Intelligent weeding control method of laser weeding robot and laser weeding robot

Abstract

The invention discloses an intelligent weeding control method of a laser weeding robot and the laser weeding robot. The method comprises the steps that ground images are collected through a binocular depth camera, and weeds and seedlings are recognized; judging the relationship between the actual density of the inter-plant weeds and a preset density threshold value; when the actual density is greater than a preset threshold value, setting a first target speed to run the laser weeding robot, and planning a strike path based on the weed position to execute point descending type strike; and when the actual density is smaller than or equal to the preset threshold value, setting a second target speed for operation and starting a surface scanning mode to cover the high-density weed area. Dynamic matching of speed grading and strike strategies is achieved through a density threshold decision mechanism, the weeding rate is guaranteed, meanwhile, the operation efficiency is optimized, the problems of low-density working condition energy waste and high-density working condition clearing efficiency reduction are solved, and the adaptability of a farmland with complex weed distribution is comprehensively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural weed control, in particular to an intelligent weed control method of a laser weeding robot and the laser weeding robot. BACKGROUND

[0002] When planting crops, a key step to ensure that crops can grow healthily and achieve high yield is to effectively control weeds. Weeds will compete with crops for nutrients, water and light in the soil, seriously interfere with the normal development process of crops, and thus cause problems of yield reduction and quality reduction. Traditionally, weeding is mainly completed by manual removal, mechanical operation or the use of chemical agents. In recent years, with the advancement of environmental protection demand and intelligent agricultural technology, laser weeding technology has gradually become a research hotspot.

[0003] In the prior art, patent CN 118235753A discloses a weeding method of a laser weeding robot, which creates a crop distribution map based on scanning data, defines weed positions and weed characteristics based on the crop distribution map and determines a weeding area, defines laser power of the laser weeding robot in combination with the weed characteristics, and adjusts a weeding strategy of the laser weeding robot based on environmental characteristics of the crop distribution map and weeding efficiency of the laser weeding robot. The weeding strategy of the patent scheme mainly includes determining the intensity, focal point and scanning path of the laser. This single mode of weeding strategy is difficult to balance weeding efficiency and weeding quality when the weed distribution density frequently changes in the field for the complex and variable field environment. Another prior art CN 117136934A provides a flame and laser coupled weeding robot with weed density detection function, which includes a mobile platform, a weed identification device, a flame and laser weeding device, identifies and classifies inter-row weed density through a deep learning model Deeplabv3+, and realizes weeding in combination with inter-plant laser weeding and inter-row variable intensity flame weeding. In the patent scheme, the flame weeding device is additionally added to cooperate with the laser weeding device for collaborative weeding, which can improve weeding efficiency and effect, but greatly increases the use cost and control difficulty. SUMMARY

[0004] The present application provides an intelligent weed control method of a laser weeding robot and the laser weeding robot, which can balance weeding effect and efficiency.

[0005] Technical scheme: To achieve the above-mentioned purpose, the intelligent weed control method of the laser weeding robot of the present application, the method comprises:

[0006] collecting images of the ground through the binocular depth camera of the laser weeding module;

[0007] identify weeds and seedlings based on the collected images;

[0008] determine the relationship between the actual density of weeds between plants in the image and a preset density threshold;

[0009] When the actual density is greater than the preset density threshold, set a first target speed as the target running speed of the laser weeding robot, and plan a striking path based on the position of weeds between plants to strike weeds between plants in a point drop mode;

[0010] When the actual density is less than or equal to the preset density threshold, set a second target speed lower than the first target speed as the target running speed of the laser weeding robot, and use a surface scanning mode to cover the area with high density of weeds.

