according to the angle difference between the current angle and the target angle. When the angle difference reaches a predetermined error range, the acquisition system enters an angle lock mode.

[0010] The method of using the image data to extract two-dimensional traits based on color index also includes the following steps: image data preprocessing, including: brightness correction, using homomorphic filtering to separate the illumination component and the reflection component of the image data; noise elimination, suppressing salt and pepper noise based on anisotropic diffusion filtering; image registration, detecting the main direction of the wheat row through Hough transform, calculating the tilt angle and performing rotation correction.

[0011] The method of using the image data to extract two-dimensional characteristics based on color index also includes the following steps: extracting green area index, including: multi-color space collaborative enhancement, introducing LAB color space to extract the first super green index I GAI Optimization is performed and HSV spatial background is removed; dynamic threshold segmentation and morphological optimization are performed using Otsu adaptive threshold segmentation and region screening; green area index calculation is performed by converting the real green ratio according to the mask coverage area and imaging resolution to calculate the green area index GAI.

[0012] The method of extracting two-dimensional traits based on color index using the image data further includes the following steps: extracting canopy yellowing index, comprising:

[0013] Multispectral feature enhancement includes: converting the image data into LAB color space, extracting and normalizing the a channel, defining an initial form of a canopy yellowing index (CYI) using the statistical extreme values ​​of the a channel at the early filling stage and the statistical extreme values ​​of the a channel at the mature stage; calculating a second super green index (ExG) based on the image data to generate an ExG image; applying Otsu threshold segmentation to the ExG image to generate a first vegetation mask, setting a fixed threshold on the a channel to generate a second vegetation mask, and merging the first vegetation mask and the second vegetation mask to generate a third vegetation mask;

[0014] Texture feature fusion includes: using the third vegetation mask to extract texture features in the wheat ear area, using the gray level co-occurrence matrix to calculate the contrast Contrast and the correlation Correlation; combining the canopy yellowing index CYI, the contrast Contrast and the correlation Correlation to construct a multiple linear regression model, and using the multiple linear regression model to predict the degree of yellowing.

[0015] The method of extracting three-dimensional canopy structure distribution characteristics using the point cloud data and the image data further includes the following steps: calculating canopy coverage, including:

[0016] Point cloud projection and rasterization include: pre-processing the point cloud data, projecting the pre-processed point cloud data onto a horizontal plane, and performing grid division; calculating the vegetation point ratio for each grid, and if the number of point clouds in a grid is less than a first set value, marking the grid as a low-density area and triggering interpolation processing;

[0017] Sliding window weighted optimization, including: using a 3×3 grid as a sliding window, performing weighted averaging of the local coverage of all grids within each sliding window; eliminating outlier grids whose local coverage is greater than a first percentage value or less than a second percentage value;

[0018] Global coverage calculation and verification includes: averaging the grid coverage of all sliding windows that have undergone weighted optimization of the sliding windows to generate a global canopy coverage; performing a linear regression analysis between the global canopy coverage and the actual measured canopy coverage to verify reliability.

[0019] The extraction of three-dimensional canopy structure distribution characteristics using the point cloud data and the image data further includes the following steps: canopy flatness calculation, including:

[0020] Point cloud segmentation and neighborhood construction: dividing the point cloud data into multiple cubic units, selecting a neighborhood radius for each cubic unit according to the growth stage, and generating a point cloud set;

[0021] Covariance matrix and PCA analysis, calculate the centroid of the point cloud set of each cubic unit, construct the covariance matrix, perform eigenvalue decomposition on the covariance matrix, generate eigenvalues ​​e1, e2 and e3 and corresponding eigenvectors, where e1≥e2≥e3;

[0022] Flatness index calculation and optimization: if the number of point clouds in a cube unit is less than the second set value, the cube unit is marked as an invalid unit, and the flatness index is defined after normalizing the eigenvalues ​​e1, e2, and e3. Setting a constraint range for the flatness index CFI, and truncating it if it exceeds the constraint range;

[0023] Spatial interpolation and visualization, for the invalid cells, inverse distance weighted interpolation is used, and the flatness index CFI is mapped according to the color gradient.

[0024] Another aspect of the present invention provides a wheat canopy geometric parameter extraction system, which adopts the above-mentioned wheat canopy geometric parameter extraction method and comprises:

[0025] A multimodal sensor integrated acquisition system, comprising a stabilized image acquisition device and multiple sensors, the multiple sensors including at least a camera, a tightly coupled lidar, and an inertial measurement unit (IMU). The stabilized image acquisition device comprises a three-axis imaging gimbal and a vibration-absorbing fixture, and the multiple sensors are mounted on the three-axis imaging gimbal.

[0026] A control module, configured to apply an ESO-LQR-based pitch axis anti-disturbance control strategy to the acquisition system;

[0027] an extraction module, configured to collect point cloud data and image data of wheat using the acquisition system, perform two-dimensional trait extraction based on color index using the image data, and perform three-dimensional canopy structure distribution trait extraction using the point cloud data and the image data;

[0028] The steps of the pitch axis anti-disturbance control strategy based on ESO-LQR specifically include: establishing a pitch axis dynamics model and performing disturbance analysis, wherein the pitch axis dynamics model includes lumped disturbances, and the lumped disturbances include low-frequency disturbances, high-frequency disturbances and coupled disturbances; expanding the lumped disturbances into system states, designing a third-order extended state observer (ESO), and generating disturbance estimates; constructing a state vector for an optimal controller (LQR), designing a cost function, and solving an optimal feedback matrix through a Riccati equation to generate a control variable; and compensating the disturbance estimate into the control variable to generate a final control input.

[0029] On the other hand, the present invention further provides a computer-readable storage medium for storing a computer program; when the computer program is executed by a processor, the method for extracting geometric parameters of wheat canopy as described above is implemented.

[0030] From the above scheme, it can be seen that the advantages of the present invention are:

[0031] A multimodal sensor integrated acquisition system is used to achieve stable and efficient data collection from multiple sensors in the field. An ESO-LQR-based pitch axis anti-disturbance control strategy is used. The extended state observer (ESO) is used to estimate and compensate for nonlinear disturbances (such as low-frequency vibration, high-frequency vibration, and coupled disturbances) in real time. Combined with state feedback optimization using the LQR optimal control law, the acquisition system's anti-disturbance capability and robustness are significantly improved.

[0032] Adopting the integrated GAI computing framework, by integrating color space conversion, dynamic threshold segmentation and morphological optimization, the accurate extraction of green areas of wheat canopy in complex field environments is achieved;

[0033] A method of crop yellowing degree converted from Lab color space is used to achieve non-destructive monitoring through threshold segmentation and texture analysis;

[0034] Using canopy cover to quantify differences in wheat canopy structure, this method combines 3D point cloud processing, rasterization analysis, and spatial optimization techniques. Through multi-level spatial optimization and physical constraints, it significantly improves the robustness and farmland applicability of canopy cover measurement, providing reliable vegetation structure parameters for precision agriculture.

