Optical fiber winding defect real-time detection system based on embedded development platform

Abstract

The invention discloses an optical fiber winding defect real-time detection system based on an embedded development platform, which relates to the technical field of optical fiber winding and detection and comprises the embedded development platform, a motor module, a tension acquisition module, an image acquisition module, an optical fiber path auxiliary module and a pay-off and take-up module. According to the method, the YOLOv3-tiny algorithm based on deep learning is adopted, rapid classification and positioning of the optical fiber winding defects can be achieved, the detection time is within 200 ms, the high requirement of an optical fiber winding machine for real-time performance is met, hardware acceleration of the YOLOv3-tiny algorithm is achieved through optimized deployment on an embedded development platform, and compared with a CPU and a GPU, the method has the advantages that the hardware speed is high, and the cost is low. When the algorithm is operated by an embedded development platform, the energy efficiency can be improved by 5.2 times and 2.1 times. The calculation efficiency is improved, and the cost and power consumption are reduced.

Description

Technical Field

[0001] The present invention relates to the field of optical fiber winding and detection technology, and in particular to a real-time optical fiber winding defect detection system based on an embedded development platform. Background Art

[0002] As fiber optic gyroscopes (FOGs) continue to improve performance, the quality requirements for fiber loops are becoming more stringent. Traditional methods for detecting fiber loop defects primarily rely on manual inspection or simple automated equipment. These methods suffer from limitations in accuracy and speed, making them unable to meet the increasing performance demands of modern FOGs. With the development of deep learning technology, deep learning-based object detection algorithms have demonstrated outstanding performance in industrial product quality inspection, particularly in fiber loop defect detection. The YOLO algorithm has become a research hotspot due to its exceptional detection speed, robustness, and accuracy. However, the convolutional neural networks used in deep learning algorithms are computationally intensive and typically require deployment on resource-rich PC platforms using CPUs or GPUs. Deploying deep learning models on resource-constrained embedded devices remains challenging, particularly in fiber loop defect detection applications that require real-time processing of large amounts of video stream data. Therefore, a real-time fiber loop defect detection system based on an embedded development platform is proposed. Summary of the Invention

[0003] In view of the deficiencies in the prior art, the present invention provides a real-time optical fiber loop defect detection system based on an embedded development platform, which solves the problems raised in the above background technology.

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: a real-time optical fiber loop defect detection system based on an embedded development platform, comprising an embedded development platform, a motor module, a tension acquisition module, an image acquisition module, an optical fiber path auxiliary module, and a pay-off and take-up module;

[0005] The embedded development platform includes an ARM processing module and an FPGA hardware acceleration module. The ARM processor is responsible for executing complex control processes and advanced processing tasks, and the FPGA hardware acceleration module is responsible for the operation of convolutional neural networks while maintaining low power consumption and high efficiency.

[0006] The motor module includes a motor and a PID controller. The motor is responsible for driving the optical fiber winding device to ensure precise positioning and uniform speed during the winding process. The PID controller dynamically adjusts the motor speed based on the difference between the actual tension feedback value and the preset target value to maintain stable optical fiber tension. Through the motor control module, winding defects caused by tension fluctuations can be effectively avoided.

[0007] The tension acquisition module includes a tension sensing module and a signal transmission module. The tension sensing module is responsible for directly measuring the actual tension generated by the optical fiber during the winding process and providing an accurate analog signal output. The signal transmission module converts the collected analog signal into a digital signal, which is processed by the ARM processing module on the embedded development platform to ensure that the optical fiber tension remains within the set range and avoid defects in the optical fiber ring caused by excessive or insufficient tension.

[0008] The image acquisition module includes a CMOS image sensor and an image preprocessing unit. The CMOS image sensor is responsible for capturing high-definition real-time images during the optical fiber winding process, providing sufficient resolution and frame rate to ensure image clarity. The raw data collected by the camera equipped with the CMOS image sensor is adjusted by the image preprocessing unit to match the input requirements of the YOLOv3-tiny algorithm.

