Multi-angle mechanical arm agricultural product bagging device based on image recognition

Abstract

The invention relates to the technical field of agricultural machinery, in particular to a multi-angle mechanical arm agricultural product bagging device based on image recognition, which comprises a bagging device, a mechanical arm, a manipulator of the mechanical arm and an AGV (Automatic Guided Vehicle) mobile trolley, the bagging device comprises a fixing device which is installed at the output end of the mechanical arm and the output end of the mechanical hand. The vacuum suction cup pipeline is installed on the fixing device, the vacuum suction cup pipeline is of a U-shaped structure, a vacuum suction cup is arranged on the vertical portion of the vacuum suction cup pipeline, and a middle clamping groove is formed in the horizontal portion of the vacuum suction cup pipeline; one end of the mechanical arm is connected with the mechanical arm connector, and the other end of the mechanical arm is connected with the rotating piece; and the meshing type mechanical arm is arranged on the upper end face of the mechanical arm, the meshing type mechanical arm is used for placing paper bags on the bagging device, grabbing mangoes and placing the mangoes into the bagging device, and the capacity of the agricultural bagging robot in automatic operation is improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural machinery technology, specifically to a multi-angle robotic arm bagging device for agricultural products based on image recognition. Background Technology

[0002] Mango bagging is a key step in orchard management, playing an irreplaceable role in preventing pests and diseases, reducing pesticide residues, and improving the appearance and quality of the fruit.

[0003] However, current mango bagging operations still rely heavily on manual labor, facing numerous serious challenges. In high-temperature environments, farmers must work with their heads tilted back for extended periods, resulting in immense physical exertion and limited effective working hours each day. This traditional method is not only labor-intensive and inefficient but also accompanied by increased occupational injury risks and a growing labor shortage. While existing auxiliary bagging devices can simplify the process of opening the netting, they still require manual handling and cannot achieve true automation. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-angle robotic arm bagging device for agricultural products based on image recognition, solving key problems in existing technologies. In mango harvesting, traditional manual picking and bagging operations are not only inefficient but also affect the continuity of work. Furthermore, traditional manual picking and bagging operations are difficult to perform in harsh environments, failing to meet the demands of efficient automated operations. This invention first finely divides the target area and calculates the optimal path, then automatically identifies the mangoes, ensuring efficient execution of the bagging action. This innovative design not only reduces the cost of manual intervention but also improves the automation level of mango harvesting, providing a new technical solution for the development of agricultural mechanization.

[0005] The technical problem solved by this invention is: The objective of this invention can be achieved through the following technical solutions: A multi-angle robotic arm bagging device for agricultural products based on image recognition includes a bagging device, a robotic arm and its manipulator, and an AGV mobile vehicle. The top surface of the AGV mobile vehicle is equipped with a robotic arm and its robotic hand, and the output end of the robotic arm and its robotic hand is equipped with a bagging device. The bagging device includes: A fixing device is installed on the output end of the robotic arm and robotic hand; The vacuum suction cup tube is installed on a fixed device and has a U-shaped structure. The vertical part of the vacuum suction cup tube is equipped with a vacuum suction cup, and the horizontal part of the vacuum suction cup tube is equipped with a middle groove. Robotic arms and their manipulators include: The robotic arm has one end connected to the robotic arm interface and the other end connected to the rotating component; The biting-tooth robotic arm is located on the upper end of the robotic arm. It is used to place paper bags in the bagging device and to grasp mangoes and put them into the bagging device.

[0006] As a further aspect of the present invention, the bagging device further includes: The bag sealing instrument is installed in the middle of the vertical part of the vacuum suction cup pipe.

[0007] As a further aspect of the present invention: the side end face of the biting robotic arm is provided with a camera for mango positioning and recognition, and is equipped with a lighting device.

[0008] As a further aspect of the present invention: the upper surface of the AGV mobile vehicle is provided with a lifting platform, a harvesting tray is placed at the rear of the AGV mobile vehicle, and a receiving device and an antenna are also provided on the AGV mobile vehicle.

[0009] As a further embodiment of the present invention: a fruit bag opening support for placing paper bag openings is provided on the side of the lifting platform.

[0010] As a further aspect of the present invention: an operating platform and a guiding device are respectively provided on the side wall of the fruit bag opening device, and the guiding device is connected to the path control system through a wireless communication module.

