Automatic Loading Method for Pallet Forklift AGV Based on QR Code-Assisted Positioning

Abstract

The present invention discloses a method for automatic loading of a pallet forklift AGV based on QR code assisted positioning. This method uses a camera to obtain images of storage pallets with QR code tags, and detects the QR code tags in the images. Based on the PnP method, the pose of the tag center in the camera coordinate system is solved, and the pose of the reference point of the storage pallet in the camera coordinate system is obtained through coordinate transformation. According to the local pose of the storage pallet and the global pose of the pallet forklift AGV, the global pose of the storage pallet is calculated. Further, the picking path from the current point of the pallet forklift AGV to the reference point of the storage pallet is obtained. The pallet forklift AGV is controlled to track the picking path, realizing the full-automatic and precise loading of the pallet forklift AGV on the storage pallet. The present invention does not require specifying the pallet color and size. Compared with the traditional technology, it has the characteristics of high pose estimation accuracy and good path tracking effect, significantly improving the success rate of pallet automatic loading, and effectively improving the handling efficiency on the premise of ensuring the safety and reliability of automatic handling.

Description

Technical Field

[0001] The present invention relates to the technical fields of autonomous navigation control of mobile robots and warehousing pallet detection, and particularly relates to an automatic loading method for a pallet forklift AGV based on QR code assisted positioning. Background Art

[0002] The development of the Internet economy has promoted the continuous transformation and upgrading of traditional logistics distribution models. How to improve logistics efficiency and reduce logistics costs has become a focus issue. Since the concept of intelligent logistics was first comprehensively proposed, logistics has been increasingly automated and intelligent in multiple links, and some highly repetitive and labor-intensive tasks have gradually been completed by intelligent robots. At the same time, the intensification of population aging has made highly automated intelligent warehousing an urgent need. The automatic handling of autonomous mobile robots represented by pallet forklift AGVs has been increasingly widely used in the logistics industry, and intelligent logistics has become the industry's vane. Intelligent logistics can improve efficiency and reduce error rates in links such as loading, handling, and sorting, and improve the level of social productivity.

[0003] In practical applications, the logistics warehousing environment is complex and changeable, with influencing factors such as sparse static feature objects, uneven light intensity, and cumulative global navigation errors. The pallet forklift AGV relies on its autonomous positioning and navigation ability to reach near the warehouse location to prepare for picking up goods. Due to the lack of the ability to automatically identify and locate the loading target, the pallet forklift AGV cannot obtain the local pose information of the warehousing pallet, and problems such as easy collision, insufficient fork picking, and even damage to goods are likely to occur during loading, posing a great safety hazard. To prevent accidents, manual supervision and adjustment are required. Although the manual burden is reduced to a certain extent, the characteristics of high handling efficiency and large load capacity of the pallet forklift AGV cannot be maximally utilized. It can be seen that the warehousing pallet detection and adaptive pose stabilization function are indispensable key technologies for the pallet forklift AGV and are crucial links for realizing unmanned operation in logistics warehousing.

[0004] Studying the warehousing pallet detection and pose estimation algorithm, as well as the adaptive pose stabilization algorithm, is of great significance for improving the intelligent level of the pallet forklift AGV, reducing the error rate and improving efficiency in links such as logistics loading, handling, and sorting, and ultimately realizing full-automatic handling. It is also one of the core issues in the research of intelligent logistics. Summary of the Invention

[0005] The object of the present invention is to overcome various difficulties in realizing automatic forklift handling and the deficiencies of existing warehousing pallet detection methods, and to provide a method for automatic loading of a pallet forklift AGV based on QR code-assisted positioning. This method involves warehousing pallet detection and adaptive pose stabilization of a mobile robot: First, detect the QR code label on the pallet; then, based on the PnP method, calculate the pose of the label center to estimate the pose of the pallet, and combine the global pose of the pallet forklift AGV to calculate and update the pallet state. For the automatic loading part of the AGV, first use a local path planning algorithm to formulate an optimal picking path and smooth it using a Bezier curve; then propose a core navigation control algorithm for path tracking; finally, implement the algorithm on the pallet forklift AGV experimental platform.