[0009] The optical fiber path auxiliary module assists in ensuring the correct trajectory of the optical fiber winding line, helps guide the optical fiber to move along the predetermined path, and achieves a precise winding process;

[0010] The pay-off and take-up modules are used to realize the pay-off and take-up functions of the optical fiber.

[0011] Optionally, the CMOS image sensor sends information, the ARM processor receives and analyzes the information, and is also responsible for transmitting the information to the pay-out and take-up modules to control the precise winding of the optical fiber ring.

[0012] Optionally, the equipment carried by the system includes a fiber optic winding platform and an optical imaging platform. The fiber optic winding platform is mainly responsible for winding the optical fiber, while the optical imaging platform is responsible for ensuring that the system can obtain a clear optical fiber contour image. The two platforms cooperate with each other to provide a reliable experimental basis for the testing of the real-time detection system for fiber optic winding defects.

[0013] The optical fiber looping platform includes a pay-out device, an optical fiber path auxiliary device, a tension detector, a take-up device and an FPGA hardware accelerator;

[0014] The optical imaging platform includes an optical imaging module;

[0015] The pay-off device is located in front of the FPGA hardware accelerator, the tension detector is located in front of the pay-off device, an optical fiber path auxiliary device is placed in front of the tension detector, a take-up device is placed in front of the optical fiber path auxiliary device, and an optical imaging module is placed in front of the take-up device.

[0016] Optionally, the pay-off device includes a pay-off motor, a pay-off wheel, a first coupling and a first bracket, a first coupling is installed on one side of the first bracket, a pay-off wheel is installed on one end of the first coupling, and a pay-off motor is installed on the other side of the first bracket, and the output end of the pay-off motor is connected to one end of the pay-off wheel.

[0017] Optionally, the tension detector includes a tension sensor and a signal transmitter, a guide wheel and a mounting frame, the signal transmitter is mounted on the mounting frame, the tension sensor is mounted on the mounting frame, the tension detector is located behind the pay-off device, and guide wheels are placed on both sides of the mounting frame.

[0018] Optionally, the fiber path assist device includes a stepper motor, a ball screw, and a screw slide. The ball screw is rotatably connected to the interior of the screw slide. A stepper motor is fixedly connected to one side of the screw slide. The output end of the stepper motor is fixedly connected to one end of the ball screw. The outer thread of the ball screw is connected to a movable cable mechanism. Every 1° of ball screw rotation is equivalent to a linear movement of approximately 11μm. The stepper motor has a step angle of 1.8°, and a single pulse can move the ball screw a distance of 1.8×11=19.8μm, which meets the requirements of the fiber loop platform.

[0019] Optionally, a dancing wheel, a spring and a second bracket are placed behind the screw slide, the dancing wheel is rotatably connected to the second bracket, a spring is installed on the second bracket, and one end of the spring is connected to the outer side of the dancing wheel.

[0020] Optionally, the take-up device includes a take-up motor, a take-up wheel, a second coupling, and a third bracket. The second coupling is mounted on the third bracket. A take-up motor is fixedly connected to one side of the third bracket. The output end of the take-up motor is fixedly connected to the take-up wheel. One end of the take-up wheel is connected to the second coupling. The take-up motor and the pay-off motor are used to wind the optical fiber around the optical fiber ring. The pay-off motor is used to release the optical fiber from the optical fiber reel. The guide wheel is used to ensure the trajectory of the winding line. The tension sensor is used to detect the current winding tension. A signal transmitter is required to convert the analog data of the tension sensor into a digital value that can be processed by the heterogeneous embedded development platform. The heterogeneous embedded development platform uses a PID algorithm to adjust the motor speed according to the current tension value to achieve a speed regulation function.

[0021] Optionally, the optical imaging module includes a microscope lens, a lifting bracket and a light source. The microscope lens is installed on the top of the lifting bracket. The light source provides sufficient light for the imaging system to illuminate the surface of the fiber optic ring and produce a clear image. Its selection and adjustment are crucial to obtaining high-quality imaging results.

