AI multi-mode intelligent shuttlecock sorting vehicle

Abstract

The invention relates to the technical field of sports equipment, and discloses an AI multi-mode intelligent badminton sorting trolley which comprises a collecting machine, a conveyor, a detecting machine, a first collecting box, a second collecting box and an intelligent sorting system, the collecting machine is arranged at the bottom of the right side of the conveyor, and the detecting machine is arranged on the left side of the conveyor. According to the AI multi-mode intelligent shuttlecock sorting vehicle, the RGB-ToF-thermal imaging multi-mode sensor fusion technology is adopted, a JetsonNano platform and an optimized YOLOv5 model are combined, high-precision and real-time recognition and positioning of shuttlecocks are achieved (the recognition accuracy reaches 90% or above, and the positioning error is smaller than or equal to 0.1 cm), the problems of complex illumination, shielding and background interference are effectively solved, and the sorting accuracy of the shuttlecocks is improved. Robustness and reliability of the system in a real training environment are remarkably improved, and low-damage conveying of badminton balls in the picking and conveying process is achieved by introducing a soft collecting mechanism with cooperation of a roller soft brush and a crawler belt and a closed-loop control strategy based on an infrared photoelectric sensor.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of sports equipment, in particular to an AI multi-modal badminton intelligent sorting vehicle. BACKGROUND

[0002] At present, badminton is one of the national sports in China, and there are more than 10 million badminton enthusiasts in China. During the badminton training process, an individual needs to consume 50-100 badminton shuttles per day. During the training or daily badminton time, picking up the badminton shuttles occupies 15%-30% of the time, which greatly consumes the training results and physical strength of athletes and ordinary badminton enthusiasts. In particular, in the training of professional badminton athletes, frequent interruption of training for picking up the badminton shuttles greatly affects the personal competition and training state of the badminton athletes.

[0003] In the badminton training scene, the problem of scattered badminton shuttles in the field is prominent. The traditional manual collection mode is time-consuming and laborious, and the recognition of the damage degree of the badminton shuttles lacks accuracy, so that a large number of badminton shuttles cannot be effectively recycled, thereby increasing the training cost and reducing the training efficiency. SUMMARY

[0004] (I) Technical problems solved In view of the shortcomings of the prior art, the application provides an AI multi-modal badminton intelligent sorting vehicle, which has the advantages of automatic sorting, multi-modal accurate recognition, efficient recycling, low badminton damage rate, intelligent path planning and remote monitoring, and solves the problems of the traditional manual collection mode being time-consuming and laborious, the recognition of the damage degree of the badminton shuttles lacking accuracy, and a large number of badminton shuttles failing to be effectively recycled, thereby increasing the training cost and reducing the training efficiency.

[0005] (II) Technical solutions In order to achieve the above-mentioned purposes of automatic sorting, multi-modal accurate recognition, efficient recycling, low badminton damage rate, intelligent path planning and remote monitoring, the application provides the following technical solutions: an AI multi-modal badminton intelligent sorting vehicle, comprising a collection machine, a conveying machine, a detection machine, a first collection box, a second collection box and an intelligent sorting system, the right bottom of the conveying machine is provided with the collection machine, the left side of the conveying machine is provided with the detection machine, and the bottom of the detection machine is provided with the first collection box and the second collection box; The detection machine includes a detection box, the left side of the conveyor is provided with a detection box, the inner side of the detection box is provided with two driving motors, the inner side of the detection box is provided with two groups of push rod pistons, the movable ends of the two groups of push rod pistons are fixedly installed with push plates, the inner side of the detection box is provided with a detection assembly, the inner side of the detection box is provided with a data line connected with the two driving motors, the outer side of the detection box is provided with a first sliding plate extending to the top of the second collection box, the outer side of the detection box is provided with a second sliding plate extending to the top of the first collection box, and the outer side of the detection box is provided with a processing assembly. The detection assembly includes a binocular RGB camera, an infrared thermal imaging sensor and a ToF sensor, the binocular RGB camera is used for capturing the color, texture and shape features of the shuttlecock, and provides basic visual information for the identification and classification of the shuttlecock, the infrared thermal imaging sensor is used for compensating for the limitations of the visual sensor under complex light or background interference, can accurately capture the small temperature difference of the shuttlecock surface and the environment background, usually ≥0.5℃, effectively detect the distance of 0.5m to 5m, and the minimum identification size is 2cm*2cm, which ensures the rapid positioning of the distribution area of the scattered shuttlecock in indoor and outdoor scenes, and the ToF sensor is used for emitting modulated light and measuring the time required for the light to be reflected back to measure the distance or depth.