[0011] Further, the number of laser weeding modules is multiple, and the final target running speed of the laser weeding robot is determined based on the minimum value of the target running speeds corresponding to all the laser weeding modules. In fact, as long as the actual density of weeds in the working area corresponding to one laser weeding module is higher than the preset density threshold, the laser weeding robot runs at the second target speed, and at this time, each laser weeding module runs according to its own weeding mode, and some laser weeding modules may run in a point drop mode and some laser weeding modules may run in a surface scanning mode.

[0012] Further, the identification of weeds and seedlings based on the collected images includes:

[0013] extracting images within the working width corresponding to the laser weeding module in the image;

[0014] using a YOLOv8-based recognition model to distinguish seedlings and weeds and locate the three-dimensional coordinates of weeds.

[0015] Further, the striking path planning based on the position of weeds between plants to strike weeds between plants in a point drop mode includes:

[0016] Based on the three-dimensional coordinates of the weeds to be cleaned and the spatial coordinate relationship between the digital micromirror and the binocular depth camera in the laser weeding module, the theoretical striking coordinates of the weeds are calculated;

[0017] Based on the theoretical striking coordinates of the weeds, the striking sequence of the weeds is planned to obtain a striking path;

[0018] Based on the real-time advancing speed of the laser weeding robot and the theoretical striking coordinate, the digital vibration mirror is controlled to dynamically deflect, and based on the laser beam focal length adjusting mechanism, the working depth of the laser beam focal point is adjusted, so as to sequentially strike the weeds and keep the continuous tracking time of each weed reaching the preset time length. The laser beam focal length adjusting mechanism adjusts the height of the laser field mirror through the motor-driven screw rod to realize the adjustment of the working depth of the laser beam focal point.

[0019] Further, the area with high density of weeds covered by the area scanning mode specifically includes:

[0020] The area coordinate parameters of the target area with the actual density of weeds greater than the preset density threshold are extracted; the target area is generally extracted into a regular shape with horizontal and vertical, and the area coordinate parameters are the coordinates of the inflection points of the target area;

[0021] The scanning path is determined according to the area coordinate parameters and the scanning radius of the area scanning mode; the scanning path can be a square wave path composed of connected horizontal paths, the width of each horizontal path is determined by the width of the target area at the position, and the interval between adjacent horizontal paths is less than the scanning radius, such as the interval between adjacent horizontal paths being equal to one-half of the scanning radius, so that the scanning area of the laser weeding module overlaps when scanning adjacent two horizontal paths, so that the ablation time of the weeds can meet the requirements;

[0022] Based on the real-time advancing speed of the laser weeding robot and the scanning path, the digital vibration mirror is controlled to dynamically deflect, so that the scanning area of the laser weeding module moves along the scanning path, and the ablation of the target area is realized.

[0023] Further, the method further includes:

[0024] Before the operation starts, the boundary coordinates of the farmland and the crop row spacing parameters are uploaded through the cloud platform;

[0025] The operation path generated by the cloud platform and the mechanical weeding module are received, and the operation path includes the logic of inter-ridge turning and head turning; the cloud platform generates the operation path based on the improved A* algorithm;

[0026] The laser weeding robot is controlled to autonomously drive along the operation path.

[0027] Further, the control of the laser weeding robot to autonomously drive along the operation path includes:

[0028] The laser weeding robot is positioned based on multi-sensor fusion data, and the deviation of the heading direction is corrected; in the scheme, the laser weeding robot has a GNSS receiver and an IMU, and the control system performs Kalman filtering fusion on the RTK positioning data output by the GNSS receiver and the inertial data output by the IMU, compensates the drift caused by satellite signal shielding in real time, and corrects the deviation of the heading direction of the robot;

[0029] The visual data collected by the panoramic camera in real time are used for visual auxiliary deviation correction, so that the laser weeding robot travels in a straight line along the ridge direction; in the scheme, the panoramic camera captures the ridge ditch image in front of 20 m, extracts the ridge center line through the U-Net semantic segmentation network, dynamically adjusts the wheel leg steering angle in combination with the PID controller, and performs visual auxiliary deviation correction.