[0006] The object of the present invention is achieved by the following technical solutions: A method for automatic loading of a pallet forklift AGV based on QR code-assisted positioning, the method comprising the following steps:

[0007] Step 1: Use a camera installed on the pallet forklift AGV to obtain the original RGB image of the warehousing pallet provided with QR code labels.

[0008] Step 2: Detect the QR code label area in the original RGB image of the warehousing pallet. If the QR code label area is not detected, continue to obtain the next frame of image.

[0009] Step 3: Based on the PnP method, calculate the six-degree-of-freedom pose of the QR code label center in the camera coordinate system, and obtain the pose of the warehousing pallet reference point in the camera coordinate system through coordinate transformation to achieve the local pose estimation of the warehousing pallet.

[0010] Step 4: Calculate the global pose of the warehousing pallet according to the local pose of the warehousing pallet obtained in Step 3 and the global pose of the pallet forklift AGV.

[0011] Step 5: Obtain the picking path from the current point of the pallet forklift AGV to the warehousing pallet reference point according to the current global pose of the warehousing pallet and the current global pose of the pallet forklift AGV.

[0012] Step 6: Control the pallet forklift AGV to track the picking path to achieve the full-automatic and precise loading of the pallet forklift AGV to the warehousing pallet.

[0013] Further, the QR code label is arranged on the outer side of the central column on the picking side of the warehousing pallet, and the point symmetric to the QR code label center on the warehousing pallet is used as the warehousing pallet reference point.

[0014] Further, in Step 2, for the original RGB image of the warehousing pallet, detect the QR code label arranged on the warehousing pallet through image grayscale conversion, adaptive threshold processing, continuous boundary segmentation, quadrilateral fitting, and decoding and matching.

[0015] Further, the continuous boundary segmentation is specifically as follows: segment the edge based on the black and white pixel information of the binary image obtained by adaptive threshold processing; use the union-find algorithm to segment the connected pixel clusters of light and dark pixels, and represent each pixel cluster with a unique ID.

[0016] Further, the quadrilateral fitting is specifically as follows: first find a small number of corner points, and then traverse all combinations of the corner points to calculate the approximate grouping, and output one or more groups of candidate quadrilaterals;

[0017] The decoding matching is specifically as follows: perform an exclusive OR comparison between the code values in the four directions included in the quadrilateral and each label code in the label cluster to filter out incorrect candidate quadrilaterals.

[0018] Further, step 3 is specifically as follows:

[0019] 3-1. For the target QR code label detected in step 2, combine the camera internal parameter matrix, the physical size of the QR code label, and the camera homography matrix to solve the six-degree-of-freedom pose information of the center of the QR code label in the camera coordinate system, that is, the coordinates in three directions in the camera coordinate system: X c 、Z c 、Y c and the Euler angles: yaw angle pitch angle and roll angle;

[0020] 3-2. Represent the pose of the center point of the QR code label in the camera coordinate system as O: After coordinate transformation, obtain the pose P of the reference point of the storage tray in the camera coordinate system: There are horizontal and vertical deviations between its reference point and the label center. Assuming the side length of the tray is a, the pose of the tray reference point is represented as:

[0021]

[0022] Further, in step 4, calculating the global pose of the storage tray is specifically as follows:

[0023] There is a pose deviation between the camera optical center and the reference point of the tray forklift AGV. Assuming the longitudinal deviation is x c 、the horizontal deviation is y c 、the angular deviation is θ; from the camera coordinate system to the forklift AGV coordinate system, the rotation matrix R is represented as:

[0024]

[0025] Let the pose of the tray forklift AGV in the world coordinate system be (X w , Y w , α), and the pose of the storage tray in the camera coordinate system be The pose calculation formula of the storage pallet in the world coordinate system is as follows:

[0026]

[0027] Where (X wp , Y wp ) is the global coordinate of the storage pallet in the world coordinate system, and φ is the global yaw angle of the storage pallet in the world coordinate system; the state of the reference point of the storage pallet in the global coordinate system is expressed as: (X wp , Y wp , φ).

