Excavator automatic grading control method and system based on visual laser fusion

Abstract

The application discloses a method and system for automatic land leveling control of excavators based on visual laser fusion, and relates to the field of automatic control of engineering machinery. The method obtains three-dimensional terrain point clouds of a working area through a binocular vision acquisition module, combines reference height signals of a laser leveling instrument and body posture data of an IMU, and constructs a global dynamic terrain model through a multi-sensor fusion algorithm. According to preset land leveling parameters, a target path for shovel operation is generated, joint control amounts of a swing arm, a dipper arm and a shovel are calculated in combination with kinematic inverse solution of the excavator, and finally, adaptive land leveling operation of the shovel is realized through a hydraulic actuator. The application solves the problems of traditional manual land leveling, such as dependence on experience, low precision and poor efficiency, significantly improves land leveling operation precision and efficiency, reduces labor intensity of operators, and is suitable for road subgrade leveling, building site excavation and leveling, farmland reconstruction and other scenes.

Description

Technical Field

[0001] This invention relates to the field of automated control technology for construction machinery, specifically to an automatic leveling control method and system for excavators based on vision-laser fusion. Background Technology

[0002] In leveling operations during engineering construction, excavators are the core equipment, and related technologies are currently mainly divided into two categories: manual operation and semi-automatic control. Manual leveling relies entirely on the operator's visual judgment and operational experience, adjusting work actions by visually observing the bucket position and terrain height. Some semi-automatic leveling solutions rely solely on a single laser sensor for single-point height control, lacking the ability to comprehensively perceive and model the overall terrain. They are mainly used in simple foundation leveling scenarios and are difficult to adapt to the needs of complex construction environments.

[0003] Existing technologies suffer from several significant drawbacks: manual operation is significantly affected by operator experience and fatigue, making it difficult to accurately control the overall flatness, resulting in large fluctuations in work quality and requiring repeated adjustments; semi-automatic solutions, lacking a global terrain perspective, are prone to incomplete leveling or over-excavation in complex terrains with local bumps and depressions. Furthermore, manual adjustment of the bucket posture exhibits noticeable lag, limiting the area worked per operation and leading to low overall efficiency. Operators must continuously monitor multiple factors, including terrain and equipment status, resulting in extremely high labor intensity during extended periods of operation. These issues severely restrict the quality and efficiency of leveling operations. Summary of the Invention

[0004] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides an automatic excavator leveling control method and system based on vision-laser fusion, which solves the problems of unstable accuracy, low efficiency, high labor intensity, and insufficient adaptability to complex terrain in traditional excavator leveling operations.

[0005] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: an automatic leveling control method for excavators based on vision-laser fusion, comprising the following steps: S1. Multi-source data synchronous acquisition: Two-dimensional images of the working area are acquired by a binocular camera installed at the front of the excavator boom and a three-dimensional terrain point cloud is generated. Ground reference height signals are acquired by a laser leveling instrument. The tilt angle and motion acceleration of the excavator body are acquired by an IMU. All sensors are acquired at the same frequency and ≥15Hz. S2. Data Preprocessing and Fusion: Denoise and downsample the point cloud data from the binocular camera, filter the laser leveling instrument signal, and compensate for drift in the IMU data; use the Kalman filter algorithm to fuse the three types of data and construct a global dynamic terrain model that includes fuser attitude correction; S3, Flat Ground Target Parameter Setting: Receives preset flat ground parameters input by the user, including target elevation H, target slope θ (0°-30°), and work area boundary coordinates, maps the parameters to the global terrain model, and generates a virtual flat ground target surface; S4. Bucket Operation Trajectory Planning: Based on the height difference between the global terrain model and the virtual flat ground target surface, an S-curve algorithm is used to plan the movement trajectory of the bucket teeth, ensuring that the trajectory meets the following requirements: horizontal movement speed of the bucket ≤ 0.8 m / s, vertical lifting acceleration ≤ 2 m / s². 2 ; S5. Inverse kinematics calculation: Based on the known parameters of the excavator, such as the link length and joint angle range, a kinematic model is established. The target trajectory coordinates of the bucket are substituted into the model to calculate the target joint angles of the boom, stick, and bucket. S6. Closed-loop execution control: Real-time acquisition of feedback signals from angle sensors of each joint, comparison with the target joint angle, and correction of control commands using PID algorithm to drive the hydraulic actuator to move the bucket along the planned trajectory until the error between the terrain of the working area and the virtual flat target surface is ≤ ±1cm.

