Wine grape inter-plant weeding device free of turning around and control method of wine grape inter-plant weeding device

Abstract

The invention discloses a turning-free inter-plant weeding device for wine grapes and a control method of the turning-free inter-plant weeding device. The turning-free inter-plant weeding device comprises GPS equipment, IMU equipment, a weeding cutter, a differential chassis and a monocular camera, the GPS devices are installed at the front end and the rear end of the central axis of the top end of the differential chassis respectively. The IMU devices are installed on the central axis of the differential chassis, and the IMU devices are tightly attached to the GPS device. The weeding cutters are installed at the front end and the rear end of the differential chassis respectively. The monocular camera is installed on the lower side of the square steel on the top of the differential chassis. By designing the weeding cutters at the front end and the rear end and combining the two-way operation capacity of the differential chassis, operation can be completed between grape rows without turning around, the operation efficiency and productivity are greatly improved, and meanwhile, the continuity and efficiency of operation are further improved by changing the working states and the advancing directions of the weeding cutters at the front end and the rear end.

Description

Technical Field

[0001] The invention belongs to the technical field of grape weeding devices, and in particular relates to a wine grape weeding device that does not require U-turns and turns, and a control method thereof. Background Art

[0002] The existing gantry-type wine grape weeding device relies entirely on manual remote control for operation and driving, and still cannot achieve unmanned operation; and the weeding device needs to make the center axis of the differential chassis accurately coincide with the center line of the grape row during driving, which is difficult to achieve precise control manually. After completing a row of operations, since the obstacle avoidance direction of the weeding tool equipment is fixed, the entire differential chassis needs to turn around. Turning around will cause great difficulties for differential chassis navigation and consume a lot of time. Therefore, operating without turning around can not only greatly improve operating efficiency and productivity, but also avoid the problem of navigation turning around, making device navigation easier. If the line break position is not accurate and smooth during line break, it will affect the accuracy of subsequent navigation, so a smooth and accurate line break control method is needed.

[0003] Generally, the automatic navigation operations of the device include navigation based on GPS, inertial navigation (IMU), laser radar (LiDAR), vision-based navigation, magnetic navigation, ultra-wideband (UWB)-based navigation, RFID-based navigation, ultrasonic-based navigation, multi-sensor fusion navigation, and map-based navigation (SLAM). Since the working environment is in an open environment and requires extremely high precision, the navigation method selects the multi-sensor fusion navigation method of GPS (RTK), IMU and vision. Summary of the invention

[0004] The present invention aims to solve the deficiencies of the prior art and provides the following solutions:

[0005] A wine grape weeding device that does not require U-turns and turns, comprising: a GPS device, an IMU device, a weeding cutter, a differential chassis, and a monocular camera;

[0006] The GPS devices are respectively installed at the front and rear ends of the top center axis of the differential chassis;

[0007] The IMU devices are respectively installed on the central axis of the differential chassis, and the IMU devices are closely attached to the GPS device;

[0008] The weeding cutters are respectively mounted at the front and rear ends of the differential chassis;

[0009] The monocular camera is installed on the lower side of the square steel on the top of the differential chassis.

[0010] The present invention also provides a control method for a wine grape weeding device that does not require U-turns or turns. The control method is used to control the weeding device, and comprises the following steps:

[0011] Get the longitude and latitude information of each row of grapes, take a longitude and latitude coordinate every preset distance, and get the longitude and latitude points of each row;

[0012] Obtain the heading angle of the differential chassis and calculate the heading error of the differential chassis;

[0013] Taking the line connecting every two of the longitude and latitude points as a reference path, and discretizing the reference path to obtain discretized longitude and latitude points;

[0014] Calculating the distance between two points in the discretized longitude and latitude points, and calculating the probability distribution of each point based on the heading error;

[0015] The probability distribution and the ideal reference distribution are optimized using KL divergence and JS divergence to obtain the optimal tracking point;

[0016] Calculating the distance between the optimal tracking point and the differential chassis, and calculating the navigation lateral error;

[0017] Using a monocular camera to obtain a wire pile image and obtain two-dimensional position information, extracting the distance between the center point of the wire pile and the vertical center line of the image, that is, obtaining a visual lateral error;

[0018] Calculating a chassis lateral error based on the navigation lateral error and the visual lateral error, and controlling the chassis lateral error to 0 by proportional control, thereby achieving navigation control of the weeding device;

[0019] After the weeding device has completed a row, before changing rows, the working states and forward directions of the two sets of weeding tool equipment are exchanged, information from the GPS device at the other end is collected, and row change navigation is performed using a cubic function curve path.