[0035] Canopy flatness is used to quantify the geometric undulation characteristics of the wheat canopy surface. Based on point cloud blocks and PCA analysis, an indicator calculation system that integrates physical properties and statistical models is constructed, providing irreplaceable decision-making support for smart agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of the wheat canopy geometric parameter extraction method of the present invention;

[0037] Figure 2A A three-dimensional diagram of the acquisition system of the present invention;

[0038] Figure 2B It is a side view of the acquisition system of the present invention;

[0039] Figure 2C It is a front view of the acquisition system of the present invention;

[0040] Figure 3 This is a structural diagram of the sensor supporting base of the present invention;

[0041] Figure 4 for Figure 1 Flowchart of step S20;

[0042] Figure 5A It is the structural principle diagram of the LQR controller in the prior art;

[0043] Figure 5B This is the principle diagram of the ESO_LQR composite control structure of the present invention;

[0044] Figure 6 for Figure 1 Another flow chart of step S20;

[0045] Figure 7 for Figure 6 Flowchart of step S202;

[0046] Figure 8 for Figure 1 Flowchart of step S30;

[0047] Figure 9 for Figure 8 Flowchart of step S300;

[0048] Figure 10 for Figure 8 Flowchart of step S310;

[0049] Figure 11 for Figure 8 Another flow chart of step S310;

[0050] Figure 12 for Figure 8 Flowchart of step S320;

[0051] Figure 13 for Figure 8 Flowchart of step S330;

[0052] Figure 14 for Figure 8 Flowchart of step S340;

[0053] Figure 15 This is a schematic diagram of canopy coverage of the present invention;

[0054] Figure 16A This is the canopy flatness color gradient map of the present invention;

[0055] Figure 16B This is the canopy flatness distribution diagram of the present invention;

[0056] Figure 17 This is a structural diagram of the wheat canopy geometric parameter extraction system of the present invention;

[0057] Figure 18 This is a structural diagram of the electronic device of the present invention;

[0058] Wherein, the reference numerals:

[0059] 100- Extraction method of wheat canopy geometric parameters; 10- Acquisition system;

[0060] 1-L-type adapter fixing plate; 2-three-axis imaging gimbal;

[0061] 20-Roll axis motor; 21-Yaw axis motor;

[0062] 22-pitch axis motor; 3-sensor bearing base;

[0063] 30-first long slot; 31-carrying plate;

[0064] 32-second long slot; 4-camera;

[0065] 5-T-type mounting component; 6-shock-absorbing ball;

[0066] 7-Inertial Measurement Unit IMU; 8-LiDAR;

[0067] 200-wheat canopy geometric parameter extraction system; 201-control module;

[0068] 202-extraction module; 300-electronic equipment;

[0069] 301-processor; 302-memory;

[0070] 303-Computer Program; 33-Sensor;

[0071] M1-moving direction of the carrier plate; M2-moving direction of the sensor;

[0072] S10~S30, S200~S240, S201, S202, S300~S303, S310~S313, S320~S322, S330~S333, S340~S344-steps. DETAILED DESCRIPTION

[0073] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments to further understand the purpose, solution and effect of the present invention, but it is not intended to limit the scope of protection of the appended claims of the present invention.

[0074] This paper focuses on phenotyping wheat breeding plots in the field, designing and developing a collaborative intelligent data acquisition system. This system uses a ground-based mobile platform with a stabilized image acquisition device carrying multiple information sensors. Area array lidar is used to obtain structural parameters such as canopy height and crop row width, while RGB cameras are used to simultaneously monitor various crop color traits.

[0075] This paper proposes a wheat field information collection method based on multimodal sensor acquisition technology, designs a collection system mounted on a self-propelled collection platform, aims to construct high-precision field three-dimensional crop maps and two-dimensional color traits while moving, and proposes a corresponding dynamic method for geometric parameters.

[0076] like Figure 1 As shown, an embodiment of the present invention provides a method 100 for extracting geometric parameters of wheat canopy, comprising the following steps:

[0077] S10: Build a multimodal sensor integrated acquisition system;

[0078] S20: Adopt the pitch axis anti-disturbance control strategy based on ESO-LQR for the acquisition system;

[0079] S30: using the acquisition system to collect point cloud data and image data of wheat, using the image data to extract two-dimensional traits based on color index, and using the point cloud data and image data to extract three-dimensional canopy structure distribution traits.

[0080] In step S10, the multimodal sensor integrated acquisition system mainly includes a stabilized image acquisition device and multiple sensors, wherein the multiple sensors include at least a camera, a tightly coupled lidar and an inertial measurement unit IMU; the stabilized image acquisition device includes a three-axis imaging gimbal and a shock-absorbing fixing device, and the multiple sensors are mounted on the three-axis imaging gimbal; the stabilized image acquisition device is arranged on the top of the self-propelled vehicle.

[0081] Specifically, if Figures 2A to 2C The figure shows the structure of the acquisition system 10, which includes an L-shaped adapter fixing plate 1, a three-axis imaging gimbal 2, a sensor supporting base 3, a camera 4, a T-shaped mounting component 5, a shock-absorbing ball 6, an inertial measurement unit (IMU) 7, and a laser radar 8. The L-shaped adapter fixing plate 1, the sensor supporting base 3, the T-shaped mounting component 5, and the shock-absorbing ball 6 constitute a shock-absorbing fixing device. The three sensors, namely the camera 4, the inertial measurement unit (IMU) 7, and the laser radar 8, are mounted on the three-axis imaging gimbal 2 via the sensor supporting base 3. The three-axis imaging gimbal 2 includes a roll axis, a pitch axis, and a yaw axis. The entire acquisition system 10 uses a ground acquisition vehicle (such as a self-propelled vehicle) as a mobile carrier and the three-axis imaging gimbal 2 as a stabilization carrier to achieve stable and efficient acquisition of multi-sensor data in the field.

[0082] During the actual construction process, self-propelled vehicles that are suitable for different field environments are selected as mobile carriers. For example, in the relatively flat and open wheat fields in the northern plains, large four-wheel drive self-propelled vehicles can be selected to ensure driving stability and endurance; in the southern hilly areas with small farmlands, small and light self-propelled vehicles are selected to facilitate flexible movement in the fields.

[0083] The mechanical architecture design of the acquisition system 10 fully integrates the differences in characteristics between ground and aerial platforms. First, ground acquisition vehicles have drawbacks such as smooth walls without mounting points, a hollow center but small interior space, and severe sensor occlusion. Therefore, this embodiment adopts the strategy of mounting the acquisition system 10 on the vehicle's outer wall. The three-axis imaging gimbal 2 should be installed at a vertical angle and connected via an L-shaped adapter plate 1. The hardware structure of the acquisition system 10 is made entirely of aluminum alloy components. The main components include the three-axis drive unit, the central control template, and the mounting structure. Among them, the drive motors of the three-axis drive units, including the roll axis motor 20, the yaw axis motor 21, and the pitch axis motor 22, are high-efficiency integrated motors with small size, high torque, and low noise. The yaw axis motor 21 can achieve 360° continuous rotation of the yaw axis. Equipped with a 17-bit magnetic encoder and a harmonic reducer (reduction ratio 1:100), it has a repeatability accuracy of ±0.05°, meeting the heading control requirements for continuous scanning between ridges in large fields. The pitch adjustment range of the pitch axis motor 22 is 0 to 90 degrees, and its large torque characteristic can drive the sensor mounted on the carrier plate to control the pitch angle.