[0006] Further, the collection machine adopts a collection device composed of a soft brush of a roller, which is driven to rotate by the motor shaft, and a fixed arc baffle on the device to sweep the shuttlecock to the conveyor.

[0007] Further, the conveyor is driven to rotate by the motor shaft, and then drives the square baffle to move upwards, so that the shuttlecock is transported upwards along the path to the other end of the track and enters the detection machine.

[0008] Further, the first collection box adopts a rectangular square box for storing shuttlecocks with low integrity, and the second collection box adopts a cylindrical box body for storing shuttlecocks with high integrity.

[0009] Further, the intelligent sorting system further includes a JetsonNano embedded AI computing platform connected with the detection assembly through a data line, which is used for receiving and fusing multi-modal data from the binocular RGB camera, the infrared thermal imaging sensor and the ToF sensor, performing real-time shuttlecock identification and positioning through the YOLOv5 target detection model, generating an optimal recovery path based on the ant colony algorithm, and controlling the collection machine and the conveyor to work cooperatively.

[0010] Further, the processing assembly includes an STM32 microcontroller and a PWM module, the STM32 microcontroller receives the badminton integrity analysis result from the detection assembly and outputs corresponding high and low level signals to the PWM module; the PWM module dynamically adjusts the duty cycle and frequency of the driving motor according to the signals to control the movement direction and speed of the push rod piston, realizing the sorting and guiding of good and bad badminton.

[0011] Further, the soft brush surface of the collecting machine drum is covered with a silica gel buffer layer, the bristles are arranged in a spiral shape, and the soft brush is driven by a high-torque DC speed reduction motor; infrared photoelectric sensors are embedded on both sides of the entrance of the collecting machine for real-time monitoring of the badminton entering state, and the soft brush speed is controlled by a STEM development board in a closed loop to synchronize the start and stop of the conveyor, preventing the entrance from being blocked.

[0012] Further, the detection machine outside is also provided with an adjustable brightness annular light supplementing lamp strip, which is controlled by a photosensitive sensor and environmental light intensity in linkage, and the light supplementing intensity is dynamically adjusted by PWM dimming technology to ensure that the detection assembly can obtain clear and low-noise badminton images under different lighting conditions.

[0013] (Three) beneficial effects Compared with the prior art, the present application provides an AI multi-modal badminton intelligent sorting vehicle, which has the following beneficial effects: 1. The AI multi-modal badminton intelligent sorting vehicle adopts RGB-ToF-thermal imaging multi-modal sensor fusion technology, combines JetsonNano platform and optimized YOLOv5 model, realizes high-precision and real-time identification and positioning of badminton (identification accuracy is more than 90%, positioning error is less than or equal to 0.1 cm), effectively overcomes the problems of complex light, shielding and background interference, and significantly improves the robustness and reliability of the system in the real training environment.

[0014] 2. The AI multi-modal badminton intelligent sorting vehicle introduces a soft brush and track belt cooperative soft collection mechanism, and a closed loop control strategy based on infrared photoelectric sensors, realizes low damage transmission of badminton in the picking and conveying process, solves the problem that traditional suction type equipment is easy to cause badminton damage, and improves the collection efficiency, supports continuous and multiple ball operation at the same time.

