A method and system for detecting fiber optic entanglement

Abstract

This application relates to the field of optical fiber entanglement detection technology, and discloses an optical fiber entanglement detection method and system. The method includes: acquiring image data of a target patch panel area under at least two different lighting conditions, performing spatial alignment and intensity correction to obtain a multi-conditional reflection observation image set; identifying port regions in the image set, determining port locations and types, and extracting optical fiber regions and constructing a set of candidate optical fiber paths; extracting reflection response features characterizing the surface state of the optical fiber based on the image response differences of the candidate optical fiber paths in the multi-conditional reflection observation image set; determining candidate abnormal regions based on the reflection response features, and combining port type, fiber exit direction, and the correspondence between the port and the candidate optical fiber paths to perform geometric morphological constraints and entanglement determination, thereby obtaining the optical fiber entanglement detection result. This invention can improve the accuracy and stability of optical fiber entanglement detection in dense patch panel scenarios.

Description

Technical Field

[0001] This application relates to the field of optical fiber winding detection technology, specifically to an optical fiber winding detection method and system. Background Technology

[0002] As data centers, communication equipment rooms, and structured cabling systems continue to expand, the number of fiber optic connections in patch panels is constantly increasing, and high-density, fine-pitch port arrangements have become common. In such scenarios, operations such as fiber optic cabling, patch cord adjustments, and port switching are performed frequently. If these tasks are still primarily performed manually, not only is the operational efficiency low, but the cabling quality is also easily affected by inaccurate port identification, incorrect fiber path judgment, or failure to detect local entanglements in a timely manner. To improve cabling efficiency and automation, fiber optic cabling robots are increasingly being applied to port identification, path positioning, and automatic insertion / removal operations.

[0003] In existing technologies, the recognition of patch panel images typically employs machine vision-based target detection, image segmentation, or feature matching methods to analyze port locations, port types, and fiber optic paths, thereby providing a basis for robots to perform automated wiring. These methods are effective in port identification, tag reading, and fiber optic contour extraction. However, in dense patch panel scenarios, multiple optical fibers often exhibit parallel contact, local overlap, obstruction, and complex background reflections. Especially when the fiber sheath surface is smooth, has weak texture, or high transparency, relying solely on color, edge, or geometric contour information from ordinary visible light images can easily lead to problems such as interrupted path recognition, confusion between adjacent fibers, and missed detection of tangled states.

[0004] Furthermore, in actual operations, fiber optic entanglement does not always manifest as obvious knots or large crossings; more often, it only presents as localized compression, abnormal contact, stress-induced bending, or slight twisting. For these early or latent anomalies, traditional two-dimensional image analysis methods can usually only obtain the planar projection shape of the fiber, making it difficult to effectively perceive changes in fiber surface pressure, normal changes, and local scattering characteristics. Therefore, it is difficult to accurately distinguish between normal bending and abnormal entanglement. On the other hand, components such as the metal frame, port adapters, and label protective layers in the patch panel can also generate complex reflections, further increasing the risk of false detection and missed detection.

[0005] Therefore, how to further improve the high-precision detection of optical fiber winding has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] The purpose of this application is to provide a method, system, and computer device for detecting optical fiber entanglement, so as to solve the problems mentioned in the background art.

[0007] According to a first aspect of this application, a method for detecting fiber optic entanglement is provided, applied to a fiber optic cabling robot, comprising the following steps: S1. Collect image data of the target patch panel area under at least two different lighting conditions, and perform spatial alignment and intensity correction to obtain a multi-conditional reflection observation image set; S2, perform port region identification on the multi-conditional reflection observation image set to determine the location and type of each port, perform initial positioning of the optical fiber region based on the port location, determine the path constraint rules of the corresponding optical fiber based on the port type, extract the optical fiber region and construct a set of candidate optical fiber paths. S3. For each optical fiber candidate path, the reflection interference corresponding to the port area is eliminated by shielding the port area or limiting the neighborhood of the optical fiber candidate path. Based on the image response differences of the optical fiber candidate paths in the multi-condition reflection observation image set, the reflection response features characterizing the surface state of the optical fiber are extracted. S4. Based on the reflection response characteristics, candidate abnormal regions are determined, and combined with the port type, fiber output direction and the correspondence between the port and the candidate optical fiber path, the candidate optical fiber path is geometrically constrained and entangled to obtain the optical fiber entanglement detection result.

[0008] In some embodiments, port region identification is performed on the multi-conditional reflectance observation image set to determine the location and type of each port, including: Feature extraction is performed on the image regions in the multi-conditional reflectance observation image set, including at least one of contour shape features, edge distribution features, array arrangement features, and marker region features. Based on the extracted features, port regions are identified from the image regions and the positions of each port are determined. The type of each port is determined based on at least one of the following: opening shape, size parameters, spacing between adjacent ports, arrangement method, and port identification information corresponding to the port area.

[0009] In some embodiments, reflection response features characterizing the surface state of the optical fiber are extracted based on the image response differences of candidate optical fiber paths in the multi-conditional reflection observation image set, including: Multiple sampling positions are selected along each candidate optical fiber path at a preset step size, and image response information, including brightness distribution, location of local bright areas, polarization response and scattering range, is obtained based on the local image regions corresponding to each sampling position under different lighting conditions. Based on the changes or differences in the image response information under different lighting conditions, the local reflection response features corresponding to each sampling position are calculated, and the local reflection response features are aggregated along the optical fiber candidate path direction to obtain the reflection response features characterizing the surface state of the corresponding optical fiber candidate path.

[0010] In some embodiments, candidate anomaly regions are determined based on the reflection response characteristics, and the candidate fiber paths are geometrically constrained and entangled in conjunction with the port type, fiber exit direction, and the correspondence between the port and the candidate fiber path to obtain fiber entanglement detection results, including: Anomaly detection is performed on the reflection response characteristics along each candidate optical fiber path at a preset step size, and the path segments whose reflection response characteristics exceed a preset anomaly threshold are identified as candidate anomaly regions. Based on the port type, determine at least one of the following for the corresponding optical fiber candidate path: allowable fiber exit form, bending range, overlap rule, or path continuity condition. Combine the fiber exit direction of the port and the correspondence between the port and the optical fiber candidate path to impose geometric constraints on the optical fiber candidate path. Based on the degree of deviation of the candidate abnormal region from the geometric constraints, the corresponding optical fiber candidate path is entangled to determine the fiber entanglement detection result.

[0011] In some embodiments, the method further includes: The system outputs the fiber optic entanglement detection results and the corresponding abnormal location information, and generates operation control instructions for the fiber optic cabling robot to execute based on the fiber optic entanglement detection results.