[0028] Furthermore, in step 5, according to the current global pose of the storage pallet and the current global pose of the pallet forklift AGV, the optimal picking path from the current point of the pallet forklift AGV to the reference point of the storage pallet is obtained and smoothed using a Bezier curve. The expression of the picking path F(t) is:

[0029] F(t) = (1 - t) 3 P 1 + 3t(1 - t) 2 P 2 + 3t 2 (1 - t)P 3 + t 3 P 4 t ∈ [0, 1]

[0030] The above is the parametric equation of a third-order Bezier curve, and the proportional coefficient t ranges from 0 to 1; P 1 is the starting point, that is, the current pose of the pallet forklift AGV, P 4 is the ending point, that is, the reference point of the storage pallet, P 2 and P 3 are control points, which are set artificially; by artificially selecting two control points for path fitting, the picking path is smoother after fitting.

[0031] Furthermore, the control points P 2 and P 3 are respectively set as the midpoint of the line connecting the current world coordinate of the pallet forklift AGV to the world coordinate of the center of the QR code label, and the midpoint of the line connecting the world coordinate of the center of the QR code label to the reference point of the storage pallet.

[0032] Furthermore, in step 6, by calculating the angular velocity and speed and sending them to the forklift chassis to control its movement, the pallet forklift AGV is made to track the picking path. Specifically:

[0033] 6-1. The angular velocity calculation formula is expressed as:

[0034] ω = K 1 Δy + K 2 β + K3 γ

[0035] where Δy is the distance from the current pose point to the tangent of the target point on the picking path, β represents the yaw angle of the current pose of the forklift, and γ represents the angle deviation between the current pose point of the forklift and the target point; the input parameters K 1 ,K 2 ,K 3 are fixed parameters, and the selection of the three parameters will affect the path tracking effect. The output result is the angular velocity ω;

[0036] 6-2. The calculation of the linear velocity is specifically as follows:

[0037] Calculate the distance d from the current point to the target point:

[0038]

[0039] where a is the acceleration, v cur and v min are the current speed and the speed lower limit respectively; assume that the projected distance from the current pose point to the target point is Δx;

[0040] If d is less than or equal to Δx, the linear velocity calculation formula is:

[0041]

[0042] If d is greater than Δx, the linear velocity calculation formula is:

[0043] v = v + aT

[0044] Here, v has a maximum speed limit v max , that is, v ≤ v max , where T is the control period.

[0045] The beneficial effects of the present invention are as follows: An automatic loading method for a pallet forklift AGV based on QR code assisted positioning provided by the present invention is divided into two parts: First, detect the QR code label on the pallet through steps such as threshold processing, continuous boundary segmentation, quadrilateral fitting, and decoding matching; then solve the pose of the label center based on the PnP method to realize the pose estimation of the pallet, and calculate and update the pallet state in combination with the global pose of the pallet forklift AGV. The AGV automatic loading part first uses a local path planning algorithm to formulate an optimal picking path and uses a Bezier curve for smoothing; then proposes a core navigation control algorithm for path tracking; finally, the algorithm is implemented on a pallet forklift AGV experimental platform. The present invention does not require specifying the pallet color and size. Compared with the traditional technology, it has the characteristics of high pose estimation accuracy and good path tracking effect, significantly improving the success rate of pallet automatic loading and effectively improving the handling efficiency on the premise of ensuring the safety and reliability of automatic handling. Description of the Drawings

[0046] Figure 1 This is a flowchart of the automatic loading method for a pallet forklift AGV based on QR code assisted positioning according to the present invention.

[0047] Figure 2 This is a result diagram of warehouse pallet detection and pose estimation.

[0048] Figure 3 This is a top view of warehouse pallet detection and pose estimation.

[0049] Figure 4 This is a schematic diagram of the pose relationship between a warehouse pallet and a forklift.

