Method to set a driving path for an autonomous mobile robot in an orchard cultivation environment

Abstract

The present invention provides a method for setting a driving path of a mobile robot that enables accurate object recognition of fruit trees present in a fruit tree cultivation environment through the matching of different image data acquired from a camera and a lidar, and enables the rapid and accurate setting of the driving posture and path of the mobile robot in the object-recognized fruit tree cultivation environment. The method for setting a driving path of the mobile robot comprises an image data acquisition step, a point cloud data acquisition step, a point of interest extraction step, and a mobile robot posture setting step. The image data acquisition step acquires image data regarding the fruit tree cultivation environment through a camera equipped on the mobile robot. The point cloud data acquisition step acquires point cloud data regarding the fruit tree cultivation environment through a lidar equipped on the mobile robot. The point of interest extraction step extracts points of interest of fruit trees present in the fruit tree cultivation environment by matching the image data and the point cloud data. The mobile robot posture setting step sets the driving position and orientation angle of the mobile robot based on the positional deviation between the extracted points of interest and the pre-stored target points of the fruit trees.

Description

Technology Field

[0001] The present invention relates to a method for setting a driving path of a mobile robot, and more specifically, to a method that enables accurate object recognition of fruit trees present in a fruit tree cultivation environment through the matching of data acquired from a camera and a LiDAR, respectively, and enables the driving posture and path of a mobile robot driving autonomously in an object-recognized fruit tree cultivation environment to be set quickly and accurately.

[0002] The present invention corresponds to the results created during the business execution process described below.

[0003] Project Name: Daejeon·Sejong·Chungnam Regional Innovation Platform Mobility ICT Project (4th Year Unit Project)

[0004] Project No.: 2024D020607007

[0005] Project Title: Development of Localization and Control Algorithms for Autonomous Orchard Robots

[0006] Supervising Agency (Specialized Institution): National Research Foundation of Korea

[0007] Research Institute: Chungnam National University

[0008] Research Period: June 1, 2024 – February 28, 2025 Background Technology

[0009] Recently, in fruit cultivation environments where large cultivation areas and large quantities of fruit trees are grown, autonomous driving devices are attracting attention in order to maximize work efficiency while reducing manpower.

[0010] Robot technology for autonomous driving in fruit cultivation environments, like autonomous vehicles in urban areas, largely requires perception technology for recognizing the surrounding environment, discrimination technology for planning a driving path from the recognized surrounding environment, and control technology for driving along the planned path. Among these, perception technology for recognizing the surrounding environment is the first technology performed for autonomous driving, and only when this perception technology is accurately preceded can the next steps, discrimination and control technology, also be accurately implemented.

[0011] Perception technologies include techniques for determining the precise location of an autonomous driving device using GPS, and techniques for recognizing the surrounding environment based on information acquired through multiple sensors, such as cameras and Inertial Measurement Units (IMUs) like LiDAR (Light Detection and Ranging).

[0012] Here, GPS only provides accuracy with a level of error of about meters, and since there are instances where signals are not received momentarily in areas surrounded by high-rise buildings or underground tunnels, it is difficult to recognize the location using GPS alone in fruit cultivation environments where there are many obstacles such as crops.

[0013] Therefore, autonomous driving devices in fruit cultivation environments mainly use a method of recognizing the surrounding environment using cameras and lidar.

[0014] Cameras have the advantage of providing intuitive and clear image information about the surrounding environment, and LiDAR has the advantage of providing high-resolution 3D image information using laser pulses. These cameras and LiDAR are being used in a way that complements each other's strengths and weaknesses through sensor fusion.

[0015] In addition, recently, there has been a trend of using deep learning and artificial intelligence in the process of recognizing the surrounding environment and planning paths to provide information required for autonomous driving more quickly and accurately. Prior art literature

[0016] Republic of Korea Published Patent Application No. 10-2024-0032191 (Published on March 12, 2024) The problem to be solved

[0017] The objective of the present invention to solve the aforementioned problems is to provide a method for setting a driving path of a mobile robot that enables accurate object recognition of a fruit tree existing in a fruit tree cultivation environment through matching different data obtained from a camera that captures images in the visible light region and a lidar that uses laser pulses, and enables fast and accurate setting of the driving posture and path of a mobile robot that drives autonomously in an object-recognized fruit tree cultivation environment.