[0015] 3. The AI multi-modal badminton intelligent sorting vehicle integrates an intelligent path planning module based on ant colony algorithm, so that the vehicle can autonomously plan the optimal recovery path in the field, flexibly avoid obstacles, efficiently traverse the whole field, control the collision rate in the recovery process to be less than 5%, greatly reduce the need for manual intervention, and improve the overall operation efficiency and system autonomy.

[0016] 4、The AI multi-modal badminton intelligent sorting vehicle fuses binocular vision, LBP texture analysis and point cloud three-dimensional reconstruction technology, realizes automatic detection and classification of badminton integrity, and realizes real-time sorting of good and bad balls to different ball storage containers by means of an STM32 microcontroller and a PWM sorting mechanism, effectively reduces the cost of using balls and resource waste, and supports remote state monitoring and abnormal intervention through a Bluetooth / Wi-Fi module, enhances the maintainability and operation convenience of the system. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 It is a schematic diagram of the badminton intelligent sorting vehicle of the present application. Figure 2 It is a front view of the badminton intelligent sorting vehicle of the present application. Figure 3 It is a top view of the badminton intelligent sorting vehicle of the present application. Figure 4 It is a structure of the present application Figure 1 It is an enlarged view of point A in the structure of the present application. Figure 5 It is a schematic diagram of the detection machine of the present application. Figure 6 It is a top view of the detection machine of the present application. Figure 7 It is a schematic diagram of the detection assembly of the present application. Figure 8 It is a flow chart of the detection process of the badminton intelligent sorting vehicle of the present application. Figure 9 It is a flow chart of the identification, path finding and grabbing process of the badminton intelligent sorting vehicle of the present application. Figure 10 It is a flow chart of the badminton integrity detection process of the present application.

[0018] In the figure: 1, collecting machine; 2, conveying machine; 3, detection machine; 301, detection box; 302, driving motor; 303, push rod piston; 304, push plate; 305, detection assembly; 3051, binocular RGB camera; 3052, infrared thermal imaging sensor; 3053, ToF sensor; 306, data line; 307, first sliding plate; 308, second sliding plate; 309, processing assembly; 4, first collecting box; 5, second collecting box. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0020] Please refer to Figures 1-10 , AI multi-modal badminton intelligent sorting vehicle, including collecting machine 1, conveyor 2, detection machine 3, first collecting box 4, second collecting box 5 and intelligent sorting system, the right bottom of the conveyor 2 is provided with the collecting machine 1, the left side of the conveyor 2 is provided with the detection machine 3, the bottom of the detection machine 3 is placed with the first collecting box 4 and the second collecting box 5; The detection machine 3 includes a detection box 301, the left side of the conveyor 2 is provided with the detection box 301, the inner side of the detection box 301 is provided with two drive motors 302, the inner side of the detection box 301 is provided with two groups of push rod pistons 303, the movable ends of the two groups of push rod pistons 303 are fixedly installed with push plates 304, the inner side of the detection box 301 is provided with a detection assembly 305, the inner side of the detection box 301 is provided with a data line 306 connected with the two drive motors 302, the outer side of the detection box 301 is provided with a first sliding plate 307 extending to the top of the second collecting box 5, the outer side of the detection box 301 is provided with a second sliding plate 308 extending to the top of the first collecting box 4, and the outer side of the detection box 301 is provided with a processing assembly 309; The detection assembly 305 includes a binocular RGB camera 3051, an infrared thermal imaging sensor 3052 and a ToF sensor 3053, the binocular RGB camera 3051 is used for capturing the color, texture and shape features of the badminton, providing basic visual information for the identification and classification of the badminton, the infrared thermal imaging sensor 3052 is used for compensating for the limitations of the visual sensor under complex lighting or background interference. It can accurately capture the small temperature difference between the surface of the badminton and the environment background, usually ≥0.5℃, the effective detection distance covers 0.5m to 5m, and the minimum identification size is 2cm×2cm, which ensures the rapid positioning of the distribution area of the scattered badminton in indoor and outdoor scenes, and the ToF sensor 3053 is used for emitting modulated light such as continuous wave CW method and measuring the time required for these lights to be reflected back to measure the distance or depth.