[0006] Preferably, in S1, the binocular camera has a resolution ≥1280×720, a baseline distance of 150-200mm, a 3D point cloud generation accuracy ≤±3mm, and a work area acquisition range of 5-15m; the laser leveling instrument is an infrared laser leveling instrument with a reference height measurement accuracy ≤±0.5mm and a signal transmission delay ≤10ms; the IMU is a six-axis IMU with a sampling rate ≥100Hz and a roll / pitch angle measurement accuracy ≤±0.1°.

[0007] Preferably, the global dynamic terrain model in S2 is represented by a raster, with a grid size of 5cm×5cm. The model update frequency is synchronized with the sensor acquisition frequency, and 80% of the effective point cloud data in the previous frame is retained during each update to reduce the amount of computation.

[0008] Preferably, the bucket operation trajectory planning in S4 also includes collision detection logic: if the distance between the bucket tip and the protruding obstacle (height > 5cm) in the terrain model in the planned trajectory is < 10cm, the trajectory height is automatically adjusted (raised by 5-10cm) while keeping the flat ground target parameters unchanged.

[0009] The excavator automatic leveling control system based on vision-laser fusion includes: Multi-source sensor module: Composed of a binocular camera, an infrared laser leveling instrument, and a six-axis IMU, used to simultaneously acquire terrain data, reference height signals, and aircraft attitude data of the work area; Data processing unit: Employs an industrial-grade ARM processor to run point cloud denoising, sensor fusion, and terrain modeling algorithms, with a processing latency of ≤80ms; Human-computer interaction unit: includes a touch screen and physical buttons, used by users to input preset flat terrain parameters, start / stop operations, and display the terrain model and operation errors in real time; Motion control unit: Stores the excavator's kinematic model and PID control algorithm, receives trajectory data from the data processing unit, and outputs joint angle control commands; Actuators include a proportional solenoid valve assembly, hydraulic cylinders, and joint angle sensors, which are used to drive the bucket movement and provide real-time position feedback.

[0010] Preferably, the human-computer interaction unit also has a work data storage function, which can record the time, work area, and average error of each leveling operation, with a storage capacity of ≥1000 records, and supports data export via USB.

[0011] Preferably, the motion control unit also includes an emergency stop module: when the sensor feedback signal is abnormal or the operating error is > ±3cm, it automatically outputs a stop command and controls the bucket to return to a safe position.

[0012] (III) Beneficial Effects This invention provides an automatic ground leveling control method and system for excavators based on vision-laser fusion. It has the following beneficial effects: 1. This invention significantly improves the accuracy of leveling operations by integrating binocular vision, laser leveling instruments, and IMU multi-sensor fusion and closed-loop control, effectively meeting the needs of high-precision construction. Automatic trajectory planning reduces the tedious process of repeated manual adjustments, significantly improving work efficiency. Operators only need to set basic parameters to achieve automated operation, eliminating the need for continuous focus on equipment and terrain, thus lowering the operational threshold and alleviating labor fatigue. Global terrain modeling and collision detection functions allow the equipment to flexibly handle complex terrains with protrusions and depressions, avoiding missed leveling or collisions, and broadening its applicability in scenarios such as roads, buildings, and farmland. Detailed Implementation