[0022] Optionally, the FPGA hardware accelerator is located behind the pay-off device. The FPGA hardware accelerator accelerates the convolution operation, that is, accelerates the YOLOv3-tiny algorithm of the defect detection module to achieve real-time classification and positioning of optical fiber winding defects.

[0023] The present invention provides a real-time optical fiber loop defect detection system based on an embedded development platform, which has the following beneficial effects:

[0024] 1. This paper uses the deep learning-based YOLOv3-tiny algorithm to rapidly classify and locate fiber winding defects, with a detection time of less than 200ms, meeting the high real-time requirements of fiber winding machines. Through optimized deployment on an embedded development platform, this paper implements hardware acceleration of the YOLOv3-tiny algorithm. Compared to CPUs and GPUs, the embedded development platform can achieve a 5.2x and 2.1x increase in energy efficiency when running this algorithm, respectively. This improves computational efficiency and reduces costs and power consumption.

[0025] 2. The present invention deploys the deep learning model on an embedded device, overcoming the challenges of deploying traditional deep learning models on resource-constrained devices. This deployment practice improves the flexibility and scalability of the system, and can also adapt to different winding requirements and conditions. Through PID algorithm and tension control, the system has better adaptability and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic diagram of the overall structure of the present invention;

[0027] Figure 2 A schematic diagram of a pay-off device according to an example of the present invention;

[0028] Figure 3 This is a schematic diagram of the tension detector module;

[0029] Figure 4 Schematic diagram of the guide wheels on both sides of the tension sensor;

[0030] Figure 5 This is a schematic diagram of the fiber optic path auxiliary module;

[0031] Figure 6 A schematic diagram of a dance wheel;

[0032] Figure 7 A schematic diagram of a wire take-up device according to an example of the present invention;

[0033] Figure 8 Schematic diagram of the camera for real-time monitoring of the loop situation;

[0034] Figure 9Schematic diagram of FPGA hardware accelerator.

[0035] In the figure: 100, pay-off device; 110, pay-off motor; 120, pay-off wheel; 130, first coupling; 140, first bracket; 200, optical imaging module; 210, microscope lens; 220, lifting bracket; 300, optical fiber path auxiliary device; 310, stepping motor; 320, ball screw; 330, screw slide; 340, dancing wheel; 341, spring; 342, second bracket; 400, tension detector; 410, tension sensor; 420, signal transmitter; 430, guide wheel; 440, mounting bracket; 500, take-up device; 510, take-up motor; 520, take-up wheel; 530, second coupling; 540, third bracket; 600, FPGA hardware accelerator. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0037] The present invention provides a technical solution: a real-time optical fiber loop defect detection system based on an embedded development platform, comprising an embedded development platform, a motor module, a tension acquisition module, an image acquisition module, an optical fiber path auxiliary module, and a pay-off and take-up module;

[0038] Embedded development platform, including ARM processing module and FPGA hardware acceleration module. The ARM processor is responsible for executing complex control processes and advanced processing tasks, and the FPGA hardware acceleration module is responsible for the operation of convolutional neural networks while maintaining low power consumption and high efficiency.

[0039] The motor module includes a motor and a PID controller. The motor is responsible for driving the fiber winding device to ensure precise positioning and uniform speed during the winding process. The PID controller dynamically adjusts the motor speed based on the difference between the actual tension feedback value and the preset target value to maintain stable fiber tension. Through the motor control module, winding defects caused by tension fluctuations can be effectively avoided.

[0040] The tension acquisition module includes a tension sensing module and a signal transmission module. The tension sensing module is responsible for directly measuring the actual tension generated by the optical fiber during the winding process and providing accurate analog signal output. The signal transmission module converts the collected analog signal into a digital signal, which is then processed by the ARM processing module on the embedded development platform to ensure that the optical fiber tension remains within the set range and avoid defects in the optical fiber loop caused by excessive or insufficient tension.

[0041] The image acquisition module includes a CMOS image sensor and an image preprocessing unit. The CMOS image sensor is responsible for capturing high-definition real-time images of the fiber winding process, providing sufficient resolution and frame rate to ensure image clarity. The raw data collected by the camera equipped with the CMOS image sensor is adjusted by the image preprocessing unit to match the input requirements of the YOLOv3-tiny algorithm.

