An adaptive anti-interference fiber optic ribbon visual inspection device and method

Abstract

This invention discloses an adaptive anti-interference optical fiber ribbon visual inspection device and method, belonging to the field of optical fiber manufacturing visual inspection technology. The device includes an integrated inspection and control cabinet, an optical fiber transmission channel, an adaptive tensioning and anti-shake fiber feeding mechanism, an image acquisition unit, a control and processing unit, an encoder assembly, an automatic light source calibration module, and a human-machine interaction and alarm unit. The adaptive tensioning and anti-shake fiber feeding mechanism eliminates high-speed fiber feeding jitter by adjusting tension in real time through floating pressure rollers; dual light sources and dual cameras perform time-division alternating synchronous stroboscopic imaging to avoid mutual interference; the control and processing unit incorporates a deep learning model to achieve integrated detection of color recognition, line sequence verification, width measurement, and appearance defects; the automatic light source calibration module stabilizes the light source brightness in real time. This invention is compatible with rapid multi-core number changeovers, achieving a color recognition accuracy of ≥99.9%, a defect detection accuracy of ≥0.01mm, and seamless integration with production line PLC and MES systems, meeting the online quality inspection requirements of ultra-high-speed optical fiber ribbon production lines.

Description

Technical Field

[0001] This invention belongs to the field of visual inspection technology for optical fiber manufacturing, specifically relating to an adaptive anti-interference optical fiber ribbon visual inspection device and inspection method. Background Technology

[0002] As a core component in optical fiber communication systems for integrating optical fibers and distinguishing fiber cores, the accuracy of fiber color, consistency of line sequence, and integrity of appearance directly determine the transmission quality, connection reliability, and lifespan of optical fiber communication. With the optical fiber manufacturing industry iterating and upgrading towards "high speed, high density, and precision," the production speed of optical fiber ribbons has increased to over 2000m / min, and fiber core specifications have expanded from the conventional 6-core, 12-core, and 24-core to ultra-dense specifications of 72-core and 96-core. At the same time, the YD / T979-2024 industry standard has imposed stricter requirements on the geometric accuracy, color sequence consistency, and defect judgment threshold of optical fiber ribbons, making traditional testing methods unable to meet the quality inspection needs of industrial production.

[0003] Currently, domestic and international optical fiber ribbon visual inspection technology has gradually completed the technological iteration from manual visual inspection and simple photoelectric inspection to machine vision inspection, forming an inspection technology system with "camera + light source" as the core. However, combined with the current high-end requirements of optical fiber manufacturing and actual application scenarios, the following defects still exist: During high-speed production, the optical fiber ribbon is prone to jitter and displacement, resulting in blurred imaging and the inability to simultaneously detect color and geometric defects; Traditional visual inspection uses fixed RGB / HSV thresholds for color judgment, which is easily affected by light source attenuation, ambient light, and ink color difference, resulting in a high false judgment rate; There is a lack of automatic light source calibration and long-term stability mechanism, and the accuracy decreases after continuous operation of the equipment; When switching between different core number optical fiber ribbons, the positioning fixture and detection parameters cannot be quickly adapted, resulting in poor compatibility; Most solutions do not consider seamless integration with production line PLC and MES systems, forming data silos; The fiber feeding mechanism mostly adopts a fixed tension design, which cannot adapt to tension fluctuations during high-speed fiber feeding, resulting in optical fiber ribbon jitter and unstable posture, affecting imaging accuracy.