[0084] The three-axis imaging gimbal 2 of the image stabilization system must have high-precision angle adjustment capabilities, with a horizontal rotation accuracy of ±0.01° and pitch and roll accuracy controlled within extremely tight limits to ensure stable data acquisition for the onboard sensors. In addition to the damping balls 6, the vibration-absorbing fixture can also utilize a special rubber cushion and spring combination to effectively absorb vehicle vibration during driving and reduce its impact on the sensors.

[0085] like Figure 3 As shown, the sensor base 3 includes a first slot 30 and a carrier plate 31 that moves along the first slot 30. The carrier plate 31 has a second slot 32 perpendicular to the first slot 30, and multiple sensors are mounted on the carrier plate 31. For example, the second slot 32 is 5 cm long, and the carrier plate 31 has a range of movement of 5 cm along the first slot 30. In this way, the sensors 33 (including the camera 4, the inertial measurement unit IMU 7, and the laser radar 8) can be placed 25 cm from the sensor base 3. 2 The position can be flexibly adjusted in the plane space to optimize the layout space and imaging space of the acquisition system 10. M1 represents the direction in which the carrier plate 31 can move, and M2 represents the direction in which the sensor 33 can move.

[0086] The stabilized image acquisition device is bolted to the center of the top front of a ground acquisition vehicle (e.g., a self-propelled vehicle). Considering the mounting position of the vehicle body and the imaging characteristics of the sensor, the stabilized image acquisition device is set at a height of 2.2 meters. The laser radar 8, inertial measurement unit IMU7, and camera 4 are all fixed to the stabilized image acquisition device via the sensor support base 3. The laser radar 8 is installed upside down, and the inertial measurement unit IMU7 is tightly attached to the bottom of the laser radar 8 to form a radar inertial measurement system. The camera 4 is installed directly below the sensor support base 3. During installation, the relative positions must be strictly calibrated to ensure the consistency and accuracy of the measurement data. After installation, professional calibration equipment and software are used to calibrate the camera 4 and the radar inertial measurement system multiple times to ensure that the error of the radar inertial measurement system is within the allowable range. For example, the laser radar 8 uses a non-repeating scanning radar with a fixed direction. It has the characteristics of high resolution and a large measurement range. In a farmland environment, it can effectively measure distances up to 100 meters and the angular resolution can be accurate to 0.5°. The mounting structure composed of the T-shaped mounting component 5 and the L-shaped adapter fixing plate 1 is 30 cm high and has a bearing capacity of 1.5 kg.

[0087] The main parameters of the acquisition system 10 are shown in Table 1. It can collect multi-dimensional sensor information. Parameters such as size, structure, and power supply are also shown in the table.

[0088] Table 1 Main parameters of acquisition system 10

[0089]

[0090] The data collection system 10 can be equipped with a high-precision Beidou positioning system, for example, with centimeter-level positioning accuracy, enabling precise recording of vehicle trajectories. Before actual operation, specialized path planning software is used to pre-set an S-shaped or round-trip data collection path based on the shape, size, and topography of the farmland. For example, a round-trip path is preferred for rectangular wheat fields to ensure comprehensive and efficient data collection. For irregularly shaped farmland, an S-shaped path is used to ensure coverage of the entire area to be measured.

[0091] The vehicle's speed is adjusted based on the crop growth conditions and sensor performance, generally within 2 m / s, with 1.5 m / s typically selected to ensure efficiency and mapping accuracy. In the early stages of crop growth, due to the shorter plants, the speed can be increased appropriately; in the later stages of crop growth, the speed is reduced to ensure data accuracy.

[0092] In step S20, a pitch-axis anti-disturbance strategy based on ESO-LQR control is used to address the vibration interference problem of the mobile platform in the field. The extended state observer (ESO) estimates nonlinear disturbances in real time and compensates them to the optimal controller LQR, significantly improving pitch-axis attitude stability and anti-disturbance capabilities.

[0093] like Figure 4As shown, step S20 further includes the following steps:

[0094] S210: Establish a pitch axis dynamic model and perform disturbance analysis. The pitch axis dynamic model includes lumped disturbances, which include low-frequency disturbances, high-frequency disturbances, and coupled disturbances.

[0095] S220: Expand the lumped disturbance into the system state, design a third-order extended state observer (ESO), and generate a disturbance estimate.

[0096] S230: Construct the state vector for the optimal controller LQR, design the cost function, and solve the optimal feedback matrix through the Riccati equation to generate the control variable;

[0097] S240: Compensate the disturbance estimation value into the control variable to generate the final control input.

[0098] Specifically, if Figure 5A As shown in FIG, the structure principle diagram of the traditional LQR controller in the prior art is shown. The traditional LQR controller is designed based on the state feedback principle, and its state equation is described as the following equations (1) and (2):

[0099]

[0100] y=Cx (2),

[0101] in, State vector, used to describe the system state such as angle and angular velocity. Control input vector, used to describe, for example, the motor drive voltage. The system state matrix is ​​used to describe the internal dynamic characteristics of the system. Input matrix, used to describe the effect of control input on the state. Output vector, used to describe the observable physical quantity of the system, sensor measurement value. The output matrix is ​​used to map the state to the observable output. The optimal feedback matrix K is obtained by solving the Riccati equation to minimize the cost function. However, the traditional LQR controller has limited ability to suppress disturbances caused by internal coupling torques (such as motor nonlinear friction and inertial coupling) in the three-axis imaging gimbal 2.

[0102] Pitch axis dynamics modeling and disturbance analysis: First, the dynamics equation is established. The pitch axis dynamics model can be simplified as follows:

[0103]

[0104] Where: J p is the pitch axis moment of inertia; τ d,p is the lumped disturbance, including external vibration, wind load moment and frame coupling torque; Cp is the pitch axis damping coefficient; θ p is the pitch angle, indicating the current pitch attitude of the system; is the first derivative of the pitch angle, i.e. the pitch angular velocity; is the second-order derivative of the pitch angle, that is, the pitch angle acceleration; K t,p is the pitch axis moment constant; u p is the control input voltage.

[0105] From the perspective of disturbance characteristics, low-frequency disturbance (<5Hz): mainly caused by carrier vibration, with an amplitude fluctuation range of ±0.6Nm; high-frequency disturbance (510Hz): motor cogging effect and gear transmission nonlinear friction, showing periodic harmonic characteristics; coupled interference: the roll axis and yaw axis motion are transmitted to the pitch axis disturbance through the stiffness of the image stabilization acquisition device frame. When collecting field information, maintaining the angle of the pitch axis is the key, so this embodiment uses the pitch axis to carry out the ESO-LQR composite control design. The ESO_LQR composite control structure principle diagram is as follows: Figure 5B .

[0106] The ESO disturbance observer converts the lumped disturbance τ d,p Expand to the system state and design the third-order extended state observer ESO as follows (4):

[0107]

[0108] Where z1 is the pitch angle θ p The estimated value of z2 is the angular velocity The estimated value of z3 is the disturbance estimate; β1, β2, and β3 are gain parameters, which are set by bandwidth parameterization (e.g., β1 = 150, β2 = 7500, β3 = 125000, corresponding to the observer bandwidth ω o =50rad / s).