[0012] According to a second aspect of this application, a fiber optic entanglement detection system is provided for use in a fiber optic cabling robot, the system comprising: The image acquisition module is used to acquire image data of the target patch panel area under at least two different lighting conditions, and perform spatial alignment and intensity correction to obtain a multi-conditional reflectance observation image set; The port identification and path construction module is used to identify port regions in the multi-conditional reflection observation image set to determine the location and type of each port, start the fiber region based on the port location, determine the path constraint rules of the corresponding fiber based on the port type, extract the fiber region and construct a set of candidate fiber paths. The reflection feature extraction module is used to exclude reflection interference corresponding to the port area by shielding the port area or limiting the neighborhood of the optical fiber candidate path for each optical fiber candidate path, and extract reflection response features that characterize the surface state of the optical fiber based on the image response differences of the optical fiber candidate paths in the multi-condition reflection observation image set. The detection and judgment module is used to determine candidate abnormal regions based on the reflection response characteristics, and to perform geometric morphological constraints and entanglement judgment on the candidate optical fiber paths in combination with the port type, fiber output direction and the correspondence between the port and the candidate optical fiber path, so as to obtain the optical fiber entanglement detection result.

[0013] In some embodiments, the port identification and path construction module is specifically used for: Feature extraction is performed on the image regions in the multi-conditional reflectance observation image set, including at least one of contour shape features, edge distribution features, array arrangement features, and marker region features. Based on the extracted features, port regions are identified from the image regions and the positions of each port are determined. The type of each port is determined based on at least one of the following: opening shape, size parameters, spacing between adjacent ports, arrangement method, and port identification information corresponding to the port area.

[0014] In some embodiments, the reflection feature extraction module is specifically used for: Multiple sampling positions are selected along each candidate optical fiber path at a preset step size, and image response information, including brightness distribution, location of local bright areas, polarization response and scattering range, is obtained based on the local image regions corresponding to each sampling position under different lighting conditions. Based on the changes or differences in the image response information under different lighting conditions, the local reflection response features corresponding to each sampling position are calculated, and the local reflection response features are aggregated along the optical fiber candidate path direction to obtain the reflection response features characterizing the surface state of the corresponding optical fiber candidate path.

[0015] In some embodiments, the detection and determination module is specifically used for: Anomaly detection is performed on the reflection response characteristics along each candidate optical fiber path at a preset step size, and the path segments whose reflection response characteristics exceed a preset anomaly threshold are identified as candidate anomaly regions. Based on the port type, determine at least one of the following for the corresponding optical fiber candidate path: allowable fiber exit form, bending range, overlap rule, or path continuity condition. Combine the fiber exit direction of the port and the correspondence between the port and the optical fiber candidate path to impose geometric constraints on the optical fiber candidate path. Based on the degree of deviation of the candidate abnormal region from the geometric constraints, the corresponding optical fiber candidate path is entangled to determine the fiber entanglement detection result.

[0016] In some embodiments, the system further includes: The result output and control module is used to output the optical fiber entanglement detection result and the corresponding abnormal location information, and generate operation control instructions for the optical fiber distribution robot to execute based on the optical fiber entanglement detection result.

[0017] According to a third aspect of this application, a computer device is provided for use in a fiber optic cabling robot, including a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, implements the method as described in any of the preceding claims.

[0018] According to a fourth aspect of this application, a computer-readable storage medium is provided, wherein computer program instructions are stored thereon; the computer program instructions, when executed by a processor, implement the method as described in any of the preceding claims.

[0019] Compared with existing technologies, this invention acquires the surface reflection response of optical fibers under multiple lighting conditions and performs joint analysis by combining port type, fiber output direction, and the correspondence between the port and the optical fiber path. This expands the traditional geometric recognition based on two-dimensional images into a detection method that integrates reflection characteristics and structural constraints. It can effectively perceive latent features such as changes in the normal direction of the optical fiber surface, local pressure, and scattering anomalies. This allows for accurate differentiation between normal bending and abnormal entanglement in dense patch panels, reducing the risk of false detection and missed detection caused by fiber contact, obstruction, or low texture. It also improves the accuracy and stability of entanglement detection and enhances the system's adaptability in complex lighting and high-density cabling scenarios. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a functional structure diagram of the fiber optic distribution robot provided in the embodiments of this application; Figure 2 A schematic flowchart illustrating an optical fiber entanglement detection method provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an optical fiber winding detection system provided in an embodiment of this application; Figure 4 This is another structural schematic diagram of an optical fiber winding detection system provided in an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0023] In this embodiment, the fiber optic distribution robot 100 is used to perform image acquisition, fiber optic identification, and automated operation in the patch panel area, such as... Figure 1 As shown, its hardware system may include a vision acquisition unit 101, a lighting control unit 102, a mechanical execution unit 103, a control processing unit 104, and a communication interface unit 105, etc.

[0024] The vision acquisition unit is used to acquire image data of the patch panel area and may include an industrial camera and an optional polarization imaging component. The industrial camera is preferably a high-resolution area array camera to ensure clear differentiation of adjacent optical fibers even in high-density port scenarios. The polarization imaging component may include a polarization filter or a polarization camera to acquire image response information under different polarization directions, thereby enhancing the ability to perceive the reflection characteristics of the optical fiber surface.

[0025] The lighting control unit is used to provide at least two different lighting conditions, which may include a ring-shaped uniform light source, a lateral directional light source, and an optional polarized light source. By controlling the activation sequence, incident angle, and light intensity parameters of different light sources, multi-condition lighting acquisition is achieved, thereby obtaining a multi-condition reflection observation image set. Preferably, the lighting control unit and the vision acquisition unit are triggered synchronously to ensure the temporal consistency of image acquisition under different lighting conditions.

[0026] The mechanical actuator is used to perform grasping, moving, and insertion / removal operations on the optical fiber, and may include a multi-degree-of-freedom robotic arm, an end effector, and a position feedback component. The end effector may employ a flexible clamping structure to avoid damaging the optical fiber; the position feedback component may include an encoder or a vision guidance module for acquiring the robotic arm's pose information in real time.

[0027] The control processing unit is used to execute image processing and entanglement detection algorithms, and may include an embedded processor, an industrial computer, or a computing platform equipped with a GPU. This unit is used to implement functions such as multi-condition image alignment and correction, port identification, fiber path extraction, reflection response feature calculation, and entanglement determination, and generates corresponding operation control instructions.

[0028] The communication interface unit is used to realize data interaction between various hardware modules and between the robot and external systems. It may include Ethernet interface, industrial bus interface or serial communication interface, etc., for transmitting image data, control commands and status information.

[0029] like Figure 2 As shown in the figure, this application discloses a method for detecting fiber optic entanglement, applied to a fiber optic cabling robot, including the following steps: S1. Collect image data of the target patch panel area under at least two different lighting conditions, and perform spatial alignment and intensity correction to obtain a multi-conditional reflection observation image set; Before the fiber optic distribution robot performs automatic plugging / unplugging or inspection operations, high-quality image information of the area to be inspected is first acquired. Considering the highly reflective components such as the metal frame and port adapters in the patch panel scenario, as well as the complex reflective characteristics that may exist on the fiber optic surface itself, a single lighting condition is insufficient to comprehensively and stably obtain the true surface condition of the fiber optic cable. Therefore, this step controls the robot's onboard lighting unit (i.e., the aforementioned lighting control unit) to continuously or sequentially capture image data of the target patch panel area under at least two different lighting conditions (e.g., changing the incident angle, intensity, or polarization state of the light source).