[0013] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Example 1: This invention provides an automatic ground leveling control method for excavators based on vision-laser fusion, comprising the following steps: S1. Multi-source synchronous data acquisition: A binocular camera installed at the front of the excavator boom acquires two-dimensional images of the working area and generates a three-dimensional terrain point cloud. A laser leveling device acquires ground reference height signals, and an IMU acquires the tilt angle and acceleration of the excavator body. All sensors acquire data at a synchronized frequency of ≥15Hz. The binocular camera in S1 has a resolution ≥1280×720, a baseline distance of 150-200mm, a three-dimensional point cloud generation accuracy ≤±3mm, and a working area acquisition range of 5-15m. The laser leveling device is an infrared laser leveling device with a reference height measurement accuracy ≤±0.5mm and a signal transmission delay ≤10ms. The IMU is a six-axis IMU with a sampling rate ≥100Hz and a roll / pitch angle measurement accuracy ≤±0.1°. A binocular camera with a resolution of 1280×720 is selected, and its baseline distance is set to 180mm (within the 150-200mm range). It is installed at the front of the excavator boom, 1.8m from the bucket. The system uses a stereo matching algorithm to convert the acquired 2D images into 3D terrain point clouds with a point cloud generation accuracy of ±2.5mm (≤±3mm). The acquisition range covers 5-15m of the work area, comprehensively capturing terrain information. An infrared laser leveling instrument is used, installed on the top of the machine, with a reference height measurement accuracy of ±0.5mm (≤±0.5mm) and a signal transmission delay controlled within 8ms (≤10ms), continuously acquiring ground reference height signals. A six-axis IMU is selected and fixed at the center of the excavator's slewing platform, with a sampling rate set to 100Hz (≥100Hz). The roll and pitch angle measurement accuracy reaches ±0.1° (≤±0.1°), acquiring the machine's tilt angle (roll angle, pitch angle) and motion acceleration in real time. Through timestamp synchronization technology, the acquisition frequencies of the binocular camera, laser leveling instrument, and IMU are uniformly set to 15Hz (≥15Hz), ensuring the spatiotemporal consistency of the data from the three types of sensors and laying the foundation for subsequent data fusion. S2. Data Preprocessing and Fusion: Noise reduction and downsampling are performed on the point cloud data from the binocular camera, the laser leveling instrument signal is filtered, and drift compensation is applied to the IMU data. A Kalman filter algorithm is used to fuse the three types of data to construct a global dynamic terrain model including fuser attitude correction. In S2, the global dynamic terrain model is represented using a rasterized grid with a grid size of 5cm × 5cm. The model update frequency is synchronized with the sensor acquisition frequency. Each update retains 80% of the effective point cloud data from the previous frame to reduce computational load. For the 3D point cloud data generated by the binocular camera, a statistical filtering algorithm is used to remove outliers caused by airborne dust and light interference, and downsampling is performed to reduce the data volume. For the reference height signal acquired by the laser leveling instrument, a Kalman filter algorithm is used to... The moving average filtering algorithm eliminates signal fluctuations caused by excavator operation vibration and electromagnetic interference, improving signal stability; drift compensation is applied to the attitude data acquired by the IMU to ensure that the roll angle and pitch angle deviations are controlled within ±0.05°; the Kalman filtering algorithm is used to fuse the preprocessed point cloud data, laser reference height signal and IMU attitude data to construct a global dynamic terrain model including fuselage attitude correction; the terrain model adopts a raster representation with a grid size of 5cm×5cm, and the model update frequency is synchronized with the sensor acquisition frequency (15Hz). Each update retains 80% of the effective point cloud data in the previous frame, which greatly reduces the amount of data processing computation while ensuring the continuity and accuracy of the model; S3. Setting Target Parameters for Flat Ground: The system receives preset flat ground parameters input by the user, including target elevation H, target slope θ (0°-30°), and work area boundary coordinates. These parameters are mapped to the global terrain model to generate a virtual flat ground target surface. The operator inputs the preset flat ground parameters through the touchscreen display of the human-machine interface unit. The target elevation H is referenced to a benchmark point set at the construction site, the target slope θ is set to 15° (within the 0°-30° range), and the work area boundary coordinates are determined by camera selection, covering the entire work surface to be flattened. After receiving these parameters, the system accurately maps them to the global dynamic terrain model. It calculates the target height of each grid cell according to the formula H_target=H+(x-x0)×tanθ (where x is the grid horizontal coordinate and x0 is the starting horizontal coordinate of the target surface), thereby generating a complete virtual flat ground target surface, ensuring a perfect match between the target surface and the preset parameters. S4. Bucket Operation Trajectory Planning: Based on the height difference between the global terrain model and the virtual flat ground target surface, an S-curve algorithm is used to plan the movement trajectory of the bucket teeth, ensuring that the trajectory meets the following requirements: horizontal movement speed of the bucket ≤ 0.8 m / s, vertical lifting acceleration ≤ 2 m / s². 2In S4, the bucket operation trajectory planning also includes collision detection logic: if the distance between the bucket tip and a protruding obstacle (height > 5cm) in the terrain model is < 10cm in the planned trajectory, the trajectory height is automatically adjusted (raised by 5-10cm) while keeping the flat ground target parameters unchanged. The global dynamic terrain model grid within the operation range is traversed, and the difference Δh between the actual height of each grid and the virtual flat ground target surface height is calculated one by one. Based on Δh, an S-curve algorithm is used to plan the movement trajectory of the bucket tip. During trajectory planning, the horizontal movement speed of the bucket is strictly controlled to be 0.6m / s (≤ 0.8m / s), and the vertical lifting acceleration is 1.5m / s² (≤ 2m / s²), avoiding collisions. To prevent decreased operational accuracy due to drastic movements; when Δh > 0 (terrain higher than target surface), control the bucket to descend to the target height of the corresponding grid and maintain a horizontal posture for leveling operations; when Δh < 0 (terrain lower than target surface), control the bucket to maintain the target height for translation to prevent over-excavation; simultaneously, activate collision detection logic to monitor the distance between the planned trajectory and protruding obstacles in the terrain model in real time. If the height of the protruding obstacle is > 5cm and the distance between the bucket tooth tip and the obstacle is < 10cm, the system automatically raises the trajectory height by 8cm (within the 5-10cm range). After the bucket passes the obstacle, immediately restore the original planned trajectory, keeping the target parameters of the flat ground unchanged throughout the process; S5. Inverse Kinematics Calculation: Based on the known parameters of the excavator, such as the link length and joint angle range, a kinematic model is established. The target trajectory coordinates of the bucket are substituted into the model to calculate the target joint angles of the boom, stick, and bucket. Based on the actual structural parameters of the excavator, the boom lengths L1, L2, and L3 are defined. A spatial rectangular coordinate system is established with the center of the excavator's slewing platform as the origin, ensuring that the axes of