[0004] Therefore, developing a visual inspection device and method for fiber optic ribbons that can solve the above-mentioned technical defects and adapt to the needs of high-end production has become an urgent technical problem to be solved in this field. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive anti-interference optical fiber ribbon visual inspection device and inspection method to solve the problems in the prior art such as optical fiber ribbon jitter during high-speed production, easy interference in color judgment, difficulty in joint detection of multiple defects, inconvenience in specification switching, lack of light source calibration and poor system integration.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: An adaptive anti-interference fiber optic ribbon visual inspection device, comprising: The integrated monitoring and control cabinet serves as the installation base for the entire device. The fiber optic transmission channel, with a wheeled structure, is installed on both sides of the integrated inspection and control cabinet to guide and transport fiber optic ribbons. An adaptive tensioning and anti-shaking fiber feeding mechanism is installed at the feed end of the optical fiber transmission channel to adaptively adjust the tension and eliminate optical fiber ribbon shaking during high-speed fiber feeding. It includes at least a floating pressure roller and an elastic connecting bracket connected to the floating pressure roller. The image acquisition unit includes a dual industrial camera assembly and a dual light source system, used to acquire clear color images of the fiber optic ribbon, and the dual light source system keeps synchronized with the shutter of the corresponding camera. A light shield, installed above the image acquisition unit, can be flipped up and down to prevent ambient light interference; The control and processing unit is connected to the dual industrial camera assembly, dual light source system, human-machine interaction and alarm unit via wiring, and is used for image processing, analysis and command issuance. The encoder assembly is installed at the output end of the fiber optic transmission channel, keeps synchronized with the movement of the fiber optic ribbon, and is connected to the control and processing unit via a high-speed optical coupler. An automatic light source calibration module is connected to the control and processing unit via a circuit and is used for automatic calibration of the light source brightness parameters. The human-machine interaction and alarm unit is installed on top of the integrated inspection and control cabinet and is used for image display, parameter setting, and alarm for non-conforming products.

[0007] Preferably, in the adaptive tensioning and anti-shaking fiber feeding mechanism, the floating pressure roller is made of silicone material, connected to the integrated inspection and control cabinet through an elastic connecting bracket, and can float up and down; when the fiber ribbon tension is too high, the floating pressure roller automatically floats upward to reduce the pressure; when the fiber ribbon tension is too low, the floating pressure roller automatically moves downward to increase the pressure, eliminating the shaking during the high-speed fiber feeding process; The dual light sources are high-brightness, high-color-rendering, and high-uniformity COB light sources, with brightness adjustable by the control and processing unit, and are installed diagonally above the detection area; the dual industrial cameras are high-resolution color industrial cameras, installed directly above the detection area, with the lens axes of the dual industrial cameras perpendicular to the surface of the fiber ribbon, and the dual industrial cameras are equipped with telecentric lenses.

[0008] Preferably, the control and processing unit includes an industrial controller and a control board, and has the following image processing and detection function modules: The image preprocessing module is used to denoise and enhance the acquired images and automatically locate the region of interest. The color recognition and line sequence verification module is used to identify the color of each optical fiber based on a lightweight convolutional neural network or object detection model, and compare it with a preset standard line sequence template. The width detection module is used to calculate the actual physical width of the fiber ribbon using an edge extraction algorithm and compare it with a standard threshold. The appearance defect detection module is used to simultaneously identify appearance defects, including ghosting, burrs, and color mixing, through a defect classification model based on transfer learning. The collaborative control module is used to realize the coordinated operation of image acquisition, image analysis and processing, alarm control, and defect statistics; it has a built-in brand specification database for multi-core fiber optic ribbons, supports manual customization of adding and modifying parameters, and realizes rapid switching configuration of multi-core fiber optic ribbon detection parameters.

[0009] Preferably, the human-machine interaction and alarm unit includes a touch screen and an audible and visual alarm; the touch screen is used to display real-time detection images, detection results and quality statistics, and supports parameter setting, manual calibration and report export; the audible and visual alarm emits an audible and visual alarm when a defective product is detected. A guide limit block is provided at the front of the optical fiber transmission channel to prevent lateral displacement of the optical fiber ribbon during the transmission process and to ensure that the optical fiber ribbon is always in the center of the detection area. The bottom of the integrated inspection and control cabinet is equipped with adjustable support feet for adjusting the levelness; the surface of the integrated inspection and control cabinet is anodized, which provides corrosion resistance and anti-interference capabilities.