[0011] As a further aspect of the present invention, the specific working process of the bagging device is as follows: Mango image information was acquired using a depth camera; Preprocess the mango image information; The Yolov11 algorithm was improved to form the Yolov11_IMO algorithm, which was then used to identify mangoes. A predetermined pixel area value of 15 is set. The area of ​​each target in the image is calculated, and targets with a pixel area less than 15 are considered interference and removed. At this point, salt-and-pepper noise still exists within the target, because the mask layer image is a binary image. Use machine vision algorithms to find the outline of the target mango and its minimum bounding rectangle; The intersection of the two diagonals of the smallest bounding rectangle of the mango outline is set as the approximate centroid of the mango.

[0012] As a further aspect of the present invention, the following working process is also included: Determine the lower boundary of the smallest bounding rectangle of the mango. The starting point of the mango bagging action is located 5cm below the lower boundary.

[0013] As a further aspect of the present invention, the specific process of forming the Yolov11_IMO algorithm is as follows: We propose the WY_conv feature extraction module; introduce the SEAM attention mechanism to improve object occlusion detection in the neck network structure; and introduce the DynamicHead detection head module to unify scale awareness, spatial awareness, and task awareness.

[0014] As a further aspect of the present invention: WY_conv consists of three branches: an average pooling branch; a max pooling branch, which aims to highlight salient features and enhance edge information; and a convolution branch, which captures a wider range of contextual information.

[0015] The beneficial effects of this invention are: This invention proposes an automated mango bagging robot for low-canopy mangoes based on image recognition and its control method. This machine, through an integrated intelligent control system, automatically bags mangoes, reducing manual intervention and improving operational efficiency. Image recognition technology and RTK navigation technology ensure the robot can complete path planning in complex environments, while the bagging device accurately handles and bags low-canopy mangoes. This technology significantly enhances the capabilities of agricultural bagging robots in automated operations, has significant application value, and can promote the further development of agricultural mechanization technology. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a structural diagram of the entire machine provided by the present invention; Figure 2 This is a structural diagram of the bagging device provided by the present invention; Figure 3 This is a schematic diagram of the structure of the robotic arm and its manipulator provided by the present invention; Figure 4 This is a structural schematic diagram of the AGV vehicle provided by the present invention.