[0021] In the case implementation, the collecting machine 1 adopts a collecting device composed of a roller soft brush, which is driven to rotate by the motor shaft, and the fixed arc baffle on the device sweeps the badminton to the conveyor 2, the surface of the roller soft brush of the collecting machine 1 is covered with a silica gel buffer layer, the bristles are arranged in a spiral shape, and are driven by a high-torque DC reduction motor; infrared photoelectric sensors are embedded on both sides of the inlet of the collecting machine 1, which are used to monitor the entering state of the badminton in real time, and the soft brush rotating speed is controlled by the STEM development board in a closed loop, and the starting and stopping of the conveyor 2 is synchronized, preventing the inlet from being blocked.

[0022] Wherein, the rotating speed of the soft brush and the starting and stopping of the track are cooperatively controlled by a central controller according to the signals of the inlet sensors; Specifically, the infrared photoelectric sensors located on both sides of the collection machine's entrance detect in real time whether there are shuttlecocks entering, and when the sensors detect that a shuttlecock has entered the collection area, they immediately send a signal to the STEM development board (central controller). The development board dynamically adjusts the speed of the high-torque DC reduction motor driving the soft brush of the roller through a PWM signal, ensuring that the soft brush gently sweeps the shuttlecock in at the optimal speed (e.g., 60-80 RPM). At the same time, the signal synchronously triggers the track motor of the conveyor_2 to start, ensuring that the shuttlecock is immediately caught by the square baffle and transported upwards after being swept in, achieving seamless connection between picking and conveying and avoiding accumulation or blockage at the entrance. Through the closed-loop cooperation control of the above-mentioned sensor feedback and motor, the system realizes real-time response to the entering state of the shuttlecock and process synchronization, not only significantly improving the collection efficiency, but also ensuring the smoothness and integrity of the shuttlecock during the transfer process, avoiding mechanical jamming or damage.

[0023] In the case implementation, the conveyor 2 rotates the track by rotating the motor shaft, and then drives the square baffle to move upwards, transporting the shuttlecock along the path upwards to the other end of the track and into the detection machine 3.

[0024] Among them, the running speed of the track can be adaptively adjusted by the path planning module of the intelligent sorting system according to the density of the shuttlecock distribution in the field; Specifically, the intelligent sorting system (running on the JetsonNano platform) judges the density of the shuttlecock distribution in the current target area in real time according to the recycling path planned by the multi-modal sensor fusion data and the ant colony algorithm, and when the system identifies a dense area of shuttlecocks, it will output a higher duty cycle signal to the track drive motor of the conveyor_2 through the PWM module, increasing the running speed of the track (e.g., by 20%) to match the faster collection rhythm at the front end, preventing the transportation channel from being congested, and vice versa, reducing the speed in sparse areas to save energy. At the same time, the inclination angle of the square baffle and the texture of the track are designed to ensure that the shuttlecock remains stable during the climbing and transportation process and does not slip or roll over. By dynamically coupling the running speed of the conveyor with the global recycling strategy, the system realizes adaptive optimization of the whole process from identification, collection to transportation, further improving the throughput and energy efficiency ratio of the overall sorting operation.

[0025] In the case implementation, the first collection box 4 is a rectangular box used to store shuttlecocks with low integrity, and the second collection box 5 is a cylindrical box used to store shuttlecocks with high integrity.