[0030] Because robots may experience slight positional shifts under different lighting conditions, precise spatial alignment is required between the acquired images. This is typically achieved using feature point matching (such as SIFT or ORB features) or phase-correlation registration methods to ensure consistent pixel coordinates for the same physical point in different images. Simultaneously, to eliminate the impact of differences in light source intensity, radiation intensity correction is also necessary to ensure comparability of grayscale values ​​between different images.

[0031] After the above processing, a set of multi-conditional reflectance observation images that are strictly aligned in space and have consistent radiation intensity is finally formed, denoted as ,in Each image corresponds to a specific lighting condition.

[0032] S2, perform port region identification on the multi-conditional reflection observation image set to determine the location and type of each port, perform initial positioning of the optical fiber region based on the port location, determine the path constraint rules of the corresponding optical fiber based on the port type, extract the optical fiber region and construct a set of candidate optical fiber paths. After obtaining the aligned multi-condition image set, the key structure in the scene, namely the fiber optic port, is analyzed.

[0033] In some embodiments, port region identification is performed on the multi-conditional reflectance observation image set to determine the location and type of each port, including: S21, extract features from the image region in the multi-conditional reflection observation image set, including at least one of contour shape features, edge distribution features, array arrangement features and marker region features, and identify port regions from the image region based on the extracted features and determine the position of each port; Specifically, since patch panel ports typically have a regular rectangular or square appearance and are arranged in an array at fixed intervals, port location can be effectively achieved through multi-level feature extraction and fusion analysis of the images. First, the image with the most uniform illumination and best contrast (e.g., an image acquired using a ring-shaped uniform light source) is selected from a multi-conditional reflectance observation image set as the main processing image to reduce the complexity of subsequent processing. This image is then preprocessed, including grayscale conversion, Gaussian filtering for noise reduction, and histogram equalization, to enhance the contrast between the port area and the background.

[0034] Based on the above preprocessing, contour shape features are extracted. The Canny operator is used to extract the edge response of the image, and morphological closing operations (such as dilation followed by erosion) are used to connect edge breaks caused by uneven lighting or dirt, forming complete edge connected regions. Subsequently, Hough transform or Douglas-Peucker contour approximation algorithm is applied to filter contours with rectangular or square geometric features from the connected regions. For each candidate contour, its shape descriptors such as area, aspect ratio, and rectangularity are calculated and compared with a preset port size range. For example, the actual physical size of the opening of a standard SC type adapter is approximately 12.7mm × 10.3mm, and its pixel size in the image needs to be converted to approximately 127 × 103 pixels according to camera calibration parameters (such as pixel equivalent 0.1mm / pixel); the LC type port is represented by a smaller rectangular or double rectangular structure. By setting a shape similarity threshold (such as aspect ratio error less than 15%, rectangularity greater than 0.85), contours that meet the conditions can be initially retained as candidate port regions.

[0035] Edge distribution features are used to verify the rationality of candidate region locations. Gradient projection is performed on the preprocessed image in both horizontal and vertical directions. The cumulative edge pixel values ​​of each row and column are calculated to generate projection curves. Since the port array has fixed row and column spacing, periodic peaks appear on the projection curves. The peak positions correspond to the gaps between ports, while the trough positions correspond to the center row or column of the ports. By analyzing the peak intervals of the projection curves, the row and column boundaries of the ports can be estimated, thus obtaining the grid position of each port. For example, for a 1U height, 24-port patch panel, it is typically arranged in two rows and 12 columns, with a row spacing of approximately 20mm and a column spacing of approximately 17.5mm. The center coordinates of the candidate regions obtained from contour detection are matched with the grid positions. If a candidate region deviates from the grid position by more than a certain tolerance (e.g., 5 pixels), it may be a false detection and requires further verification using other features.

[0036] In addition, patch panel ports are usually marked with numbers, barcodes, or color blocks above or below them. These markings have a fixed spatial relationship with the ports (e.g., the marking is directly above the port, and the center-to-center distance is fixed at 8mm). The marking regions in the image are detected using OCR recognition or template matching methods. For example, a template library containing digits 0-9 and common letters (e.g., A, B) is pre-built. Normalized cross-correlation matching is performed within a preset search window above the port area. A valid marking is identified when the correlation coefficient exceeds 0.7. Based on the identified marking position and combined with a priori spatial offset (e.g., the vertical offset of the marking center relative to the port center is -50 pixels), the precise coordinates of the port center point are calculated. This method is particularly suitable for situations where port edges are blurred or contour extraction is incomplete, significantly improving positioning robustness.

[0037] Finally, the final port location is output using a fusion decision-making mechanism, specifically: For each candidate port, a weighted confidence score is calculated by combining its contour shape matching degree, edge projection matching degree, and correlation with the identified region. For example, the contour shape matching degree has a weight of 0.4, the edge projection matching degree has a weight of 0.3, and the identification correlation degree has a weight of 0.3. When the confidence score exceeds a preset threshold (e.g., 0.85), the region is confirmed as a valid port, and its center point pixel coordinates are output. For candidate regions with low confidence, images under other lighting conditions (e.g., side-illuminated port depth information) from a multi-conditional reflectance image set can be used for verification to further improve the reliability of recognition.

[0038] S22, based on at least one of the following: opening shape, size parameters, spacing between adjacent ports, arrangement method, and port identification information corresponding to the port area, determine the type of each port.

[0039] Because different types of optical fibers (such as SC, LC, and FC) have different physical dimensions, fiber exit methods, and allowable bending radii, these parameters directly affect the formulation of subsequent path constraint rules. This step achieves accurate port type identification through multi-dimensional feature fusion, as detailed below: Preliminary classification is performed based on the opening morphology of the port regions. For each port region identified in step S21, its magnified local image is extracted, and the internal geometry of the opening is analyzed using Hough circle detection or morphological skeleton extraction methods. For example, SC-type ports are typically rectangular openings with beveled guide grooves, and the circular end face of the ceramic ferrule is visible inside the opening. A circular region with a diameter of approximately 2.5 mm can be identified using Hough circle detection. LC-type ports appear as two closely spaced small circular openings (duplex LC) or a single small circular opening (simplex LC), with a smaller opening diameter (approximately 1.25 mm), and the center distance between the two circular holes in a duplex LC is approximately 6.25 mm. FC-type ports show shadows or edge features of threaded structures on their periphery, which can be identified by analyzing the texture directionality around the opening. A lightweight convolutional neural network (such as MobileNetV2) can be pre-trained to classify the opening region images, outputting a probability distribution of SC, LC, FC, etc., and selecting the type with the highest probability as the initial label.

