A ball picking device based on YOLO technology

CN224853763UActive Publication Date: 2026-10-16GUANGDONG INST OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202522044748.8
Authority / Receiving Office
CN · China
Patent Type
Utility models(China)
Current Assignee / Owner
Filing Date
2025-09-23
Publication Date
2026-10-16
Estimated Expiration
2035-09-23

AI Technical Summary

Technical Problem

[0006]为解决上述所述的装置智能化程度不足的技术问题,本实用新型采用技术方案的基本构思是:

Benefits of technology

[0017] 1. Improve the intelligence level of the robot to analyze and process real-time video of the court, detect the position and trajectory of the ping-pong ball in real time, and use a precise motion control system to enable the robot to quickly move to the position of the ping-pong ball and accurately pick it up. Utilize the YOLO convolutional neural network to analyze the numerical target and improve the recognition accuracy of the ping-pong ball.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN224853763U_ABST
    Figure CN224853763U_ABST
Patent Text Reader

Abstract

The utility model relates to intelligent robot technical field discloses a kind of ball picking device based on YOLO technology, including: robot body, robot body is rectangular;Intelligent identification structure, intelligent identification structure includes first intelligent identification part and second intelligent identification part, first intelligent identification part includes the support frame of fixed installation in the top of robot body, the top of support frame is provided with 3D depth camera, the front side of robot body is provided with slot, and custom version laser radar is installed in slot;Second intelligent identification part includes the fixed base plate of fixed installation in the bottom of robot body, reach to the real-time video of court is analyzed and handled, the position and motion trajectory of table tennis are detected in real time, and through accurate motion control system, robot is quickly moved to the position of table tennis, accurate ball picking is carried out, using YOLO convolution neural network, numerical target is analyzed, improve the identification accuracy of table tennis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This utility model belongs to the field of intelligent robot technology, specifically, it relates to a ball-picking device based on YOLO technology. Background Technology

[0002] During table tennis training and competitions, a large number of table tennis balls are scattered throughout the court. Manually picking up the balls is not only inefficient, but also consumes a lot of manpower and time.

[0003] Utility model CN214913178U discloses an intelligent ping-pong ball picking device, relating to the field of ping-pong ball picking robot technology. It includes a vehicle body and a motion mechanism located at the bottom of the vehicle body. The vehicle body has a cavity containing a storage basket that is movably mounted through it. The storage basket contains multiple sets of elastic strips. Stepper motors are symmetrically arranged on both sides of the storage basket on the vehicle body. The output ends of the stepper motors are connected to threaded rods, each threaded rod having a threaded block fitted onto it. The threaded blocks are fixedly connected to both sides of the storage basket. Beneficial effects: It has the advantages of low noise, low energy consumption, and simple structure. It can pick up multiple ping-pong balls at once, greatly improving the picking efficiency. It has extremely high robustness. Its advantages lie in its high structural reliability, simple manufacturing process, low production cost, safe use, environmental friendliness, strong practicality, and ease of popularization.

[0004] However, the aforementioned patents still have the following problems: Although existing ball-collecting robots can complete the task of collecting balls to a certain extent, they are insufficient in terms of the accuracy of ping-pong ball recognition, the precision of field map construction, the rationality of path planning, and the ability to avoid obstacles, and the use of the device is inconvenient.

[0005] In view of this, this utility model is proposed. Utility Model Content

[0006] To address the technical problem of insufficient intelligence in the aforementioned devices, the basic concept of the technical solution adopted by this utility model is as follows:

[0007] A ball-picking device based on YOLO technology, comprising:

[0008] The robot body is rectangular.

[0009] The intelligent recognition structure includes a first intelligent recognition unit and a second intelligent recognition unit. The first intelligent recognition unit includes a support frame fixedly installed on the top of the robot body, with a 3D depth camera installed on the top of the support frame. A slot is opened on the front side of the robot body, and a customized LiDAR is installed in the slot. The second intelligent recognition unit includes a fixed base plate fixedly installed on the bottom of the robot body. A set of first connecting plates is fixedly installed on both outer walls of the fixed base plate. A first fixing plate is fixedly installed on the top of the first connecting plate. A Hall encoder geared motor is fixedly installed on one outer wall of the first fixing plate. A first rotating shaft is rotatably installed inside the first connecting plate, and the output shaft of the Hall encoder geared motor is fixedly connected to one end of the first rotating shaft.