[0042] Fiber path auxiliary module, which assists in the correct trajectory of the fiber winding line, helps guide the fiber to move along the predetermined path, and achieves a precise winding process;

[0043] The pay-off and take-up module is used to realize the pay-off and take-up functions of the optical fiber.

[0044] The equipment carried by the system includes a pay-out device 100, an optical imaging module 200, a fiber path auxiliary device 300, a tension detector 400, a take-up device 500, and an FPGA hardware accelerator 600. The pay-out device 100 is located in front of the FPGA hardware accelerator 600, the tension detector 400 is located in front of the pay-out device 100, the fiber path auxiliary device 300 is placed in front of the tension detector 400, the take-up device 500 is placed in front of the fiber path auxiliary device 300, and the optical imaging module 200 is placed in front of the take-up device 500.

[0045] The function of the take-up device 500 is to wind the optical fiber on the optical fiber ring, while the function of the pay-out device 100 is to release the optical fiber on the optical fiber reel. The optical fiber released by the pay-out device 100 moves along a predetermined path through the optical fiber path auxiliary device 300 to ensure the trajectory of the loop line and realize the wiring function. The tension detector 400 detects the tension of the loop. In the meantime, the tension sensor 410 is required to measure the actual tension. The signal transmitter 420 then converts the analog data of the tension sensor 410 into a digital quantity that can be processed by the embedded development platform. The embedded development platform uses an algorithm to adjust the motor speed according to the current tension value, thereby realizing the speed regulation function. The function of the optical imaging module 200 is to obtain clear images in real time. The FPGA hardware accelerator 600 is responsible for the operation of the convolutional neural network, and is combined with the ARM processor to form an embedded development platform to realize the real-time classification and positioning of light winding defects;

[0046] Depend on Figure 2As shown, the pay-off device 100 includes a pay-off motor 110, a pay-off wheel 120, a first coupling 130 and a first bracket 140. The first coupling 130 is installed on one side of the first bracket 140, the pay-off wheel 120 is installed at one end of the first coupling 130, and the pay-off motor 110 is installed on the other side of the first bracket 140. The output end of the pay-off motor 110 is connected to one end of the pay-off wheel 120. The rotation of the pay-off motor 110 will drive the pay-off wheel 120 to operate synchronously, thereby realizing the pay-off function of the optical fiber.

[0047] Depend on Figure 3 As shown, the tension detector 400 includes a tension sensor 410, a signal transmitter 420, guide wheels 430, and a mounting bracket 440. The signal transmitter 420 is mounted on the mounting bracket 440, and the tension sensor 410 is mounted on the mounting bracket 440. The tension detector 400 is located behind the pay-off device 100, with guide wheels 430 placed on both sides of the mounting bracket 440. The tension sensor 410 measures the tension on the optical fiber in real time. The tension value data signal it emits is an analog quantity, which is then converted into a digital quantity by the rear signal transmitter 420 and transmitted to the embedded development platform. The two guide wheels 430 ensure the trajectory of the loop line.

[0048] Depend on Figure 5 As shown, the optical fiber path auxiliary device 300 includes a stepper motor 310, a ball screw 320 and a screw slide 330. The ball screw 320 is rotatably connected inside the screw slide 330, and one side of the screw slide 330 is fixedly connected to the stepper motor 310. The output end of the stepper motor 310 is fixedly connected to one end of the ball screw 320. The outer side of the ball screw 320 is threadedly connected to a movable wiring mechanism, whose function is to move the wiring mechanism to complete the optical fiber wiring during the optical fiber winding process. After each turn of winding is completed, the stepper motor 310 sends a pulse, which can move the ball screw 320 forward on the screw slide 330 by a distance of an optical fiber diameter, and then start winding the next turn, and repeat this cycle.