[0040] A refined identification process is performed by combining dimensional parameters with the spacing between adjacent ports. Camera calibration parameters are used to convert the pixel width and height of the port area into actual physical dimensions, which are then compared with a pre-stored library of standard port dimensions. For example, SC ports typically have a width of 12.5-13.0 mm and a height of 10.0-10.5 mm; LC ports typically have a width of 5.0-6.0 mm and a height of 5.0-6.0 mm. If the measured value falls within a certain type of dimensional tolerance range (e.g., ±5%), the matching score for that type is increased. Simultaneously, the center-to-center distance between adjacent ports is measured: for LC duplex ports, the center-to-center distance between two small openings within the same port should be 6.25 mm; for SC ports, the center-to-center distance between adjacent ports (belonging to different port units) is typically 17.5 mm. If the center-to-center distance between two adjacent circular openings is close to 6.25 mm, it can be identified as an LC duplex port; if the center-to-center distance between adjacent rectangular areas is close to 17.5 mm, it is more likely to be an SC port. By analyzing the consistency between the arrangement method (simplex / duplex) and the spacing, the fuzzy results of the opening shape classification can be effectively corrected.

[0041] Port identification information is used for auxiliary verification. The identification area near the port not only contains numerical codes, but sometimes also directly prints the port type name (such as "SC" or "LC") or uses specific colors (such as blue for SC / UPC and green for SC / APC). OCR is used to recognize the text content in the identification area; if "SC" is recognized, it is directly confirmed as type SC; if "LC" is recognized, it is confirmed as type LC. Furthermore, color histogram analysis of the identification area extracts the dominant color tone, and comparison with a preset color-type mapping table (such as blue → SC, green → SC / APC, beige → LC) provides additional type evidence.

[0042] The final port type is determined by combining various information sources using a weighted fusion or decision tree model. For example, if the opening shape is classified as SC and the size matching degree is high (≥0.8), it is directly determined to be SC; if the opening shape is LC but the size matching degree is low (<0.5), the spacing is checked to see if it is 6.25mm, and if so, it is determined to be LC duplex; if the identification result conflicts with the shape classification, the identification information is used first. After determining the port type, the path constraint rules corresponding to that type are automatically associated, including fiber outer diameter, minimum bending radius (e.g., 30mm for SC type, 20mm for LC type), typical fiber exit direction (e.g., SC is mostly straight out, LC allows lateral bending), and overlap tolerance, etc.

[0043] After port identification and type determination, the precise location of each port is used as the starting point to search for and extract the connected fiber optic regions. Specifically, based on the continuous characteristics of optical fibers in images—that is, linear structures with a certain width (usually 20-30 pixels) and maintaining consistent grayscale or texture along the extension direction—image segmentation and path tracing algorithms are used for fiber extraction. For example, using the center of the fiber outlet as the seed point, dynamic programming or... The path search algorithm, guided by the fiber exit direction at the port, iteratively extends along the direction of least gradient change in the image until it reaches another port or exceeds a preset search range, thereby obtaining a binary mask covering the entire fiber region and its central skeleton line. For cases of breakage or branching due to occlusion or close proximity, the algorithm can generate multiple candidate paths, forming a set of fiber candidate paths. , in This represents the total number of candidate paths.

[0044] Simultaneously, based on each port type, a pre-set path constraint rule base is automatically associated to assign corresponding physical constraint parameters to each candidate fiber path. This path constraint rule base, indexed by port type, stores the geometric and physical characteristics of various fiber types, including but not limited to: fiber outer diameter (e.g., 3.0mm for SC type, 2.0mm for LC type), minimum allowable bending radius (e.g., 30mm for SC type, 20mm for LC type), allowable fiber exit form (e.g., SC type is mostly straight exit, allowing deviation from the normal ±15°; LC type allows more flexible lateral bending), overlap tolerance (e.g., an overlap angle exceeding 45° or an overlap length exceeding 10mm is considered abnormal), and path continuity requirements (e.g., a path interruption length exceeding 20mm needs to be marked as suspicious).

[0045] S3. For each optical fiber candidate path, the reflection interference corresponding to the port area is eliminated by shielding the port area or limiting the neighborhood of the optical fiber candidate path. Based on the image response differences of the optical fiber candidate paths in the multi-condition reflection observation image set, the reflection response features characterizing the surface state of the optical fiber are extracted. To eliminate the interference of strong reflections in the port area on the extraction of fiber surface features, this step uses one or a combination of the following two methods: (1) Shielding the port area. Based on the precise location of each port and its corresponding binary mask, all pixels within the mask coverage area are set to invalid values ​​(e.g., grayscale values ​​are set to zero or marked as ignored areas) during image processing. Since port adapters, ceramic ferrules, and metal frames typically have high reflectivity, they are prone to saturated highlights or specular reflections under various lighting conditions. If these areas are not shielded, they will introduce a large amount of noise in subsequent reflection response feature calculations, and may even drown out the weak reflection signal of the optical fiber itself. By using pixel-level mask shielding, it can be ensured that all subsequent feature extraction operations only apply to non-port areas, thereby eliminating the interference of port reflections.

[0046] (2) Limit the neighborhood of candidate fiber paths. Strictly limit the analysis scope to each candidate fiber path. Within a narrow neighborhood. In specific implementation, based on the central skeleton line of the path, a preset width of pixels is extended to both sides (for example, the extension width is set to 15-20 pixels according to the actual diameter of the fiber, which is sufficient to cover the radial cross-section of the fiber), forming a strip-shaped region of interest. The width design of this region needs to meet the following requirements: it should be slightly larger than the projection width of the fiber in the image (usually 20-30 pixels) to completely contain the fiber body; at the same time, it should be as compact as possible to exclude other adjacent fibers, port edges, or background structures. For parts of the path where the direction changes due to curvature, the neighborhood rotates dynamically accordingly, always maintaining cross-sectional coverage orthogonal to the local path direction.

[0047] The above two methods can be used alone or in combination. Preferably, the port area is first shielded by method (1), and then the remaining fiber path after shielding is subject to the neighborhood limitation of method (2), so as to purify the analysis area to the greatest extent and ensure that the reflection response features extracted in subsequent steps S31 and S32 truly reflect the surface state of the fiber itself, rather than pseudo features contaminated by port reflection.

[0048] In some embodiments, reflection response features characterizing the surface state of the optical fiber are extracted based on the image response differences of candidate optical fiber paths in the multi-conditional reflection observation image set, including: S31, select multiple sampling positions along each optical fiber candidate path with a preset step size, and obtain image response information based on the local image region corresponding to each sampling position under different lighting conditions, including brightness distribution, location of local bright areas, polarization response and scattering range. Specifically, for the set of candidate fiber paths Each path in Sampling points are set along its extension direction with a fixed step size (e.g., every 5 pixels) to form a sampling point sequence. It should be noted that the choice of step size needs to balance computational efficiency and the precision of feature representation. A step size that is too large will miss subtle local changes, while a step size that is too small will introduce redundant calculations and increase noise sensitivity.

[0049] For each sampling point In the multi-conditional reflectance observation image set In each image, a local image window (e.g., 15×15 pixels) centered at that point is extracted. This window size should be large enough to cover the radial cross-section of the fiber while maintaining sensitivity to local variations. From each local window, the following image response information is extracted: Brightness distribution: Calculates the mean, standard deviation, and grayscale histogram distribution of pixels within the calculation window. Under different lighting conditions, the brightness distribution of a normal optical fiber surface should exhibit similar statistical characteristics, while abnormal areas may show significantly enhanced or weakened reflection under a certain lighting direction due to changes in the surface normal.