[0010] Furthermore, this invention also mentions an adaptive anti-interference fiber optic ribbon visual inspection method, which utilizes the adaptive anti-interference fiber optic ribbon visual inspection device described above, and includes the following steps: Step S1: System initialization and specification switching: After the system is powered on, the detection software will be started automatically to perform function initialization; if the brand and specification of the fiber optic ribbon currently being produced are inconsistent with the system preset, the detection parameters of the corresponding brand and specification will be switched through the MES system automatic switching, barcode scanner scanning switching or manual switching. Step S2: Encoder Synchronous Triggering and Image Acquisition: When the fiber optic ribbon moves, the encoder rotates synchronously. The control board acquires the encoder signal and controls the dual light source and dual industrial cameras to take pictures alternately in time division according to the encoder pulse count. Step S3: Image preprocessing: Gaussian filtering and histogram equalization are performed on the acquired color image to reduce noise and enhance it, and the region of interest of the fiber optic color band is automatically located. Step S4: Multi-dimensional detection and analysis: Simultaneously perform color recognition and line sequence verification, width detection and appearance defect detection on the preprocessed image; Step S5: Result Judgment and Output: Determine whether the fiber optic ribbon is qualified based on the detection and analysis results. If it is not qualified, trigger an audible and visual alarm, save the defect image, and mark the defect location, defect type, and detection time information.

[0011] Preferably, in step S2, the photo interval between the two sets of light sources and the industrial camera is determined by the encoder pulse number N, N = (LPR / L) × S, where LPR is the number of lines in one revolution of the encoder, L is the circumference of the encoder synchronization wheel, and S is the length of the fiber optic ribbon captured by the industrial camera each time; the distance between the two sets of light sources and the industrial camera is an even number of S lengths to ensure that the fiber optic ribbon is not missed or overlapped in the photo; the light sources and the industrial camera shutter are synchronized with each other and the strobe frequency is adjusted according to the transport speed of the fiber optic ribbon.

[0012] Preferably, the color recognition and line sequence verification in step S4 specifically includes: Lightweight convolutional neural networks or YOLO series target detection models are used to classify and locate the color of a single optical fiber region, and obtain the color category and spatial coordinates of each optical fiber. The fiber positions are sorted according to preset horizontal or vertical arrangement rules to generate an actual color sequence; The actual color sequence is compared with the standard line sequence template bit by bit. The abnormal identification results are filtered by confidence threshold and the results of misalignment, missed detection and wrong color detection are output.

[0013] Preferably, the width detection in step S4 specifically includes: The edge contours on both sides of the color band are located by Canny edge detection or subpixel-level edge extraction algorithm, and noise interference and false edges are removed by combining morphological operations. The actual pixel width of the color band is calculated using the minimum bounding rectangle or contour fitting algorithm, and the pixel values ​​are converted into physical dimensions using camera calibration parameters; The physical dimensions are compared with the standard width threshold, and the output results are used to detect width deviations, unevenness, and missing widths.

[0014] Preferably, the appearance defect detection in step S4 specifically includes: Collect OK and NG sample images of appearance defects, and train the defect classification model through few-shot transfer learning; The trained defect classification model is used to simultaneously identify defects such as ghosting, burrs, color mixing, coating defects, uneven single fiber spacing, and edge distortion in optical fiber ribbons.

[0015] Preferably, the method further includes an automatic light source calibration step: real-time acquisition of the light source power supply current value; when the acquired value deviates from the preset value, automatic adjustment of the current output of the constant current source of the light source is performed to ensure the stability of the light source brightness.

[0016] The beneficial technical effects of this invention are as follows: 1. High adaptability, meeting the needs of ultra-high speed and ultra-dense scenarios: Through dual-camera dual-light source synchronous stroboscopic imaging + adaptive tensioning and image stabilization mechanism, it is adapted to ultra-high speed production lines, supports the detection of 72-core, 96-core and above ultra-dense fiber ribbons, and shortens the adaptation time of multiple ribbon specifications to within 20 seconds, greatly improving production efficiency and solving the problem of insufficient adaptability of existing technologies.