[0109] LQR optimal control law, constructing the state vector The cost function is as follows (5):

[0110]

[0111] Where q1 is the pitch angle θ p The weight coefficient is used to adjust the angle tracking accuracy; q2 is the pitch angular velocity The weight coefficient is used to adjust the response speed; r is the control input u p The weight coefficient is used to suppress excessive control amount; t is time.

[0112] The optimal feedback matrix is ​​solved by the Riccati equation, K = [k1, k2], and the control input after compensation is the following equation (6):

[0113]

[0114] Among them, K t,p is the pitch axis moment constant; J p is the pitch axis moment of inertia (kg·m 2 );θ ref is the target pitch angle (rad).

[0115] like Figure 6 As shown, step S20 further includes the following steps: angle initialization S200, including: manual angle initialization S201 and automatic angle initialization S202, wherein:

[0116] Step S201 includes: controlling the pitch axis angle by sending a PWM signal or an S.Bus signal via the remote controller, and setting and locking the pitch axis and yaw axis angles according to the angle values ​​fed back by the inertial measurement unit (IMU);

[0117] Step S202 includes: setting the target angle of the pitch axis, obtaining the pitch angle of the laser radar or the current angle of the pitch axis, and automatically adjusting the pitch angle of the acquisition system according to the angle difference between the pitch angle of the laser radar or the current angle and the target angle. When the angle difference reaches a predetermined error range, the angle lock mode is entered.

[0118] Specifically, in step S201, the angle manual initialization control method uses the PWM signal of S.Bus for angle control, and a manual remote control method can be used to fix a specific angle. By observing the angle feedback value of the inertial measurement unit IMU7 attached to the surface of the lidar 8, the pitch and yaw angles of the sensor can be controlled. At the same time, the sensor supporting base 3 has a feedback function, which can compensate for the existing instantaneous offset.

[0119] In step S202, the automatic initialization of the angle only requires inputting the acquisition parameters, and the acquisition system 10 will automatically rotate to the appropriate acquisition position and enter the angle locking mode. Figure 7As shown, it is a flow chart of automatic angle initialization control, specifically: start; power on the acquisition system 10; read the current pitch angle θ1 of the laser radar 8; set the installation height of the acquisition system 10, the height of the plant (i.e., wheat), the field of view size, and the field of view acquisition length; calculate the target pitch angle θ2; calculate the difference between the target pitch angle θ2 and the current pitch angle θ1; judge whether θ2-θ1 is greater than zero; when θ2-θ1 is greater than zero, the acquisition system 10 issues a control instruction, and the three-axis imaging gimbal 2 "raises its head"; when θ2-θ1 is less than or equal to zero, the acquisition system 10 issues a control instruction, and the three-axis imaging gimbal 2 "lowers its head"; calculate whether the angle difference between the target pitch angle θ2 and the current pitch angle θ1 reaches a predetermined error range of 5%; when the angle difference reaches within the predetermined error range of 5%, enter the gimbal angle lock mode, otherwise return to calculating the difference between the target pitch angle θ2 and the current pitch angle θ1; end.

[0120] The current pitch angle θ1 is calculated by the inertial measurement unit IMU7 attached to the top of the laser radar 8 and sent to the microcontroller via the serial port of the acquisition computer. The target pitch angle θ2 is the aforementioned θ ref The control instructions use the S.BUS protocol output by the microcontroller. An Arduino UNO is used for S.BUS data parsing and command redefinition. The microcontroller is configured for a 100k baud rate, 8 data bits, even parity, 2 stop bits, and no flow control. The protocol format consists of 25 bytes, with data 1 through 22 representing 11-bit values ​​for 16 channels. During hardware circuit construction, the S.BUS signal must be inverted before being connected to the microcontroller for proper reading.

[0121] The angle automatic control adopts the PID (Proportion Integration Differentiation) control strategy to perform feedback control on the pitch angle. The microcontroller sends an angle rotation instruction to the pitch axis motor 22 of the three-axis imaging gimbal 2 to complete the response and angle locking. This process can solve the problem of inaccurate initial angle setting of the remote control.

[0122] In order to collect data on the wheat area to be measured ahead, the laser radar 6 is mounted obliquely downward through the three-axis imaging gimbal 2, and has the function of freely switching the collection angle. By setting the initial value of the collection angle and utilizing the angle locking mode of the three-axis imaging gimbal 2, the locked angle can be used for stable collection in the subsequent collection process.

[0123] In step S30,

[0124] Based on the collected point cloud and image raw data, this study addresses the low efficiency of three-dimensional crop phenotypic feature analysis and insufficient multidimensional data fusion in complex field environments. By studying multidimensional data preprocessing techniques and image color feature extraction algorithms, we achieve two-dimensional crop trait extraction based on color index. Based on the characteristics of wheat point clouds, we conduct an in-depth analysis of the impact of search radius selection on phenotypic imaging and use point cloud feature extraction methods to mine and verify wheat three-dimensional traits.

[0125] Dynamic extraction of canopy geometry includes two-dimensional trait extraction based on color index and three-dimensional canopy structural distribution trait extraction. Through multi-dimensional analysis of point cloud and image data, precise quantification of geometric parameters such as coverage (CC), flatness (CFI), and yellowing index (CYI) is achieved.

[0126] Before extracting 3D canopy structure and distribution characteristics using point cloud and image data, the point cloud data must be preprocessed. This includes transforming the point cloud data from the radar coordinate system to the world coordinate system using a coordinate transformation formula based on the pre-acquisition IMU attitude information. During the region of interest (ROI) extraction phase, the point cloud data is clipped using a straight-through filter, using the geographic information of the farmland boundary and the range of the lidar-collected point cloud data. For example, for a specific field, the x-axis clipping range is set to [50,50] and the y-axis clipping range is set to [50,50] to remove invalid point cloud data outside the field boundary.

[0127] In the Octree-based downsampling method, the number of octree segments and the number of points per child node are appropriately set based on the initial density of the point cloud data and subsequent processing requirements. During the noise filtering phase, the statistical filtering parameter n and the standard deviation threshold multiple n_sigma are determined through multiple experiments and data analysis.

[0128] Point cloud normal vector calculation is based on PCA analysis. During the calculation process, the neighborhood radius r is appropriately selected based on the distribution characteristics of the crop point cloud. For example, for wheat crops, the neighborhood radius r is set to 0.15 in the early growth stage. In the later growth stage, due to the denser crop canopy, the neighborhood radius r is appropriately adjusted to 0.3 to ensure accurate normal vector calculation.

[0129] like Figure 8 and Figure 9 As shown, the two-dimensional trait extraction based on color index further includes: S300: image data preprocessing, including:

[0130] S301: Brightness correction, using homomorphic filtering to separate the illumination component and the reflection component of the image data;

[0131] S302: noise elimination, suppressing salt and pepper noise based on anisotropic diffusion filtering;

[0132] S303: Image registration, detecting the main direction of the wheat rows through Hough transform, calculating the tilt angle and performing rotation correction.