[0049] Depend on Figure 6 As shown, a dancing wheel 340, a spring 341 and a second bracket 342 are placed behind the screw slide 330. The dancing wheel 340 is rotatably connected to the second bracket 342. The spring 341 is installed on the second bracket 342. One end of the spring 341 is connected to the outer side of the dancing wheel 340. When the tension of the optical fiber changes greatly, the elastic force generated by the spring 341 overcomes the change in the tension of the optical fiber, causing the dancing wheel 340 to swing slightly, playing a buffering role, helping to maintain a stable tension of the optical fiber during the winding process and preventing uneven winding due to sudden tension changes.

[0050] Depend on Figure 7As shown, the take-up device 500 includes a take-up motor 510, a take-up wheel 520, a second coupling 530, and a third bracket 540. The second coupling 530 is mounted on the third bracket 540. One side of the third bracket 540 is fixedly connected to the take-up motor 510. The output end of the take-up motor 510 is fixedly connected to the take-up wheel 520. One end of the take-up wheel 520 is connected to the second coupling 530. The rotation of the take-up motor 510 drives the take-up wheel 520 to operate synchronously, thereby realizing the optical fiber pay-off function.

[0051] Depend on Figure 8 As shown, the optical imaging module 200 includes a microscope lens 210, a lifting bracket 220, and a light source. The microscope lens 210 is mounted on top of the lifting bracket 220. The microscope lens 210 is responsible for magnifying the contour details of the optical fiber surface and projecting them into the camera. The lifting bracket 220 provides precise control and adjustment of the microscope lens 210, ensuring clear images in various situations and providing stable support to prevent vibration from affecting imaging quality. The light source provides sufficient light for the imaging system to illuminate the surface of the optical fiber ring and produce a clear image. Its selection and adjustment are crucial to achieving high-quality imaging results. These three core components work together to ensure clear and accurate images.

[0052] Depend on Figure 9 As shown, the FPGA hardware accelerator 600 is located behind the pay-off device 100. The FPGA hardware accelerator 600 accelerates the convolution operation, that is, accelerates the YOLOv3-tiny algorithm of the defect detection module, realizes the real-time classification and positioning of the optical fiber winding defects, improves the computing efficiency, and reduces the computing power consumption.

[0053] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A real-time optical fiber loop defect detection system based on an embedded development platform, characterized by: It includes embedded development platform, motor module, tension acquisition module, image acquisition module, optical fiber path auxiliary module and pay-off and take-up module; The embedded development platform includes an ARM processing module and an FPGA hardware acceleration module. The ARM processor is responsible for executing complex control processes and advanced processing tasks, and the FPGA hardware acceleration module is responsible for the operation of convolutional neural networks while maintaining low power consumption and high efficiency. The motor module includes a motor and a PID controller. The motor is responsible for driving the optical fiber winding device to ensure precise positioning and uniform speed during the winding process. The PID controller dynamically adjusts the motor speed based on the difference between the actual tension feedback value and the preset target value to maintain stable optical fiber tension. Through the motor control module, winding defects caused by tension fluctuations can be effectively avoided. The tension acquisition module includes a tension sensing module and a signal transmission module. The tension sensing module is responsible for directly measuring the actual tension generated by the optical fiber during the winding process and providing an accurate analog signal output. The signal transmission module converts the collected analog signal into a digital signal, which is processed by the ARM processing module on the embedded development platform to ensure that the optical fiber tension remains within the set range and avoid defects in the optical fiber ring caused by excessive or insufficient tension. The image acquisition module includes a CMOS image sensor and an image preprocessing unit. The CMOS image sensor is responsible for capturing high-definition real-time images during the optical fiber winding process, providing sufficient resolution and frame rate to ensure image clarity. The raw data collected by the camera equipped with the CMOS image sensor is adjusted by the image preprocessing unit to match the input requirements of the YOLOv3-tiny algorithm. The optical fiber path auxiliary module assists in ensuring the correct trajectory of the optical fiber winding line, helps guide the optical fiber to move along the predetermined path, and achieves a precise winding process; The pay-off and take-up modules are used to realize the pay-off and take-up functions of the optical fiber.