[0050] Location of local highlight areas: Connectivity analysis is performed on pixels within the window whose grayscale values ​​exceed the global mean by a certain multiple (e.g., 1.5 times) to identify the center coordinates of the highlight areas. Under different lighting conditions, the positional offset of the highlight areas reflects the change in the curvature of the fiber surface. For example, in the left-hand illumination image... Mid-high light appears at the left edge of the fiber, while the right side illuminates the image. The mid-highlight shifts to the right edge; this symmetrical shift is a typical feature of a normal cylindrical surface. If the highlight position remains fixed or the shift is abnormal, it may indicate surface cladding or distortion.

[0051] Polarization response: If the vision acquisition unit is equipped with a polarization imaging component, it acquires image responses under different polarization directions (e.g., 0°, 45°, 90°, 135°). Parameters such as the degree of polarization and polarization angle are calculated, as these parameters are extremely sensitive to the stress state of the fiber surface. For example, scattered light from a normal fiber surface typically maintains certain polarization characteristics, while in stressed areas, changes in surface microstructure may lead to enhanced depolarization effects.

[0052] Scattering range: The spatial distribution range of gray values ​​within the analysis window, used to calculate the full width at half maximum (FWHM) or energy diffusion radius. When an optical fiber is bent, microcracks or deformations may occur on its surface, leading to an increase in the scattering range.

[0053] S32, based on the changes or differences in the image response information under different lighting conditions, calculate the local reflection response features corresponding to each sampling position, and aggregate each of the local reflection response features along the optical fiber candidate path direction to obtain the reflection response features characterizing the surface state of the corresponding optical fiber candidate path.

[0054] With dual lighting conditions (left-side lighting) and right-side lighting For example, for sampling points Calculate the following differences: Brightness difference ratio: Calculation and The ratio of the average brightness of a local window ,in To prevent division by zero, a small constant is used. This ratio reflects the reflection symmetry of the fiber surface to illumination from different directions. On a normal cylindrical surface, It should be close to 1.0; if A value significantly greater than 1 or less than 1 indicates that the surface normal has been deflected, causing positive reflection of illumination in one direction and shadow in another. This is a typical characteristic of fiber bending or twisting.

[0055] Specular offset: Calculation and Euclidean distance at the center of the local highlighted area For a normal optical fiber, the highlights under dual illumination should appear at the two edges of the fiber, with an offset approximately equal to the pixel width of the fiber diameter; if... A value much smaller than normal indicates planarization and cladding on the fiber surface; if If the value is much larger than the normal value, it may be accompanied by abnormal distortion.

[0056] Polarization difference: Calculates the difference in degree of polarization or polarization angle under different lighting conditions. Stress concentration regions often lead to significant changes in polarization characteristics. Exceeding a certain threshold can be used as an indication of an anomaly.

[0057] Scattering range ratio: Calculation and The ratio of the mid-scattering range (e.g., full width at half maximum) to the mid-scattering range. This reflects the directional changes in scattering characteristics.

[0058] Based on this, the above-mentioned multiple differential features are combined into a multidimensional local reflection response feature vector, namely: This vector condenses the sampling points Information on the physical state of the fiber optic surface at that location.

[0059] After calculating the local features of all sampling points, feature aggregation is performed along the candidate fiber path. A preferred approach is to concatenate the local feature vectors of each sampling point in path order to form a two-dimensional feature matrix. ,in The number of sampling points. This represents the local feature dimension. It should be understood that this matrix fully records the surface state change trajectory of the entire optical fiber path from the starting point to the ending point.

[0060] Understandably, traditional features mainly rely on the contrast between the optical fiber and the background, which is prone to failure when the optical fiber is close to the background or the background is complex; while the reflection response features extracted in this step Based on the physical response of the optical fiber's surface to multidirectional illumination, even if the fiber is similar in color to the background or partially obscured, the reflection differences caused by changes in its surface normal can still be effectively detected. For example, when two optical fibers are closely attached, the boundary may be difficult to distinguish visually. However, the surface normal changes due to mutual compression in the attached area, resulting in a different reflection pattern under left and right side illumination compared to normally separated fibers, which can be effectively captured by the reflection response characteristics. In this way, this step transforms latent anomalies that are difficult to perceive in traditional two-dimensional image analysis into quantifiable feature representations, thereby improving the accuracy of subsequent entanglement determination.

[0061] S4. Based on the reflection response characteristics, candidate abnormal regions are determined, and combined with the port type, fiber output direction and the correspondence between the port and the candidate optical fiber path, the candidate optical fiber path is geometrically constrained and entangled to obtain the optical fiber entanglement detection result.

[0062] Since reflection response characteristics can sensitively reflect microscopic physical changes on the fiber surface, but not all changes correspond to actual entanglement faults—for example, normal fiber bending can also cause surface normal deflection, resulting in similar reflection response anomalies—this step introduces geometric constraints to combine reflection response characteristics with the physical structural rules of the fiber, enabling accurate determination of entanglement.

[0063] In some embodiments, candidate anomaly regions are determined based on the reflection response characteristics, and the candidate fiber paths are geometrically constrained and entangled in conjunction with the port type, fiber exit direction, and the correspondence between the port and the candidate fiber path to obtain fiber entanglement detection results, including: S41, along each optical fiber candidate path, perform anomaly detection on the reflection response characteristics at a preset step size, and determine the path segment where the reflection response characteristics exceed the preset anomaly threshold as a candidate anomaly region. For each fiber candidate path The corresponding reflection response feature matrix ,in The number of sampling points. To determine the dimension of local reflection response characteristics, anomaly detection is performed segment by segment along the path direction. Considering that fiber surface anomalies typically manifest as abrupt changes in reflection response in local areas, a sliding window analysis method can be used: setting the length to... sampling points (e.g.) The detection window is set in steps. (For example Slide along the path and perform statistical analysis on the feature submatrix within each window.

[0064] For each window, calculate the mean of the feature vectors of all sampling points within it. and standard deviation And the global statistics of the mean and standard deviation of the entire path. A comparison is then made. Preferably, the anomaly metric is Mahalanobis distance:

[0065] in This is the covariance matrix of the global features. Mahalanobis distance can effectively measure the deviation of the current window feature distribution from the global normal distribution, and automatically takes into account the correlation between each feature dimension.

[0066] when Exceeding the preset abnormal threshold When this occurs, the path segment corresponding to the center of the window is marked as a candidate abnormal region. Threshold Calibration can be achieved through offline experiments: calculate the statistical distribution of Mahalanobis distance on a large number of normal fiber optic samples and take its 95th percentile as the benchmark threshold; or adopt an adaptive threshold strategy to dynamically adjust according to the overall noise level of the current scene.