[0133] Specifically, to address issues such as uneven lighting and noise interference in field images, a multi-stage enhancement and correction process is adopted, including:

[0134] Brightness correction uses homomorphic filtering to separate the image illumination component and reflection component, suppressing the interaction between canopy overexposure and soil shadow, as shown in the following formula (7):

[0135]

[0136] Where: γ H is the high-frequency gain of the image after Fourier transform; γ L is the base frequency gain of the image after Fourier transformation; D0 is the filter cutoff frequency; u, v are the coordinate variables in the frequency domain; c is the steepness coefficient of the ramp function; n is the order of the filter; and γ H =0.5,γ L =0.25, D0=20.

[0137] Noise elimination, based on anisotropic diffusion filtering (PeronaMalik model) to suppress salt and pepper noise, uses a gradient threshold function, as shown in the following formula (8):

[0138]

[0139] Among them, the derivative function With parameter K=30, this noise removal method can preserve sufficient texture details while improving the sharpness of the leaf edges.

[0140] Image registration targets the characteristics of wheat row planting, detects the main direction of the field row through Hough transform, and calculates the tilt angle θ RGB And perform rotation correction. Tilt angle θ RGB The calculation formula is as follows (9):

[0141]

[0142] Where Δy i , Δx i is the vertical and horizontal displacement (pixels) of the i-th wheat row in the image detected by Hough transform.

[0143] like Figure 10 As shown, the two-dimensional trait extraction based on the color index further includes S310: Green Area Index (GAI) extraction, including:

[0144] S311: Multi-color space collaborative enhancement, introducing LAB color space for the first super green index I GAI Optimize and perform HSV space background removal;

[0145] S312: Dynamic threshold segmentation and morphological optimization, using Otsu adaptive threshold segmentation and region screening for dynamic threshold segmentation and morphological optimization;

[0146] S313: Green area index calculation: Calculate the green area index GAI by converting the actual green ratio according to the mask coverage area and the imaging resolution.

[0147] The Green Area Index (GAI) is defined as the total unilateral area of ​​green vegetation elements per unit horizontal surface area. It is a core agronomic parameter that characterizes the density of green tissue cover in crop canopies. It is defined as the proportion of the projected area of ​​green photosynthetic organs (leaves, stems, etc.) of the crop per unit land area. Traditional measurement relies on destructive sampling and manual measurement, which is inefficient and difficult to achieve dynamic field monitoring. With the development of phenotyping platforms and the popularization of imaging technology, computer vision-based rapid GAI extraction technology has become a hot topic in precision agriculture research.

[0148] This paper adopts the integrated GAI computing framework, and realizes the accurate extraction of green areas of wheat canopy in complex field environment by integrating color space conversion, dynamic threshold segmentation and morphological optimization. Figure 11 .

[0149] After the image data is input, the multi-color space collaborative enhancement mechanism is first used. Multi-channel feature extraction is used to introduce the dynamic response characteristics of the a channel (red and green axis) of the LAB color space based on the green characteristics of wheat in the early growth period, and an improved super green index is constructed as shown in the following formula (10):

[0150] I GAI =2.5L-1.5a-0.2b (10),

[0151] Where: L is the brightness component in the LAB color space (L channel); a is the red-green axis component in the LAB color space (a channel); b is the yellow-blue axis component in the LAB color space (b channel);

[0152] Where L represents the overall brightness and darkness information of the image. The weighting coefficient of 2.5 is used to enhance the illumination characteristics of the wheat leaves and highlight the brightness contrast of the green area. a reflects the color change characteristics from red (negative value) to green (positive value). In formula (10), the negative weight of 1.5 is used to suppress non-green interference (such as soil or yellowed areas) and enhance the dynamic response sensitivity of the green wheat leaves. b represents the color change characteristics from yellow (positive value) to blue (negative value). In formula (10), a small negative weight of 0.2 is used to weaken the yellow background noise (such as early yellowing wheat leaves) and optimize the yellow-green gradient separation effect.

[0153] Then, the HSV space background is removed: combining the double threshold constraints of hue (H) and saturation (S), the background mask is defined as follows (11):

[0154]

[0155] Where H is the hue component of the HSV color space; S is the saturation component of the HSV color space.

[0156] HSV spatial background removal can effectively separate withered tissue (H<30°) and moist soil area (V<0.1) in wheat plants.

[0157] After completing the spatial color enhancement, due to the complex field environment, there is a data fluctuation area between the soil and the leaf gap. The Otsu adaptive threshold segmentation and region screening method can be used to perform dynamic threshold segmentation and morphological optimization to effectively remove the residual soil noise. Among them, the threshold is determined by the Otsu adaptive method. GAI The grayscale image was binarized and the leaf gaps were filled with morphological closing operations (5×5 ellipse kernel). Area filtering was performed using the region screening method to address soil residual noise, and the main canopy structure was retained through connected domain analysis, where the threshold was a single connected domain area ≥ 100 pixels.

[0158] The GAI index is calculated by converting the true green ratio based on the mask coverage area and imaging resolution, as shown in the following formula (12):

[0159]

[0160] Where, S pixel is the actual area of ​​a single pixel (based on camera calibration parameters), d GSD is the ground sampling distance, θ is the pitch angle of the camera 4, which can be obtained by the early acquisition system 10 (eliminating perspective distortion); W image , H image is the width and height of the image (pixels); Mask final is the total number of pixels of the final vegetation mask.

[0161] like Figure 12 As shown, the two-dimensional trait extraction based on color index also includes S320: canopy yellowing index extraction, including:

[0162] S321: Multispectral feature enhancement, including: converting image data into LAB color space, extracting and normalizing the a channel, defining the initial form of the canopy yellowing index (CYI) using the statistical extreme values ​​of the a channel at the early filling stage and the statistical extreme values ​​of the a channel at the mature stage; calculating the second super green index (ExG) based on the image data to generate an ExG image; applying Otsu threshold segmentation to the ExG image to generate a first vegetation mask, setting a fixed threshold on the a channel to generate a second vegetation mask, and merging the first vegetation mask and the second vegetation mask to generate a third vegetation mask;

[0163] S322: Texture feature fusion, including: using the third vegetation mask to extract texture features in the wheat ear area, using the gray-level co-occurrence matrix to calculate the contrast and correlation; combining the canopy yellowing index (CYI), contrast, and correlation to construct a multiple linear regression model, and using the multiple linear regression model to predict the degree of yellowing.

[0164] In research on wheat color traits based on RGB images, wheat at different growth stages exhibits intuitive color differences, manifested as a shift from green to yellow. To distinguish and quantify these differences, a method for non-destructively monitoring the degree of yellowing in late wheat growth stages was employed using Lab color space conversion, followed by threshold segmentation and texture analysis. The core principle is to capture the color migration patterns and structural changes of wheat ears from the filling stage to maturity. The specific process is as follows:

[0165] The first step is color space conversion and feature enhancement. The RGB image is converted to Lab color space, and the sensitivity of the a channel (red and green axis) to the yellowing process of wheat is used to construct the basic response parameters. Normalization is performed to eliminate the interference of light differences and define the Canopy Yellowing Index (CYI) as follows (13):

[0166]

[0167] Where: a max is the statistical extreme value of channel a at the initial grouting stage; a min is the statistical extreme value of channel a in the mature period.