[0050] Figure 5 This is a schematic diagram of the path tracking algorithm principle. Specific implementation mode

[0051] The technical solution of the present invention will be further clearly and completely described below with reference to the accompanying drawings. However, the described embodiments are only a part of the embodiments of the present invention, not all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the protection scope of the present invention.

[0052] As Figure 1 shown, the automatic loading method for a pallet forklift AGV based on QR code assisted positioning provided in this embodiment includes the following steps:

[0053] Step 1: Use the camera installed on the pallet forklift AGV to obtain the original RGB image of the warehouse pallet with QR code labels.

[0054] In one embodiment, the implementation of step 1 is specifically as follows:

[0055] 1-1: The camera is installed at the end of the right fork leg of the pallet forklift AGV. The camera uses Realsense D435. The real-time transmission speed of the RGB camera of the Realsense D435 camera reaches 30fps, and the image resolution supports 1920×1080. The firmware version of the Realsense D435 camera is Signed_Image_UVC_5_12_14_50.bin. The camera uses USB3.1 for data transmission with the controller.

[0056] 1-2: The QR code labels arranged on the warehouse pallet are selected from the Tag36h11 family in the Apriltag reference library, with a total of 587 different labels, and the label size is 60mm×60mm.

[0057] 1-3: The QR code labels are arranged on the outer side of the warehouse pallet column, preferably on the outer side of the central column of the picking side, asFigure 2 As shown, the point symmetric to the center of the QR code label on the storage tray is used as the storage tray reference point. For example, Figure 3 in which P is the reference point;

[0058] 1-4. Set parameters such as the image format, resolution, and frame rate of the D435 camera to a unified format. After the D435 camera performs hardware initialization and software initialization, turn on the D435 camera and start acquiring the RGB image containing the storage tray.

[0059] Step 2. For the original RGB image of the storage tray, detect the QR code label area through methods such as image grayscale conversion, adaptive threshold processing, continuous boundary segmentation, quadrilateral fitting, and decoding matching. If the QR code label area is not detected, continue to acquire the next frame of image; Figure 2 For example, it is an example of the detection result; in one embodiment, the implementation of step 2 is specifically as follows:

[0060] (1) Convert the RGB image to grayscale;

[0061] (2) Adaptive threshold processing: Use the adaptive threshold method to process the input grayscale image into a binary image. Calculate the threshold by taking the maximum value max and the minimum value min in a 4×4 pixel block, and form a binary image according to the threshold. The threshold calculation formula is:

[0062] T = (max + min) / 2

[0063] (3) Continuous boundary segmentation: To improve the segmentation accuracy, segment the edge based on the generated black and white pixel information. Use the Union-Find algorithm to segment the connected pixel clusters of light and dark pixels, and represent each pixel cluster with a unique ID;

[0064] (4) Quadrilateral fitting: First find a small number of corner points, and then traverse all combination methods of the corner points to calculate the approximate grouping. The quadrilateral fitting step outputs one or several groups of candidate quadrilaterals, and many quadrilateral structures in the environment, including switches, tray grids, label single bits, etc. can be found;

[0065] (5) Fast decoding matching: XOR compare the code values in four directions contained in the quadrilateral with each label code in the label cluster respectively, and filter out the wrong candidate quadrilaterals.

[0066] Step 3. Based on the PnP (Perspective-n-Point) method, solve the six-degree-of-freedom pose of the QR code label center in the camera coordinate system, and obtain the pose of the storage tray reference point in the camera coordinate system through coordinate transformation, so as to realize the local pose estimation of the storage tray. As shown in Figure 3 For example; in one embodiment, the implementation of step 3 is specifically as follows:

[0067] (1) PnP is a method for solving the 3D to 2D point pair motion, which can calculate the pose of the camera according to the spatial point coordinates and their projection positions. After detecting the target QR code label through the above step 2, it is necessary to solve the 6DOF pose information of the center of the QR code label in the camera coordinate system by combining the camera internal parameter matrix, the physical size of the QR code label, and the homography matrix of the camera, that is, the coordinates in three directions in the camera coordinate system: X c 、Z c 、Y c and Euler angles: yaw angle pitch angle and roll angle.

