Anti-resonance optical fiber identification and high-precision axis alignment method based on end face characteristics

Abstract

The invention discloses an anti-resonance optical fiber identification and high-precision axis alignment method based on end face characteristics, and relates to the technical field of optical fiber fusion splicers, and the method comprises the following steps: S1, collecting end face images of multiple types of anti-resonance optical fibers, the multiple types of anti-resonance optical fibers including a basic type and a complex nested type; preprocessing the acquired image, wherein the preprocessing comprises graying, Gaussian blur denoising, Sobel operator gradient calculation, binaryzation and morphological operation; s2, preprocessing the acquired image, including graying, Gaussian blur denoising, Sobel operator gradient calculation, binaryzation and morphological operation; according to the anti-resonance optical fiber identification and high-precision axis alignment method based on the end face features, the end face feature differences of anti-resonance optical fibers of different structures are deeply excavated, a two-stage classification framework of large class distinguishing-subdivision identification is innovatively constructed, and the adaptability and accuracy of classification identification of the multi-structure anti-resonance optical fibers are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of optical fiber fusion splicer technology, specifically to a method for anti-resonant optical fiber identification and high-precision alignment based on end-face features. Background Technology

[0002] Hollow-core antiresonant fiber, with its characteristics of low transmission loss, high power carrying capacity, and wide-band transmission, has become a key transmission medium in fields such as high-power laser transmission and deep space communication. However, the splicing quality of hollow-core antiresonant fiber directly depends on the accurate identification of the fiber type and the alignment accuracy. If the type identification is incorrect or the alignment is off, it will lead to optical signal leakage, a sharp increase in transmission loss, or even splicing failure, which will seriously affect the stability of the system.

[0003] Existing fiber optic fusion splicing equipment is mainly compatible with traditional types such as ordinary solid-core fiber and polarization-maintaining fiber, and its compatibility with hollow-core anti-resonant fiber is significantly insufficient. For example, Chinese patent publication number CN116883387A discloses a fusion splicer alignment technology that can only cover conventional fibers and cannot distinguish between hollow-core anti-resonant fibers with different structures such as Kagome type and nested type. When dealing with hollow-core anti-resonant fiber, current market equipment often lacks feature extraction logic for its "central transmission hollow core + cladding hollow core unit" structure, resulting in missing classification and identification and incorrect alignment reference.

[0004] From the perspective of end-face inspection technology, manual inspection of hollow anti-resonant fibers relies on professional experience, which is inefficient and prone to misjudgment. Although semi-automated inspection introduces software analysis, it still requires manual intervention to adjust the identification parameters of the cladding hollow core, resulting in poor flexibility when adapting to hollow anti-resonant fibers with different structures. Existing automatic inspection technology can only be adapted to single-structure hollow fibers. For the topological differences of Kagome-type single-ring hollow fibers and nested multi-ring hollow fibers, it lacks hierarchical feature extraction and classification logic, making it difficult to meet the technical requirement of "accurate classification first, and then high-precision axis alignment".

[0005] In summary, existing technologies have shortcomings in the accuracy of multi-structure classification and identification of hollow anti-resonant optical fibers and the adaptability of alignment algorithms. There is an urgent need for an algorithm for identification and alignment of hollow anti-resonant optical fibers based on end-face features. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method for identifying and aligning anti-resonant optical fibers based on end-face features. This method solves the problems mentioned in the background section regarding the insufficient adaptability of multi-structure classification and identification of hollow anti-resonant optical fibers (such as Kagome type, nested type, double nested type, etc.) and the poor splicing quality caused by alignment reference deviation.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for anti-resonant fiber identification and high-precision axis alignment based on end-face features, comprising the following steps: S1: Acquire end-face images of various types of anti-resonant optical fibers, including basic and complex nested types; preprocess the acquired images, including grayscale conversion, Gaussian blur denoising, Sobel gradient calculation, binarization, and morphological operations. S2: Preprocess the acquired image, including grayscale conversion, Gaussian blur denoising, Sobel operator gradient calculation, binarization, and morphological operations; S3: Perform contour detection on the preprocessed image to obtain the contour set of the fiber end face, central transmission hollow core and cladding hollow core unit; S4: Classification and recognition based on hierarchical contour features: Screen contours that meet the conditions in terms of area and topology, and fit the large contour of the fiber end face. S5: If a single-ring cladding hollow core outline is detected, it is determined to be a basic anti-resonant fiber; if a multi-ring nested cladding hollow core outline is detected, the nesting level and hollow core arrangement characteristics are combined to complete the subdivision identification of complex nested anti-resonant fibers. S6: Based on the classification and recognition results, the corresponding geometric parameter fitting strategy is adapted. The basic anti-resonant fiber is based on the contour fitting of the central transmission hollow core and the single-ring cladding hollow core, while the complex nested anti-resonant fiber is based on the fitting of the feature regions of the central transmission hollow core and the multi-ring nested cladding hollow core. S7: Construct a high-precision alignment model based on the fitted geometric parameters, calculate the alignment reference, and complete the alignment.