[0026] Among them, the first collection box 4 and the second collection box 5 are both equipped with independent capacity sensors and detachable structures, and the box body design takes into account the physical properties of the shuttlecock accumulation and the convenience of storage and retrieval. Specifically, the first collecting box 4 is designed with a larger opening and a flat inner wall for accommodating badminton shuttlecocks with low integrity due to broken feathers, deformed heads, etc. These irregularly shaped balls are more stable when placed flat. The second collecting box 5 has an arc-shaped inner wall and a vertical cylindrical structure, optimizing the storage space for high-integrity badminton shuttlecocks (regular shape), allowing them to naturally arrange in order based on their conical structure, improving space utilization. Each collecting box is equipped with a pressure or infrared photoelectric sensor to monitor the quantity of badminton shuttlecocks in real time. When the quantity reaches a preset threshold (e.g., 80%), the system will notify the user through sound and light prompts or wireless signals to replace it in time. Both collecting boxes are designed with a buckle or sliding rail for easy disassembly, emptying, and resetting by the user. Through this targeted design based on the physical state differences of badminton shuttlecocks, combined with inventory monitoring and user-friendly disassembly structure, the system not only effectively separates and manages good and bad badminton shuttlecocks, but also greatly improves the convenience of equipment maintenance and user experience, ensuring the continuity and efficiency of the sorting process.

[0027] In the case implementation, the intelligent sorting system also includes a JetsonNano embedded AI computing platform connected to the detection component 305 through a data line 306 for receiving and fusing multi-modal data from the binocular RGB camera 3051, infrared thermal imaging sensor 3052, and ToF sensor 3053. The YOLOv5 target detection model is used for real-time badminton recognition and positioning, and the ant colony algorithm is used to generate the optimal recovery path to control the cooperation of the collecting machine 1 and the transport machine 2.

[0028] The fusion of multi-modal data uses a combination of feature-level and decision-level fusion strategies, and an adaptive light compensation mechanism is set to cope with changes in environmental lighting. Specifically, the system first performs feature-level fusion: the color and texture features (analyzed by LBP) extracted by the binocular RGB camera, the accurate three-dimensional point cloud depth information obtained by the ToF sensor, and the temperature features captured by the infrared thermal imaging sensor are mapped to the same spatiotemporal coordinate system through calibration and registration on the JetsonNano platform, generating an enhanced feature vector containing visual, geometric, and thermal radiation information. Then, decision-level fusion is performed: the YOLOv5 model performs target detection and classification based on the fused features, and the ant colony algorithm receives the coordinates of all identified badminton shuttlecocks, combines the current position of the car and obstacle information (from the ToF sensor), and calculates the globally optimal recovery path. To ensure the quality of sensor data, the system monitors the ambient light intensity through a photosensitive sensor. When it falls below a threshold, it automatically triggers a ring-shaped light compensation strip and dynamically adjusts the brightness through PWM dimming technology to ensure clear RGB images with low noise. Through the above multi-level data fusion and adaptive environment perception mechanism, the system significantly improves the perception accuracy and robustness of the target in a complex and dynamic badminton training field, providing a reliable data foundation for efficient autonomous navigation and recovery operations.

[0029] In the case implementation, the processing component 309 includes an STM32 microcontroller and a PWM module. The STM32 microcontroller receives the badminton integrity analysis results from the detection component 305 and outputs corresponding high and low level signals to the PWM module. The PWM module dynamically adjusts the duty cycle and frequency of the driving motor 302 according to the signals to control the movement direction and speed of the push rod piston 303, realizing the sorting and guiding of good and bad shuttlecocks.

[0030] The integrity analysis results are derived from a three-dimensional model analysis process based on point cloud comparison and ICP algorithm, and the PWM module integrates a PID control algorithm to realize precise adjustment of the motor drive. Specifically, the detection component 305 performs three-dimensional scanning on the shuttlecock entering the detection cabin to generate its point cloud model. The algorithm in the processing component 309 compares the point cloud with a pre-stored "intact shuttlecock" standard three-dimensional model using the Iterative Closest Point (ICP) algorithm. The distance deviation between the points is calculated to quantitatively evaluate the integrity indicators such as feather damage and head deformation. The classification signal of "good ball" or "bad ball" is output. The STM32 microcontroller receives this signal and converts it into a high or low level instruction to control the action of a certain push rod piston. The PWM module then calls the pre-set PID control algorithm to dynamically calculate and output the optimal PWM duty cycle and frequency required for the driving motor (for example, high speed and high duty cycle for the good ball channel, and low speed and low duty cycle for the bad ball channel), thereby accurately controlling the extension speed and force of the push rod piston to stably push the shuttlecock into the corresponding slide channel. By introducing three-dimensional model comparison for objective quality evaluation and combining closed-loop PID control to execute sorting actions, the system realizes high accuracy, high consistency, and soft and precise actions in sorting good and bad shuttlecocks, effectively replacing subjective human judgment and avoiding secondary damage during the sorting process.