[0020] Preferably, the method for calculating the heading error includes:

[0021] Get the relative heading angle of the IMU device;

[0022] Before the weeding device starts working, the differential chassis is allowed to move straight for 1 m, the longitude and latitude coordinates of the start and end points are recorded, and the longitude and latitude coordinates of the start and end points are inversely solved using Vincenty to obtain the azimuths of the start and end points;

[0023] The heading angle is calculated based on the relative heading angle and the start and end point azimuths:

[0024] θ=θ v -θ i

[0025] Where θ represents the heading angle, θ v Indicates the azimuth of the start and end points, θ i Represents the relative heading angle;

[0026] Calculate the heading error based on the heading angle:

[0027] e h =θ f -θ

[0028] Among them, e h represents the heading error, θ f Indicates the first azimuth angle between the differential chassis and the target point.

[0029] Preferably, the method for calculating the probability distribution of each point comprises:

[0030] Calculate the distance v1(s) from the discretized longitude and latitude point to the differential chassis using Vincenty, record the heading error as v2(s), and record the curvature from the differential chassis to the reference path as v3(s);

[0031] Use the sigmoid function to normalize the distance v1(s), heading error v2(s) and curvature v3(s) to obtain the feature vector v(s) = [v1(s), v2(s), v3(s)] of each point;

[0032] The Softmax function is used to convert the feature vector into the probability distribution p(s)=[p1(s), p2(s), p3(s)].

[0033] Preferably, the method for obtaining the optimal tracking point includes:

[0034] Set the ideal reference distribution p*(s) and construct an intermediate distribution between the probability distribution and the ideal reference distribution m ;

[0035] Using KL divergence, constructing KL divergence functions based on the probability distribution, the ideal reference distribution and the intermediate distribution m: KL(p||m) and KL(p*||m);

[0036] Using JS divergence, the JS divergence function is calculated based on the KL divergence function:

[0037]

[0038] Based on the JS divergence function, the existing reference path is dynamically evaluated to obtain the optimal tracking point (φ path (s*),λ path (s*)).

[0039] Preferably, the method for calculating the navigation lateral error includes:

[0040] Calculate the great circle distance between the optimal tracking point and the differential chassis, the orientation of the reference path at the optimal tracking point, and the second azimuth angle from the reference point to the differential chassis using Vincenty;

[0041] Calculate the navigation lateral error based on the great circle distance, the heading and the second azimuth:

[0042] e g =d geo ·sin(α g -θ path (s * ))

[0043] Among them, e g represents the lateral navigation error, d geo represents the great circle distance between the optimal tracking point and the differential chassis, θ path (s) represents the direction of the reference path at the optimal tracking point, α g Indicates the second azimuth angle from the reference point to the differential chassis.

[0044] Preferably, the method for calculating the chassis lateral error includes:

[0045] e=μ g e g +(1-μ g ) s +ε

[0046] Among them, μ g represents GPS confidence, ε represents inherent error, e s Represents the lateral visual error.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The present invention realizes the completion of the operation between the grape rows without turning around by designing the front and rear double-end weeding cutters and combining the two-way operation capability of the differential chassis, which greatly improves the operation efficiency and productivity; adopts the multi-sensor fusion technology of GPS equipment (RTK), IMU equipment and monocular camera to realize high-precision centimeter-level positioning and navigation; through the optimization of KL divergence and JS divergence between probability distribution and ideal reference distribution, dynamically evaluates and selects the optimal tracking point, which effectively solves the error accumulation problem caused by factors such as noise and complex terrain in path tracking, and improves the stability and accuracy of path tracking; after a row of operations is completed, the present invention adopts a cubic function curve for line-changing path planning to ensure that the line-changing process is smooth and natural, avoiding the problems of inaccurate path and uneven driving in the traditional line-changing method. At the same time, by exchanging the working state and forward direction of the front and rear weeding cutters, the continuity and efficiency of the operation are further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0050] Figure 1 A schematic diagram of the structure of a weeding device according to an embodiment of the present invention;