[0008] Preferably, a 24-bit RGB camera is used with a 5.2x magnifying lens tube when acquiring images, the image resolution is set to 1280×720, the image is a three-channel color image, and the camera acquisition direction is perpendicular to the optical fiber centerline to ensure the accuracy and consistency of end face image acquisition.

[0009] Preferably, the specific method for preprocessing the acquired image in step S2 includes: Grayscale processing: The grayscale value of the image is calculated using the weighted average method, and the formula is as follows: ; Where Igray is the pixel value of the grayscale image, and B, G, and R are the pixel values ​​of the blue, green, and red channels of the original color image, respectively; Gaussian blur denoising: A 5×5 Gaussian filter is used, and the Gaussian kernel function formula is:

[0010] Image smoothing and noise reduction are achieved through convolution operations; where... Gaussian kernel function in coordinates The value at that position is used for smoothing filtering of the image. Gaussian standard deviation controls the diffusion degree of the Gaussian kernel, affecting the effect of image denoising and smoothing. : The coordinates of the pixels within the kernel relative to the center, with values ​​of -2, -1, 0, 1, 2, using a 5×5 Gaussian filter; Gradient calculation: The horizontal and vertical gradients are calculated using a 3×3 Sobel operator. Gradient magnitude: And normalize the gradient magnitude; Binarization: A fixed threshold method is used, with a threshold value of T=30. The formula is as follows:

[0011] Morphological operations: Using a 3×3 matrix of all 1s as structuring elements, perform two dilation operations followed by one erosion operation, as shown in the following formula: Expansion:

[0012] corrosion:

[0013] The resulting morphologically processed image is used to eliminate isolated noise points and connect broken contours.

[0014] Preferably, in step S3, the contour detection employs a contour retrieval algorithm, with a tree-structure retrieval mode and a simple contour approximation method. The algorithm focuses on extracting the outer contour of the end face, the central transmission hollow core contour, the cladding hollow core unit contour, and the pipe wall connection feature contour to form a complete contour set.

[0015] Preferably, the specific method for classification and recognition in S4 is as follows: End face large contour screening and fitting: Traverse the contour set, screen the outer contour that satisfies the largest area and sufficient number of contour points, and use the ellipse fitting algorithm to fit the large contour of the fiber end face. Cladding hollow core topology analysis: Within the large contour of the end face, based on the contour area, spacing and arrangement rules, the contour of the cladding hollow core unit is selected and its arrangement topology is analyzed. Hollow core characteristic parameter calculation: Calculate the roundness of the hollow cladding profile. (Identifying circular hollow cores), arrangement density and ring number characteristics, where S is the contour area and L is the contour perimeter; Classification and determination: If the cladding hollow core is arranged in a single ring (such as a Kagome-type multi-circular hole single ring array), it is determined to be a basic anti-resonant fiber; if the cladding hollow core is arranged in a multi-ring nested pattern (such as a nested double / multi-ring hollow core structure), it is determined to be a complex nested anti-resonant fiber based on the nesting level and hollow core size difference characteristics, and further subdivision identification is completed.

[0016] Preferably, the geometric parameter fitting in S6 adopts a contour fitting algorithm, specifically as follows: Basic anti-resonant fiber fitting: An ellipse fitting algorithm is used for the central transmission hollow core profile, and a batch ellipse fitting algorithm is used for the single-ring cladding hollow core profile. The formula is as follows:

[0017] Where (x0, y0) are the coordinates of the center of the fitted ellipse, and a and b are the major and minor axes of the ellipse, respectively; Fitting of complex nested anti-resonant optical fibers: an ellipse fitting algorithm is used for the central transmission hollow core profile, a ring-by-ring fitting strategy is used for the multi-ring nested cladding hollow core feature region, each ring cladding hollow core is fitted with an ellipse, and a polygon approximation fitting is used for the tube wall connection feature profile.