[0031] In the case implementation, the detection machine 3 also has an adjustable brightness ring-shaped light supplementing lamp strip outside. It is controlled by a photosensitive sensor and environmental light intensity, dynamically adjusts the light supplementing intensity through PWM dimming technology, and ensures that the detection component 305 can obtain clear and low-noise shuttlecock images under different lighting conditions.

[0032] The ring-shaped light supplementing lamp strip uses multiple high color rendering index (CRI>90) LED lamp beads arranged in a ring and integrated around the sensor module. The brightness adjustment follows a pre-set environmental light-light supplementing mapping curve. Specifically, the light-sensitive sensor collects the ambient light intensity (unit: Lux) in real time and transmits it to the central processor (JetsonNano). The built-in light compensation control algorithm in the processor calculates the ideal light compensation intensity required to offset the lack of ambient light and optimize image contrast based on the pre-set brightness mapping curve. Then, the algorithm generates a PWM signal with a corresponding duty cycle through a general-purpose timer (such as TIMx), which controls the current of the LED beads in the ring-shaped light strip through the LED driving circuit, thereby achieving stepless dimming. The mapping curve is designed to provide strong light compensation in low-light environments (such as <200Lux) to eliminate shadows and noise, while maintaining minimal light compensation or turning off in normal lighting (such as >500Lux) to save energy and avoid overexposure. Through this closed-loop PWM dimming control based on real-time feedback of ambient light, the system can automatically provide stable and uniform auxiliary lighting for the binocular RGB camera under various complex lighting conditions such as early morning, evening, cloudy day, or uneven indoor lighting, thereby fundamentally ensuring the input data quality of subsequent image recognition, texture analysis, and three-dimensional point cloud reconstruction, which is a key basic link to improve the recognition accuracy and robustness of the entire system.

[0033] In implementation, the following steps are taken: 1) First, start the system and initialize each module: turn on the device power, start the JetsonNano embedded AI computing platform, load the multi-modal sensor (binocular RGB camera 3051, ToF sensor 3053, infrared thermal imaging sensor 3052) driver, complete the space-time calibration and data synchronization of each sensor, at the same time, initialize the PWM module, motor drive and ring-shaped light compensation light strip control unit of the STM32 microcontroller, the system performs self-checking, confirms that the collection machine 1, transport machine 2, detection machine 3 and each sensor are working normally, and enters the standby state; 2) Then, autonomous navigation and shuttlecock collection: the JetsonNano platform fuses multi-modal sensor data, runs the YOLOv5 model to recognize and locate the scattered shuttlecocks in the field in real time, generates the three-dimensional coordinates of each ball, plans the current optimal global recycling path based on the ant colony algorithm, and controls the mobile chassis to travel along the planned path. When the infrared photoelectric sensor detects that the shuttlecock enters the collection area, the soft brush is triggered at a preset speed to gently sweep the ball into the detection machine 3 inlet, and the transport machine 2 caterpillar belt is started to smoothly transport the shuttlecock upward to the detection machine 3 inlet; 3) Complete detection and intelligent sorting of badminton: After the badminton enters the closed cabin of the detection machine 3, the binocular RGB camera 3051 and the ToF sensor 3053 synchronously collect the RGB image and depth information thereof, the JetsonNano platform accurately analyzes the completeness of the badminton by means of LBP texture analysis and point cloud three-dimensional reconstruction technology, combined with the fast ICP algorithm and the standard intact model, and classifies it as "good ball" or "bad ball", and the classification result is sent to the STM32 in the form of electrical signal, and the STM32 controls the corresponding PWM output according to the signal to drive the electric push rod piston 303 of a specific side to act, push the badminton into the corresponding slide, and finally fall into the first collection box 4 or the second collection box 5; 4) Finally complete the work and enter standby or cycle mode: when the system judges that the badminton in the current area has been recycled or reaches the preset working time through the sensor or user instruction, the moving chassis returns to the charging pile or standby point, and the system can send the work data (such as the number of recycled balls, classification statistics, and path efficiency) to the user terminal through the Bluetooth / Wi-Fi module, and if it is set to cycle mode, the trolley will start scanning and path planning again after a short pause, and enter the next round of collection and sorting work until the task is completed.