[0051] Figure 2 Schematic diagram of the control flow of the weeding device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0054] Embodiment 1

[0055] In this embodiment, a weeding device for wine grapes without turning around includes: a GPS device, an IMU device, a weeding tool, a differential chassis, and a monocular camera. The GPS device is respectively installed at the front and rear ends of the top center axis of the differential chassis; the IMU device is respectively installed on the center axis of the differential chassis, and the IMU device is closely attached to the GPS device; the weeding tool is respectively installed at the front and rear ends of the differential chassis; and the monocular camera is installed on the lower side of the square steel at the top of the differential chassis.

[0056] In this embodiment, a set of GPS devices is installed on the front and rear of the top center axis of the differential chassis, the IMU is installed on the center axis of the differential chassis and close to any GPS, and the monocular camera is installed on the lower side of the square steel on the top of the differential chassis. During operation, only the GPS signal in the forward direction of the differential chassis is received. When a line change is required, switch to receive another GPS signal.

[0057] The weeding tool equipment is installed at the front and rear ends of the differential chassis, so that the obstacle avoidance directions of the two sets of weeding tool equipment are opposite. During the operation, only one set of weeding tool equipment that meets the obstacle avoidance direction is in working state, and the other set remains in obstacle avoidance state. When the operation of a row of grape vines is completed, the working state of the two sets of weeding tool equipment is swapped to immediately enter the next row of operation without the need for complicated U-turn navigation calculations. This not only avoids technical difficulties but also saves time and improves productivity.

[0058] Embodiment 2

[0059] In this embodiment, a control method of a wine grape weeding device without turning around comprises the following steps:

[0060] S1. Obtain the longitude and latitude information of each row of grapes, take a longitude and latitude coordinate every preset distance, and obtain the longitude and latitude points of each row.

[0061] In this embodiment, a drone is used to scan the longitude and latitude information of each row of grapes before the operation. A longitude and latitude coordinate is taken every 2.5 m to obtain the longitude and latitude points of each row.

[0062] S2. Obtain the heading angle of the differential chassis and calculate the heading error of the differential chassis.

[0063] The method for calculating the heading error includes: obtaining the relative heading angle of the IMU device. Since the absolute heading angle of the IMU device is unstable and cannot meet the requirements of high-precision operation, the relative heading angle of the IMU device is used. The characteristic of the relative heading angle of the IMU device is that it is 0° in any direction. Therefore, before the weeding device starts working, the differential chassis is allowed to go straight for 1m, and the longitude and latitude coordinates of the start and end points are recorded. The longitude and latitude coordinates of the start and end points are inversely solved using Vincenty to obtain the azimuth of the start and end points. The heading angle is calculated based on the relative heading angle and the azimuth of the start and end points:

[0064] θ=θ v -θ i

[0065] Where θ represents the heading angle, θ v Indicates the azimuth of the start and end points, θ i Represents the relative heading angle; before starting navigation at each point, first calculate the difference between the azimuth angle between the differential chassis and the target point and the heading angle of the differential chassis, that is, calculate the heading error based on the heading angle:

[0066] e h =θ f -θ

[0067] Among them, e h represents the heading error, θ f Indicates the first azimuth angle between the differential chassis and the target point. Whenever the heading error angle to the target point is greater than 3 degrees, the differential chassis is controlled to rotate in situ to make the heading error angle less than or equal to 3 degrees.

[0068] S3. Take the line connecting every two longitude and latitude points as a reference path, and discretize the reference path to obtain discretized longitude and latitude points.

[0069] During navigation, the differential chassis is controlled by lateral error in order to maintain straight driving. The lateral error is calculated by GPS information, IMU information and visual information. The longitude and latitude coordinates of every two grape rows are regarded as a path, and the reference path is discretized into a series of longitude and latitude points (φ path (s), λ path (s)), s is the path parameter, representing the sequence number of the longitude and latitude points.