[0018] The problems that the present invention aims to solve are not limited to those mentioned above, and other unmentioned problems will be clearly understood by those skilled in the art to which the present invention belongs from the description below. means of solving the problem

[0019] A method for setting a driving path of a mobile robot autonomously driving in a fruit tree cultivation environment according to an embodiment of the present invention for solving the above-mentioned problem comprises: an image data acquisition step of acquiring image data of a fruit tree cultivation environment through a camera equipped in the mobile robot; a point cloud data acquisition step of acquiring point cloud data of a fruit tree cultivation environment through a LiDAR equipped in the mobile robot; a point of interest extraction step of extracting a point of interest of a fruit tree existing in the fruit tree cultivation environment by matching the image data and the point cloud data; and a mobile robot attitude setting step of setting the driving position and orientation angle of the mobile robot based on the position deviation between the extracted point of interest and a pre-stored target point of the fruit tree.

[0020] In a method for setting a driving path of a mobile robot autonomously driving in a fruit tree cultivation environment according to an embodiment of the present invention, the interest point extraction step comprises an image bounding box extraction step for extracting an image bounding box corresponding to the interest point from the image data, and a point cloud bounding box extraction step for projecting the point cloud data onto the image data to extract a point cloud bounding box that matches the image bounding box, and the center point of the point cloud bounding box can be set as the interest point.

[0021] In a method for setting a driving path of a mobile robot driving autonomously in a fruit cultivation environment according to an embodiment of the present invention, the image bounding box extraction step can extract the image bounding box from the image data using a deep learning-based object detection algorithm.

[0022] In a method for setting a driving path for a mobile robot autonomously driving in a fruit cultivation environment according to an embodiment of the present invention, the point of interest extraction step further includes an external parameter calculation step for calculating an external parameter to synchronize a reference point of the camera coordinate system of the image data and the LiDAR coordinate system of the point cloud data, and the point cloud bounding box extraction step aligns the LiDAR coordinate system and the camera coordinate system by reflecting the calculated external parameter, and can project the point cloud data with aligned coordinate systems onto the image data.

[0023] In a method for setting a driving path of a mobile robot autonomously driving in a fruit tree cultivation environment according to an embodiment of the present invention, the point cloud bounding box extraction step may include an effective point cloud bounding box extraction step for each of the extracted point cloud bounding boxes, wherein an outlier region having a relatively low occupancy rate of point cloud data is removed to extract an effective point cloud bounding box having an area smaller than the area of ​​the image bounding box.

[0024] In a method for setting a driving path for a mobile robot driving autonomously in a fruit cultivation environment according to an embodiment of the present invention, the point cloud bounding box extraction step may further include a valid point cloud bounding box noise removal step, wherein a linear regression model is generated using training data based on the area of ​​the extracted valid point cloud bounding box, and a point cloud bounding box whose result value output through the generated linear regression model is greater than a preset reference value is identified as an outlier and removed.

[0025] In a method for setting a driving path of a mobile robot autonomously driving in a fruit cultivation environment according to an embodiment of the present invention, the mobile robot attitude setting step may include: a comparison plane acquisition step of acquiring a comparison plane in which the point of interest exists; a plane alignment step of moving the comparison plane in which the point of interest exists in a straight line or rotating it with respect to a reference plane in which the target point exists, so that the point of interest and the target point coincide; and a mobile robot control step of moving the mobile robot in a straight line corresponding to the straight movement of the comparison plane to set a driving position, and rotating the mobile robot in a rotational movement corresponding to the rotational movement of the comparison plane to set a direction angle.

[0026] In a method for setting a driving path for a mobile robot driving autonomously in a fruit cultivation environment according to an embodiment of the present invention, in the step of obtaining a comparison plane, the comparison plane may be a plane in which a point of interest exists that is randomly searched using a RANSAC (random sample consensus) algorithm. Effects of the invention

[0027] According to the present invention, accurate object recognition of fruit trees present in a fruit tree cultivation environment is possible through the matching of different data obtained from a camera and a lidar, and from this, the driving posture and path of a mobile robot driving autonomously in a fruit tree cultivation environment can be set quickly and accurately.