[0010] In a preferred embodiment of the present invention, the second intelligent recognition unit further includes a four-way encoder motor drive module disposed inside the robot body. A data cable is fixedly installed between the four-way encoder motor drive module and the Hall encoder reduction motor. A wheel is fixedly installed on the outer wall of the first rotating shaft, and an air cushion is provided on the outer wall of the wheel.

[0011] In a preferred embodiment of this utility model, a third fixing plate is fixedly installed on one outer wall of the fixed base plate, a fourth fixing plate is fixedly installed on the top of the third fixing plate, two sets of support plates are fixedly installed on the top of the fourth fixing plate, a limit plate is sleeved between the two sets of support plates, four sets of limit bolts are spirally installed inside the limit plate, a high-definition touch LCD screen is fixedly installed on one outer wall of the limit plate, a data cable is fixedly installed on one outer wall of the high-definition touch LCD screen, and one end of the data cable is fixedly installed inside the robot body.

[0012] In a preferred embodiment of this utility model, a multi-functional expansion module is fixedly installed on the top of the robot body. The multi-functional expansion module has an onboard IMU sensor to detect the robot's body posture in real time.

[0013] In a preferred embodiment of this utility model, a first electrical control box is fixedly installed on the top of the robot body. A first limiting groove is provided inside the first electrical control box. A second fixing plate is provided inside the first limiting groove. A bracket is fixedly installed on one outer wall of the second fixing plate. A second rotating shaft is rotatably installed inside the bracket. A triangular protective sleeve is fixedly installed on the outer wall of the second rotating shaft. Two sets of second wheels are provided on the outer wall of the robot body. A wheel hub is fixedly installed inside the second wheel. A first limiting hole is provided inside the wheel hub. The second rotating shaft is rotatably installed inside the first limiting hole.

[0014] In a preferred embodiment of this utility model, sixteen sets of first arc-shaped fixing plates are fixedly installed on the outer wall of the second wheel, and second arc-shaped fixing plates are fixedly installed on the outer wall of the first arc-shaped fixing plates. A set of ball-collecting net rope is sleeved between every two sets of first arc-shaped fixing plates. Three sets of limiting blocks are fixedly installed on one side of the outer wall of the second wheel, and a set of protective net rope is sleeved between the three sets of limiting blocks.

[0015] In a preferred embodiment of this utility model, a voice control module is fixedly installed on the inner bottom of the support frame, a first speaker is fixedly installed on the top of the voice control module, and a set of second speakers is provided on the three outer walls of the voice control module.

[0016] Compared with the prior art, the present invention has the following advantages:

[0017] 1. Improve the intelligence level of the robot to analyze and process real-time video of the court, detect the position and trajectory of the ping-pong ball in real time, and use a precise motion control system to enable the robot to quickly move to the position of the ping-pong ball and accurately pick it up. Utilize the YOLO convolutional neural network to analyze the numerical target and improve the recognition accuracy of the ping-pong ball.

[0018] 2. To achieve the purpose of touch screen control of the device, improve the efficiency of issuing commands, expand the applicability of the device, meet the usage needs of different brand chips and systems, and improve the ease of use of the device.

[0019] 3. To achieve the purpose of picking up ping-pong balls and to control the movement of the device through voice, thereby improving the flexibility of the device and optimizing the user experience.

[0020] The specific embodiments of this utility model will be described in further detail below with reference to the accompanying drawings. Attached Figure Description

[0021] In the attached diagram:

[0022] Figure 1 This is a schematic diagram of the structure of this utility model;

[0023] Figure 2 This is a schematic diagram showing the disassembled bracket and second wheel of this utility model.

[0024] Figure 3 This is a schematic diagram of the first arc-shaped fixing plate structure of this utility model;

[0025] Figure 4 This is a schematic diagram of the first limiting groove structure of this utility model;

[0026] Figure 5 This is a schematic diagram of the wheel structure of this utility model;

[0027] Figure 6 This is a schematic diagram of the second fixing plate structure of this utility model;

[0028] Figure 7 This is a schematic diagram of the customized lidar structure of this utility model.