the coordinate system are consistent with the excavator's body posture. The target trajectory coordinates (x, y, z) of the bucket tooth tips are substituted one by one into the preset excavator kinematic model. Through trigonometric function relationships, the target values ​​of the boom joint angle α, stick joint angle β, and bucket joint angle γ are derived and calculated. During the calculation process, it is ensured that each joint angle is within the allowable angle range of the equipment to avoid exceeding mechanical limits. S6. Closed-Loop Execution Control: Real-time acquisition of feedback signals from joint angle sensors at each joint, comparison with target joint angles, and correction of control commands using a PID algorithm. This drives the hydraulic actuator to move the bucket along the planned trajectory until the error between the terrain of the work area and the virtual flat ground target surface is ≤±1cm. Joint angle sensors are installed at each joint axis of the boom, stick, and bucket to acquire the actual angle of each joint in real time at a frequency of 100Hz. The acquired data is fed back to the motion control unit in real time. The motion control unit compares the actual joint angles with the calculated target joint angles to obtain the angle deviation value. This deviation value is input to the PID controller, which calculates the control increment based on the deviation and then corrects the opening of the proportional solenoid valve. After receiving the control command, the proportional solenoid valve precisely adjusts the oil volume and pressure of the hydraulic cylinder, driving the boom, stick, and bucket to move according to the target joint angle. This closed-loop control process continues to cycle until the error between the actual height of all grids in the work area and the height of the virtual flat ground target surface is ≤±1cm. At this point, the system determines that the leveling operation is complete and automatically stops the actuator. Example 2: The excavator automatic leveling control system based on vision-laser fusion includes: Multi-source sensor module: Composed of a binocular camera, an infrared laser leveling instrument, and a six-axis IMU, it is used to simultaneously collect terrain data, reference height signals, and aircraft attitude data of the work area. The source sensor module consists of a binocular camera, an infrared laser leveling instrument, and a six-axis IMU, and the three work together to complete data acquisition. The binocular camera uses a GigE Ethernet data interface, with a resolution of 1280×720, a baseline distance of 180mm, and a 3D point cloud generation accuracy of ±2.5mm. The acquisition range for the working area is 5-15m, and it can stably output 3D terrain data for the working area. The infrared laser leveling instrument uses RS485 signal transmission, with a laser wavelength of 635nm, an effective measurement distance of 100m, a reference height measurement accuracy of ≤±0.5mm, and a signal transmission delay of ≤10ms, providing a precise height reference for the system. The six-axis IMU uses an SPI data interface, with a sampling rate of 100Hz, a roll and pitch angle measurement accuracy of ≤±0.1°, and an acceleration measurement range of ±16g. It can capture the excavator's attitude changes and motion acceleration in real time, providing data support for terrain model correction. Data Processing Unit: Employing an industrial-grade ARM processor, this unit runs point cloud denoising, sensor fusion, and terrain modeling algorithms with a processing latency of ≤80ms. Specifically, it uses an industrial-grade ARM Cortex-A9 processor with a clock speed of 2.0GHz (≥2.0GHz), paired with 2GB DDR3 memory and 16GB Flash storage, providing powerful data processing and storage capabilities. This unit runs core algorithms such as point cloud denoising, Kalman filtering, and terrain modeling in real time. Through a multi-task parallel processing mechanism, the overall data processing latency is controlled to 70ms (≤80ms), enabling rapid fusion processing of multi-source data and the construction and updating of a global dynamic terrain model, providing timely and accurate terrain data support for subsequent trajectory planning. The human-machine interface (HMI) unit includes a touchscreen display and physical buttons for users to input preset leveling parameters, start / stop operations, and view the terrain model and work area in real time. The HMI unit also has data storage capabilities, recording the time, area, and average error of each leveling operation, with a storage capacity of ≥1000 records. It supports data export via USB. The hardware of the HMI unit includes a 7-inch touchscreen display with a resolution of 1024×600, facilitating operation. The four physical buttons correspond to start, stop, pause, and parameter saving functions, meeting the needs of emergency operations and quick access to frequently used functions. The core functions include two main categories: display and operation. The display function supports 2D / 3D switching of the global terrain model, real-time operation error value display, and preset parameter display, allowing operators to intuitively grasp the operation status. The operation function supports inputting preset leveling parameters (target elevation, slope) via the touch screen and controlling the start, stop, and pause of the operation via physical buttons. It also has an operation data storage function, which can automatically record key data such as the time, operation area, and average error of each leveling operation, with a storage capacity of up to 1200 records (≥1000 records). It supports exporting the stored data via USB interface, which is convenient for subsequent operation analysis, data traceability, and construction acceptance. Motion Control Unit: Stores the excavator's kinematic model and PID control algorithm, receives trajectory data from the data processing unit, and outputs joint angle control commands. The motion control unit also includes an emergency stop module: when sensor feedback signals are abnormal or the operating error is >±3cm, it automatically outputs a stop command, controlling the bucket to return to a safe position. The motion control unit uses a dedicated PLC for construction machinery as its core controller, with 32 / 16 input / output points, supporting pulse output and analog control, adapting to the control requirements of hydraulic actuators. This unit has a built-in pre-calibrated excavator kinematic model and PID control algorithm, with a control cycle set to 8ms (≤10ms). It can quickly receive the bucket target trajectory data output by the data processing unit, calculate and output joint angle control commands for the boom, stick, and bucket. Simultaneously, it integrates an emergency stop module to monitor the feedback signals of each sensor in real time. If sensor abnormalities such as IMU data drift >0.5° are detected, or the operating error is >±3cm, the emergency stop module immediately and automatically outputs a stop command, controlling the bucket to quickly rise to a safe position 35cm (≥30cm) above the ground, preventing equipment damage or serious substandard work quality. The actuator, consisting of a proportional solenoid valve assembly, hydraulic cylinders, and joint angle sensors, drives the bucket movement and provides real-time position feedback. The proportional solenoid valve assembly has a response time of 40ms (≤50ms), an operating voltage of 24V, and a control accuracy of ±0.3% (≤±0.5%), accurately responding to control commands from the motion control unit and adjusting the flow and pressure of the hydraulic oil. The hydraulic cylinders are functionally divided into boom cylinders, stick cylinders, and bucket cylinders, each equipped with a displacement sensor with a displacement measurement accuracy of ±0.1mm, providing real-time feedback on the cylinder's extension and retraction status. The joint angle sensors are installed at the joint axes of the boom, stick, and bucket, with a measurement range of 0°-360° and a measurement accuracy of ±0.05°, acquiring the actual angle of each joint in real-time and feeding it back to the motion control unit, forming a complete closed-loop control circuit.