[0028] The effects of the present invention are not limited to the effects described above, and should be understood to include all effects that can be inferred from the configuration of the invention described in the detailed description or claims of the present invention. Brief explanation of the drawing

[0029] FIG. 1 is a planar example diagram showing when a mobile robot has a normal driving posture (a) and when it has an abnormal driving posture (b) in a fruit tree cultivation environment according to an embodiment of the present invention. FIG. 2 is a block flowchart illustrating a method for setting a driving path of a mobile robot according to an embodiment of the present invention. Figure 3 is a block flowchart showing the point of interest extraction step of Figure 2. Figure 4 is a block flowchart showing the point cloud bounding box extraction step of Figure 3. Figure 5 is an example diagram showing the image bounding box extraction step. Figure 6 is an example diagram showing the point cloud bounding box extraction step. Figure 7 is an example diagram showing the effective point cloud bounding box noise removal step. Figure 8 is a block flowchart showing the attitude setting step of the mobile robot of Figure 2. Figure 9 is an example diagram illustrating the posture setting steps of a mobile robot. Specific details for implementing the invention

[0030] The present invention will be described below with reference to the attached drawings. Of course, the present invention may be implemented in various different forms not illustrated and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification have been given similar reference numerals.

[0031] Throughout the specification, when it is stated that a part is "connected (connected, in contact, combined)" with another part, this includes not only cases where they are "directly connected," but also cases where they are "indirectly connected" with other members interposed between them. Furthermore, when it is stated that a part "includes" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but rather allows for the inclusion of additional components.

[0032] The terms used in this specification are used merely to describe specific embodiments and are not intended to limit the invention. Additionally, singular expressions include plural expressions unless the context clearly indicates otherwise. Furthermore, terms such as “comprising,” “having,” and “having” in this specification are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0033] Embodiments of the present invention will be described in detail below with reference to the attached drawings.

[0034] FIG. 1 is a planar example diagram showing when a mobile robot has a normal driving posture (a) and when it has an abnormal driving posture (b) in a fruit tree cultivation environment according to an embodiment of the present invention.

[0035] Referring to FIG. 1, the present invention can provide a driving posture and path of a mobile robot (10) that drives autonomously in a fruit tree cultivation environment.

[0036] The fruit tree cultivation environment can basically have multiple fruit trees (1) spaced apart in the longitudinal and transverse directions.

[0037] In this fruit cultivation environment, the autonomous mobile robot (10) can detect the trunk of the fruit tree (1), which is the target point (TP).

[0038] And, as shown in FIG. 1 (a), the mobile robot (10) can be configured to drive autonomously along a driving line (DL) that coincides with the center line of a groove extending in the longitudinal direction of the fruit cultivation environment.

[0039] For example, as shown in (b) of FIG. 1, if the mobile robot (10) is positioned apart from the target driving line (DL), the mobile robot (10) can be positioned correctly on the target driving line (DL) based on detection data obtained from the camera and lidar. In addition, if the aiming angle (10A) of the mobile robot (10) is not parallel to the driving line (DL) and is twisted so as not to maintain the target aiming angle, the aiming angle (10A) of the mobile robot (10) can be rotated to be parallel to the driving line (DL) based on detection data obtained from the camera and lidar so that the mobile robot (10) looks at the target aiming angle.

[0040] Meanwhile, map information regarding the fruit tree cultivation environment may be stored in advance in the mobile robot (10), and such map information may include coordinate information of target points (TP) arranged in the longitudinal and transverse directions on a reference plane viewed from above regarding the fruit tree cultivation environment, as in (a) of FIG. 1.

[0041] FIG. 2 is a block flowchart showing a method for setting a driving path of a mobile robot according to an embodiment of the present invention, FIG. 3 is a block flowchart showing the point of interest extraction step of FIG. 2, and FIG. 4 is a block flowchart showing the point cloud bounding box extraction step of FIG. 3.

[0042] Referring to FIGS. 1 to 4, the method for setting a driving path of a mobile robot autonomously driving in a fruit tree cultivation environment according to the present embodiment includes an image data acquisition step (S100), a point cloud data acquisition step (S200), a point of interest extraction step (S300), and a mobile robot attitude setting step (S400).

[0043] The image data acquisition step (S100) is a step of acquiring image data of a fruit tree cultivation environment through a camera equipped on a mobile robot (10).

[0044] The point cloud data acquisition step (S200) is a step of acquiring point cloud data about the fruit cultivation environment through a lidar equipped on the mobile robot (10).

[0045] The point of interest extraction step (S300) is a step of extracting points of interest (IP: see FIG. 9) of fruit trees by matching image data and point cloud data.