[0029] In the diagram: 10. Robot body; 11. Support frame; 12. 3D depth camera; 13. Customized LiDAR; 14. Fixed base plate; 15. First connecting plate; 16. First fixing plate; 17. Hall encoder geared motor; 18. First rotating shaft; 19. First wheel; 20. Air cushion; 21. Multifunctional expansion module; 22. First electrical control box; 23. Bracket; 24. Second fixing plate; 25. Second rotating shaft; 26. Triangular protective sleeve; 27. 28. Wheel hub; 29. ​​Second wheel; 30. First arc-shaped fixing plate; 31. Second arc-shaped fixing plate; 32. Ball-collecting net rope; 33. Limiting block; 34. Protective net rope; 35. First limiting hole; 36. First limiting groove; 37. Voice control module; 38. First horn; 39. Second horn; 40. Third fixing plate; 41. Fourth fixing plate; 42. Support plate; 43. Limiting bolt; 44. High-definition touch LCD screen; 45. Data cable. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of this utility model clearer, the technical solutions in the embodiments will be clearly and completely described below with reference to the accompanying drawings. The following embodiments are used to illustrate this utility model.

[0031] Example 1: A ball-picking device based on YOLO technology, specifically as follows... Figure 1 , Figure 3 , Figure 4 and Figure 5As shown, the robot includes a rectangular robot body 10 and an intelligent recognition structure, comprising a first intelligent recognition unit and a second intelligent recognition unit. The first intelligent recognition unit includes a support frame 11 fixedly mounted on the top of the robot body 10, with a 3D depth camera 12 mounted on the top of the support frame 11. A slot is formed on the front side of the robot body 10, and a customized LiDAR 13 is installed in the slot. The second intelligent recognition unit includes a fixed base plate 14 fixedly mounted on the bottom of the robot body 10. A set of first connecting plates 15 are fixedly mounted on both outer walls of the fixed base plate 14. A first fixing plate 16 is fixedly mounted on the top of the first connecting plate 15. A Hall encoder reduction motor 17 is fixedly mounted on one outer wall of the first fixing plate 16. A first rotating shaft 18 is rotatably mounted inside the first connecting plate 15, and the output shaft of the Hall encoder reduction motor 17 is fixedly connected to one end of the first rotating shaft 18. The intelligent recognition structure enables rapid identification and pickup of ping-pong balls. Infrared radar identification and scanning are performed by a customized LiDAR 13, real-time video analysis and processing of the court is performed by a 3D depth camera 12, the first connecting plate 15 is fixed by a fixed base plate 14, and the Hall encoder reduction motor 17 is started to drive the first rotating shaft 18 to rotate.

[0032] Specifically, such as Figure 1 , Figure 3 , Figure 4 and Figure 5 As shown, the second intelligent recognition unit also includes a four-way encoder motor drive module located inside the robot body 10. A data cable is fixedly installed between the four-way encoder motor drive module and the Hall encoder reduction motor 17. A wheel 19 is fixedly installed on the outer wall of the first rotating shaft 18, and an air cushion 20 is provided on the outer wall of the wheel 19. When the first rotating shaft 18 rotates, it drives the wheel 19 to rotate, driving the car to move. The motor is precisely controlled by the four-way encoder motor drive module.

[0033] Based on the above, the structure of the robot body 10, support frame 11, 3D depth camera 12, customized LiDAR 13, fixed base plate 14, first connecting plate 15, first fixed plate 16, Hall encoder reduction motor 17, first rotating shaft 18, wheel 19 and air cushion 20 improves the robot's intelligence level, enabling it to analyze and process real-time video of the court, detect the position and trajectory of the ping-pong ball in real time, and use a precise motion control system to quickly move the robot to the position of the ping-pong ball for accurate ball pickup. By using the YOLO convolutional neural network to analyze the numerical target, the accuracy of ping-pong ball recognition is improved.

[0034] Example 2: Based on Example 1, specifically as follows... Figure 1 , Figure 3 and Figure 6As shown, a third fixing plate 39 is fixedly installed on one outer wall of the fixed base plate 14. A fourth fixing plate 40 is fixedly installed on the top of the third fixing plate 39. Two sets of support plates 41 are fixedly installed on the top of the fourth fixing plate 40. A limit plate 42 is sleeved between the two sets of support plates 41. Four sets of limit bolts 43 are spirally installed inside the limit plate 42. A high-definition touch screen 44 is fixedly installed on one outer wall of the limit plate 42. A data cable 45 is fixedly installed on one outer wall of the high-definition touch screen 44. One end of the data cable 45 is fixedly installed inside the robot body 10. The support plate 41 is fixed by the third fixing plate 39 and the fourth fixing plate 40. The limit plate 42 is fixed to the support plate 41 by the limit bolts 43. The device is controlled by the high-definition touch screen 44. The high-definition touch screen 44 supports Raspberry Pi, NVIDIA, Ubuntu images, and secondary screens of computer systems.