[0018] Preferably, the high-precision alignment reference calculation method for the shaft model in S7 is as follows: The center of the fitted ellipse of the central transmission hollow core is used as the position alignment reference, and the main direction of the cladding hollow core arrangement is used as the angle alignment reference. Position alignment error:

[0019] Pixels; of which , These are the coordinates of the center of the hollow core of each of the two optical fibers; Angle alignment error:

[0020] in , These are the main direction angles of the hollow cladding arrangement of the two optical fibers.

[0021] An anti-resonant fiber optic identification and high-precision alignment system based on end-face features includes an image acquisition module, a preprocessing module, a contour detection module, a classification and recognition module, a parameter fitting module, and an alignment calculation module.

[0022] Preferably, the image acquisition module includes a camera and an optical magnification component, used to acquire end-face images of various types of anti-resonant optical fibers and transmit them to the preprocessing module, with the acquisition direction perpendicular to the optical fiber's central axis; The preprocessing module sequentially performs grayscale conversion, Gaussian blur denoising, Sobel gradient calculation, binarization, and morphological operations on the image, and outputs the preprocessed image to the contour detection module. The contour detection module uses a contour retrieval algorithm to extract the outer contour of the end face, the central transmission hollow contour, the cladding hollow unit contour, and the pipe wall connection feature contour, and transmits the contour set to the classification and recognition module. The classification and recognition module is based on the "cladding hollow core topology" judgment logic to distinguish between basic and complex nested anti-resonant optical fibers, and completes the subdivision recognition of complex nested fibers, and outputs the classification results to the parameter fitting module. The parameter fitting module adapts the type fitting strategy according to the classification results and outputs key fitting parameters such as the center transmission hollow core and cladding hollow core to the axis calculation module. The alignment calculation module calculates the position alignment error and angle alignment error based on the fitting parameters and outputs the alignment result.

[0023] This invention provides a method for anti-resonant fiber identification and high-precision axis alignment based on end-face features, which has the following advantages: 1. This anti-resonant fiber identification and high-precision alignment method based on end-face features, by deeply mining the differences in end-face features of anti-resonant fibers with different structures (such as cladding hollow core topology, nesting levels, size distribution, etc.), innovatively constructs a two-level classification framework of "major category distinction - sub-level identification", which significantly improves the adaptability and accuracy of classification and identification of multi-structure anti-resonant fibers. It uses image processing technology to realize automatic classification and identification and high-precision alignment of Kagome type, nested type, double nested type and other anti-resonant fibers, and solves the problems of insufficient adaptability of existing fusion splicing equipment to anti-resonant fibers, low classification and identification efficiency and large alignment deviation.

[0024] 2. This anti-resonant fiber identification and high-precision alignment method based on end-face features boasts advantages such as strong adaptability, high identification accuracy, high automation, and good scalability. By comprehensively covering basic and complex nested anti-resonant fiber structures, it eliminates the need for separate algorithm development for individual structures, achieving full-type compatibility. Through hierarchical feature extraction and multi-dimensional topology matching, the classification accuracy can reach over 99%, and the alignment accuracy (position error ±1.4μm, angle error ±0.1°) meets the requirements for low-loss fusion splicing. The entire process, from image acquisition, feature extraction, classification recognition to alignment calculation, is fully automated, requiring no manual intervention and significantly improving fusion efficiency and reliability. If a new anti-resonant fiber structure type is added, it can be quickly adapted by supplementing the feature matching library, providing flexible support for the technological iteration of fusion splicing equipment. It provides core technical support for the full-type adaptation and automated upgrade of fiber optic fusion splicing equipment to anti-resonant fibers, and can be widely applied in fiber optic fusion splicing equipment in fields such as high-power laser transmission, broadband communication, and deep space exploration, demonstrating significant practical value and promising prospects for widespread application. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the anti-resonant fiber identification and high-precision axis alignment method based on end-face features of the present invention. Figure 2 This is a schematic diagram of the optical fiber end face image acquired by the present invention; Figure 3 This is a schematic diagram of the image after preprocessing gradient calculation according to the present invention; Figure 4 This is a schematic diagram of the contour after scanning using the connected component scanning algorithm of this invention; Figure 5 This is a schematic diagram of the hollow fiber end face detected by the least squares ellipse fitting algorithm of this invention. Figure 6 This is a schematic diagram of the output results after the hollow fiber end face detection and identification of the present invention. Detailed Implementation