[0017] 2. Comprehensive detection, high accuracy, and strong anti-interference ability: It adopts a color gamut clustering adaptive color judgment algorithm + composite defect joint judgment logic to realize the integrated detection of color, line sequence, width, and appearance defects. The color judgment accuracy is ≥99.9%, the defect detection accuracy is ≥0.01mm, the missed detection rate and false judgment rate are ≤0.1%, and it has strong resistance to interference from ambient light and light source attenuation.

[0018] 3. Stable long-term operation and low maintenance cost: Built-in automatic light source calibration function realizes real-time automatic calibration, avoids accuracy drift, reduces the frequency of manual calibration, reduces maintenance costs, ensures production continuity, and solves the problem of poor stability of existing technologies.

[0019] 4. High level of intelligence, adaptable to intelligent manufacturing: It adopts a small sample transfer learning algorithm to reduce sample dependence, and there is no need to retrain the model when changing production lines or specifications; it achieves seamless integration with PLC and MES systems, generates quality statistical reports, forms closed-loop management, adapts to the needs of industrial intelligent manufacturing, and solves the problem of poor integration of existing technologies.

[0020] 5. Reasonable structure, controllable cost, and easy to popularize: The optimized fiber feeding mechanism and overall structure eliminate fiber jitter, the equipment is compact, and installation and maintenance are convenient; the use of domestic high-end vision hardware to replace imports reduces equipment investment costs, making it easy for small and medium-sized enterprises to popularize and apply, and solving the problem of high cost of existing technologies.

[0021] 6. Low overall cost and high practicality: One set of equipment can simultaneously complete the integrated inspection of color, line sequence, width and appearance defects, without the need for multiple sets of equipment, which greatly reduces equipment investment and operating costs; the inspection process is in line with the actual production logic and can be directly connected to the existing fiber optic ribbon production line without the need for large-scale modification of the production line, making it highly practical. Attached Figure Description

[0022] Figure 1 This is a block diagram illustrating the electrical connection principle of the present invention.

[0023] Figure 2 This is a left-side view structural diagram of the device of the present invention.

[0024] Figure 3 This is a rear view structural diagram of the device of the present invention.

[0025] Figure 4 This is a right-side view of the device of the present invention.

[0026] Figure 5 This is a flowchart of the process of this invention; The components are as follows: 1-Integrated inspection and control cabinet; 2-Supporting foot; 3-Guide limit block; 4-Transmission wheel; 5-Tension floating wheel; 6-Industrial camera; 7-Light source; 8-Light shield; 9-Encoder; 10-Touch LCD screen; 11-LCD bracket; 12-Audio and visual alarm; 13-Power switch; 14-Power indicator light; 15-Power supply socket; 16-Ventilation panel; 17-Industrial controller; 18-Control board; 19-Switching power supply; 20-Cooling fan; 21-Power supply filter; 22-Distributor. Detailed Implementation

[0027] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Device Examples like Figures 1 to 4 As shown, an adaptive anti-interference optical fiber ribbon visual inspection device is installed on the optical fiber ribbon production line, located after the optical fiber ribbon coating and curing processes and before the fiber winding process.

[0028] Figures 2 to 4 This is a structural diagram of the device of the present invention. The integrated inspection and control cabinet 1 is placed after the coating and curing processes and before the take-up process in the optical fiber ribbon production line. The height of the four support feet 2 at the bottom is adjusted to ensure that the integrated inspection and control cabinet 1 is level and meets the height requirements of the optical fiber ribbon in the production line. The optical fiber ribbon is placed directly below the industrial camera 6 and the light source 7 via the guide limit block 3, the conveyor wheel 4, and the tension floating wheel 5. When adjusting the position of the industrial camera 6 and the light source 7, the light shield 8 is flipped upwards, and then flipped downwards after the adjustment is completed.