[0015] Example 3: The invention will be further explained below using a roadbed leveling operation scenario as an example: (1) Preparation for operation: Park the excavator in the roadbed construction area (100m long and 12m wide), set a benchmark point at the construction site (elevation H0=35.2m); install multi-source sensor modules: fix the binocular camera at the front end of the boom, install the laser leveling instrument on the top of the machine body and align it with the benchmark point, and fix the IMU at the center of the slewing platform; the user inputs preset parameters through the human-machine interaction unit: target elevation H=35.2m (aligned with the benchmark point), target slope θ=2% (road cross slope, higher on the left and lower on the right), and the working range is the full width of the roadbed (12m).

[0016] (2) After the data acquisition and fusion system is started, the binocular camera acquires two-dimensional images of the roadbed at a frequency of 15Hz and generates a three-dimensional point cloud (the point cloud density is 100 points per square meter); the laser leveling instrument synchronously acquires the reference height signal (stable at 35.2m), and the IMU acquires the fuselage attitude (roll angle 0.2°, pitch angle 0.1°, and the deviation after compensation is <0.05°); the data processing unit completes point cloud denoising, signal filtering and Kalman fusion within 75ms to construct a rasterized (5cm×5cm) global terrain model. The model shows a local bulge on the left side of the roadbed (the highest point is 35.4m) and a local depression on the right side (the lowest point is 35.0m).