[0026] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0027] Please see Figures 1 to 6 This invention provides a technical solution: a method for identifying and aligning anti-resonant optical fibers based on end-face features, comprising the following steps: S1: Acquire end-face images of various types of anti-resonant optical fibers, including basic types (such as Kagome type) and complex nested types (such as nested type, double nested type, etc.); perform preprocessing on the acquired images, including grayscale conversion, Gaussian blur denoising, Sobel operator gradient calculation, binarization, and morphological operations. S2: Preprocess the acquired image, including grayscale conversion, Gaussian blur denoising, Sobel operator gradient calculation, binarization, and morphological operations; S3: Perform contour detection on the preprocessed image to obtain the contour set of the fiber end face, central transmission hollow core and cladding hollow core unit; S4: Classification and recognition based on hierarchical contour features: Screen contours that meet the conditions in terms of area and topology, and fit the large contour of the fiber end face. S5: If a single-ring cladding hollow core outline (such as a Kagome-type multi-circular hole array) is detected, it is determined to be a basic anti-resonant fiber; if a multi-ring nested cladding hollow core outline (such as a nested multi-ring hole structure) is detected, the nesting level and hollow core arrangement features are combined to complete the subdivision identification of complex nested anti-resonant fibers. S6: Based on the classification and recognition results, the corresponding geometric parameter fitting strategy is adapted. The basic anti-resonant fiber is based on the contour fitting of the central transmission hollow core and the single-ring cladding hollow core, while the complex nested anti-resonant fiber is based on the fitting of the feature regions of the central transmission hollow core and the multi-ring nested cladding hollow core. S7: Construct a high-precision alignment model based on the fitted geometric parameters, calculate the alignment reference, and complete the alignment.

[0028] A 24-bit RGB camera with a 5.2x magnifying lens was used to acquire images. The image resolution was set to 1280×720, and the image was a three-channel color image. The camera acquisition direction was perpendicular to the optical fiber centerline to ensure the accuracy and consistency of the end face image acquisition.

[0029] The specific methods for preprocessing the acquired images in S2 include: Grayscale processing: The grayscale value of the image is calculated using the weighted average method, and the formula is as follows: ; Where Igray is the pixel value of the grayscale image, and B, G, and R are the pixel values ​​of the blue, green, and red channels of the original color image, respectively; Gaussian blur denoising: A 5×5 Gaussian filter is used, and the Gaussian kernel function formula is:

[0030] Image smoothing and noise reduction are achieved through convolution operations; where... Gaussian kernel function in coordinates The value at that position is used for smoothing filtering of the image. Gaussian standard deviation controls the diffusion degree of the Gaussian kernel, affecting the effect of image denoising and smoothing. : The coordinates of the pixels within the kernel relative to the center, with values ​​of -2, -1, 0, 1, 2 (due to the use of a 5×5 Gaussian filter). Gradient calculation: The horizontal and vertical gradients are calculated using a 3×3 Sobel operator. Gradient magnitude: And normalize the gradient magnitude; Binarization: A fixed threshold method is used, with a threshold value of T=30. The formula is as follows:

[0031] Morphological operations: Using a 3×3 matrix of all 1s as structuring elements, perform two dilation operations followed by one erosion operation, as shown in the following formula: Expansion:

[0032] corrosion:

[0033] The resulting morphologically processed image is used to eliminate isolated noise points and connect broken contours.

[0034] In S3, contour detection employs a contour retrieval algorithm with a tree-structure retrieval mode and a simple contour approximation method. The algorithm focuses on extracting the outer contour of the end face, the central transmission hollow core contour, the cladding hollow core unit contour, and the pipe wall connection feature contour to form a complete contour set.

[0035] The specific method for classification and recognition in S4 is as follows: End face large contour screening and fitting: Traverse the contour set, screen the outer contour that satisfies the largest area and sufficient number of contour points, and use the ellipse fitting algorithm to fit the large contour of the fiber end face. Cladding hollow core topology analysis: Within the large contour of the end face, based on the contour area, spacing and arrangement rules, the contour of the cladding hollow core unit is selected and its arrangement topology is analyzed. Hollow core characteristic parameter calculation: Calculate the roundness of the hollow cladding profile. (Identifying circular hollow cores), arrangement density and ring number characteristics, where S is the contour area and L is the contour perimeter; Classification and determination: If the cladding hollow core is arranged in a single ring (such as a Kagome-type multi-circular hole single ring array), it is determined to be a basic anti-resonant fiber; if the cladding hollow core is arranged in a multi-ring nested pattern (such as a nested double / multi-ring hollow core structure), it is determined to be a complex nested anti-resonant fiber based on the nesting level and hollow core size difference characteristics, and further subdivision identification is completed.