[0018] Figure 5 This is the AGV control flowchart provided by the present invention; Figure 6 This is a flowchart of the mango bagging machine robotic arm and its robotic glove bagging process provided by the present invention; In the diagram: 1. Bagging device; 2. Robotic arm and its manipulator; 3. AGV mobile trolley; 101. Intermediate slot; 102. Vacuum suction cup; 103. Fixing device; 104. Vacuum suction cup pipe; 105. Sealing instrument; 201. Robotic arm interface; 202. Gear; 203. Gear-type manipulator; 204. Robotic arm; 205. Rotating component; 206. Camera; 207. Gripping claw; 208. Lighting device; 301. Motor; 302. Receiving device; 303. Spring; 304. Harvesting tray; 305. Antenna; 306. Lifting platform; 307. Operating end platform; 308. Fruit bag opening device; 309. Guiding device. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0020] Example 1 like Figure 1-4 As shown in the figure, the multi-angle robotic arm agricultural product bagging device based on image recognition provided by the embodiment of the present invention specifically includes a bagging device 1, a robotic arm and its robotic hand 2, and an AGV mobile trolley 3; A robotic arm and its robotic hand 2 are installed on the top surface of the AGV mobile vehicle 3, and a bagging device 1 is installed at the output end of the robotic arm and its robotic hand 2. Specifically, the bagging device 1 includes: The fixing device 103 can be a fixing plate, which is installed on the output end of the robotic arm and its robotic hand 2, in order to cooperate with the robotic arm and its robotic hand 2 to connect and rotate the subsequent work. Vacuum suction cup pipe 104 is installed on the fixing device 103 and has a U-shaped structure. A vacuum suction cup 102 is provided on the vertical part of the vacuum suction cup pipe 104 and an intermediate slot 101 is provided on the horizontal part of the vacuum suction cup pipe 104. The bag sealing instrument 105 is installed in the middle of the vertical part of the vacuum suction cup pipe 104; When the bagging device 1 is working, it accurately places the paper bag in the middle slot 101 and fixes the tail of the paper bag. Then, it transmits vacuum through the vacuum suction cup pipe 104, so that the vacuum suction cup 102 has a suction effect. The lower part of the vacuum suction cup 102 fixes the tail of the paper bag, and the upper part has two pairs of vacuum suction cups 102 on both sides to open the sides of the paper bag. The robotic arm and its robotic hand 2 put the grasped mango into the bag opening. The mango will open the middle paper bag due to gravity and put the mango in. Finally, the sealing instrument 105 seals the bag opening using the principle of a stapler. Secondly, specifically, the robotic arm and its robotic hand 2 include: The robotic arm 204 has one end connected to the robotic arm interface 201 and the other end connected to the rotating component 205. The rotating component 205 is used to facilitate the rotation of the robotic arm 203 to complete the grasping operation, and the rotating component 205 is connected to the fixing device 103 of the bagging device 1. The robotic arm interface 201 is used to connect the circuit and control switch, and is connected to the AGV mobile trolley 3. A toothed robotic arm 203 is mounted on the upper end face of the robotic arm 204. The toothed robotic arm 203 is used to place paper bags in the bagging device 1 and to grasp mangoes and place them into the bagging device 1. A camera 206 for mango positioning and recognition is set on the side end face of the toothed robotic arm 203, which can accurately identify and grasp mangoes. The toothed robotic arm 203 is equipped with a lighting device 208, which allows bagging operations to be performed at night. Among them, the biting robot 203 mainly includes two sets of symmetrically arranged gears 202 and gripping claws 207. One set of gears 202 is connected to the output end of the drive motor, and the gripping claws 207 are connected to the gears 202 through a connecting rod. By controlling the gears 202, the gripping claws 207 in the biting robot 203 will be driven to perform grasping operations. When the robotic arm and its manipulator 2 are working, the upper end face of the robotic arm 204 is equipped with a biting manipulator 203 for placing paper bags in the bagging device 1 and grasping mangoes, placing the mangoes into the bagging device 1. The side end face of the biting manipulator 203 is equipped with a camera 206 for mango positioning and recognition, which can accurately identify and grasp mangoes. After the camera 206 recognizes the mango, the gear 202 drives the gripper 207 in the biting manipulator 203 to perform the grasping operation. The biting manipulator 203 is equipped with a lighting device 208, allowing bagging operations to be performed at night. Thirdly, specifically, the AGV mobile trolley 3 is equipped with a motor 301 for providing power for operation and travel, and a spring 303 is installed above the tires of the AGV mobile trolley 3. A lifting platform 306 is provided on the upper surface of the AGV mobile trolley 3. A harvesting tray 304 is placed at the rear of the AGV mobile trolley 3. A receiving device 302 and an antenna 305 are also installed on the AGV mobile trolley 3. A robotic arm and its robotic hand 2 are installed on the upper surface of the lifting platform 306. A fruit bag opening device 308 for holding paper bag openings is installed on the side of the lifting platform 306. An operating end platform 307 and a guide device 309 are respectively installed on the side wall of the fruit bag opening device 308. The guide device 309 is connected to the path control system through a wireless communication module. When the AGV mobile vehicle 3 is in operation, it is driven by the motor 301 and connects to the path control system via a wireless communication module according to the guidance device 309. The system feeds back the acquired infrared light, magnetic field strength, and movement frequency information of the AGV mobile vehicle 3 to the path control system to control its travel path and obstacle avoidance. To prevent excessive vibration during the movement of the AGV mobile vehicle 3 in the planting base from damaging the robotic arm 204 and the lifting platform 306, a spring 303 is installed above the tires of the AGV mobile vehicle 3 to buffer vibration. A harvesting tray 304 is placed at the rear of the AGV mobile vehicle 3 to hold the bagged mangoes. A lifting platform 306 is located on the upper surface of the AGV mobile vehicle 3. A robotic arm and its robotic hand 2 are mounted on the upper surface of the lifting platform 306. The ends of the robotic arm and its robotic hand 2 are connected to the lifting platform, which is used to adjust the lifting height of the robotic arm. A fruit bag opening device 308 is installed on its side to hold the paper bag opening. The operating platform 307 is placed next to the fruit bag opening device 308 and connected to the AGV mobile trolley 3. The operating platform 307 is located next to the AGV mobile trolley 3, and the horizontal edge of the operating platform 307 is equipped with a frame to prevent the operating end from falling off during movement. The receiving device 302 and the antenna 305 cooperate to transmit signals to the operating platform 307, thereby synchronously feeding back the mango bagging information.