[0017] (3) The trajectory planning and execution system generates a virtual flat target surface based on the target slope (2%). The left elevation of the target surface is 35.26m (the cross slope difference of the 12m wide road is 0.24m), and the right elevation is 35.02m. The bucket trajectory is planned: starting from the starting point on the left side of the roadbed (x=0, y=0), the bucket tooth tip descends to 35.26m and moves horizontally to the right. During the process, the height is automatically adjusted according to the terrain model (the left side protrusion descends 0.14m to level it, and the right side depression is kept at 35.02m for translation). The motion control unit calculates the inverse kinematic solution: the boom joint angle α=45°, the stick joint angle β=120°, and the bucket joint angle γ=95°, and outputs control commands. The actuator response: the proportional solenoid valve adjusts the hydraulic cylinder oil volume, drives the boom, stick, and bucket to move along the trajectory, the joint angle sensor provides real-time feedback (feedback frequency 100Hz), and the PID controller corrects the deviation (maximum deviation 0.5°, correction time <0.3s).

[0018] (4) After the roadbed leveling operation is completed, the roadbed flatness is checked with a high-precision level: 100 points are measured, the average error is 0.8cm, the maximum error is 1.2cm, which meets the accuracy requirements of municipal roadbed construction (≤±2cm); Operation time statistics: Traditional manual leveling takes 40 minutes, while this system only takes 25 minutes, which improves efficiency by 37.5%; The operator only needs to operate when the initial parameters are input, and no intervention is required at other times, which significantly reduces labor intensity.