[0030] Based on the collected data of the laser radar, a three-dimensional point cloud in the field is constructed, and obstacles in the field are detected in real time and obstacle avoidance motion is performed on the obstacles.

[0031] Further, the method further comprises:

[0032] The range increasing system preferentially uses battery power, and automatically switches to a hybrid mode when the battery power is lower than 60%. In the hybrid mode, the diesel generator starts to generate electricity and supplements the battery power.

[0033] A laser weeding robot comprises:

[0034] A walking frame;

[0035] A laser weeding module having a binocular depth camera, a laser module with a digital galvanometer, and a laser beam focal length adjusting mechanism;

[0036] A mechanical weeding module having a hoeing shovel and a driving mechanism capable of driving the hoeing shovel to be retracted and extended; the mechanical weeding module and the laser weeding module are staggered in the left-right direction;

[0037] A sensor group for positioning the laser weeding robot and collecting surrounding environment data; specifically, the sensor group comprises a GNSS receiver, an IMU, a panoramic camera, and a laser radar;

[0038] A range increasing system comprising a battery, a diesel generator, and a charger connected between the diesel generator and the battery;

[0039] A control system for implementing the intelligent weeding control method described above.

[0040] Beneficial effects: the intelligent weeding control method of the laser weeding robot and the laser weeding robot have the following beneficial effects:

[0041] (1) In the present application, by setting a preset density threshold as the decision basis for operation mode switching, the speed classification and dynamic matching of the striking strategy are carried out accordingly, which optimizes the operation efficiency while ensuring the weed control rate, avoids energy waste when the actual density of interplant weeds is low, and avoids the problem of affecting the running speed of the machine when the interplant weed density is high, thereby improving the adaptability of the complex weed distribution in the farmland scene as a whole.

[0042] (2) The working width of the mechanical weeding module on both sides of the laser weeding module can determine the working width of the laser weeding module, and by excluding the working area of the mechanical weeding module, the subsequent image recognition difficulty can be reduced, and the waste of computing power can be reduced.

[0043] (3) By establishing the spatial mapping relationship calculation theory of weed three-dimensional coordinates and digital micromirror, the millimeter-level positioning accuracy of digital micromirror deflection is ensured by combining path planning and real-time speed dynamic compensation; at the same time, by presetting the tracking time length, the laser energy can continuously act on the weeds, realizing the accurate ablation of single weed in 0.4-0.8s in the moving state, avoiding missing or damaging the seedlings, and significantly improving the interplant weeding effectiveness and operation efficiency.

[0044] (4) The face scanning mode defines the target area by extracting the regularized inflection point coordinates of the high-density weed area, and combines the half radius interval overlap design of the square wave scanning path to ensure that the laser coverage has no dead angle and the ablation time meets the standard; at the same time, based on the real-time speed of the robot, the digital micromirror deflection is dynamically regulated, so that the scanning area accurately tracks the path trajectory, and by making the interval between adjacent transverse paths less than the scanning radius, the ablation time can be guaranteed, realizing high killing rate in high-density weed area. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a side view structure diagram of the laser weeding robot;

[0046] Figure 2 is a perspective structure diagram of the laser weeding robot;

[0047] Figure 3 is a top view structure diagram of the laser weeding robot;

[0048] Figure 4 is Figure 2 is an enlarged structure diagram of part A in the present application;

[0049] Figure 5 is a first state structure diagram of the spade part in the preferred embodiment;

[0050] Figure 6 is a second state structure diagram of the spade part in the preferred embodiment;

[0051] Figure 7 is a flowchart of the intelligent weeding control method.