[0034] In summary, the AI multi-modal badminton intelligent sorting vehicle realizes high-precision and real-time identification and positioning of badminton (identification accuracy is more than 90%, positioning error is less than or equal to 0.1 cm) by adopting RGB-ToF-thermal imaging multi-modal sensor fusion technology, combined with the JetsonNano platform and the optimized YOLOv5 model, effectively overcoming the problems of complex light, occlusion and background interference, significantly improving the robustness and reliability of the system in the real training environment, introducing the soft brush and track belt cooperative gentle collection mechanism, and the closed-loop control strategy based on infrared photoelectric sensor, realizing low-damage transmission of badminton in the picking and conveying process, solving the problem of damage to badminton caused by traditional suction type equipment, and improving the collection efficiency, supporting continuous and multiple ball operation at the same time.

[0035] In addition, by integrating the intelligent path planning module based on ant colony algorithm, the trolley can autonomously plan the optimal recovery path in the field, flexibly avoid obstacles, and efficiently traverse the entire field, with a collision rate of less than 5% in the recovery process, greatly reducing the need for manual intervention, improving the overall operation efficiency and system autonomy, realizing automatic detection and classification of badminton completeness by fusing binocular vision, LBP texture analysis and point cloud three-dimensional reconstruction technology, and real-time sorting of good and bad balls to different ball storage containers by means of STM32 microcontroller and PWM sorting mechanism, effectively reducing the cost of using balls and resource waste, while supporting remote state monitoring and abnormal intervention through Bluetooth / Wi-Fi module, enhancing the maintainability and operation convenience of the system.

[0036] It has to be noted that, in the present document, the terms "first", "second", etc. merely serve to identify a subject or action, without necessarily requiring or implying any such actual relationship or order between such subjects or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0037] While embodiments of the application have been shown and described, it is to be understood that the application is not limited to the details of the embodiments described, since numerous further modifications and changes can be apparent to one skilled in the art without departing from the spirit and scope of the application, which is defined by the appended claims and their equivalents.

Claims

1. An AI multimodal badminton shuttlecock intelligent sorting vehicle, comprising a collector (1), a transport vehicle (2), an inspection machine (3), a first collection box (4), a second collection box (5), and an intelligent sorting system, characterized in that: A collector (1) is provided on the bottom right side of the transport machine (2), and a detector (3) is provided on the left side of the transport machine (2). A first collection box (4) and a second collection box (5) are placed at the bottom of the detector (3). The detection machine (3) includes a detection box (301). The detection box (301) is located on the left side of the transport machine (2). Two drive motors (302) are located inside the detection box (301). Two sets of push rod pistons (303) are located inside the detection box (301). Push plates (304) are fixedly installed on the movable ends of the two sets of push rod pistons (303). A detection component (305) is located inside the detection box (301). A data cable (306) connected to the two drive motors (302) is located inside the detection box (301). A first sliding plate (307) extending to the top of the second collection box (5) is located outside the detection box (301). A second sliding plate (308) extending to the top of the first collection box (4) is located outside the detection box (301). A processing component (309) is located outside the detection box (301). The detection component (305) includes a binocular RGB camera (3051), an infrared thermal imaging sensor (3052), and a ToF sensor (3053). The binocular RGB camera (3051) is used to capture the color, texture, and shape features of badminton shuttlecocks, providing basic visual information for the identification and classification of badminton shuttlecocks. The infrared thermal imaging sensor (3052) is used to compensate for the limitations of visual sensors under complex lighting or background interference. It can accurately capture the small temperature difference between the surface of the badminton shuttlecock and the background environment, usually ≥0.5℃, with an effective detection distance covering 0.5m to 5m and a minimum recognition size of 2cm×2cm, ensuring rapid location of the distribution area of ​​scattered badminton shuttlecocks in indoor and outdoor scenes. The ToF sensor (3053) is used to emit modulated light (such as the continuous wave CW method) and measure the time required for this light to be reflected back to measure distance or depth.