[0067] As a supplementary or alternative approach, a pre-trained anomaly detection model (such as a support vector machine or an isolated forest) can be used to classify the feature vectors of each window and output the anomaly probability. The model takes the local reflectance response features of each sampling point within the window as input and, after training, can identify patterns that significantly deviate from the normal sample distribution. When the anomaly probability exceeds a preset threshold (e.g., 0.8), the segment is identified as a candidate anomaly region.

[0068] Using the above method, a set of candidate anomaly regions is generated along each fiber optic path. Each region corresponds to a fiber optic segment that may have surface anomalies, and its start and end coordinates and anomaly measurement values ​​are recorded.

[0069] S42, based on the port type, determine at least one of the allowed fiber exit form, bending range, overlap rule or path continuity condition of the corresponding optical fiber candidate path, and combine the fiber exit direction of the port and the correspondence between the port and the optical fiber candidate path to impose geometric constraints on the optical fiber candidate path. Candidate anomalous regions only indicate anomalies in the reflection response, but are insufficient to distinguish between normal bending and genuine entanglement. For example, the natural bending of an optical fiber when it bends around an obstacle also causes changes in the surface normal, resulting in similar reflection response characteristics. Therefore, this step aims to construct a personalized geometric constraint framework for each candidate optical fiber path as a benchmark for subsequent entanglement determination. This framework is constructed based on the following two types of input information: The first type of input consists of general rules determined by the port type. Based on the aforementioned determined port type, the physical constraint parameters corresponding to that type of optical fiber are retrieved from a pre-built rule database. This rule database is pre-constructed based on international optical fiber standards or offline experimental calibration results, and stores the geometric characteristic parameters of various optical fibers using port type as an index. These parameters include: permissible fiber exit form (e.g., the permissible angle range of deviation from the normal for straight exit). Bending range (minimum allowable bending radius) ), overlap rules (overlap angle threshold with adjacent optical fibers) and overlap length threshold ) and path continuity conditions (maximum allowable interruption length) For example, SC type optical fiber can be configured. Because of their smaller diameter, fiber optic cables allow for more flexible parameter settings, such as... .

[0070] The second type of input consists of specific information about the current path, including the port fiber exit direction and the correspondence between the port and candidate fiber paths. Among these, the port fiber exit direction vector... The path is obtained by analyzing the port opening direction or adapter installation direction; the correspondence between the port and the path is established by calculating the Euclidean distance between the path endpoint and the center point of each port. If the distance is less than a preset threshold (such as half the port spacing), the endpoint is considered to be associated with the port, and then it is determined whether the path is a complete link connecting two ports, a free end connected to a port at one end, or an invalid path that is eliminated.

[0071] Based on the two types of inputs mentioned above, a candidate fiber path is generated for the current fiber optic path. Geometric constraint framework It includes the following quantization parameters: Maximum allowable deviation angle from starting direction Based on the fiber outlet direction at the port, the angle between the average direction of the initial segment of the path and the fiber outlet direction must not exceed this value. Minimum allowable bending radius throughout the entire process The radius of curvature at any point on the path must not be less than this value. If the two ends of the path connect to different port types, the more stringent value shall apply. ; Overlap Angle Threshold With overlap length threshold When the current path has a spatially adjacent region with other paths, the overlap angle and overlap length must not exceed the corresponding thresholds. Maximum allowed interrupt length If there are broken segments on the path skeleton line with a length exceeding this value, they are considered discontinuous.

[0072] S43, based on the degree of deviation of the candidate abnormal region under the condition of satisfying the geometric shape constraint, the corresponding optical fiber candidate path is entangled to determine the entanglement result of the optical fiber.

[0073] For each candidate abnormal region Extract the path segment corresponding to this region and calculate its geometric parameters: local curvature radius. The angle between the path direction and the fiber output direction at the port The overlap angle with adjacent optical fiber candidate paths and overlap length Path interruption length The above geometric parameters are then compared with the geometric shape constraint framework. The corresponding thresholds are compared item by item, and the judgment rules are as follows: Determining the bending range: If If so, the area is determined to be an excessively bent type of winding.

[0074] Starting point direction determination: If the candidate abnormal region is located at the beginning of the path (distance from the port center is less than a preset value), and If the reflection response characteristics of this area are abnormal, it is determined to be a twisted winding at the fiber outlet.

[0075] Overlap rule determination: If the candidate anomaly region spatially overlaps with the candidate anomaly region of the adjacent fiber candidate path, and If so, it is determined to be a multi-fiber overlapping type of winding.

[0076] Continuity determination: If the candidate anomaly region is located in the middle of the path, and the region exists... If the fractured segment shows surface stress concentration in the reflection response characteristics, it is determined to be a local entanglement.

[0077] For edge cases that cannot be directly determined by a single rule, a weighted scoring mechanism is adopted: the deviation of each geometric parameter from the constraint is quantified into a score and assigned different weights (e.g., bending radius violation weight 0.5, direction deviation weight 0.3, overlap degree weight 0.2), and the overall entanglement probability is calculated. .like If the threshold value exceeds the preset threshold (e.g., 0.7), it is determined to be entangled.

[0078] The final output includes the entanglement detection results for each candidate fiber path, including the entanglement status label (normal / entangled), the specific location of the entanglement area, the entanglement type (excessive bending, fiber outlet twisting, multi-fiber overlap, etc.), and the corresponding anomaly measurement value.

[0079] In some embodiments, continue to refer to Figure 2 As shown, the method further includes: S5, output the optical fiber entanglement detection result and the corresponding abnormal location information, and generate operation control instructions for the optical fiber wiring robot to execute based on the optical fiber entanglement detection result.

[0080] First, the entanglement detection results are correlated with anomaly location information and output. The anomaly location information needs to be transformed from the image pixel coordinate system to the robot's base coordinate system so that the mechanical execution unit can accurately reach it. Specifically, for each entanglement area, the midpoint pixel coordinates are taken, and combined with camera intrinsic parameters and depth information, they are back-projected to a 3D point in the camera coordinate system. Then, using hand-eye calibration parameters, this point is transformed to the robot's base coordinate system to obtain precise spatial coordinates. These coordinates are then integrated with entanglement status labels, entanglement type, and other information to form a complete detection result data package.

[0081] Next, differentiated operation control instructions are generated based on the entanglement detection results. If no entanglement is detected, a regular operation instruction is generated, and the robot performs inspection or insertion / removal tasks according to a preset path. If entanglement is detected, corresponding instructions are generated according to the type and severity of the entanglement: for severe excessive bending, an alarm and avoidance instruction is generated, the area is marked as an obstacle, and manual handling is reported; for fiber outlet twisting, a local verification instruction is generated, and the vision unit is invoked for fine-tuning detection; for multiple fiber overlaps, if the robot has untangling capabilities, an untangling action sequence is generated; otherwise, it is marked as requiring manual handling.

[0082] The generated instructions are sent to the mechanical execution unit via a communication interface, and execution status feedback is received, enabling dynamic adjustment and anomaly handling. Optionally, all detection results and instruction execution records are stored in a local database or uploaded to the cloud for subsequent operation and maintenance analysis.