[0021] Example 2 like Figure 5-6 As shown in the figure, the present invention provides a multi-angle robotic arm agricultural product bagging device based on image recognition. The specific working process of the bagging device is as follows: Step S1: Acquire mango image information using a depth camera; Step S2 involves preprocessing the mango image information, including cropping and filtering. The cropped image size is 640*640. In this step, an adaptive Gaussian filter is used, which is an improved Gaussian filter algorithm that dynamically adjusts the filtering parameters according to the local features of the image. This solves the edge blurring problem caused by the use of the same parameters globally in standard Gaussian filtering, ensuring the removal of noise in the image while preserving the image edges and details as much as possible.

[0022] The guiding principle of the adaptive Gaussian filter is to strike a balance between the smoothest possible result and the best possible result that preserves detail. It uses an energy function of the form: This function has two terms, the first of which is... The second item is The sigma that is the minimum sum of these two terms is the desired value.

[0023] epsilon is the residual. Assuming Gaussian smoothing of a two-dimensional image can be expressed by the following formula: At pixel (x, y), the smoothed gray value at (x, y) is obtained by applying Gaussian smoothing. Use the grayscale value of this pixel from the original image. minus The residual of that pixel can be obtained.

[0024] The first term, 'c', is a constant term. Therefore, to achieve its minimum value, 'sigma' in this term should be as large as possible. That is, this term encourages 'sigma' to increase, aiming for the smoothest possible result. The second term, 'epsilon', is the square of the residual. To minimize the sum, the residual must be as small as possible. That is, after Gaussian smoothing, the gray value at point (x, y) should not change significantly. In this case, 'sigma' should be minimized.

[0025] Therefore, the first term encourages sigma to increase, while the second term encourages sigma to decrease. Thus, this formula allows for the adaptive selection of an appropriate sigma, achieving a trade-off between smooth results and preservation of detail.

[0026] Step S3: Improve the Yolov11 algorithm to form the Yolov11_IMO algorithm, and use the Yolov11_IMO algorithm to identify mangoes; In step S3, the specific process of forming the Yolov11_IMO algorithm is as follows: We propose the WY_conv feature extraction module; introduce the SEAM attention mechanism to improve object occlusion detection in the neck network structure; and introduce the DynamicHead detection head module to unify scale awareness, spatial awareness, and task awareness. The WY_conv feature extraction module is proposed. WY_conv consists of three branches: an average pooling branch, designed to preserve overall background information and smooth noise; a max pooling branch, designed to highlight salient features and enhance edge information; and a convolution branch, which captures a wider range of contextual information. This allows for the fusion of features at different levels of abstraction. The model achieves fast downsampling while preserving multi-scale features, balancing computational efficiency and feature representation capability, thus achieving both lightweight performance and accuracy for mobile vision tasks. By combining depthwise separable convolutions and residual connections, the outputs of convolutions at different depths are combined through point-to-point (1x1) convolutions, and then two fully connected networks are used to fuse the information of each channel to enhance the connection between all channels. Through multi-scale feature fusion and channel-space joint enhancement, the feature response of the unoccluded area is strengthened, and the loss of the occluded area is compensated. In addition, the DynamicHead detection module is introduced to unify scale perception, spatial perception, and task perception.

[0027] 1. Scale-aware attention: This module focuses on the feature hierarchy (L), forming a three-dimensional tensor F∈RL×S×C by scaling the feature pyramid to the same scale, and then using it as input to the dynamic head, enabling the detection head to handle multiple objects of different scales coexisting in the image, while taking into account that objects usually present different shapes, rotations and positions at different viewpoints.

[0028] 2. Spatial Awareness Attention: Multiple dynamic headblocks (Dy-Headblocks) containing scale awareness, spatial awareness, and task awareness attention are stacked sequentially. This module focuses on spatial location (S), paying attention to discriminative regions in the image by aggregating features, applying attention to each spatial location, and adaptively aggregating multiple feature levels to learn more discriminative representations.

[0029] 3. Task-Aware Attention: Ultimately, the output of the dynamic head can be used for representations of different tasks and object detection, such as classification, center / box regression, etc. It focuses on channels (C), dynamically turning feature channels on or off to support different tasks.

[0030] Step S4: Use white to represent mangoes and black to represent the background, so that you can see the position, shape and size of each instance in the image.