[0035] Specifically, such as Figure 1 , Figure 3 and Figure 7 As shown, a multi-functional expansion module 21 is fixedly installed on the top of the robot body 10. The multi-functional expansion module 21 has an onboard built-in IMU sensor to detect the robot's body posture in real time. Multimedia expansion is achieved through the multi-functional expansion module 21.

[0036] Based on the above, the structure consisting of a fixed base plate 14, a multi-functional expansion module 21, a third fixed plate 39, a fourth fixed plate 40, a support plate 41, a limiting plate 42, a limiting bolt 43, a high-definition touch LCD screen 44, and a data cable 45 achieves the purpose of touch screen control of the device, improves the efficiency of the device issuing commands, expands the applicability of the device, meets the usage requirements of different brand chips and systems, and improves the convenience of using the device.

[0037] Example 3: Based on Examples 1 and 2, specifically as follows... Figure 1 , Figure 2 and Figure 4 As shown, a first electrical control box 22 is fixedly installed on the top of the robot body 10. A first limiting groove 35 is formed inside the first electrical control box 22. A second fixing plate 24 is disposed inside the first limiting groove 35. A bracket 23 is fixedly installed on one outer wall of the second fixing plate 24. A second rotating shaft 25 is rotatably installed inside the bracket 23. A triangular protective sleeve 26 is fixedly installed on the outer wall of the second rotating shaft 25. Two sets of second wheels 28 are provided on the outer wall of the robot body 10. A wheel hub 27 is fixedly installed inside the second wheel 28. A first limiting hole 34 is formed inside the wheel hub 27. The second rotating shaft 25 is rotatably installed inside the first limiting hole 34. The second fixing plate 24 is installed inside the first limiting groove 35, thus limiting the second fixing plate 24 to the inside of the first electrical control box 22.

[0038] Specifically, such as Figure 1 and Figure 2 As shown, sixteen sets of first arc-shaped fixing plates 29 are fixedly installed on the outer wall of the second wheel 28. Second arc-shaped fixing plates 30 are fixedly installed on the outer wall of the first arc-shaped fixing plates 29. A ball-collecting net rope 31 is sleeved between every two sets of first arc-shaped fixing plates 29. Three sets of limiting blocks 32 are fixedly installed on one side of the outer wall of the second wheel 28. A protective net rope 33 is sleeved between the three sets of limiting blocks 32. When the trolley moves, the second arc-shaped fixing plate 30 is in close contact with the ground, causing the second wheel 28 to roll on the outer wall of the second axle 25. The ball-collecting net rope 31 picks up the ping-pong ball, and the limiting blocks 32 and the protective net rope 33 protect the side of the second wheel 28 from falling off.

[0039] Specifically, such as Figure 1 and Figure 3 As shown, a voice control module 36 is fixedly installed on the inner bottom of the support frame 11, a first speaker 37 is fixedly installed on the top of the voice control module 36, and a set of second speakers 38 are provided on the three outer walls of the voice control module 36. The device is controlled by voice through the voice control module 36, voice commands are received through the first speaker 37, and voice feedback is broadcast through the second speakers 38. The voice control module 36 has a built-in iFlytek six-channel microphone array.

[0040] In summary, through the structure of the robot body 10, the first electrical control box 22, the bracket 23, the second fixing plate 24, the second rotating shaft 25, the triangular protective sleeve 26, the wheel hub 27, the second wheel 28, the first arc-shaped fixing plate 29, the second arc-shaped fixing plate 30, the ball-collecting net rope 31, the limiting block 32, the protective net rope 33, the first limiting hole 34, the first limiting groove 35, the voice control module 36, the first speaker 37, and the second speaker 38, the robot achieves the purpose of picking up ping-pong balls. The robot moves via voice control, employing a combination of voice recognition technology and the Cartographer mapping algorithm to accurately construct a 2D or 3D map of the sports field. Simultaneously, AI algorithms are used for path planning, and obstacles are avoided using radar or infrared sensors. This allows for the planning of local routes, improving ball-collecting efficiency, enhancing the flexibility of the device, and optimizing the user experience.