[0052] In the figure: 1 - walking frame; 2 - laser weeding module; 21 - binocular depth camera; 22 - laser module; 23 - laser beam focal length adjusting mechanism; 3 - mechanical weeding module; 31 - hoe; 311 - hoe body; 312 - widening part; 313 - connecting rod; 314 - rotary drive motor; 32 - driving mechanism; 4 - GNSS receiver; 5 - IMU; 6 - panoramic camera; 7 - laser radar; 81 - battery; 82 - diesel generator. DETAILED DESCRIPTION

[0053] The application will be further described below in conjunction with the accompanying drawings.

[0054] As Figures 1 to 3 shown, the laser weeding robot of the application comprises:

[0055] a walking frame 1;

[0056] a laser weeding module 2, as Figure 4 shown, having a binocular depth camera 21, a laser module 22 with a laser generator and a digital galvanometer, and a laser beam focal length adjusting mechanism 23; the laser weeding module 2 is used for inter-plant weeding.

[0057] a mechanical weeding module 3 having a plurality of hoes 31 arranged left and right and a driving mechanism 32 capable of driving the hoes 31 to be retracted and extended; the mechanical weeding module 3 is staggered with the laser weeding module 2 in the left-right direction; the hoes 31 are mainly used for inter-row weeding;

[0058] a sensor group for positioning the laser weeding robot and collecting surrounding environment data; specifically, the sensor group comprises a GNSS receiver 4, an IMU 5, a panoramic camera 6, and a laser radar 7;

[0059] a range extending system 8 comprising a battery 81, a diesel generator 82, and a charger connecting the diesel generator 82 and the battery 81;

[0060] a control system for implementing an intelligent weeding control method as follows.

[0061] The intelligent weeding control method of the laser weeding robot as Figure 7 shown, the method comprises the following steps S101-S105:

[0062] Step S101, collecting images of the ground by the binocular depth camera 21 of the laser weeding module 2;

[0063] Step S102, identifying weeds and seedlings based on the collected images;

[0064] Step S103, judging the relationship between the actual density of weeds between plants in the image and the preset density threshold; in the embodiment, the preset density threshold is 5 plants / m 2 ;

[0065] Step S104, when the actual density is greater than the preset density threshold, setting a first target speed as the target running speed of the laser weeding robot, and planning a striking path based on the position of weeds between plants to perform point reduction type striking on weeds between plants; in the embodiment, the first target speed is 0.6 m / s;

[0066] Step S105, when the actual density is less than or equal to the preset density threshold, setting a second target speed lower than the first target speed as the target running speed of the laser weeding robot, and using a surface scanning mode to cover the area with high density of weeds. In the embodiment, the second target speed is 0.4 m / s.

[0067] In the present application, by setting a preset density threshold as the decision basis for switching the operation mode, the speed classification and the dynamic matching of the striking strategy are performed accordingly, which optimizes the operation efficiency while ensuring the weeding rate, avoids energy waste when the actual density of weeds between plants is low, and avoids the problem of affecting the running speed of the machine when the density of weeds between plants is high, thereby improving the adaptability of the complex weed distribution in the farmland scene as a whole.

[0068] Preferably, the number of laser weeding modules 2 is multiple, and the final target running speed of the laser weeding robot is determined based on the minimum value of the target running speeds corresponding to all the laser weeding modules 2. That is, in actual operation, as long as the actual density of weeds in the operation area corresponding to one laser weeding module 2 is higher than the preset density threshold, the laser weeding robot runs at the second target speed, at this time, among the laser weeding modules 2, each laser weeding module 2 runs according to its own weeding mode, and some laser weeding modules 2 may run in the point reduction type striking mode, and some laser weeding modules 2 may run in the surface scanning mode.

[0069] Preferably, in the step S102, the identification of weeds and seedlings based on the collected image comprises:

[0070] Step S201, extracting the image corresponding to the working width of the laser weeding module 2 in the image;

[0071] Step S202, using a YOLOv8-based recognition model to distinguish seedlings and weeds and locate the three-dimensional coordinates of the weeds.