[0031] Step S5: Set the pixel area to a predetermined value of 15. Calculate the area of ​​each target in the image, and remove targets with a pixel area less than 15 as interference. At this point, salt-and-pepper noise still exists inside the target. Since the mask layer image is a binary image, an adaptive Gaussian filtering method is used to reduce noise in the image.

[0032] Step S6: Use machine vision algorithms to find the outline of the target mango and its minimum bounding rectangle; The intersection of the two diagonals of the smallest bounding rectangle of the mango outline is set as the approximate centroid of the mango.

[0033] In step S7, the position of the lower boundary of the smallest bounding rectangle of the mango is determined, and the starting point of the mango bagging action is located 5cm below the lower boundary position.

[0034] like Figure 5As shown, the AGV (Automated Guided Vehicle) control process is also included, as follows: Initial navigation: The AGV navigates to a designated fruit tree marker in the orchard based on the preset navigation map and target point information.

[0035] Perform bagging operation: After the AGV reaches the target fruit tree marker, it performs the bagging operation. The system records the operation status as "Bagging successful".

[0036] Movement within the area: After bagging the current fruit tree, the AGV moves to the next fruit tree marker within the same area according to the planned path. The initial navigation and bagging operation are repeated.

[0037] Area completion judgment: After the AGV processes the last fruit tree marker in the current area and successfully bags the fruit, the system will judge: "Current area complete?" If the judgment result is no, it will continue to move to the next marker in the same area. If the judgment result is yes, it will proceed to the next step.

[0038] Cross-regional transfer: After confirming that all tasks in the current area are completed, the AGV will navigate to the new work area. Upon arrival in the new area, the above steps will be completed.

[0039] like Figure 6 The process flow of the robotic arm and its robotic glove bag is shown below: Image recognition and positioning: After the AGV stably docks at the target fruit tree marker, it activates its onboard machine vision system. The system acquires and analyzes images of the fruit tree area, with the core task being to identify and accurately locate the spatial position (X, Y, Z coordinates) of the target mango.

[0040] Adaptability assessment: The vision system calculates the current height difference between the target mango and the bagging actuator on the AGV (usually located at the end of the lifting platform 306).

[0041] The system determines whether the height is sufficient (i.e., whether the bagging mechanism can directly reach the target mango). If yes (height sufficient): execute the bag removal / bag opening operation. If no (height insufficient): the process proceeds to adjust the height of the lifting platform 306.

[0042] Lifting platform 306 height adjustment: If the height is insufficient, the system controls the AGV's lifting platform 306 to lift.

[0043] Lifting platform 306 is raised to an appropriate height: Based on the visual positioning results, the system calculates and instructs the lifting platform 306 to be raised to an estimated appropriate height so that the bagging mechanism can reach the target mango.

[0044] Lifting platform 306 rises to stop: Lifting platform 306 performs a lifting action until it reaches the commanded height and then stops.

[0045] Looping judgment: After the lifting platform 306 stops, the process returns to step 2 ("Is the height sufficient?"), and the vision system checks again whether the current height meets the requirements. This process is repeated until the height is adjusted to the correct position (judged as "yes").

[0046] Bag retrieval / opening operation: Once the height requirement is met (judged as "yes"), the AGV's bagging mechanism (such as a robotic arm, pneumatic device, etc.) begins to operate. The mechanism completes the preparatory actions of bag retrieval (removing the fruit bag from the storage device) and bag opening (opening the fruit bag opening to a suitable state).

[0047] Bagging Action Completion and Result Determination: The bagging actuator performs the final bagging action, placing the fruit bag over the target mango and potentially sealing it (depending on the mechanism design). After the action is completed, the system uses preset sensors (such as visual confirmation, force feedback, and positioning sensors) to determine the result: "Was the bagging successful?" If yes (bagging successful): the system records that the mango bagging task was successful.

[0048] The system updates the backend terminal: Successful information (including mango location, timestamp, etc.) is sent to the backend management system in real time for recording and monitoring. Once the process is complete, the AGV will move to the next marker point in the same area according to the main path planning process.

[0049] Trigger alarm interruption process: If not (bagging failure): The system immediately triggers an audible and visual alarm and / or a remote alarm to notify of the abnormal status.

[0050] Manual intervention to troubleshoot: The process is interrupted, awaiting on-site inspection by operators. Possible faults include: mechanism jamming, depletion of fruit bags, severe misalignment of mango positioning, and unexpected obstacles. Personnel must intervene to diagnose and repair the fault or manually complete the bagging process.