[0068] (2) Through coordinate transformation, the pose of the reference point of the storage tray in the camera coordinate system is obtained, and the local pose estimation of the storage tray is realized. In the actual logistics application scenario, only the lateral offset X c 、longitudinal offset Z c and yaw angle need to be considered for the relative pose relationship. The pose of the center point of the QR code label in the camera coordinate system is represented as O: After coordinate transformation, the pose of the reference point of the storage tray in the camera coordinate system is obtained as P: There are lateral and longitudinal deviations between its reference point and the label center. Assuming the side length of the tray is a, the pose of the tray reference point P can be expressed as:

[0069]

[0070] Step 4. Calculate the global pose of the storage tray according to the local pose of the storage tray and the global pose of the tray forklift AGV obtained in step 3, as shown in Figure 4 ; In one embodiment, the implementation of step 4 is as follows:

[0071] There is a pose deviation between the camera optical center O C and the reference point O F of the tray forklift AGV. Assuming the longitudinal deviation is x c 、the lateral deviation is y c 、and the angular deviation is θ; from the camera coordinate system to the forklift AGV coordinate system, the rotation matrix R is expressed as:

[0072]

[0073] The tray forklift AGV can usually achieve positioning under the global map based on the lidar sensor, and its global real-time pose can be obtained through the reserved interface. Let the pose of the tray forklift AGV in the world coordinate system be (X w , Y w , α), and the pose of the storage tray in the camera coordinate system is Then the pose calculation formula of the storage tray in the world coordinate system is as follows:

[0074]

[0075] Where (X wp , Y wp ) is the global coordinate of the storage tray in the world coordinate system, and φ is the global yaw angle of the storage tray in the world coordinate system; the state of the reference point of the storage tray in the global coordinate system is expressed as: (X wp , Y wp , φ).

[0076] Step 5: Obtain the picking path from the current point of the pallet forklift AGV to the reference point of the storage tray according to the current global pose of the storage tray and the current global pose of the pallet forklift AGV; in one embodiment, use the local path tracking algorithm to obtain the optimal picking path from the current point of the pallet forklift AGV to the reference point of the storage tray, and use the Bezier curve for smoothing processing. The expression of the picking path F(t) is:

[0077] F(t) = (1 - t) 3 P 1 + 3t(1 - t) 2 P 2 + 3t 2 (1 - t)P 3 + t 3 P 4 t ∈ [0, 1]

[0078] The above formula is the parametric equation of the third-order Bezier curve, and the proportional coefficient t takes values from 0 to 1; P 1 is the starting point, that is, the current pose of the pallet forklift AGV, P 4 is the end point, that is, the reference point of the storage tray, P 2 and P 3 are the control points, which are set manually. Preferably, they are the midpoints of the line connecting the current world coordinate of the pallet forklift AGV to the world coordinate of the center of the QR code label, and the midpoint of the line connecting the world coordinate of the center of the QR code label to the reference point of the storage tray; by manually selecting two control points for path fitting, the picking path is smoother after fitting.

[0079] Step 6: Control the pallet forklift AGV to track the picking path to achieve the fully automatic and precise loading of the pallet forklift AGV on the storage tray. Specifically: as Figure 5 shown, calculate the angular velocity and speed and send them to the forklift chassis to control its movement to achieve the pallet forklift AGV tracking the picking path:

[0080] 6-1. The angular velocity calculation formula is expressed as:

[0081] ω = K 1 Δy + K2 β + K 3 γ

[0082] where Δy is the distance from the current pose point to the tangent of the target point on the picking path, β represents the yaw angle of the current pose of the forklift, and γ represents the angular deviation between the current pose point of the forklift and the target point; the input parameter K 1 , K 2 , K 3 are fixed parameters. The selection of the three parameters will affect the path tracking effect, and the output result is the angular velocity ω;

[0083] 6 - 2. The calculation of the linear velocity is specifically as follows:

[0084] Calculate the distance d from the current point to the target point:

[0085]

[0086] where a is the acceleration, v cur and v min are the current speed and the lower speed limit respectively; assume that the projected distance from the current pose point to the target point is Δx;

[0087] If d is less than or equal to Δx, the linear velocity calculation formula is:

[0088]

[0089] If d is greater than Δx, the linear velocity calculation formula is:

[0090] v = v + aT

[0091] Here, v has a maximum speed limit v max , that is, v ≤ v max , where T is the control period.