[0072] The working width of the mechanical weeding module 3 on both sides of the laser weeding module 2 can determine the working width of the laser weeding module 2, and by excluding the working area of the mechanical weeding module 3, the subsequent image recognition difficulty can be reduced, and the waste of computing power can be reduced.

[0073] Preferably, in the step S104, the striking path planning based on the positions of the weeds between the plants is performed to perform point reduction striking on the weeds between the plants, including the following steps S301-S303:

[0074] In step S301, the theoretical striking coordinates of the weeds are calculated based on the three-dimensional coordinates of the weeds to be cleaned and the spatial coordinate relationship between the digital micromirror and the binocular depth camera 21 in the laser weeding module 2;

[0075] In step S302, the striking sequence of the weeds is planned based on the theoretical striking coordinates of the weeds to obtain a striking path;

[0076] In step S303, the digital micromirror is controlled to dynamically deflect based on the real-time advancing speed of the laser weeding robot and the theoretical striking coordinates, and the laser beam focal point working depth is adjusted based on the laser beam focal length adjusting mechanism 23 to sequentially strike the weeds and keep the continuous tracking time of each weed to reach a preset time length. In this embodiment, the preset time length is 0.4-0.8s. The laser beam focal length adjusting mechanism 23 adjusts the height of the laser field lens by a motor-driven screw to adjust the laser beam focal point working depth.

[0077] By establishing the spatial mapping relationship between the three-dimensional coordinates of the weeds and the digital micromirror to calculate the theoretical striking coordinates, combining path planning and real-time speed dynamic compensation, the millimeter-level positioning accuracy of the digital micromirror deflection is ensured; at the same time, by presetting the tracking time, the laser energy can continuously act on the weeds, realizing 0.4-0.8s accurate ablation of single weed in a moving state, avoiding missed striking or damaging seedlings, and significantly improving the inter-plant weeding effectiveness and working efficiency.

[0078] Preferably, in the step S105, the target area with an actual density greater than a preset density threshold is extracted to obtain the region coordinate parameters of the target area; the target area is generally extracted into a regular shape with horizontal and vertical, and the region coordinate parameters are the inflection point coordinates of the target area.

[0079] In step S401, the region coordinate parameters of the target area with an actual density greater than a preset density threshold are extracted; the target area is generally extracted into a regular shape with horizontal and vertical, and the region coordinate parameters are the inflection point coordinates of the target area.

[0080] Step S402, determining a scanning path according to the region coordinate parameter and the scanning radius of the area scanning mode; the scanning path can be a square wave path composed of connected lateral paths, the width of each lateral path is determined by the width of the target region at the position, and the interval between adjacent lateral paths is less than the scanning radius, such as the interval between adjacent lateral paths being equal to one half of the scanning radius, so that the scanning region overlaps when the laser weeding module 2 scans adjacent two lateral paths, so as to ensure that the ablation time of weeds meets the requirements;

[0081] Step S403, based on the real-time advancing speed of the laser weeding robot and the scanning path, controlling the dynamic deflection of the digital vibration mirror, so that the scanning region of the laser weeding module 2 moves along the scanning path, realizing the coverage ablation of the target region.

[0082] The area scanning mode defines the target region by extracting the regularized inflection point coordinates of the high-density weed area, and combines the half-radius interval overlap design of the square wave scanning path to ensure that the laser coverage has no dead angle and the ablation time meets the standard; at the same time, based on the real-time speed of the robot, the deflection of the digital vibration mirror is dynamically adjusted, so that the scanning region accurately tracks the path trajectory, and by making the interval between adjacent lateral paths less than the scanning radius, the ablation time is guaranteed, and high killing rate of high-density weed area is realized.

[0083] Since the laser weeding module is difficult to remove the roots, preferably, the hoe 31 comprises a hoe body 311, further comprises a widening part 312 capable of rotating relative to the hoe body 311, and a driving mechanism for driving the widening part 312 to rotate relative to the hoe body 311, the driving mechanism comprising a connecting rod 313 connected to the widening part 312 and a rotary driving motor 314 for driving the connecting rod 313 to rotate.