[0036] The geometric parameter fitting in S6 adopts a contour fitting algorithm, specifically: Basic anti-resonant fiber fitting: An ellipse fitting algorithm is used for the central transmission hollow core profile, and a batch ellipse fitting algorithm is used for the single-ring cladding hollow core profile. The formula is as follows:

[0037] Where (x0, y0) are the coordinates of the center of the fitted ellipse, and a and b are the major and minor axes of the ellipse, respectively; Fitting of complex nested anti-resonant optical fibers: an ellipse fitting algorithm is used for the central transmission hollow core profile, a ring-by-ring fitting strategy is used for the multi-ring nested cladding hollow core feature region, each ring cladding hollow core is fitted with an ellipse, and a polygon approximation fitting is used for the tube wall connection feature profile.

[0038] The alignment reference calculation method for the high-precision shaft model in S7 is as follows: The center of the fitted ellipse of the central transmission hollow core is used as the position alignment reference, and the main direction of the cladding hollow core arrangement is used as the angle alignment reference. Position alignment error:

[0039] Pixels; of which , These are the coordinates of the center of the hollow core of each of the two optical fibers; Angle alignment error:

[0040] in , These are the main direction angles of the hollow cladding arrangement of the two optical fibers.

[0041] This embodiment has the following advantages: High adaptability: It fully covers anti-resonant optical fibers with various structures, including basic and complex nested types, without the need to develop separate algorithms for a single structure, achieving full compatibility; High recognition accuracy: Through hierarchical feature extraction and multi-dimensional topology matching, the classification and recognition accuracy can reach over 99%, and the axis accuracy (position error ±1.4μm, angle error ±0.1°) meets the requirements for low-loss welding. High degree of automation: The entire process from image acquisition, feature extraction, classification and recognition to axis calculation is automated, requiring no manual intervention, which greatly improves welding efficiency and reliability; Good scalability: If a new anti-resonant fiber structure type is added, it can be quickly adapted by supplementing the feature matching library, providing flexible support for the technological iteration of fusion splicing equipment.

[0042] It is especially suitable for automatic classification, identification and fusion alignment of multi-structure hollow anti-resonant optical fibers (such as Kagome type, nested type, double nested type, etc.), and can support the full-type adaptation of optical fiber fusion splicing equipment to hollow anti-resonant optical fibers.

[0043] A system for identifying and aligning anti-resonant optical fibers based on end-face features, comprising an image acquisition module, a preprocessing module, a contour detection module, a classification and recognition module, a parameter fitting module, and an alignment calculation module; The image acquisition module includes a camera and an optical magnification component deployed perpendicular to the optical fiber's central axis. It is used to acquire end-face images of various types of anti-resonant optical fibers and transmit them to the preprocessing module. The acquisition direction is perpendicular to the optical fiber's central axis. The preprocessing module sequentially performs grayscale conversion, Gaussian blur denoising, Sobel gradient calculation, binarization, and morphological operations on the image, and outputs the preprocessed image to the contour detection module. The contour detection module uses a contour retrieval algorithm to extract the outer contour of the end face, the central transmission hollow contour, the cladding hollow unit contour, and the pipe wall connection feature contour, and transmits the contour set to the classification and recognition module. The classification and recognition module includes a major category differentiation unit (to determine basic / complex nested type) and a sub-category recognition unit (to distinguish Kagome type, nested type, double nested type, etc.). Based on the judgment logic of "cladding hollow core topology", it distinguishes between basic and complex nested anti-resonant optical fibers, completes the sub-category recognition of complex nested type, and outputs the classification results to the parameter fitting module. The parameter fitting module adapts the type fitting strategy according to the classification results and outputs key fitting parameters such as the center transmission hollow core and cladding hollow core to the axis calculation module. The alignment calculation module calculates the position alignment error and angle alignment error based on the fitting parameters, constructs a high-precision alignment model, completes the alignment calculation of anti-resonant optical fibers with different structures, and outputs the alignment results.