2. The real-time optical fiber loop defect detection system based on an embedded development platform according to claim 1, characterized in that: The CMOS image sensor sends information, the ARM processor receives and analyzes the information, and is also responsible for transmitting the information to the pay-out and take-up modules to control the precise winding of the optical fiber ring.

3. The real-time optical fiber loop defect detection system based on an embedded development platform according to claim 2, characterized in that: The equipment carried by the system includes a pay-out device (100), an optical imaging module (200), an optical fiber path auxiliary device (300), a tension detector (400), a take-up device (500) and an FPGA hardware accelerator (600). The pay-out device (100) is located in front of the FPGA hardware accelerator (600), the tension detector (400) is located in front of the pay-out device (100), the optical fiber path auxiliary device (300) is placed in front of the tension detector (400), the take-up device (500) is placed in front of the optical fiber path auxiliary device (300), and the optical imaging module (200) is placed in front of the take-up device (500).

4. The real-time optical fiber loop defect detection system based on an embedded development platform according to claim 3, characterized in that: The pay-off device (100) comprises a pay-off motor (110), a pay-off wheel (120), a first coupling (130) and a first bracket (140), wherein the first coupling (130) is mounted on one side of the first bracket (140), the pay-off wheel (120) is mounted on one end of the first coupling (130), the pay-off motor (110) is mounted on the other side of the first bracket (140), and the output end of the pay-off motor (110) is connected to one end of the pay-off wheel (120).

5. The real-time optical fiber loop defect detection system based on an embedded development platform according to claim 4, characterized in that: The tension detector (400) comprises a tension sensor (410), a signal transmitter (420), a guide wheel (430), and a mounting frame (440). The signal transmitter (420) is mounted on the mounting frame (440), and the tension sensor (410) is mounted on the mounting frame (440). The tension detector (400) is located behind the pay-off device (100), and guide wheels (430) are placed on both sides of the mounting frame (440).

6. The real-time optical fiber loop defect detection system based on an embedded development platform according to claim 5, characterized in that: The optical fiber path auxiliary device (300) comprises a stepper motor (310), a ball screw (320) and a screw slide (330); the ball screw (320) is rotatably connected inside the screw slide (330); one side of the screw slide (330) is fixedly connected to the stepper motor (310); the output end of the stepper motor (310) is fixedly connected to one end of the ball screw (320); and the outer side of the ball screw (320) is threadedly connected to a movable cable arrangement mechanism.

7. The real-time optical fiber loop defect detection system based on an embedded development platform according to claim 6, characterized in that: A dancing wheel (340), a spring (341) and a second bracket (342) are placed behind the screw slide (330); the dancing wheel (340) is rotatably connected to the second bracket (342); a spring (341) is installed on the second bracket (342); one end of the spring (341) is connected to the outer side of the dancing wheel (340).

8. The real-time optical fiber loop defect detection system based on an embedded development platform according to claim 7, characterized in that: The wire-taking device (500) comprises a wire-taking motor (510), a wire-taking wheel (520), a second coupling (530), and a third bracket (540); the second coupling (530) is mounted on the third bracket (540); one side of the third bracket (540) is fixedly connected to the wire-taking motor (510); the output end of the wire-taking motor (510) is fixedly connected to the wire-taking wheel (520); and one end of the wire-taking wheel (520) is connected to the second coupling (530).

9. The real-time optical fiber loop defect detection system based on an embedded development platform according to claim 8, characterized in that: The optical imaging module (200) comprises a microscope lens (210), a lifting bracket (220) and a light source. The microscope lens (210) is mounted on the top of the lifting bracket (220). The light source provides sufficient light for the imaging system to illuminate the surface of the optical fiber ring and produce a clear image. The selection and adjustment of the light source are crucial for obtaining high-quality imaging results.

10. The real-time optical fiber loop defect detection system based on an embedded development platform according to claim 9, characterized in that: The FPGA hardware accelerator (600) is located behind the pay-off device (100). The FPGA hardware accelerator (600) accelerates convolution operations, that is, accelerates the YOLOv3-tiny algorithm of the defect detection module, thereby achieving real-time classification and positioning of optical fiber winding defects.

Images