[0083] Reference Figure 3 As shown in the figure, this application embodiment also provides an optical fiber entanglement detection system 200, applied to an optical fiber distribution robot, the system comprising: Image acquisition module 201 is used to acquire image data of the target patch panel area under at least two different lighting conditions, and perform spatial alignment and intensity correction to obtain a multi-condition reflection observation image set; The port identification and path construction module 202 is used to identify port regions in the multi-conditional reflection observation image set to determine the location and type of each port, start the optical fiber region based on the port location, determine the path constraint rules of the corresponding optical fiber based on the port type, extract the optical fiber region and construct a set of optical fiber candidate paths. The reflection feature extraction module 203 is used to exclude reflection interference corresponding to the port area by shielding the port area or limiting the neighborhood of the optical fiber candidate path for each optical fiber candidate path, and extract reflection response features that characterize the surface state of the optical fiber based on the image response differences of the optical fiber candidate paths in the multi-condition reflection observation image set. The detection and judgment module 204 is used to determine candidate abnormal regions based on the reflection response characteristics, and combine the port type, fiber output direction and the correspondence between the port and the candidate optical fiber path to perform geometric morphological constraints and entanglement judgment on the candidate optical fiber path, so as to obtain the optical fiber entanglement detection result.

[0084] In some embodiments, the port identification and path construction module 202 is specifically used for: Feature extraction is performed on the image regions in the multi-conditional reflectance observation image set, including at least one of contour shape features, edge distribution features, array arrangement features, and marker region features. Based on the extracted features, port regions are identified from the image regions and the positions of each port are determined. The type of each port is determined based on at least one of the following: opening shape, size parameters, spacing between adjacent ports, arrangement method, and port identification information corresponding to the port area.

[0085] In some embodiments, the reflection feature extraction module 203 is specifically used for: Multiple sampling positions are selected along each candidate optical fiber path at a preset step size, and image response information, including brightness distribution, location of local bright areas, polarization response and scattering range, is obtained based on the local image regions corresponding to each sampling position under different lighting conditions. Based on the changes or differences in the image response information under different lighting conditions, the local reflection response features corresponding to each sampling position are calculated, and the local reflection response features are aggregated along the optical fiber candidate path direction to obtain the reflection response features characterizing the surface state of the corresponding optical fiber candidate path.

[0086] In some embodiments, the detection and determination module 204 is specifically used for: Anomaly detection is performed on the reflection response characteristics along each candidate optical fiber path at a preset step size, and the path segments whose reflection response characteristics exceed a preset anomaly threshold are identified as candidate anomaly regions. Based on the port type, determine at least one of the following for the corresponding optical fiber candidate path: allowable fiber exit form, bending range, overlap rule, or path continuity condition. Combine the fiber exit direction of the port and the correspondence between the port and the optical fiber candidate path to impose geometric constraints on the optical fiber candidate path. Based on the degree of deviation of the candidate abnormal region from the geometric constraints, the corresponding optical fiber candidate path is entangled to determine the fiber entanglement detection result.

[0087] In some embodiments, refer to Figure 4 As shown, the system also includes: The result output and control module 205 is used to output the optical fiber winding detection result and the corresponding abnormal position information, and generate operation control instructions for the optical fiber distribution robot to execute based on the optical fiber winding detection result.

[0088] Since the system embodiment and the method embodiment are based on the same technical concept, and employ a combination of multi-illumination condition fusion analysis and geometric constraint verification, they can effectively detect latent characteristics such as changes in the normal direction of the fiber surface, local pressure, and scattering anomalies. In dense patch panel scenarios, they can accurately distinguish between normal bending and abnormal entanglement, reducing the risk of false detections and missed detections caused by fiber contact, obstruction, or low texture. Therefore, their technical effects are similar to those of the method embodiment, and will not be repeated here. Those skilled in the art can construct and operate this system based on the description of the method embodiment.

[0089] This application also provides a computer device for use in a fiber optic cabling robot, including a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the method described in any of the preceding claims.

[0090] Specifically, the computer device, serving as the control and processing core of the fiber optic cabling robot, can be implemented using an embedded industrial computer, a PC-based controller, or a GPU-accelerated computing platform to meet the computational performance and real-time requirements of tasks such as image processing, feature extraction, model inference, and motion control. This computer device connects to various functional units via an internal bus, enabling centralized control and data interaction with the vision acquisition unit, lighting control unit, mechanical execution unit, and communication interface unit.

[0091] In terms of hardware architecture, the computer device includes at least a processor, memory, storage interface, communication interface, and power management module. The processor can employ a heterogeneous computing architecture combining a multi-core CPU and a GPU (such as the NVIDIA Jetson series). The CPU handles logic control and instruction scheduling, while the GPU handles image processing and parallel acceleration computation for deep learning models. The memory includes volatile memory (such as DDR4 RAM) and non-volatile memory (such as SSD or eMMC). Volatile memory loads runtime programs and data, while non-volatile memory stores the operating system, control programs, calibration parameters, and historical detection logs for long-term storage. The storage interface supports the expansion of external storage devices, facilitating data backup and system upgrades. The communication interface may include a Gigabit Ethernet interface, a CAN bus interface, an RS485 interface, and a digital I / O interface for real-time data exchange with other robot units and the upper-level management system. The power management module provides stable operating voltage for all components and supports low-power sleep and wake-up functions.

[0092] At the software level, the computer device runs a real-time operating system or a general-purpose operating system, on which a fiber optic entanglement detection and control program is deployed. After being loaded into memory, this control program is executed by the processor to implement all the steps of the aforementioned method: including controlling the synchronous triggering of the illumination unit and the vision acquisition unit to acquire a multi-conditional reflection observation image set; performing port identification and path construction on the images; extracting reflection response features; determining entanglement; and finally generating operation control commands to send to the mechanical execution unit. Intermediate data generated during program execution (such as port locations, candidate paths, reflection features, and abnormal areas) is temporarily stored in memory, and after the task is completed, key results (such as entanglement detection reports, abnormal images, and operation logs) are stored in non-volatile memory or uploaded to a cloud management platform.

[0093] Furthermore, to meet the computing power requirements of different scenarios, the computer device can be designed with a modular structure, supporting the expansion of dedicated acceleration cards (such as FPGAs or TPUs) via PCIe or Mini-PCIe interfaces to accelerate image preprocessing or neural network inference. In a distributed architecture, some computing tasks (such as deep learning model inference) can also be offloaded to edge servers, and the computer device can work collaboratively with the edge servers through communication interfaces to achieve computing resource sharing.

[0094] It is understood that those skilled in the art can select appropriate hardware platforms and software configurations according to actual application needs. As long as they can execute the aforementioned method steps, the fiber optic entanglement detection function described in the embodiments of this application can be realized. The computer device not only acts as the "brain" of the robot to complete data processing and decision-making, but also ensures the level of intelligence and automation of fiber optic cabling operations through closed-loop interaction with each execution unit.

[0095] This application also provides a computer-readable storage medium storing computer program instructions; when executed by a processor, the computer program instructions implement the method described in any of the preceding embodiments.