2. The AI ​​multimodal badminton shuttlecock intelligent sorting vehicle according to claim 1, characterized in that: The collecting machine (1) is a collecting device composed of roller brushes. The brushes are rotated by the rotation of the motor shaft, and the shuttlecocks are swept to the transport machine (2) by the fixed arc baffle on the device.

3. The AI ​​multimodal badminton shuttlecock intelligent sorting vehicle according to claim 1, characterized in that: The transport machine (2) drives the track to rotate by the rotation of the motor shaft, which in turn drives the square baffle to move up and transport the badminton shuttlecock along the path to the other end of the track and into the inspection machine (3).

4. The AI ​​multimodal badminton shuttlecock intelligent sorting vehicle according to claim 1, characterized in that: The first collection box (4) is a rectangular box used to store badminton shuttlecocks with low integrity, and the second collection box (5) is a cylindrical box used to store badminton shuttlecocks with high integrity.

5. The AI ​​multimodal badminton shuttlecock intelligent sorting vehicle according to claim 1, characterized in that: The intelligent sorting system also includes a JetsonNano embedded AI computing platform, which is connected to the detection component (305) via a data cable (306). It is used to receive and fuse multimodal data from a binocular RGB camera (3051), an infrared thermal imaging sensor (3052), and a ToF sensor (3053). It performs real-time badminton shuttlecock identification and positioning through the YOLOv5 target detection model and generates the optimal recycling path based on the ant colony algorithm, controlling the collector (1) and the transport machine (2) to work together.

6. The AI ​​multimodal badminton shuttlecock intelligent sorting vehicle according to claim 1, characterized in that: The processing component (309) includes an STM32 microcontroller and a PWM module. The STM32 microcontroller receives the badminton shuttlecock integrity analysis result from the detection component (305) and outputs the corresponding high and low level signals to the PWM module. The PWM module dynamically adjusts the duty cycle and frequency of the drive motor (302) according to the signal to control the movement direction and speed of the push rod piston (303) and realize the sorting guidance of good and bad badminton shuttlecocks.

7. The AI ​​multimodal badminton shuttlecock intelligent sorting vehicle according to claim 3, characterized in that: The roller brush of the collector (1) is covered with a silicone buffer layer, the bristles are arranged in a spiral shape and driven by a high-torque DC geared motor; infrared photoelectric sensors are embedded on both sides of the inlet of the collector (1) to monitor the badminton shuttlecock entering the machine in real time, and the brush speed is controlled in a closed loop by the STEM development board to synchronize with the start and stop of the transport machine (2) to prevent the inlet from being blocked.

8. The AI ​​multimodal badminton shuttlecock intelligent sorting vehicle according to claim 4, characterized in that: The detector (3) is also equipped with an adjustable brightness ring light strip on the outside. It is controlled by a photosensitive sensor in conjunction with the ambient light intensity. The light intensity is dynamically adjusted by PWM dimming technology to ensure that the detector component (305) can obtain clear, low-noise badminton images under different lighting conditions.

Images