[0041] Working Principle: The customized LiDAR 13 emits infrared light for radar recognition and scanning, quickly acquiring information about the surrounding environment and detecting the approximate position and distance of the ping-pong ball and other objects on the court. This provides the robot with preliminary environmental perception data, assisting it in planning its movement path and locating the ping-pong ball. The 3D depth camera 12, fixedly mounted on the top of the support frame 11, performs real-time video analysis of the court. It can acquire the depth information of the court, construct a 3D scene model of the court, and more accurately identify the position, size, and spatial relationship of the ping-pong ball with other objects, improving the accuracy and reliability of ping-pong ball recognition. The fixed base plate 14 is fixed to the bottom of the robot body 10, providing a fixed foundation for the first connecting plate 15. After the Hall encoder reduction motor 17 on the first connecting plate 15 starts, it drives the first rotating shaft 18 to rotate. The wheels 19 on the outer wall of the first rotating shaft 18 rotate accordingly, driving the robot to move. The four-way encoder motor drive module is located inside the robot body 10 and is connected to the Hall encoder reduction motor 17 via a data cable. This allows for precise control of the motor, enabling the robot to move and turn accurately to reach the location of the ping-pong ball. A support plate 41 is fixed to one side of the base plate 14 via a third fixing plate 39 and a fourth fixing plate 40. A limiting plate 42 is fitted between the two sets of support plates 41 and fixed by four sets of limiting bolts 43. A high-definition touchscreen LCD 44 is mounted on the limiting plate 42 and connected to the inside of the robot body 10 via a data cable 45. Users can send operation commands to the robot via the high-definition touch screen 44, such as start, stop, and movement direction control. The high-definition touch screen 44 supports Raspberry Pi, NVIDIA, Ubuntu images, and secondary screens for computer systems, facilitating system operation and data display as needed. The multi-functional expansion module 21, fixed to the top of the robot body 10, has an onboard IMU sensor. The IMU sensor can detect the robot's posture in real time, such as tilt angle and acceleration, and feed this data back to the robot's control system. This allows the robot to adjust its movement strategy based on its posture, ensuring stability during movement and ball retrieval. The multi-functional expansion module 21 can also be expanded with multimedia capabilities to meet more functional requirements. A first limiting groove 35 is formed inside the first electrical control box 22 on the top of the robot body 10, and a second fixing plate 24 is installed within the first limiting groove 35 for limiting. A bracket 23 on the second fixing plate 24 rotatably mounts a second rotating shaft 25, and a triangular protective sleeve 26 is fixed to the outer wall of the second rotating shaft 25. The inner hubs 27 of the two sets of second wheels 28 on the outer wall of the robot have first limiting holes 34. The second rotating shaft 25 is rotatably installed in the first limiting holes 34 to provide rotational support for the second wheels 28, ensuring that the second wheels 28 can rotate stably and assist the robot in moving. Sixteen sets of first arc-shaped fixing plates 29 are fixed to the outer wall of the second wheels 28. Second arc-shaped fixing plates 30 are fixed to the outer wall of the first arc-shaped fixing plates 29. A ball-collecting net rope 31 is sleeved between every two sets of first arc-shaped fixing plates 29.As the robot moves, the second arc-shaped fixing plate 30 is in close contact with the ground, causing the second wheel 28 to roll on the outer wall of the second axle 25. The ball-collecting net rope 31 moves accordingly, catching ping-pong balls on the ground and collecting them. A protective net rope 33 is fitted between the three sets of limiting blocks 32 on one side of the outer wall of the second wheel 28 to prevent the collected ping-pong balls from falling off. A voice control module 36 is fixed to the bottom of the inner side of the support frame 11, with a first speaker 37 installed on its top and a second speaker 38 on each of the three outer walls. The voice control module 36 has a built-in iFlytek six-channel microphone array, which can accurately receive voice commands from the surrounding environment. After the user issues a voice command, the first speaker 37 listens to the voice command and transmits it to the voice control module 36. The voice control module 36 analyzes and processes the command, controls the robot to perform the corresponding operation, and broadcasts voice feedback through the second speaker 38 to inform the user of the operation status.