[0070] S4. Calculate the distance between two points in the discretized longitude and latitude points, and calculate the probability distribution of each point based on the heading error.

[0071] The optimal tracking point is selected, but the traditional cost function design is usually a simple weighted sum of several error indicators (such as position error, heading error, environmental complexity, etc.), which has the limitations of linear combination, difficulty in weight selection, lack of high-level structure measurement and other disadvantages. To overcome the above problems, the multi-dimensional indicators of each point are mapped into a probability distribution.

[0072] In this embodiment, the method for calculating the probability distribution of each point includes: using Vincenty to calculate the distance v1(s) from the discretized longitude and latitude point to the differential chassis, recording the heading error as v2(s), and recording the curvature from the differential chassis to the reference path as v3(s); using the sigmoid function to normalize the distance v1(s), the heading error v2(s) and the curvature v3(s) to obtain the feature vector v(s) = [v1(s), v2(s), v3(s)] of each point; and using the Softmax function to convert the feature vector into a probability distribution p(s) = [p1(s), p2(s), p3(s)].

[0073] S5. Use KL divergence and JS divergence to optimize the probability distribution and the ideal reference distribution to obtain the optimal tracking point.

[0074] Simple Euclidean distance or KL divergence (Kullback-Leibler divergence) to measure the difference between distributions will have problems such as asymmetry of KL divergence and numerical instability of KL divergence. In order to overcome the above problems, Jensen-Shannon divergence (JS divergence) is combined as a tool to measure distribution differences. When selecting from many candidate path points, JS divergence can be used to quantify the "goodness" of each point - that is, how close the feature distribution of the point is to the ideal feature distribution designed in advance. The process of selecting path points has a clear quantitative standard, rather than just qualitative or empirical judgment. Moreover, this method of comparing and measuring the actual feature distribution with the ideal distribution using Jensen-Shannon divergence can adapt to the changes in different vineyard environments. In vineyards with flat terrain, the "distance" factor is given priority to make the path shorter; in vineyards with large terrain fluctuations, the proportion of features such as "curvature" or "heading error" is increased. By setting different proportion structures of the ideal distribution, strategies and preferences can be flexibly adjusted without changing the framework of the entire algorithm. Only by changing the target distribution, navigation strategy switching and scene adaptation can be achieved. JS divergence is not only symmetrical, but also always bounded in value. It can also produce more stable results at the position where the distribution is 0, and is not prone to numerical overflow or undefined situations. This robustness is very important in the actual navigation process, and can avoid algorithm crashes caused by sensor data fluctuations or incomplete features. In real-time navigation, the environment and conditions are constantly changing. By continuously calculating the JS divergence of the actual distribution and the ideal distribution, the navigation system can dynamically evaluate the pros and cons of existing path points, and then quickly adjust decisions when necessary. This dynamic adjustment mechanism ensures that navigation decisions always move towards the preset optimization goals. And after converting the feature weighting problem into distribution comparison, we can not only know which point has better features, but also know at what level the goodness is reflected - by looking at the differences between the distributions, we can clearly point out which feature dimensions cause the gap.

[0075] In this embodiment, the method for obtaining the optimal tracking point includes: setting an ideal reference distribution p*(s)=[0.8, 0.1, 0.1], and constructing an intermediate distribution m between the probability distribution and the ideal reference distribution; using KL divergence, constructing KL divergence functions based on the probability distribution, the ideal reference distribution and the intermediate distribution m: KL(p‖m) and KL(p*‖m); using JS divergence, calculating JS divergence function based on KL divergence function:

[0076]

[0077]

[0078]

[0079] Based on the JS divergence function, each discrete path point is calculated JSD(p, p*). The smaller the JSD(p, p*) value is, the closer the point is to the set ideal state. Among all the calculated JSD(p, p*), the point corresponding to the smallest value is found to obtain the optimal tracking point (φ path (s*),λ path (s*)).

[0080] S6. Calculate the distance between the optimal tracking point and the differential chassis, and calculate the navigation lateral error.