[0084] The widening part 312 can be switched between the storage state and the working state by the driving mechanism, as shown in Figure 5 In the storage state, the widening part 312 is located at the rear side of the hoe body 311, as shown in Figure 6 In the working state, at least part of the widening part 312 extends outward from one side of the hoe body 311 to intervene between the two adjacent plants to remove the roots between the plants, and the connecting rod 313 is perpendicular to the widening part 312 in the working state. In actual use, an image acquisition unit can be provided for each weeding component 32, and the control system controls the switching state of the widening part 312 based on the image acquired by the image acquisition unit. With the above structure, the hoe 31 can not only weed between rows, but also clean the roots between plants, further improving the weeding effect.

[0085] Preferably, the method further comprises steps S501-S503:

[0086] Step S501, before starting the work, upload the farmland boundary coordinates and crop row spacing parameters through the cloud platform;

[0087] Step S502, receive the work path generated by the cloud platform and the mechanical weeding module 3's retracting and deploying timing, the work path includes inter-ridge turn and head turning logic; the cloud platform generates the work path based on the improved A* algorithm;

[0088] Step S503, control the laser weeding robot to drive autonomously along the work path, during the driving process of the laser weeding robot, control the laser weeding module 2 to operate based on the above steps S101-S102, and control the mechanical weeding module 3 to lift up and put down based on the retracting and deploying timing, that is, control the driving mechanism 32 to retract and deploy the spade 31.

[0089] Receive farmland information through the cloud platform and generate the optimal work path, realize automatic path planning and autonomous navigation, improve work efficiency and accuracy, reduce manual intervention, and adapt to complex farmland environment.

[0090] Preferably, the control of the laser weeding robot to drive autonomously along the work path in the above step S503 includes the following steps S601-S603:

[0091] Step S601, position the laser weeding robot based on multi-sensor fusion data and correct the heading deviation; in this embodiment, the laser weeding robot has a GNSS receiver 4 and an IMU 5, the control system performs Kalman filter fusion on the RTK positioning data output by the GNSS receiver 4 and the inertial data output by the IMU 5, compensates for the drift caused by satellite signal blockage in real time, and corrects the robot heading deviation;

[0092] Step S602, based on the visual data collected by the panoramic camera 6 in real time, perform visual auxiliary correction to make the laser weeding robot drive straight along the ridge direction; in this embodiment, the panoramic camera 6 captures the image of the front 20m ridge and ditch, extracts the ridge center line through the U-Net semantic segmentation network, dynamically adjusts the wheel leg steering angle combined with the PID controller, and performs visual auxiliary correction;

[0093] Step S603, construct a three-dimensional point cloud in the field based on the collected data of the laser radar 7, and detect and avoid obstacles in real time.

[0094] Preferably, the method further comprises:

[0095] The control range increasing system 8 preferentially uses the battery 81 for power supply, and automatically switches to the hybrid mode when the battery 81 is below 60%. In the hybrid mode, the diesel generator 82 starts power generation and supplements the battery 81.

[0096] The above merely describes the preferred embodiments of the present application, and it should be pointed out that those skilled in the art can make several improvements and refinements without departing from the principles of the present application, and these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. An intelligent weed control method for a laser weeding robot, the method comprising: The binocular depth camera of the laser weeding module is used to collect images of the ground; Identify weeds and seedlings based on collected images; Characterized in that the method further comprises: Determine the relationship between the actual density of weeds between plants in the image and the preset density threshold; When the actual density is greater than the preset density threshold, a first target speed is set as the target operating speed of the laser weeding robot, and an attack path is planned based on the positions of the weeds between the plants, so as to perform a point-down attack on the weeds between the plants; When the actual density is less than or equal to the preset density threshold, a second target speed lower than the first target speed is set as the target operating speed of the laser weeding robot, and a surface scanning mode is used to cover areas with high weed density.