[0092] The present invention does not need to specify the pallet color and size. Compared with the traditional technology, it has the characteristics of high pose estimation accuracy and good path tracking effect, significantly improving the success rate of automatic pallet loading. On the premise of ensuring the safety and reliability of automatic handling, the handling efficiency is effectively improved.

[0093] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non - exclusive inclusion, so that a process, method, commodity or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or elements inherent to such a process, method, commodity or device. Without more limitations, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, commodity or device including the said element.

[0094] The above description is of specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the acts or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0095] The terms used in one or more embodiments of this specification are for the purpose of describing particular embodiments only and are not intended to limit one or more embodiments of this specification. The singular forms "a", "the", and "said" as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0096] It should be understood that although the terms first, second, third, etc. may be used in one or more embodiments of this specification to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of one or more embodiments of this specification, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "upon" or "in response to determining".

[0097] The above description is only the preferred embodiments of one or more embodiments of this specification and is not intended to limit one or more embodiments of this specification. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of one or more embodiments of this specification shall be included within the scope of protection of one or more embodiments of this specification.

Claims

1. An automatic loading method for a pallet forklift AGV based on QR code assisted positioning, characterized in that, it includes: Step 1: Use the camera installed on the pallet forklift AGV to obtain the original RGB image of the storage pallet with QR code labels; Step 2: Detect the QR code label area in the original RGB image of the storage pallet. If the QR code label area is not detected, continue to obtain the next frame of image; for the original RGB image of the storage pallet, detect the QR code labels arranged on the storage pallet through image grayscale conversion, adaptive threshold processing, continuous boundary segmentation, quadrilateral fitting, and decoding matching; the specific process of the quadrilateral fitting is as follows: first find a small number of corner points, and then traverse all combination methods of the corner points to calculate approximate groupings, and output one or more groups of candidate quadrilaterals; The specific process of the decoding matching is as follows: respectively perform exclusive OR comparison on the code values in four directions contained in the quadrilateral with each label code in the label cluster to filter out incorrect candidate quadrilaterals; Step 3: Based on the PnP method, solve the six-degree-of-freedom pose of the center of the QR code label in the camera coordinate system, and obtain the pose of the reference point of the storage pallet in the camera coordinate system through coordinate transformation to realize the local pose estimation of the storage pallet; the pallet forklift AGV realizes positioning under the global map based on the lidar sensor; Step 4: According to the local pose of the storage pallet and the global pose of the pallet forklift AGV obtained in Step 3, calculate the global pose of the storage pallet; the specific process of calculating the global pose of the storage pallet is as follows: There is a pose deviation between the optical center of the camera and the reference point of the pallet forklift AGV. Assume that the longitudinal deviation is x c and the lateral deviation is y c and the angular deviation is θ. From the camera coordinate system to the forklift AGV coordinate system, the rotation matrix R is expressed as: Let the pose of the pallet forklift AGV in the world coordinate system be (X w , Y w , α), and the pose of the storage pallet in the camera coordinate system be Then the calculation formula for the pose of the storage pallet in the world coordinate system is: where (X wp , Y wp ) is the global coordinate of the storage pallet in the world coordinate system, and φ is the global yaw angle of the storage pallet in the world coordinate system; the state of the reference point of the storage pallet in the global coordinate system is expressed as: (X wp , Y wp , φ); Step 5: According to the current global pose of the storage pallet and the current global pose of the pallet forklift AGV, obtain the picking path from the current point of the pallet forklift AGV to the reference point of the storage pallet; Step 6: Control the pallet forklift AGV to track the picking path to realize the fully automatic and accurate loading of the pallet forklift AGV for the storage pallet.