[0044] In summary, this anti-resonant fiber identification and high-precision axis alignment method based on end-face features includes the following steps: Step 1: Image acquisition of the anti-resonant fiber end face An OV5640 camera was used to capture images of the end face of the multi-structure anti-resonant optical fiber. The camera has a resolution of 5 megapixels and a minimum pixel size of 1.4μm. The acquisition direction is set perpendicular to the central axis of the optical fiber to ensure the accuracy and consistency of image acquisition. The camera, together with an optical magnification component, can make full use of its resolution and pixel size characteristics to effectively capture clear details of the optical fiber end face (such as the outline of the central transmission hollow core, the shape of the cladding hollow core unit, and the tube wall connection structure). Step 2: Image Preprocessing The acquired images were processed sequentially including grayscale conversion, Gaussian blur denoising, Sobel operator gradient calculation, binarization, and morphological operations. Grayscale conversion: Converting a color image to a grayscale image simplifies subsequent calculations; Gaussian blur denoising: suppresses image noise while preserving the feature details of the transmission hollow core, cladding hollow core, and tube wall at the end face center; Sobel operator gradient calculation: enhances the contrast between fiber end face edge, hollow core boundary and tube wall profile; Binarization: Converts an image into a black-and-white binary form to highlight target features such as central transmission hollow cores and cladding hollow cores; Morphological operations: Optimize the connectivity and integrity of the binarized image, eliminate isolated noise points and connect broken hollow boundaries, laying the foundation for feature extraction; Step 3: Preliminary Feature Extraction Preliminary features of the fiber end face outer contour, central transmission hollow core contour, and cladding hollow core unit are extracted from the preprocessed image, including contour area, arrangement density, and topological distribution, which are used to preliminarily distinguish the structural categories of anti-resonant fibers. Step 4: Differentiate between anti-resonant fiber types (basic type vs. complex nested type) Based on the topological characteristics of the hollow cladding core, the fiber is classified as follows: if the hollow cladding core is arranged in a single ring (such as a Kagome-type multi-circular hole single-ring array), it is classified as a basic anti-resonant fiber; if the hollow cladding core is arranged in a multi-ring nested pattern (such as a nested or double-nested multi-ring hole structure), it is classified as a complex nested anti-resonant fiber. Step 5: Feature Extraction and Segmentation Recognition For basic antiresonant fibers: extract the shape, size, spacing of the cladding core, and contour feature points of the central transmission core (such as core boundary feature points, tube wall connection feature points, etc.), and construct an identification model based on these features to achieve accurate identification of basic antiresonant fibers such as Kagome type; for complex nested antiresonant fibers: extract the nesting level, inter-ring spacing, core size differences, and contour feature points of the central transmission core of the cladding core, and construct a multi-dimensional feature matching model to achieve accurate subdivision and identification of complex antiresonant fibers such as nested and double-nested types; Step 6: Output classification results The output anti-resonant fiber classification and identification results clearly distinguish specific categories such as basic type (e.g., Kagome type) and complex nested type (e.g., nested type, double nested type); Step 7: Adapting geometric parameters and high-precision axis alignment Based on the classification results, the corresponding geometric parameter fitting strategy is adapted: Basic anti-resonant fiber: geometric parameters are fitted based on the contours of the central transmission hollow core and the single-ring cladding hollow core to construct an alignment model; Complex nested anti-resonant fiber: geometric parameters are fitted by combining the feature regions of the central transmission hollow core and the multi-ring nested cladding hollow core to construct a high-precision alignment model, thus achieving high-precision alignment of anti-resonant fibers with different structures. Preferred technical details: In step 1, the camera can be a megapixel-level industrial camera, with a customized optical lens barrel, to ensure that details such as the central transmission hollow core, cladding hollow core, and tube wall structure of the optical fiber end face are clearly distinguishable, and the image resolution is not less than 1280×720. In step 2, the Sobel operator gradient calculation is performed using the following method: This type of operator enhances the horizontal gradient features and highlights the details of the cladding hollow boundary and tube wall profile; In the detailed identification step 5, a multi-dimensional feature matching library is constructed by matching the topological structure of feature points (such as Kagome-type single-ring multi-circle hole distribution and nested multi-ring nested distribution) and size parameter differences, and the identification accuracy can reach more than 99%. In step 7, the high-precision alignment error is controlled within ±1.4μm and the angle alignment error is controlled within ±0.1°, which meets the low-loss and high-precision requirements of anti-resonant fiber fusion splicing. In step 1, a megapixel industrial camera is used, along with a customized optical lens barrel, to ensure that details such as the fiber core and stress zone on the fiber end face are clearly distinguishable, with an image resolution of no less than 1280×720; in the Sobel operator gradient calculation in step 2, the following is used... The type of operator enhances the gradient features in the horizontal direction and highlights the edge contour of the fiber end face; in the polarization-maintaining fiber subdivision identification in step S5, a multi-dimensional feature matching library is constructed by matching the shape of feature points (such as the "double circle" stress area of ​​the panda type and the "bow tie" stress area of ​​the bow tie type) and the difference in gray scale distribution, and the identification accuracy can reach more than 99%. In step 7, the high-precision alignment error is controlled within ±1.4μm and the angle alignment error is controlled within ±0.1°, which meets the low-loss and high-precision requirements of anti-resonant fiber fusion splicing.