[0046] The point of interest (IP) may be a pillar of the fruit tree (1) and may be a center point of the pillar of the fruit tree (1).

[0047] The point of interest extraction step (S300) may include an image bounding box extraction step (S310) and a point cloud bounding box extraction step (S330).

[0048] The image bounding box extraction step (S310) is a step of extracting an image bounding box corresponding to a point of interest (IP) from image data acquired from a camera.

[0049] FIG. 5 is an example diagram illustrating an image bounding box extraction step. As shown in FIG. 5, in the image bounding box extraction step (S310), a deep learning-based object detection algorithm can be used in conjunction with a known image processing program to extract an image bounding box in the shape of a square box corresponding to a point of interest (IP), which is a pillar area of ​​a fruit tree, from 2D image data.

[0050] The point cloud bounding box extraction step (S330) is a step of projecting point cloud data acquired from the lidar onto image data acquired from the camera to extract a point cloud bounding box that matches the image bounding box.

[0051] Figure 6 is an example diagram showing the point cloud bounding box extraction step. The LiDAR emits laser pulses around the mobile robot and detects light reflected back from the surrounding fruit cultivation environment of the mobile robot, thereby obtaining data in the form of a point cloud, in which a large number of points are clustered, as 3D information about the surrounding fruit cultivation environment.

[0052] Unlike image data, which contains 2D coordinate information corresponding to pixels, point cloud data can contain not only 3D coordinate information but also intensity information. For example, in point cloud data, objects closer to the viewer have a higher number of points, are denser, and have less noise due to strong reflection signals. Conversely, objects further away have fewer points, are looser, and have more noise due to weaker reflection signals.

[0053] As shown in Fig. 6, the point cloud data obtained in this way can be projected onto image data to extract a point cloud bounding box (green) that matches the image bounding box (red). Based on the extracted point cloud bounding box, as well as intensity information such as the size (area) of the box and the occupancy, density, and sharpness of the points inside the box, XYZ coordinate information in 3D space can be verified.

[0054] The point of interest extraction step (S300) may further include an extrinsic parameter calculation step (S320).

[0055] The external parameter calculation step (S320) may be performed prior to the point cloud bounding box extraction step (S330). The external parameter calculation step (S320) is a step for calculating external parameters to synchronize the reference point of the camera coordinate system of the image data and the LiDAR coordinate system of the point cloud data.

[0056] Since the camera and lidar equipped on the mobile robot (10) have different mounting positions, it is necessary to synchronize the sensing reference points of the camera and lidar, and to calculate an external parameter corresponding to the deviation of the sensing reference point. That is, the external parameter corresponds to a parameter representing the correlation between the 2D coordinates of the camera and the 3D coordinates of the lidar.

[0057] As such, the external parameter calculation step (S320) is a process for synchronizing the camera coordinate system and the LiDAR coordinate system, and can be described as a calibration process that unifies the different coordinate systems of image data and point cloud data. In other words, the external parameter may correspond to the deviation of a specific point element existing in the image data based on the camera coordinate system and a specific point element existing in the point cloud data based on the LiDAR coordinate system. Consequently, the external parameter can be obtained by calculating a transformation matrix (translation / rotation matrix) to align the specific point element existing in the image data based on the camera coordinate system with the specific point element existing in the point cloud data based on the LiDAR coordinate system, and the camera coordinate system and the LiDAR coordinate system can be aligned through a calibration process that reflects the external parameter obtained in this way.

[0058] In addition, in the alignment process that aligns the camera coordinate system with the lidar coordinate system, not only the extrinsic parameters between the camera and lidar described above but also the intrinsic parameters of the camera may be required. Intrinsic parameters are used to correct deviations caused by internal camera elements, such as the principal point or focal length of the lens. When multiple cameras are provided, unique intrinsic parameters for each camera can be calculated, and image data acquired from each camera can be synchronized based on these parameters. However, when a single monocular camera is used, the process of calculating the intrinsic parameters mentioned above may be excluded.

[0059] Consequently, when a single monocular camera and lidar are provided, the camera coordinate system and lidar coordinate system are aligned through a matching process that reflects external parameters between the camera and lidar, and in this state, by projecting image data onto point cloud data or projecting point cloud data onto image data, point cloud bounding boxes that exactly match a large number of points of interest (IP) existing in the fruit cultivation environment can be effectively extracted.