2. The automatic loading method for a pallet forklift AGV based on QR code assisted positioning according to claim 1, characterized in that, the QR code labels are arranged on the outer side of the central column on the picking side of the storage pallet, and the point symmetric to the center of the QR code label on the storage pallet is used as the reference point of the storage pallet.

3. The automatic loading method for a pallet forklift AGV based on QR code assisted positioning according to claim 1, characterized in that, the specific process of the continuous boundary segmentation is as follows: segment the edge based on the black and white pixel information of the binary image obtained by the adaptive threshold processing; use the union find algorithm to segment the connected pixel clusters of the bright and dark pixels, and use a unique ID to represent each pixel cluster.

4. The automatic loading method for a pallet forklift AGV based on QR code assisted positioning according to any one of claims 1-3, characterized in that, the specific process of Step 3 is as follows: 3-1. Based on the target QR code label detected in Step 2, combined with the camera internal parameter matrix, the physical size of the QR code label, and the camera homography matrix, solve the six-degree-of-freedom pose information of the center of the QR code label in the camera coordinate system, that is, the coordinates in three directions in the camera coordinate system: X c , Z c , Y c and the Euler angles: yaw angle pitch angle and roll angle; 3-2. Represent the pose of the center point of the QR code label in the camera coordinate system as O: After coordinate transformation, obtain the pose P of the reference point of the storage tray in the camera coordinate system: There are horizontal and vertical deviations between its reference point and the label center. Assuming the side length of the tray is a, the pose of the tray reference point is represented as:

5. The automatic loading method for a pallet forklift AGV based on QR code assisted positioning according to claim 1, characterized in that, in Step 5, according to the current global pose of the storage pallet and the current global pose of the pallet forklift AGV, obtain the optimal picking path from the current point of the pallet forklift AGV to the reference point of the storage pallet, and perform smoothing processing using the Bezier curve. The expression of the picking path F(t) is: F(t) = (1 - t) 3 P 1 + 3t(1 - t) 2 P 2 + 3t 2 (1 - t)P 3 + t 3 P 4 t ∈ [0, 1] The above equation is the parametric equation of a third-order Bézier curve, and the proportional coefficient t ranges from 0 to 1; P 1 is the starting point, that is, the current pose of the pallet forklift AGV, and P 4 is the end point, that is, the reference point of the storage pallet, and P 2 and P 3 are control points, which are set artificially; by artificially selecting two control points for path fitting, the picking path is smoother after fitting.

6. The automatic loading method of the pallet forklift AGV based on QR code assisted positioning according to claim 5, characterized in that, Control point P 2 and P 3 are respectively set as the midpoint of the line connecting the current world coordinates of the pallet forklift AGV to the world coordinates of the center of the QR code label, and the midpoint of the line connecting the world coordinates of the center of the QR code label to the reference point of the storage pallet.

7. The automatic loading method of the pallet forklift AGV based on QR code assisted positioning according to claim 1, characterized in that, in step 6, by calculating the angular velocity and speed and sending them to the forklift chassis to control its movement, so as to realize that the pallet forklift AGV tracks the picking path, specifically: 6-1. The calculation formula of the angular velocity is expressed as: ω = K 1 Δy + K 2 β + K 3 γ where Δy is the distance from the current pose point to the tangent of the target point on the picking path, β represents the yaw angle of the current pose of the forklift, and γ represents the angular deviation between the current pose point of the forklift and the target point; the input parameters K 1 , K 2 , K 3 are fixed parameters, and the selection of the three parameters will affect the path tracking effect. The output result is the angular velocity ω; 6-2. The calculation of the linear velocity is specifically: Calculate the distance d from the current point to the target point: where a is the acceleration, v cur and v min are the current speed and the lower limit of speed respectively; assume that the projection distance from the current pose point to the target point is Δx; If d is less than or equal to Δx, the calculation formula of the linear velocity is: If d is greater than Δx, the calculation formula of the linear velocity is: v = v + aT Here, v has the maximum speed v max There is a limit, i.e., v ≤ v max , where T is the control period.

Images