[0045] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for identifying and aligning anti-resonant optical fibers with high precision based on end-face features, characterized in that: Includes the following steps: S1: Acquire end-face images of various types of anti-resonant optical fibers, including basic and complex nested types; preprocess the acquired images, including grayscale conversion, Gaussian blur denoising, Sobel gradient calculation, binarization, and morphological operations. S2: Preprocess the acquired image, including grayscale conversion, Gaussian blur denoising, Sobel gradient calculation, binarization, and morphological operations; S3: Perform contour detection on the preprocessed image to obtain the contour set of the fiber end face, central transmission hollow core and cladding hollow core unit; S4: Classification and recognition based on hierarchical contour features: Screen contours that meet the conditions in terms of area and topology, and fit the large contour of the fiber end face. S5: If a single-ring cladding hollow core outline is detected, it is determined to be a basic anti-resonant fiber; if a multi-ring nested cladding hollow core outline is detected, the nesting level and hollow core arrangement characteristics are combined to complete the subdivision identification of complex nested anti-resonant fibers. S6: Based on the classification and recognition results, the corresponding geometric parameter fitting strategy is adapted. The basic anti-resonant fiber is based on the contour fitting of the central transmission hollow core and the single-ring cladding hollow core, while the complex nested anti-resonant fiber is based on the fitting of the feature regions of the central transmission hollow core and the multi-ring nested cladding hollow core. S7: Construct a high-precision alignment model based on the fitted geometric parameters, calculate the alignment reference, and complete the alignment.

2. The anti-resonant fiber identification and high-precision axis alignment method based on end-face features according to claim 1, characterized in that: A 24-bit RGB camera with a 5.2x magnifying lens was used to acquire images. The image resolution was set to 1280×720, and the image was a three-channel color image. The camera acquisition direction was perpendicular to the optical fiber centerline to ensure the accuracy and consistency of the end face image acquisition.

3. The anti-resonant fiber identification and high-precision axis alignment method based on end-face features according to claim 1, characterized in that: The specific methods for preprocessing the acquired images in S2 include: Grayscale processing: The grayscale value of the image is calculated using the weighted average method, and the formula is as follows: ; Where Igray is the pixel value of the grayscale image, and B, G, and R are the pixel values ​​of the blue, green, and red channels of the original color image, respectively; Gaussian blur denoising: A 5×5 Gaussian filter is used, and the Gaussian kernel function formula is: ； Image smoothing and noise reduction are achieved through convolution operations; where... Gaussian kernel function in coordinates The value at that position is used for smoothing filtering of the image. Gaussian standard deviation controls the diffusion degree of the Gaussian kernel, affecting the effect of image denoising and smoothing. : The coordinates of the pixels within the kernel relative to the center, with values ​​of -2, -1, 0, 1, 2, using a 5×5 Gaussian filter; Gradient calculation: The horizontal and vertical gradients are calculated using a 3×3 Sobel operator. Gradient magnitude: And normalize the gradient magnitude; Binarization: A fixed threshold method is used, with a threshold value of T=30. The formula is as follows: ； Morphological operations: Using a 3×3 matrix of all 1s as structuring elements, perform two dilation operations followed by one erosion operation, as shown in the following formula: Expansion: ; corrosion: ; The resulting morphologically processed image is used to eliminate isolated noise points and connect broken contours.