[0168] Color space fusion: Synchronously extract the ExG (Extra Green Index) of the RGB image and the a channel of the Lab space. Dynamic threshold segmentation: For the ExG image, the Otsu algorithm is used to adaptively determine the vegetation mask. For the a channel, a threshold of a≥5 (empirical value, to suppress soil background) is set.

[0169] At the same time, in order to continuously transform the yellowing degree, a multivariate linear regression model is constructed by combining the color index CYI and texture features. The gray-level co-occurrence matrix texture parameters need to be extracted from the wheat ear area, including contrast (Contrast) as shown in the following formula (14) and correlation (Correlation) as shown in the following formula (15):

[0170] Contrast=∑ i,j |ij| 2 P(i,j) (14),

[0171] Contrast=∑ i,j |ij| 2 P(i,j) (15),

[0172] Where P(i, j) is the probability distribution of the gray-level co-occurrence matrix, μ and σ are the mean and standard deviation, respectively. Gray-level co-occurrence matrix (GLCM): Window size: 15 × 15 pixels (corresponding to the actual field size of 5 cm × 5 cm).

[0173] Combined color index CYI origin With the texture features, a multiple linear regression model is constructed as follows (16):

[0174] CYI=0.65CYI origin +0.28Contrast-0.07Correlation (16),

[0175] Multivariate Linear Regression Model Optimization: Training Data: 200 wheat canopy samples (covering the grain filling to maturity stages) were collected and manually labeled with yellowing levels (0-1). Model Fitting: Ridge Regression was used to prevent overfitting, resulting in the weight coefficients shown above.

[0176] The multivariate linear regression model determined the weight coefficients through principal component analysis, which can quantify the continuous change of the yellowness degree from 0 (early grain filling) to 1 (full maturity).

[0177] In step S30, if Figure 13 , using point cloud data and image data to extract three-dimensional canopy structure distribution characteristics, also includes the following steps: S330: canopy coverage calculation, including:

[0178] S331: Point cloud projection and rasterization, including: pre-processing the point cloud data, projecting the pre-processed point cloud data onto a horizontal plane, and performing grid division; calculating the vegetation point ratio for each grid; if the number of point clouds in a grid is less than a first set value, marking the grid as a low-density area and triggering interpolation processing;

[0179] S332: Sliding window weighted optimization, including: using a 3×3 grid as a sliding window, performing weighted averaging of local coverage of all grids within each sliding window; and removing outlier grids whose local coverage is greater than a first percentage value or less than a second percentage value;

[0180] S333: Global coverage calculation and verification, including: averaging the grid coverage of all sliding windows that have undergone sliding window weighted optimization to generate global canopy coverage; performing linear regression analysis on the global canopy coverage and the actual measured canopy coverage to verify reliability.

[0181] like Figure 14 As shown, the extraction of three-dimensional canopy structure distribution characteristics using point cloud data and image data also includes the following steps: S340: canopy flatness calculation, including:

[0182] S341: Point cloud segmentation and neighborhood construction: divide the point cloud data into multiple cubic units, select the neighborhood radius for each cubic unit according to the growth stage, and generate a point cloud set;

[0183] S342: Covariance matrix and PCA analysis, calculate the centroid of the point cloud set of each cubic unit, construct the covariance matrix, perform eigenvalue decomposition on the covariance matrix, generate eigenvalues ​​e1, e2 and e3 and corresponding eigenvectors, where e1≥e2≥e3;

[0184] S343: Calculation and optimization of the flatness index. If the number of point clouds in a cube unit is less than the second set value, the cube unit is marked as an invalid unit. The flatness index is defined after normalizing the eigenvalues ​​e1, e2, and e3. Set a constraint range for the flatness index CFI, and truncate if it exceeds the constraint range;

[0185] S344: Spatial interpolation and visualization. For invalid cells, inverse distance weighted interpolation is used, and the flatness index CFI is mapped according to the color gradient.

[0186] Specifically, during the growth process of wheat crops, each breeding plot has different growth conditions, which are reflected not only in the vertical structure but also in the horizontal structure of the canopy. To evaluate the canopy distribution differences of crop growth, canopy cover and canopy flatness are used to quantify the canopy structure differences between crops. Based on the geometric characteristics of LiDAR point clouds, this embodiment constructs a dual-index calculation system that integrates physical properties and statistical models, including:

[0187] (1) Calculation of canopy cover (CC)

[0188] CC quantification is based on point cloud classification and rasterization. First, the lidar point cloud is subjected to region clipping and elevation threshold segmentation to extract crop points. This includes:

[0189] Grid Cell Division: The 3D point cloud is projected onto a horizontal plane to generate a 5cm x 5cm grid cell. (This size has been verified in field trials to balance accuracy and computational efficiency.) The average wheat row spacing is 15-20cm, and a 5cm grid cell can capture coverage differences of 2-3 plants within a single row.

[0190] Point cloud classification rules: Vegetation point determination: elevation threshold segmentation method, set points with H ≥ 5 cm above the ground as vegetation points (excluding soil surface noise).

[0191] Density correction: If the total number of points n in the grid is less than 10 (the first set value), it is marked as a low confidence area and the interpolation process is triggered.

[0192] The proportion of vegetation points in each grid CC i The calculation formula is as follows (17):

[0193]

[0194] Where, is the number of plant point clouds; is the total number of point clouds in the region.

[0195] Then, the sliding window weighted optimization is performed, including:

[0196] Window size: A 3×3 grid (15 cm × 15 cm) sliding window was used to simulate the distribution characteristics of wheat populations. The weight of high-density areas was close to 1, and the weight of low-density areas was inherited from the adjacent 8 grids using bilinear interpolation (interpolation radius of 15 cm).

[0197] Abnormal filtering: outlier grids with CC>95% (first percentage value) or CC<5% (second percentage value) are eliminated (possibly caused by point cloud noise or plant lodging).

[0198] Finally, the global coverage calculation is performed, including:

[0199] The global coverage is obtained by integrating the local grid results through sliding window weighted averaging, as shown in formula (18):

[0200]

[0201] Where M is the number of grids; w i Adjust weights for density.

[0202] The weight w i The grid point density is dynamically adjusted, and the low-density area is optimized by neighboring interpolation. Figure 15 Schematic diagram of canopy coverage CC. Different colors represent point clouds of crops at different locations. Red and blue represent areas with concentrated crop coverage and areas with sparse crop coverage, respectively.

[0203] Manual sampling comparison: randomly select 10 1m 2 The projected area ratio of plants was manually counted and linear regression (R 2 ≥0.92).

[0204] (2) Calculation of canopy flatness index (CFI)

[0205] The quantification of canopy flatness (CFI) relies on principal component analysis (PCA) to reveal the geometric distribution characteristics of the canopy surface. The specific steps include:

[0206] Point cloud segmentation and domain construction include: analysis unit division: dividing the canopy point cloud into 20 cm × 20 cm cubic units (covering a single plant and up to 3 wheat bushes); neighborhood radius selection: in the early growth stage (plant height < 30 cm): neighborhood radius r = 0.1 m (to avoid cross-plant interference); in the middle and late growth stage (plant height ≥ 30 cm): neighborhood radius r = 0.2 m (to adapt to the close connection characteristics of the canopy).

[0207] Covariance matrix and PCA analysis include: calculating the centroid of the point cloud set P in each analysis unit, constructing the covariance matrix, solving Cov, and obtaining the eigenvalues ​​e1, e2, e3 and the corresponding eigenvectors.