[0081] The method for calculating the navigation lateral error includes: using Vincenty to calculate the great circle distance between the optimal tracking point and the differential chassis, the direction of the reference path at the optimal tracking point, and the second azimuth from the reference point to the differential chassis; and calculating the navigation lateral error based on the great circle distance, the direction, and the second azimuth:

[0082] e g =d geo ·sin(α g -θ path (s * ))

[0083] Among them, e g represents the lateral navigation error, d geo represents the great circle distance between the optimal tracking point and the differential chassis, θ path (s) represents the direction of the reference path at the optimal tracking point, α g Indicates the second azimuth angle from the reference point to the differential chassis.

[0084] In this embodiment, the calculation method of the second azimuth angle is:

[0085] α g =arctan2(cos(φ2)sin(Δλ), cos(φ1)sin(φ2)-sin(φ1)cos(φ2)cos(Δλ))

[0086] Among them, the differential chassis position is (φ1, λ1), the reference point position is (φ2, λ2), φ1 and φ2 represent the latitude of the differential chassis position and the reference point position respectively, λ1 and λ2 represent the longitude of the differential chassis position and the reference point position respectively, Δλ represents the difference between the longitudes of the two points, and arctan2 represents a two-parameter inverse tangent function to ensure the correct direction of the azimuth angle.

[0087] S7. Use a monocular camera to obtain the image of the wire pile and obtain the two-dimensional position information, extract the distance between the center point of the wire pile and the vertical center line of the image, and obtain the visual lateral error.

[0088] In this embodiment, a monocular camera uses Hough garden transform to detect the wooden poles of the grape row, obtain the two-dimensional position information of the poles, and extract the distance between the center point of the wooden poles and the vertical center line of the image, that is, the lateral error e based on vision is obtained. s .

[0089] S8. Calculate the chassis lateral error based on the navigation lateral error and the visual lateral error, and use proportional control to set the chassis lateral error to 0, thereby achieving navigation control of the weeding device.

[0090] Methods for calculating chassis lateral error include:

[0091] e=μ g e g +(1-μ g ) s +ε

[0092] Among them, μ g represents GPS confidence, ε represents inherent error, e s Represents the lateral visual error.

[0093] S9. After the weeding device completes a row of work and before changing rows, the working status and forward direction of the two sets of weeding tool equipment are exchanged, the information of the GPS device at the other end is collected, and the row change navigation is performed using the cubic function curve path.

[0094] Specifically: During the line-changing process, the differential chassis has a sudden change in angular acceleration, and the cubic function has continuous function values ​​and derivatives throughout the entire trajectory. This means that the curve is not only smooth (no sudden angle changes), but also the speed and acceleration can be adjusted through derivatives. The line-changing driving path is fitted to a cubic function curve, the coordinate origin is the perpendicular line between the center line of the grape row and the midpoint of the perpendicular line from the differential chassis to the grape row, the distance from the transition point to the X-axis is 1 / 8 of the distance from the differential chassis to the grape row, and the distance from the transition point to the Y-axis is 1 / 4 of the distance between the two grape rows. The starting point and the end point are known, and the cubic function passing through the origin has 3 known coordinate points, and the expression of the cubic function can be determined. The cubic function curve path is divided into several straight paths with a distance of 0.1m, and the chassis control is performed using the method of controlling the GPS lateral deviation, which can complete accurate and smooth line-changing operations.

[0095] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A wine grape weeding device that does not require U-turns, characterized in that: include: GPS equipment, IMU equipment, weeding knives, differential chassis and monocular cameras; The GPS devices are respectively mounted on the front and rear ends of the top center axis of the differential chassis; The IMU devices are respectively installed on the central axis of the differential chassis, and the IMU devices are closely attached to the GPS device; The weeding cutters are respectively mounted at the front and rear ends of the differential chassis; The monocular camera is installed on the lower side of the square steel on the top of the differential chassis.