2. The intelligent weed control method of the laser weeding robot according to claim 1, characterized in that: There are multiple laser weeding modules, and the final target running speed of the laser weeding robot is determined based on the minimum value of the target running speeds corresponding to all the laser weeding modules.

3. The intelligent weed control method of the laser weeding robot according to claim 1, characterized in that: The identification of weeds and seedlings based on the collected images includes: extracting an image corresponding to a working width of the laser weeding module from the image; A YOLOv8-based recognition model is used to distinguish seedlings from weeds and locate the three-dimensional coordinates of weeds.

4. The intelligent weed control method of the laser weeding robot according to claim 1, characterized in that: The method of performing strike path planning based on the positions of weeds between plants and performing point-to-point strike on weeds between plants includes: Calculate the theoretical striking coordinates of the weeds based on the three-dimensional coordinates of the weeds to be cleaned and the spatial coordinate relationship between the digital galvanometer and the binocular depth camera in the laser weed removal module; Based on the theoretical attack coordinates of weeds, the attack sequence of weeds is planned and the attack path is obtained; Based on the real-time forward speed and theoretical striking coordinates of the laser weeding robot, the dynamic deflection of the digital galvanometer is controlled, and the laser beam focus operating depth is adjusted based on the laser beam focal length adjustment mechanism, so as to strike the weeds in sequence and maintain the continuous tracking time of each weed to a preset time.

5. The intelligent weed control method of the laser weeding robot according to claim 1, characterized in that: The area where weeds are densely distributed using the area scanning mode specifically includes: Extracting regional coordinate parameters of a target area where the actual density of weeds is greater than a preset density threshold; Determining a scanning path according to the area coordinate parameters and a scanning radius of a surface scanning mode; Based on the real-time forward speed of the laser weeding robot and the scanning path, the digital galvanometer is controlled to dynamically deflect, so that the scanning area of ​​the laser weeding module moves along the scanning path, thereby achieving coverage and ablation of the target area.

6. The intelligent weed control method of the laser weeding robot according to claim 1, characterized in that: The method further comprises: Before the operation begins, the coordinates of the farmland boundary and crop row spacing parameters are uploaded through the cloud platform; Receive the operation path generated by the cloud platform and the retraction and deployment sequence of the mechanical weeding module, the operation path including the inter-ridge return and field headland turning logic; Control the laser weeding robot to move autonomously along the operation path.

7. The intelligent weed control method of the laser weeding robot according to claim 6, characterized in that: The method of controlling the laser weeding robot to autonomously travel along an operation path includes: Position the laser weeding robot based on multi-sensor fusion data and correct the heading deviation; Based on the visual data collected in real time by the panoramic camera, visual-assisted deviation correction is performed to enable the laser weeding robot to move in a straight line along the ridge direction; Based on the data collected by LiDAR, a three-dimensional point cloud of the field is constructed to detect field obstacles in real time and perform obstacle avoidance movements.

8. The intelligent weed control method of the laser weeding robot according to claim 1, characterized in that: The method further comprises: The range-extending system is controlled to give priority to battery power supply, and automatically switches to hybrid mode when the battery power level is less than 60%.

9. A laser weeding robot, characterized in that: It includes: Walking frame; A laser weeding module, which has a binocular depth camera, a laser module with a digital galvanometer, and a laser beam focus adjustment mechanism; A mechanical weeding module having a hoe and a drive mechanism capable of retracting and extending the hoe; the mechanical weeding module and the laser weeding module are staggered in the left-right direction; A sensor group, used to locate the laser weeding robot and collect data about the surrounding environment; a range-extending system, which includes a battery, a diesel generator, and a charger connecting the diesel generator and the battery; A control system for implementing the intelligent weed control method according to any one of claims 1 to 8.

Images