[0208] The calculation and optimization of the flatness index include: after solving the eigenvalues ​​e1≥e2≥e3, the flatness index CFI is defined as the following formula (19):

[0209]

[0210] Noise suppression strategy, valid point cloud threshold: if the number of points n in a unit is less than 20, it is marked as invalid data (to avoid eigenvalue distortion caused by sparse point cloud); eigenvalue normalization: divide e1, e2, e3 by the point cloud spatial scale to eliminate the absolute scale shadow; output range correction: enforce the constraint CFI∈[0,1]: if the calculated result exceeds the range, it will be truncated.

[0211] This formula (19) quantifies the degree of canopy surface undulation by normalizing the variance difference between the secondary and weakest directions. When e1 is significantly greater than e2 and e3, the CFI approaches zero, indicating a planar canopy distribution. Conversely, if the difference between e2 and e3 increases, the CFI increases, reflecting three-dimensional heterogeneity. This method shares the same mathematical kernel as the least squares planar fitting method but extends it to three dimensions to capture complex canopy structure.

[0212] Spatial interpolation and visualization, including: filling in gaps: for invalid cells (n<20), inverse distance weighted interpolation (IDW) is used; visualization mapping: mapping CFI values ​​according to color gradients (red: CFI=1, canopy undulations are severe; blue: CFI=0, completely flat). Figure 16A and Figure 16B This is a schematic diagram of canopy flatness. Flatness is divided using a color gradient. The value is normalized between 0 and 1. Red and blue represent the endpoints of areas with high and low flatness, respectively.

[0213] The following is a system embodiment corresponding to the above method embodiment. This embodiment can be implemented in conjunction with the above embodiment. The relevant technical details mentioned in the above embodiment are still valid in this embodiment and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above embodiment.

[0214] like Figure 17 FIG. 2 is a structural diagram of a wheat canopy geometric parameter extraction system 200 according to another embodiment of the present invention, comprising:

[0215] A multimodal sensor integrated acquisition system 10, comprising a stabilized image acquisition device and multiple sensors, the multiple sensors including at least a camera 4, a tightly coupled lidar 8, and an inertial measurement unit (IMU) 7; the stabilized image acquisition device comprising a three-axis imaging gimbal 2 and a vibration-absorbing fixture, the multiple sensors being mounted on the three-axis imaging gimbal 2;

[0216] A control module 201 is configured to apply an ESO-LQR-based pitch axis anti-disturbance control strategy to the acquisition system 10;

[0217] The extraction module 202 is used to collect point cloud data and image data of wheat using the acquisition system, perform two-dimensional trait extraction based on color index using the image data, and perform three-dimensional canopy structure distribution trait extraction using the point cloud data and image data.

[0218] The steps of the pitch axis anti-disturbance control strategy based on ESO-LQR specifically include:

[0219] Establish a pitch axis dynamic model and perform disturbance analysis. The pitch axis dynamic model includes lumped disturbances, which include low-frequency disturbances, high-frequency disturbances, and coupled disturbances.

[0220] The lumped disturbance is expanded into the system state, and a third-order extended state observer (ESO) is designed to generate disturbance estimates.

[0221] Construct the state vector for the optimal controller LQR, design the cost function, and solve the optimal feedback matrix through the Riccati equation to generate the control variable;

[0222] The disturbance estimate is compensated into the control variable to generate a final control input.

[0223] In addition, technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, the specific working process of the wheat canopy geometric parameter extraction system 200 can refer to the corresponding process in the aforementioned wheat canopy geometric parameter extraction method 100 embodiment, and will not be repeated here.

[0224] Furthermore, another embodiment of the present invention also discloses an electronic device 300, such as Figure 18 The contents of the drawings should not be considered as any limitation on the scope of application of the present invention.

[0225] Figure 18 Schematic diagram of the structure of an electronic device 300 provided in an embodiment of the present invention. The electronic device 300 may specifically include: at least one processor 301 and at least one memory 302. The memory 302 is used to store a computer program 303, which is loaded and executed by the processor 301 to implement the relevant steps in the wheat canopy geometric parameter extraction method 100 disclosed in any of the above embodiments, such as Figure 1 Steps shown.

[0226] In addition, the memory 302 as a carrier for resource storage can be a read-only memory, a random access memory, a magnetic disk or an optical disk, etc., and the storage method can be temporary storage or permanent storage.

[0227] In addition to the computer program that can be used to complete the wheat canopy geometric parameter extraction method 100 executed by the electronic device 300 disclosed in any of the aforementioned embodiments, the computer program 303 can further include computer programs that can be used to complete other specific tasks.

[0228] Another embodiment of the present invention further discloses a computer storage medium, wherein the computer storage medium stores computer executable instructions. When the computer executable instructions are loaded and executed by a processor, the steps of the wheat canopy geometric parameter extraction method 100 disclosed in any of the above embodiments are implemented. Figure 1Steps shown.

[0229] In summary, the present invention has the following advantages:

[0230] Flexible equipment and easy acquisition: The acquisition system 10 of the present invention is compact and flexible, and can be easily mounted on various field platforms. No additional field targets are required during acquisition, which reduces the complexity of pre-production preparation and the resources required for point cloud matching. This greatly improves acquisition efficiency and makes the entire acquisition process more convenient and efficient.

[0231] Accurate and reliable trait extraction: The proposed method for extracting wheat canopy geometric parameters is 100% accurate, enabling stable and accurate acquisition of wheat canopy information. This is crucial for assessing wheat growth status, helping farmers identify growth anomalies promptly, thus providing a basis for precision agriculture operations and improving wheat yield and quality.

[0232] Highly adaptable and stable algorithms: The algorithms used for point cloud data processing and canopy information extraction are highly adaptable to complex field environments. They effectively overcome challenges such as ground undulation, crop obstruction, and noise interference. Even in the face of unexpected conditions such as sensor errors, they maintain stable operation, ensuring reliable data processing results.

[0233] Wide range of applications: This invention is not limited to wheat; its principles and methods can be extended to monitoring the growth of a variety of crops. This system can be used to obtain three-dimensional plant information in scenarios such as corn and rice cultivation, orchard and tea plantation management, and urban greening plant monitoring, providing innovative solutions for plant growth research and management in various fields.

[0234] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which fall within the scope of protection of the present invention.

Claims

1. A method for extracting geometric parameters of wheat canopy, characterized in that: Constructing a multimodal sensor integrated acquisition system, the acquisition system including a stabilized image acquisition device and multiple sensors, the multiple sensors including at least a camera, a tightly coupled lidar, and an inertial measurement unit (IMU); the stabilized image acquisition device including a three-axis imaging gimbal and a vibration-absorbing fixture, the multiple sensors being mounted on the three-axis imaging gimbal; An ESO-LQR-based pitch axis anti-disturbance control strategy is adopted for the acquisition system; Using the acquisition system to collect point cloud data and image data of wheat, using the image data to perform two-dimensional trait extraction based on color index, and using the point cloud data and the image data to perform three-dimensional canopy structure distribution trait extraction; The pitch axis anti-disturbance control strategy based on ESO-LQR further includes the following steps: Establishing a pitch axis dynamics model and performing disturbance analysis, wherein the pitch axis dynamics model includes lumped disturbances, and the lumped disturbances include low-frequency disturbances, high-frequency disturbances, and coupled disturbances; Expanding the lumped disturbance into a system state, designing a third-order extended state observer (ESO) to generate a disturbance estimate; Construct the state vector for the optimal controller LQR, design the cost function, and solve the optimal feedback matrix through the Riccati equation to generate the control variable; The disturbance estimation value is compensated into the control variable to generate a final control input.