[0060] Meanwhile, the point cloud bounding box extraction step (S330) may include an effective point cloud bounding box extraction step (S331) and an effective point cloud bounding box noise removal step (S332).

[0061] The valid point cloud bounding box extraction step (S331) may be a step of extracting a valid point cloud bounding box (green box in FIG. 6) having an area smaller than the area of ​​the image bounding box (red box in FIG. 6) by removing outlier areas with a relatively low occupancy rate (density, quantity) of the point cloud data for each of the previously extracted point cloud bounding boxes.

[0062] Figure 7 is an example diagram illustrating a valid point cloud bounding box noise removal step. Referring to Figure 7, the valid point cloud bounding box noise removal step (S332) may be a step of generating a linear regression model using training data based on the area of ​​the extracted valid point cloud bounding box, and identifying and removing the corresponding point cloud bounding box as an outlier when the result value output through the generated linear regression model is greater than a preset reference value.

[0063] After undergoing this valid point cloud bounding box noise removal step (S332), the valid point cloud bounding box that remains can be set as the point of interest (IP).

[0064] FIG. 8 is a block flowchart showing the attitude setting step of the mobile robot of FIG. 2, and FIG. 9 is an example diagram explaining the attitude setting step of the mobile robot. FIG. 9 (a) is a diagram showing a reference plane having a target point, FIG. 9 (b) is a diagram showing a comparison plane having a point of interest, and FIG. 9 (c) is a diagram showing the process of aligning the comparison plane with respect to the reference plane.

[0065] Referring to FIGS. 1, FIGS. 2, FIGS. 8 and FIGS. 9, the mobile robot posture setting step (S400) is a step of setting the driving position and orientation angle (10A) of the mobile robot (10) based on the position deviation between the extracted point of interest (IP) and the target point (TP) of the fruit (1) that was stored in advance.

[0066] The mobile robot attitude setting step (S400) may include a comparison plane acquisition step (S410), a plane alignment step (S420), and a mobile robot control step (S430).

[0067] As shown in Fig. 9(b), the comparison plane acquisition step (S410) may be a step of acquiring a comparison plane (CP) in which a point of interest (IP) corresponding to an effective point cloud bounding box exists.

[0068] Here, the comparison plane (CP) may correspond to a plane containing points of interest (IP) corresponding to valid point cloud bounding boxes randomly searched using the RANSAC (random sample consensus) algorithm. When the comparison plane (CP) is obtained using this RANSAC algorithm, consistency can be improved during the subsequent alignment process between data points using the ICP (iterative closest point) algorithm.

[0069] As shown in (c) of FIG. 9, the plane alignment step (S420) may be a step of aligning a reference plane (RP) where a target point (TP) exists and a comparison plane (CP) where a point of interest (IP) exists.

[0070] That is, the plane alignment step (S420) can linearly or rotationally move the comparison plane (CP) where the point of interest (IP) exists relative to the reference plane (RP) where the target point (TP) exists, so that the point of interest (IP) and the target point (TP) coincide.

[0071] In this plane alignment step (S420), an iterative closest point (ICP) algorithm may be used. That is, the ICP algorithm repeatedly performs the process of finding the matching point and calculating the transformation matrix (translation / rotation matrix) so that the target point (TP) and the point of interest (IP) are optimally matched and aligned. Through the execution of this ICP algorithm, the alignment of the target point (TP) and the point of interest (IP) can be significantly improved.

[0072] The mobile robot control step (S430) can adjust and set the driving position of the mobile robot (10) so as to align with the driving line (DL) by moving the mobile robot (10) in a straight line corresponding to the straight movement of the comparison plane (CP). Additionally, the mobile robot control step (S430) can adjust and set the orientation angle (10A) of the mobile robot (10) by rotating the mobile robot (10) in correspondence with the rotational movement of the comparison plane (CP).

[0073] That is, as shown in (b) of FIG. 1, if the mobile robot (10) is positioned apart from the target driving line (DL), the position of the mobile robot (10) can be moved in a straight line (d) to be correctly positioned on the target driving line (DL). In addition, if the orientation angle (10A) of the mobile robot (10) is not parallel to the driving line (DL) and is positioned at an angle, the orientation angle (10A) of the mobile robot (10) can be rotated (a) parallel to the driving line (DL) to be directed toward the target orientation angle.