[0029] When the fiber optic ribbon moves, the encoder 9 moves synchronously with the conveyor wheel 4. The real-time acquired image of the fiber optic ribbon is displayed on the touch screen LCD 10, which is fixed to the top of the integrated control cabinet 1 via the LCD bracket 11. The audible and visual alarm 12 is fixed to the right rear of the touch screen LCD 10. The right panel of the integrated control cabinet 1 has a power switch 13, a power indicator light 14, and a power socket 15. Ventilation panels 16 are fixed to the lower left and right sides, respectively. The lower half of the integrated control cabinet 1 houses electrical components such as an industrial controller 17, a control board 18, a switching power supply 19, a cooling fan 20, a power supply filter 21, and a splitter 22.

[0030] Method Implementation Examples Figure 5This is a flowchart of the workflow of the present invention. After the system is powered on, the detection software is automatically started to initialize various functions. If the brand and specification of the currently produced optical fiber ribbon are inconsistent with those of the system, a brand and specification switch is required. Brand switching can be performed automatically through brand and specification information sent by the MES system, automatically by scanning the product process card QR code on-site with a barcode scanner, or manually by using brand and specification information from the system software's brand library. The switching time should not exceed 20 seconds.

[0031] When the fiber optic ribbon is in motion, the encoder rotates synchronously, and the control board collects the encoder signals. Every N encoder pulses, the first light source and the first industrial camera are triggered to take a picture; every 2N encoder pulses, the second light source and the second industrial camera are triggered to take a picture. The two sets of light sources and industrial cameras take pictures alternately in a time-sharing manner. The number of pulses between the two pictures is N = (LPR / L) × S, where LPR is the number of lines in one revolution of the encoder, L is the circumference of one revolution of the encoder's synchronous wheel, and S is the length of the fiber optic ribbon captured by each industrial camera. The distance between the two sets of industrial cameras is an even number of S lengths to ensure that the fiber optic ribbon is not missed or overlapped in the picture.

[0032] Images captured by the camera are transmitted to the industrial controller, which uses a Gaussian algorithm for noise reduction and a histogram equalization algorithm for enhancement to improve image contrast and highlight the color characteristics of the fiber optic ribbon. Traditional and AI-based inspection tools are used to analyze and determine for defects such as color difference, line sequence, appearance, and width anomalies. If a defect is found, an audible and visual alarm is triggered, and the defect image is saved, marking the defect location, defect type, and detection time information on the image.

[0033] During the production of optical fiber ribbons, if the fiber tension is too high, the tensioning wheel will automatically float upwards; conversely, if the tension is too low, the tensioning wheel will automatically float downwards, ensuring smooth and vibration-free transmission of the optical fiber ribbon. Simultaneously, the control board collects the power supply current of the light source in real time. If the collected value deviates from the set value, the output current of the constant current source of the light source is automatically adjusted to ensure the stability of the light source brightness and the stability of the detection accuracy.

[0034] This invention's device employs an integrated design of dual cameras, dual light sources, synchronous stroboscopic imaging, and an adaptive tensioning and anti-shake mechanism. This effectively reduces issues such as fiber feeding jitter, blurring, and color crosstalk between adjacent fiber cores on high-speed production lines. It utilizes an AI deep learning model for detection, replacing traditional fixed threshold color judgment and improving resistance to ambient light and light source attenuation interference. A combined defect judgment logic is designed to simultaneously detect color, line sequence, width, and appearance defects, increasing detection accuracy to over 99.9% while keeping the missed detection rate and false judgment rate below 0.1%. A built-in automatic light source calibration function adjusts the light source brightness in real time, avoiding accuracy drift, reducing manual calibration frequency, lowering maintenance costs, and ensuring long-term stable operation. An expansion interface allows seamless integration with production line PLCs, MES systems, and barcode scanners, generating quality statistical reports and achieving closed-loop management of detection, statistics, and optimization, thus meeting the needs of intelligent manufacturing.