[0051] After troubleshooting: The operator confirms that the troubleshooting has been resolved or that the manual operation has been completed, and the system is reset or the process continues according to the instructions.

[0052] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A multi-angle robotic arm bagging device for agricultural products based on image recognition, characterized in that, Includes a bagging device (1), a robotic arm and its robotic hand (2), and an AGV mobile trolley (3); The top surface of the AGV mobile vehicle (3) is equipped with a robotic arm and its robotic hand (2), and the output end of the robotic arm and its robotic hand (2) is equipped with a bagging device (1). The bagging device (1) includes: A fixing device (103) is installed on the output end of the robotic arm and its manipulator (2); Vacuum suction cup pipe (104) is installed on the fixing device (103) and the vacuum suction cup pipe (104) has a U-shaped structure. A vacuum suction cup (102) is provided on the vertical part of the vacuum suction cup pipe (104) and a middle slot (101) is provided on the horizontal part of the vacuum suction cup pipe (104). The robotic arm and its manipulator (2) include: A robotic arm (204) has one end connected to a robotic arm interface (201) and the other end connected to a rotating component (205); A biting manipulator (203) is set on the upper end face of the robotic arm (204). The biting manipulator (203) is used to place paper bags in the bagging device (1) and to grab mangoes and put the mangoes into the bagging device (1).

2. The multi-angle robotic arm agricultural product bagging device based on image recognition according to claim 1, characterized in that, The bagging device (1) also includes: The bag sealing instrument (105) is installed in the middle of the vertical part of the vacuum suction cup pipe (104).

3. The multi-angle robotic arm agricultural product bagging device based on image recognition according to claim 2, characterized in that, The bite-type robotic arm (203) is equipped with a camera (206) for mango positioning and recognition on its side end face, and is also equipped with a lighting device (208).

4. The multi-angle robotic arm agricultural product bagging device based on image recognition according to claim 1, characterized in that, The upper surface of the AGV mobile vehicle (3) is provided with a lifting platform (306), a harvesting tray (304) is placed behind the AGV mobile vehicle (3), and a receiving device (302) and an antenna (305) are also provided on the AGV mobile vehicle (3).

5. A multi-angle robotic arm agricultural product bagging device based on image recognition according to claim 4, characterized in that, A fruit bag opening device (308) for placing paper bag openings is provided on the side of the lifting platform (306).

6. The multi-angle robotic arm agricultural product bagging device based on image recognition according to claim 5, characterized in that, The side wall of the fruit bag opening device (308) is provided with an operating end platform (307) and a guide device (309). The guide device (309) is connected to the path control system through a wireless communication module.

7. The multi-angle robotic arm agricultural product bagging device based on image recognition according to claim 1, characterized in that, The specific working process of the bagging device (1) is as follows: Mango image information was acquired using a depth camera; Preprocess the mango image information; The Yolov11 algorithm was improved to form the Yolov11_IMO algorithm, which was then used to identify mangoes. The preset pixel area is set to 15. The area of ​​each target in the image is calculated, and targets with a pixel area less than 15 are considered interference and removed. At this point, salt-and-pepper noise still exists inside the target because the mask layer image is a binary image. Use machine vision algorithms to find the outline of the target mango and its minimum bounding rectangle; The intersection of the two diagonals of the smallest bounding rectangle of the mango outline is set as the approximate centroid of the mango.

8. A multi-angle robotic arm agricultural product bagging device based on image recognition according to claim 7, characterized in that, It also includes the following work processes: Determine the lower boundary of the smallest bounding rectangle of the mango. The starting point of the mango bagging action is located 5cm below the lower boundary.

9. A multi-angle robotic arm agricultural product bagging device based on image recognition according to claim 7, characterized in that, The specific process of forming the Yolov11_IMO algorithm is as follows: We propose the WY_conv feature extraction module; introduce the SEAM attention mechanism to improve object occlusion detection in the neck network structure; and introduce the DynamicHead detection head module to unify scale awareness, spatial awareness, and task awareness.

10. A multi-angle robotic arm agricultural product bagging device based on image recognition according to claim 9, characterized in that, WY_conv consists of three branches, which are average pooling branches; Max pooling branch aims to highlight salient features and enhance edge information; convolution branch captures a wider range of contextual information.

Images