2. The method for extracting wheat canopy geometric parameters according to claim 1, characterized in that: The pitch axis anti-disturbance control strategy based on ESO-LQR is adopted for the acquisition system, and further comprises the following steps: Angle initialization, including manual angle initialization and automatic angle initialization, where: The manual angle initialization includes: controlling the pitch axis angle by sending a PWM signal or an S.Bus signal through the remote control, and setting and locking the angles of the pitch axis and the yaw axis according to the angle value fed back by the inertial measurement unit IMU; The automatic angle initialization includes: setting a target angle of the pitch axis, obtaining the current angle of the pitch axis, and automatically adjusting the angle of the pitch axis according to the angle difference between the current angle and the target angle. When the angle difference reaches a predetermined error range, the acquisition system enters the angle lock mode.

3. The wheat canopy geometric parameter extraction method according to claim 1, characterized in that: The method of using the image data to extract two-dimensional properties based on color index further includes the following steps: image data preprocessing, including: Brightness correction, using homomorphic filtering to separate the illumination component and the reflection component of the image data; Noise removal, suppressing salt and pepper noise based on anisotropic diffusion filtering; Image registration, detecting the main direction of the wheat rows by Hough transform, calculating the tilt angle and performing rotation correction.

4. The method for extracting wheat canopy geometric parameters according to claim 1, characterized in that: The method of using the image data to extract two-dimensional properties based on color index further includes the following steps: extracting green area index, comprising: Multi-color space synergistic enhancement, introducing LAB color space for the first super green index I GAI Optimize and perform HSV space background removal; Dynamic threshold segmentation and morphological optimization, using Otsu adaptive threshold segmentation and region screening for dynamic threshold segmentation and morphological optimization; Green area index calculation: The green area index GAI is calculated by converting the actual green ratio based on the mask coverage area and imaging resolution.

5. The method for extracting geometric parameters of wheat canopy according to claim 1, characterized in that: The method of extracting two-dimensional traits based on color index using the image data further includes the following steps: extracting canopy yellowing index, comprising: Multispectral feature enhancement, including: Convert the image data into LAB color space, extract the a channel and normalize the a channel, and define the initial form of the canopy yellowing index CYI using the statistical extreme value of the a channel at the early filling stage and the statistical extreme value of the a channel at the mature stage; Calculating a second super green index ExG based on the image data to generate an ExG image; Applying Otsu threshold segmentation to the ExG image to generate a first vegetation mask, setting a fixed threshold on the a channel to generate a second vegetation mask, and merging the first vegetation mask and the second vegetation mask to generate a third vegetation mask; Texture feature fusion, including: Extracting texture features in the wheat ear region using the third vegetation mask, and calculating contrast and correlation using a gray level co-occurrence matrix; A multiple linear regression model is constructed by combining the canopy yellowing index CYI, the contrast Contrast and the correlation Correlation, and the multiple linear regression model is used to predict the yellowing degree.

6. The method for extracting wheat canopy geometric parameters according to claim 1, characterized in that: The method of extracting three-dimensional canopy structure distribution characteristics using the point cloud data and the image data further includes the following steps: calculating canopy coverage, including: Point cloud projection and rasterization, including: Preprocessing the point cloud data, projecting the preprocessed point cloud data onto a horizontal plane, and performing grid division; Calculate the vegetation point ratio for each grid. If the number of point clouds in a grid is less than the first set value, mark the grid as a low-density area and trigger interpolation processing. Sliding window weighted optimization, including: Using a 3×3 grid as a sliding window, performing a weighted average of the local coverage of all grids within each sliding window; Eliminate outlier grids whose local coverage is greater than a first percentage value or less than a second percentage value; Global coverage calculation and verification, including: Averaging the grid coverage of all sliding windows that have undergone weighted optimization of the sliding windows to generate a global canopy coverage; A linear regression analysis was performed between the global canopy cover and the actual measured canopy cover to verify the reliability.

7. The method for extracting geometric parameters of wheat canopy according to claim 1, characterized in that: The extraction of three-dimensional canopy structure distribution characteristics using the point cloud data and the image data further includes the following steps: canopy flatness calculation, including: Point cloud segmentation and neighborhood construction: dividing the point cloud data into multiple cubic units, selecting a neighborhood radius for each cubic unit according to the growth stage, and generating a point cloud set; Covariance matrix and PCA analysis, calculate the centroid of the point cloud set of each cubic unit, construct the covariance matrix, perform eigenvalue decomposition on the covariance matrix, generate eigenvalues ​​e1, e2 and e3 and corresponding eigenvectors, where e1≥e2≥e3; Flatness index calculation and optimization: if the number of point clouds in a cube unit is less than the second set value, the cube unit is marked as an invalid unit, and the flatness index is defined after normalizing the eigenvalues ​​e1, e2, and e3. Setting a constraint range for the flatness index CFI, and truncating it if it exceeds the constraint range; Spatial interpolation and visualization, for the invalid cells, inverse distance weighted interpolation is used, and the flatness index CFI is mapped according to the color gradient.

8. A wheat canopy geometric parameter extraction system, characterized in that: The method for extracting geometric parameters of wheat canopy according to any one of claims 1 to 7 comprises: A multimodal sensor integrated acquisition system, comprising a stabilized image acquisition device and multiple sensors, the multiple sensors including at least a camera, a tightly coupled lidar, and an inertial measurement unit (IMU). The stabilized image acquisition device comprises a three-axis imaging gimbal and a vibration-absorbing fixture, and the multiple sensors are mounted on the three-axis imaging gimbal. A control module, configured to apply an ESO-LQR-based pitch axis anti-disturbance control strategy to the acquisition system; an extraction module, configured to collect point cloud data and image data of wheat using the acquisition system, perform two-dimensional trait extraction based on color index using the image data, and perform three-dimensional canopy structure distribution trait extraction using the point cloud data and the image data; The steps of the pitch axis anti-disturbance control strategy based on ESO-LQR specifically include: Establishing a pitch axis dynamics model and performing disturbance analysis, wherein the pitch axis dynamics model includes lumped disturbances, and the lumped disturbances include low-frequency disturbances, high-frequency disturbances, and coupled disturbances; Expanding the lumped disturbance into a system state, designing a third-order extended state observer (ESO) to generate a disturbance estimate; Construct the state vector for the optimal controller LQR, design the cost function, and solve the optimal feedback matrix through the Riccati equation to generate the control variable; The disturbance estimation value is compensated into the control variable to generate a final control input.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the method for extracting wheat canopy geometric parameters according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that Used to store a computer program; when the computer program is executed by a processor, the method for extracting wheat canopy geometric parameters as described in any one of claims 1 to 7 is implemented.

Images