[0096] The above description represents the preferred embodiments of the present invention. It should be noted that, for those skilled in the art, various improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for detecting fiber optic entanglement, applied to a fiber optic cabling robot, characterized in that, Includes the following steps: Image data of the target patch panel area under at least two different lighting conditions were collected, and spatial alignment and intensity correction were performed to obtain a multi-conditional reflectance observation image set; Port region identification is performed on the multi-conditional reflection observation image set to determine the location and type of each port, and the fiber region is initially located based on the port location. The path constraint rules of the corresponding fiber are determined based on the port type, the fiber region is extracted, and a set of candidate fiber paths is constructed. For each optical fiber candidate path, reflection interference corresponding to the port area is eliminated by shielding the port area or limiting the neighborhood of the optical fiber candidate path. Based on the image response differences of the optical fiber candidate paths in the multi-condition reflection observation image set, reflection response features characterizing the surface state of the optical fiber are extracted. Based on the reflection response characteristics, candidate abnormal regions are determined, and combined with the port type, fiber output direction, and the correspondence between the port and the candidate fiber path, the candidate fiber path is geometrically constrained and entangled to obtain the fiber entanglement detection result.

2. The fiber optic entanglement detection method according to claim 1, characterized in that, Port region identification is performed on the multi-conditional reflectance observation image set to determine the location and type of each port, including: Feature extraction is performed on the image regions in the multi-conditional reflectance observation image set, including at least one of contour shape features, edge distribution features, array arrangement features, and marker region features. Based on the extracted features, port regions are identified from the image regions and the positions of each port are determined. The type of each port is determined based on at least one of the following: opening shape, size parameters, spacing between adjacent ports, arrangement method, and port identification information corresponding to the port area.

3. The fiber optic entanglement detection method according to claim 1, characterized in that, Based on the differences in image response of candidate fiber paths in the multi-conditional reflection observation image set, reflection response features characterizing the fiber surface state are extracted, including: Multiple sampling positions are selected along each candidate optical fiber path at a preset step size, and image response information, including brightness distribution, location of local bright areas, polarization response and scattering range, is obtained based on the local image regions corresponding to each sampling position under different lighting conditions. Based on the changes or differences in the image response information under different lighting conditions, the local reflection response features corresponding to each sampling position are calculated, and the local reflection response features are aggregated along the optical fiber candidate path direction to obtain the reflection response features characterizing the surface state of the corresponding optical fiber candidate path.

4. The optical fiber entanglement detection method according to claim 1, characterized in that, Based on the aforementioned reflection response characteristics, candidate anomaly regions are determined. Combined with the port type, fiber exit direction, and the correspondence between the port and the candidate fiber path, geometric constraints and entanglement determination are applied to the candidate fiber path to obtain fiber entanglement detection results, including: Anomaly detection is performed on the reflection response characteristics along each candidate optical fiber path at a preset step size, and the path segments whose reflection response characteristics exceed a preset anomaly threshold are identified as candidate anomaly regions. Based on the port type, determine at least one of the following for the corresponding optical fiber candidate path: allowable fiber exit form, bending range, overlap rule, or path continuity condition. Combine the fiber exit direction of the port and the correspondence between the port and the optical fiber candidate path to impose geometric constraints on the optical fiber candidate path. Based on the degree of deviation of the candidate abnormal region from the geometric constraints, the corresponding optical fiber candidate path is entangled to determine the fiber entanglement detection result.

5. The fiber optic entanglement detection method according to claim 1, characterized in that, The method further includes: The system outputs the fiber optic entanglement detection results and the corresponding abnormal location information, and generates operation control instructions for the fiber optic cabling robot to execute based on the fiber optic entanglement detection results.

6. A fiber optic entanglement detection system, applied to a fiber optic cabling robot, characterized in that, The system includes: The image acquisition module is used to acquire image data of the target patch panel area under at least two different lighting conditions, and perform spatial alignment and intensity correction to obtain a multi-conditional reflectance observation image set; The port identification and path construction module is used to identify port regions in the multi-conditional reflection observation image set to determine the location and type of each port, start the fiber region based on the port location, determine the path constraint rules of the corresponding fiber based on the port type, extract the fiber region and construct a set of candidate fiber paths. The reflection feature extraction module is used to exclude reflection interference corresponding to the port area by shielding the port area or limiting the neighborhood of the optical fiber candidate path for each optical fiber candidate path, and extract reflection response features that characterize the surface state of the optical fiber based on the image response differences of the optical fiber candidate paths in the multi-condition reflection observation image set. The detection and judgment module is used to determine candidate abnormal regions based on the reflection response characteristics, and to perform geometric morphological constraints and entanglement judgment on the candidate optical fiber paths in combination with the port type, fiber output direction and the correspondence between the port and the candidate optical fiber path, so as to obtain the optical fiber entanglement detection result.

7. The optical fiber winding detection system according to claim 6, characterized in that, The port identification and path construction module is specifically used for: Feature extraction is performed on the image regions in the multi-conditional reflectance observation image set, including at least one of contour shape features, edge distribution features, array arrangement features, and marker region features. Based on the extracted features, port regions are identified from the image regions and the positions of each port are determined. The type of each port is determined based on at least one of the following: opening shape, size parameters, spacing between adjacent ports, arrangement method, and port identification information corresponding to the port area.

8. The optical fiber winding detection system according to claim 6, characterized in that, The port identification and path construction module is specifically used for: Feature extraction is performed on the image regions in the multi-conditional reflectance observation image set, including at least one of contour shape features, edge distribution features, array arrangement features, and marker region features. Based on the extracted features, port regions are identified from the image regions and the positions of each port are determined. The type of each port is determined based on at least one of the following: opening shape, size parameters, spacing between adjacent ports, arrangement method, and port identification information corresponding to the port area.

9. The optical fiber winding detection system according to claim 6, characterized in that, The detection and determination module is specifically used for: Anomaly detection is performed on the reflection response characteristics along each candidate optical fiber path at a preset step size, and the path segments whose reflection response characteristics exceed a preset anomaly threshold are identified as candidate anomaly regions. Based on the port type, determine at least one of the following for the corresponding optical fiber candidate path: allowable fiber exit form, bending range, overlap rule, or path continuity condition. Combine the fiber exit direction of the port and the correspondence between the port and the optical fiber candidate path to impose geometric constraints on the optical fiber candidate path. Based on the degree of deviation of the candidate abnormal region from the geometric constraints, the corresponding optical fiber candidate path is entangled to determine the fiber entanglement detection result.

10. The optical fiber winding detection system according to claim 6, characterized in that, The system also includes: The result output and control module is used to output the optical fiber entanglement detection result and the corresponding abnormal location information, and generate operation control instructions for the optical fiber distribution robot to execute based on the optical fiber entanglement detection result.

11. A computer device used in a fiber optic cabling robot, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the method as described in any one of claims 1-5.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions; when executed by a processor, the computer program instructions implement the method as described in any one of claims 1-5.

Images