[0042] It is understood that this utility model has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of this utility model. Furthermore, under the teachings of this utility model, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of this utility model. Therefore, this utility model is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of this utility model.

Claims

1. A ball-picking device based on YOLO technology, characterized in that, include: The robot body (10) is rectangular; The intelligent recognition structure includes a first intelligent recognition unit and a second intelligent recognition unit. The first intelligent recognition unit includes a support frame (11) fixedly installed on the top of the robot body (10). A 3D depth camera (12) is installed on the top of the support frame (11). A slot is opened on the front side of the robot body (10), and a customized laser radar (13) is installed in the slot. The second intelligent recognition unit includes a fixed base plate (14) fixedly installed on the bottom of the robot body (10). A set of first connecting plates (15) are fixedly installed on both outer walls of the fixed base plate (14). A first fixed plate (16) is fixedly installed on the top of the first connecting plate (15). A Hall encoder geared motor (17) is fixedly installed on one outer wall of the first fixed plate (16). A first rotating shaft (18) is rotatably installed inside the first connecting plate (15). The output shaft of the Hall encoder geared motor (17) is fixedly connected to one end of the first rotating shaft (18).

2. The ball-picking device based on YOLO technology according to claim 1, characterized in that, The second intelligent recognition unit also includes a four-way encoder motor drive module located inside the robot body (10). A data cable is fixedly installed between the four-way encoder motor drive module and the Hall encoder reduction motor (17). A wheel (19) is fixedly installed on the outer wall of the first rotating shaft (18), and an air cushion (20) is provided on the outer wall of the wheel (19).

3. The ball-picking device based on YOLO technology according to claim 1, characterized in that, A third fixing plate (39) is fixedly installed on one side of the outer wall of the fixed base plate (14). A fourth fixing plate (40) is fixedly installed on the top of the third fixing plate (39). Two sets of support plates (41) are fixedly installed on the top of the fourth fixing plate (40). A limit plate (42) is sleeved between the two sets of support plates (41). Four sets of limit bolts (43) are spirally installed inside the limit plate (42). A high-definition touch LCD screen (44) is fixedly installed on one side of the outer wall of the limit plate (42). A data cable (45) is fixedly installed on one side of the outer wall of the high-definition touch LCD screen (44). One end of the data cable (45) is fixedly installed inside the robot body (10).

4. The ball-picking device based on YOLO technology according to claim 1, characterized in that, The top of the robot body (10) is fixedly equipped with a multi-functional expansion module (21). The multi-functional expansion module (21) has an onboard IMU sensor to detect the robot's body posture in real time.

5. The ball-picking device based on YOLO technology according to claim 1, characterized in that, The top of the robot body (10) is fixedly installed with a first electrical control box (22). The first electrical control box (22) has a first limiting groove (35) inside. The first limiting groove (35) has a second fixing plate (24) inside. A bracket (23) is fixedly installed on one side of the outer wall of the second fixing plate (24). A second rotating shaft (25) is rotatably installed inside the bracket (23). A triangular protective sleeve (26) is fixedly installed on the outer wall of the second rotating shaft (25). Two sets of second wheels (28) are provided on the outer wall of the robot body (10). A wheel hub (27) is fixedly installed inside the second wheel (28). A first limiting hole (34) is opened inside the wheel hub (27). The second rotating shaft (25) is rotatably installed inside the first limiting hole (34).

6. The ball-picking device based on YOLO technology according to claim 5, characterized in that, Sixteen sets of first arc-shaped fixing plates (29) are fixedly installed on the outer wall of the second wheel (28). Second arc-shaped fixing plates (30) are fixedly installed on the outer wall of the first arc-shaped fixing plates (29). A set of ball-collecting net rope (31) is sleeved between every two sets of first arc-shaped fixing plates (29). Three sets of limiting blocks (32) are fixedly installed on one side of the outer wall of the second wheel (28). A set of protective net rope (33) is sleeved between the three sets of limiting blocks (32).

7. The ball-picking device based on YOLO technology according to claim 1, characterized in that, A voice control module (36) is fixedly installed on the bottom inner side of the support frame (11), a first speaker (37) is fixedly installed on the top of the voice control module (36), and a set of second speakers (38) are provided on the three outer walls of the voice control module (36).