[0019] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An automatic ground leveling control method for excavators based on vision-laser fusion, characterized in that, Includes the following steps: S1. Multi-source data synchronous acquisition: Two-dimensional images of the working area are acquired by a binocular camera installed at the front of the excavator boom and a three-dimensional terrain point cloud is generated. Ground reference height signals are acquired by a laser leveling instrument. The tilt angle and motion acceleration of the excavator body are acquired by an IMU. All sensors are acquired at the same frequency and ≥15Hz. S2. Data Preprocessing and Fusion: Denoise and downsample the point cloud data from the binocular camera, filter the laser leveling instrument signal, and compensate for drift in the IMU data; use the Kalman filter algorithm to fuse the three types of data and construct a global dynamic terrain model that includes fuser attitude correction; S3, Flat Ground Target Parameter Setting: Receives preset flat ground parameters input by the user, including target elevation H, target slope θ (0°-30°), and work area boundary coordinates, maps the parameters to the global terrain model, and generates a virtual flat ground target surface; S4. Bucket Operation Trajectory Planning: Based on the height difference between the global terrain model and the virtual flat ground target surface, an S-curve algorithm is used to plan the movement trajectory of the bucket teeth, ensuring that the trajectory meets the following requirements: horizontal movement speed of the bucket ≤ 0.8 m / s, vertical lifting acceleration ≤ 2 m / s². 2 ; S5. Inverse kinematics calculation: Based on the known parameters of the excavator, such as the link length and joint angle range, a kinematic model is established. The target trajectory coordinates of the bucket are substituted into the model to calculate the target joint angles of the boom, stick, and bucket. S6. Closed-loop execution control: Real-time acquisition of feedback signals from angle sensors of each joint, comparison with the target joint angle, and correction of control commands using PID algorithm to drive the hydraulic actuator to move the bucket along the planned trajectory until the error between the terrain of the working area and the virtual flat target surface is ≤ ±1cm.

2. The automatic ground leveling control method for excavators based on vision-laser fusion according to claim 1, characterized in that: The binocular camera in S1 has a resolution ≥1280×720, a baseline distance of 150-200mm, a 3D point cloud generation accuracy ≤±3mm, and a work area acquisition range of 5-15m; the laser leveling instrument is an infrared laser leveling instrument with a reference height measurement accuracy ≤±0.5mm and a signal transmission delay ≤10ms; the IMU is a six-axis IMU with a sampling rate ≥100Hz and a roll / pitch angle measurement accuracy ≤±0.1°.

3. The automatic ground leveling control method for excavators based on vision-laser fusion according to claim 1, characterized in that: The global dynamic terrain model in S2 is represented by a rasterized grid with a grid size of 5cm×5cm. The model update frequency is synchronized with the sensor acquisition frequency. Each update retains 80% of the effective point cloud data from the previous frame to reduce computational load.

4. The automatic ground leveling control method for excavators based on vision-laser fusion according to claim 1, characterized in that: The bucket operation trajectory planning in S4 also includes collision detection logic: if the distance between the bucket tip and the protruding obstacle (height > 5cm) in the terrain model in the planned trajectory is < 10cm, the trajectory height is automatically adjusted (raised by 5-10cm) while keeping the flat ground target parameters unchanged.

5. An automatic leveling control system for excavators based on vision-laser fusion, characterized in that, include: Multi-source sensor module: Composed of a binocular camera, an infrared laser leveling instrument, and a six-axis IMU, used to simultaneously acquire terrain data, reference height signals, and aircraft attitude data of the work area; Data processing unit: Employs an industrial-grade ARM processor to run point cloud denoising, sensor fusion, and terrain modeling algorithms, with a processing latency of ≤80ms; Human-computer interaction unit: includes a touch screen and physical buttons, used by users to input preset flat terrain parameters, start / stop operations, and display the terrain model and operation errors in real time; Motion control unit: Stores the excavator's kinematic model and PID control algorithm, receives trajectory data from the data processing unit, and outputs joint angle control commands; Actuators include a proportional solenoid valve assembly, hydraulic cylinders, and joint angle sensors, which are used to drive the bucket movement and provide real-time position feedback.

6. The excavator automatic leveling control system based on vision-laser fusion according to claim 5, characterized in that: The human-computer interaction unit also has a work data storage function, which can record the time, work area and average error of each leveling operation, with a storage capacity of ≥1000 records, and supports data export via USB.

7. The excavator automatic leveling control system based on vision-laser fusion according to claim 5, characterized in that: The motion control unit also includes an emergency stop module: when the sensor feedback signal is abnormal or the operating error is greater than ±3cm, it automatically outputs a stop command and controls the bucket to return to a safe position.