2. A control method for a wine grape weeding device that does not require U-turns, the control method being used to control the weeding device according to claim 1, characterized in that: The following steps are involved: Get the longitude and latitude information of each row of grapes, take a longitude and latitude coordinate every preset distance, and get the longitude and latitude points of each row; Obtain the heading angle of the differential chassis and calculate the heading error of the differential chassis; Taking the line connecting every two of the longitude and latitude points as a reference path, and discretizing the reference path to obtain discretized longitude and latitude points; Calculating the distance between two points in the discretized longitude and latitude points, and calculating the probability distribution of each point based on the heading error; The probability distribution and the ideal reference distribution are optimized using KL divergence and JS divergence to obtain the optimal tracking point; Calculating the distance between the optimal tracking point and the differential chassis, and calculating the navigation lateral error; Using a monocular camera to obtain a wire pile image and obtain two-dimensional position information, extracting the distance between the center point of the wire pile and the vertical center line of the image, that is, obtaining a visual lateral error; Calculating a chassis lateral error based on the navigation lateral error and the visual lateral error, and proportionally controlling the chassis lateral error to be 0, thereby achieving navigation control of the weeding device; After the weeding device has completed a row and before changing rows, the working status and forward direction of the two sets of weeding tool equipment are exchanged, information from the GPS device at the other end is collected, and row change navigation is performed using a cubic function curve path.

3. The control method of the wine grape weeding device without turning around according to claim 2, characterized in that: The method for calculating the heading error includes: Get the relative heading angle of the IMU device; Before the weeding device starts working, the differential chassis is allowed to move straight for 1 m, the longitude and latitude coordinates of the start and end points are recorded, and the longitude and latitude coordinates of the start and end points are inversely solved using Vincenty to obtain the azimuths of the start and end points; The heading angle is calculated based on the relative heading angle and the start and end point azimuths: θ=θ v -θ i Where θ represents the heading angle, θ v represents the azimuth of the start and end points, θi represents the relative heading angle; Calculate the heading error based on the heading angle: e h =θ f -θ Among them, e h represents the heading error, θ f Indicates the first azimuth angle between the differential chassis and the target point.

4. The control method of the wine grape weeding device without turning around according to claim 2, characterized in that: The method of calculating the probability distribution of each point includes: Calculate the distance v1(s) from the discretized longitude and latitude point to the differential chassis using Vincenty, record the heading error as v2(s), and record the curvature from the differential chassis to the reference path as v3(s); Use the sigmoid function to normalize the distance v1(s), heading error v2(s) and curvature v3(s) to obtain the feature vector v(s) = [v1(s), v2(s), v3(s)] of each point; The Softmax function is used to convert the feature vector into the probability distribution p(s)=[p1(s), p2(s), p3(s)].

5. The control method of the wine grape weeding device without turning around according to claim 4, characterized in that: The method for obtaining the optimal tracking point includes: Setting the ideal reference distribution p*(s) and constructing an intermediate distribution m between the probability distribution and the ideal reference distribution; Using KL divergence, constructing KL divergence functions based on the probability distribution, the ideal reference distribution and the intermediate distribution m: KL(p||m) and KL(p*||m); Using JS divergence, the JS divergence function is calculated based on the KL divergence function: Based on the JS divergence function, the existing reference path is dynamically evaluated to obtain the optimal tracking point (φ path (s*),λ path (s*)).

6. The control method of the wine grape weeding device without turning around according to claim 2, characterized in that: The method for calculating the navigation lateral error includes: Calculate the great circle distance between the optimal tracking point and the differential chassis, the orientation of the reference path at the optimal tracking point, and the second azimuth angle from the reference point to the differential chassis using Vincenty; Calculate the navigation lateral error based on the great circle distance, the orientation and the second azimuth: e g =d geo ·sin(a g -θ path (s * )) Among them, e g represents the lateral navigation error, d geo represents the great circle distance between the optimal tracking point and the differential chassis, θ path (s * ) represents the direction of the reference path at the optimal tracking point, α g Indicates the second azimuth angle from the reference point to the differential chassis.

7. The control method of the wine grape weeding device without turning around according to claim 6, characterized in that: The method for calculating the chassis lateral error includes: e=μ g e g +(1-m g )e s +e Among them, μ g represents GPS confidence, ε represents inherent error, e s Represents the lateral visual error.

Images