[0035] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. An adaptive anti-interference fiber optic ribbon visual inspection device, characterized in that, include: The integrated monitoring and control cabinet serves as the installation base for the entire device. The fiber optic transmission channel, with a wheeled structure, is installed on both sides of the integrated inspection and control cabinet to guide and transport fiber optic ribbons. An adaptive tensioning and anti-shaking fiber feeding mechanism is installed at the feed end of the optical fiber transmission channel to adaptively adjust the tension and eliminate optical fiber ribbon shaking during high-speed fiber feeding. It includes at least a floating pressure roller and an elastic connecting bracket connected to the floating pressure roller. The image acquisition unit includes a dual industrial camera assembly and a dual light source system, used to acquire clear color images of the fiber optic ribbon, and the dual light source system keeps synchronized with the shutter of the corresponding camera. A light shield, installed above the image acquisition unit, can be flipped up and down to prevent ambient light interference; The control and processing unit is connected to the dual industrial camera assembly, dual light source system, human-machine interaction and alarm unit via wiring, and is used for image processing, analysis and command issuance. The encoder assembly is installed at the output end of the fiber optic transmission channel, keeps synchronized with the movement of the fiber optic ribbon, and is connected to the control and processing unit via a high-speed optical coupler. An automatic light source calibration module is connected to the control and processing unit via a circuit and is used for automatic calibration of the light source brightness parameters. The human-machine interaction and alarm unit is installed on top of the integrated inspection and control cabinet and is used for image display, parameter setting, and alarm for non-conforming products.

2. The adaptive anti-interference fiber optic ribbon visual inspection device according to claim 1, characterized in that, In the adaptive tensioning and anti-shaking fiber feeding mechanism, the floating pressure roller is made of silicone material and is connected to the integrated inspection and control cabinet through an elastic connecting bracket, and can float up and down; when the fiber ribbon tension is too high, the floating pressure roller automatically floats upward to reduce the pressure; when the fiber ribbon tension is too low, the floating pressure roller automatically moves downward to increase the pressure and eliminate the shaking during the high-speed fiber feeding process; The dual light sources are high-brightness, high-color-rendering, and high-uniformity COB light sources, with brightness adjustable by the control and processing unit, and are installed diagonally above the detection area; the dual industrial cameras are high-resolution color industrial cameras, installed directly above the detection area, with the lens axes of the dual industrial cameras perpendicular to the surface of the fiber ribbon, and the dual industrial cameras are equipped with telecentric lenses.

3. The adaptive anti-interference fiber optic ribbon visual inspection device according to claim 1, characterized in that, The control and processing unit includes an industrial controller and a control board, and has the following image processing and detection function modules: The image preprocessing module is used to denoise and enhance the acquired images and automatically locate the region of interest. The color recognition and line sequence verification module is used to identify the color of each optical fiber based on a lightweight convolutional neural network or object detection model, and compare it with a preset standard line sequence template. The width detection module is used to calculate the actual physical width of the fiber ribbon using an edge extraction algorithm and compare it with a standard threshold. The appearance defect detection module is used to simultaneously identify appearance defects, including ghosting, burrs, and color mixing, through a defect classification model based on transfer learning. The collaborative control module is used to realize the coordinated operation of image acquisition, image analysis and processing, alarm control, and defect statistics; it has a built-in brand specification database for multi-core fiber optic ribbons, supports manual customization of adding and modifying parameters, and realizes rapid switching configuration of multi-core fiber optic ribbon detection parameters.

4. The adaptive anti-interference fiber optic ribbon visual inspection device according to claim 1, characterized in that, The human-machine interaction and alarm unit includes a touch screen and an audible and visual alarm; the touch screen is used to display real-time detection screen, detection results and quality statistics, and supports parameter setting, manual calibration and report export; the audible and visual alarm sounds an audible and visual alarm when a defective product is detected. A guide limit block is provided at the front of the optical fiber transmission channel to prevent lateral displacement of the optical fiber ribbon during the transmission process and to ensure that the optical fiber ribbon is always in the center of the detection area. The bottom of the integrated inspection and control cabinet is equipped with adjustable support feet for adjusting the levelness; the surface of the integrated inspection and control cabinet is anodized, which provides corrosion resistance and anti-interference capabilities.