[0074] As a result, as shown in FIG. 1 (a), the mobile robot (10) is able to achieve continuous autonomous driving while maintaining a uniform spacing and orientation angle for the fruit trees (1) arranged at intervals in the longitudinal and transverse directions in a cultivation environment.

[0075] The method described above is depicted as a series of numbered steps for clarity, but the numbering does not specify the order of each step. Some of these steps may be skipped, performed in parallel, or executed without strictly maintaining a sequential order. However, generally, the above method proceeds according to the numbered order of the indicated steps.

[0076] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0077] The scope of the present invention is defined by the claims set forth below, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present invention. Explanation of the symbols

[0078] 10: Mobile robot 10A: Directional angle DL: Driving line TP: Target Point IP: Point of Interest

Claims

Claim 1 A method for setting a driving path of a mobile robot autonomously driving in a fruit tree cultivation environment, characterized by comprising: an image data acquisition step of acquiring image data of a fruit tree cultivation environment through a camera equipped on a mobile robot; a point cloud data acquisition step of acquiring point cloud data of the fruit tree cultivation environment through a LiDAR equipped on the mobile robot; a point of interest extraction step of extracting a point of interest of a fruit tree existing in the fruit tree cultivation environment by matching the image data and the point cloud data; and a mobile robot attitude setting step of setting the driving position and orientation angle of the mobile robot based on the position deviation between the extracted point of interest and a pre-stored target point of the fruit tree. Claim 2 A method for setting a driving path of a mobile robot autonomously driving in a fruit tree cultivation environment, wherein, in claim 1, the interest point extraction step comprises an image bounding box extraction step for extracting an image bounding box corresponding to the interest point from the image data, and a point cloud bounding box extraction step for projecting the point cloud data onto the image data to extract a point cloud bounding box that matches the image bounding box, and the center point of the point cloud bounding box is set as the interest point. Claim 3 A method for setting a driving path of a mobile robot autonomously driving in a fruit cultivation environment, wherein, in paragraph 2, the image bounding box extraction step is characterized by extracting the image bounding box from the image data using a deep learning-based object detection algorithm. Claim 4 A method for setting a driving path of a mobile robot autonomously driving in a fruit cultivation environment, wherein, in paragraph 2, the interest point extraction step further includes an external parameter calculation step for calculating an external parameter to synchronize a reference point of the camera coordinate system of the image data and the LiDAR coordinate system of the point cloud data, and the point cloud bounding box extraction step aligns the LiDAR coordinate system and the camera coordinate system by reflecting the calculated external parameter, and projects the point cloud data with aligned coordinate systems onto the image data. Claim 5 A method for setting a driving path of a mobile robot autonomously driving in a fruit tree cultivation environment, wherein, in paragraph 2, the point cloud bounding box extraction step comprises a valid point cloud bounding box extraction step for each of the extracted point cloud bounding boxes, wherein an outlier region having a relatively low occupancy rate of point cloud data is removed to extract a valid point cloud bounding box having an area smaller than the area of ​​the image bounding box. Claim 6 A method for setting a driving path of a mobile robot autonomously driving in a fruit tree cultivation environment, wherein, in claim 5, the point cloud bounding box extraction step further includes a valid point cloud bounding box noise removal step, wherein a linear regression model is generated using training data based on the area of ​​the extracted valid point cloud bounding box, and point cloud bounding boxes whose output value through the generated linear regression model is greater than a preset reference value are identified as outliers and removed. Claim 7 A method for setting a driving path of a mobile robot autonomously driving in a fruit tree cultivation environment, wherein, in claim 1, the mobile robot attitude setting step comprises: a comparison plane acquisition step for acquiring a comparison plane in which the point of interest exists; a plane alignment step for linearly moving or rotating the comparison plane in which the point of interest exists relative to a reference plane in which the target point exists, so that the point of interest and the target point coincide; and a mobile robot control step for setting a driving position by linearly moving the mobile robot in correspondence with the linear movement of the comparison plane, and setting a direction angle by rotating the mobile robot in correspondence with the rotational movement of the comparison plane. Claim 8 A method for setting a driving path for a mobile robot autonomously driving in a fruit cultivation environment, characterized in that, in the step of obtaining the comparison plane, the comparison plane is a plane in which a point of interest exists that was randomly searched using the RANSAC (random sample consensus) algorithm.

Images