4. The anti-resonant fiber identification and high-precision axis alignment method based on end-face features according to claim 1, characterized in that: In S3, contour detection employs a contour retrieval algorithm with a tree-structure retrieval mode and a simple contour approximation method. The algorithm focuses on extracting the outer contour of the end face, the central transmission hollow core contour, the cladding hollow core unit contour, and the pipe wall connection feature contour to form a complete contour set.

5. The anti-resonant fiber identification and high-precision axis alignment method based on end-face features according to claim 1, characterized in that: The specific method for classification and recognition in S4 is as follows: End face large contour screening and fitting: Traverse the contour set, screen the outer contour that satisfies the largest area and sufficient number of contour points, and use the ellipse fitting algorithm to fit the large contour of the fiber end face. Cladding hollow core topology analysis: Within the large contour of the end face, based on the contour area, spacing and arrangement rules, the contour of the cladding hollow core unit is selected and its arrangement topology is analyzed. Hollow core characteristic parameter calculation: Calculate the roundness of the hollow cladding profile. (Identifying circular hollow cores), arrangement density and ring number characteristics, where S is the contour area and L is the contour perimeter; Classification and determination: If the cladding hollow core is arranged in a single ring (such as a Kagome-type multi-circular hole single ring array), it is determined to be a basic anti-resonant fiber; if the cladding hollow core is arranged in a multi-ring nested pattern (such as a nested double / multi-ring hollow core structure), it is determined to be a complex nested anti-resonant fiber based on the nesting level and hollow core size difference characteristics, and further subdivision identification is completed.

6. The anti-resonant fiber identification and high-precision axis alignment method based on end-face features according to claim 1, characterized in that: The geometric parameter fitting in S6 adopts a contour fitting algorithm, specifically: Basic anti-resonant fiber fitting: An ellipse fitting algorithm is used for the central transmission hollow core profile, and a batch ellipse fitting algorithm is used for the single-ring cladding hollow core profile. The formula is as follows: ； Where (x0, y0) are the coordinates of the center of the fitted ellipse, and a and b are the major and minor axes of the ellipse, respectively; Fitting of complex nested anti-resonant optical fibers: an ellipse fitting algorithm is used for the central transmission hollow core profile, a ring-by-ring fitting strategy is used for the multi-ring nested cladding hollow core feature region, each ring cladding hollow core is fitted with an ellipse, and a polygon approximation fitting is used for the tube wall connection feature profile.

7. The anti-resonant fiber identification and high-precision axis alignment method based on end-face features according to claim 1, characterized in that: The alignment reference calculation method for the high-precision shaft model in S7 is as follows: The center of the fitted ellipse of the central transmission hollow core is used as the position alignment reference, and the main direction of the cladding hollow core arrangement is used as the angle alignment reference. Position alignment error: ； Pixels; of which , These are the coordinates of the center of the hollow core of each of the two optical fibers; Angle alignment error: ； in , These are the main direction angles of the hollow cladding arrangement of the two optical fibers.

8. A system for identifying and aligning anti-resonant optical fibers based on end-face features, used in the method for identifying and aligning anti-resonant optical fibers based on end-face features as described in any one of claims 1-7, characterized in that: It includes an image acquisition module, a preprocessing module, a contour detection module, a classification and recognition module, a parameter fitting module, and an axis calculation module.

9. The anti-resonant fiber optic identification and high-precision axis alignment system based on end-face features according to claim 8, characterized in that: The image acquisition module includes a camera and an optical magnification component, used to acquire end-face images of various types of anti-resonant optical fibers and transmit them to the preprocessing module. The acquisition direction is perpendicular to the central axis of the optical fiber. The preprocessing module sequentially performs grayscale conversion, Gaussian blur denoising, Sobel gradient calculation, binarization, and morphological operations on the image, and outputs the preprocessed image to the contour detection module. The contour detection module uses a contour retrieval algorithm to extract the outer contour of the end face, the central transmission hollow contour, the cladding hollow unit contour, and the pipe wall connection feature contour, and transmits the contour set to the classification and recognition module. The classification and recognition module is based on the "cladding hollow core topology" judgment logic to distinguish between basic and complex nested anti-resonant optical fibers, and completes the subdivision recognition of complex nested fibers, and outputs the classification results to the parameter fitting module. The parameter fitting module adapts the type fitting strategy according to the classification results and outputs key fitting parameters such as the center transmission hollow core and cladding hollow core to the axis calculation module. The alignment calculation module calculates the position alignment error and angle alignment error based on the fitting parameters and outputs the alignment result.

Images