5. An adaptive anti-interference fiber optic ribbon visual inspection method, characterized in that, The adaptive anti-interference fiber optic ribbon visual inspection device according to any one of claims 1 to 4 includes the following steps: Step S1: System initialization and specification switching: After the system is powered on, the detection software will be started automatically to perform function initialization; if the brand and specification of the fiber optic ribbon currently being produced are inconsistent with the system preset, the detection parameters of the corresponding brand and specification will be switched through the MES system automatic switching, barcode scanner scanning switching or manual switching. Step S2: Encoder Synchronous Triggering and Image Acquisition: When the fiber optic ribbon moves, the encoder rotates synchronously. The control board acquires the encoder signal and controls the dual light source and dual industrial cameras to take pictures alternately in time division according to the encoder pulse count. Step S3: Image preprocessing: Gaussian filtering and histogram equalization are performed on the acquired color image to reduce noise and enhance it, and the region of interest of the fiber optic color band is automatically located. Step S4: Multi-dimensional detection and analysis: Simultaneously perform color recognition and line sequence verification, width detection and appearance defect detection on the preprocessed image; Step S5: Result Judgment and Output: Determine whether the fiber optic ribbon is qualified based on the detection and analysis results. If it is not qualified, trigger an audible and visual alarm, save the defect image, and mark the defect location, defect type, and detection time information.

6. The adaptive anti-interference fiber optic ribbon visual inspection method according to claim 5, characterized in that, In step S2, the photo interval between the two sets of light sources and the industrial camera is determined by the encoder pulse number N, N = (LPR / L) × S, where LPR is the number of lines in one revolution of the encoder, L is the circumference of the encoder synchronization wheel, and S is the length of the fiber optic ribbon captured by the industrial camera each time. The distance between the two sets of light sources and the industrial camera is an even number of S lengths to ensure that the fiber optic ribbon is not missed or overlapped in the photo. The light sources and the industrial camera shutter are synchronized with each other and the strobe frequency is adjusted according to the fiber optic ribbon's transport speed.

7. The adaptive anti-interference fiber optic ribbon visual inspection method according to claim 5, characterized in that, The color recognition and line sequence verification in step S4 specifically include: Lightweight convolutional neural networks or YOLO series target detection models are used to classify and locate the color of a single optical fiber region, and obtain the color category and spatial coordinates of each optical fiber. The fiber positions are sorted according to preset horizontal or vertical arrangement rules to generate an actual color sequence; The actual color sequence is compared with the standard line sequence template bit by bit. The abnormal identification results are filtered by confidence threshold and the results of misalignment, missed detection and wrong color detection are output.

8. The adaptive anti-interference fiber optic ribbon visual inspection method according to claim 5, characterized in that, The width detection in step S4 specifically includes: The edge contours on both sides of the color band are located by Canny edge detection or subpixel-level edge extraction algorithm, and noise interference and false edges are removed by combining morphological operations. The actual pixel width of the color band is calculated using the minimum bounding rectangle or contour fitting algorithm, and the pixel values ​​are converted into physical dimensions using camera calibration parameters; The physical dimensions are compared with the standard width threshold, and the output results are used to detect width deviations, unevenness, and missing widths.

9. The adaptive anti-interference fiber optic ribbon visual inspection method according to claim 5, characterized in that, The visual defect detection in step S4 specifically includes: Collect OK and NG sample images of appearance defects, and train the defect classification model through few-shot transfer learning; The trained defect classification model is used to simultaneously identify defects such as ghosting, burrs, color mixing, coating defects, uneven single fiber spacing, and edge distortion in optical fiber ribbons.

10. The adaptive anti-interference fiber optic ribbon visual inspection method according to claim 5, characterized in that, It also includes an automatic light source calibration step: real-time acquisition of the light source power supply current value, and automatic adjustment of the current output of the constant current source of the light source when the acquired value deviates from the preset value, so as to ensure the stability of the light source brightness.

Images