A method of visual activation of a fabric of optical fibers

Abstract

The application relates to the technical field of optical fiber fabric image processing, in particular to a kind of optical fiber fabric visual activation method;The method comprises using binocular industrial camera to collect optical fiber fabric image sequence, carries out gray normalization, contrast enhancement and directionality filter processing through structure guide type multi-scale image enhancement network, and strengthens optical fiber edge feature;Based on the optical fabric structure semantic analysis method, a light depth network is used to identify structure units and construct a direction perception semantic topology map, path tracking and function labeling are realized;Combined with multi-scale stereo matching and three-dimensional point cloud projection, three-dimensional reconstruction with millimeter-level precision is completed by fusing the semantic topology map;According to the model, the laser path is planned, and the gradient power control strategy is used to dynamically compensate the light attenuation of the optical fiber;Defects are detected in real time during laser engraving and automatic engraving is performed.The application finally realizes high-precision and high-efficiency optical fiber fabric activation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of optical fiber fabric image processing, and particularly relates to a kind of optical fiber fabric visual activation method. BACKGROUND

[0002] With the vigorous development of flexible optoelectronics and intelligent textile technology, optical fiber fabric has broad application prospects in light-emitting clothing, medical monitoring, human-computer interaction and other fields due to its unique light guiding characteristics. The core process of this technology requires the removal of optical fiber cladding to activate the light guiding function, but in practical applications, it still faces multiple challenges. The flexible nature of optical fiber fabric leads to fiber deformation and positioning deviation during processing. The complex spatial topological relationship formed by the intersection of warp and weft yarns and optical fibers in the microstructure requires higher three-dimensional geometric mapping. The influence of the inherent light attenuation effect of optical fibers on light emission uniformity during laser engraving also needs to be addressed.

[0003] Chinese invention patent with publication number CN118314138B discloses a laser processing method and system based on machine vision, which includes analyzing the video stream obtained by machine vision to detect visual features representing the area to be processed. By comparing the geometric shape of the visual features with the predefined shape template, the real-time outer contour is identified, the position information and coordinate mapping of the area to be processed on the surface of the object to be processed are calculated to compensate for the visual deviation. A three-dimensional visual model is constructed to optimize the focal length setting and path planning of the laser on the surface of the object to be processed, generate a laser processing trajectory document, control the laser to position to the initial processing point, and start the laser processing sequence. Through real-time video analysis and optical sensor feedback, the real-time image of the workpiece in the processing environment can be accurately captured, and the workpiece contour and posture can be identified and adjusted according to the image processing algorithm, improving the precision and adaptive ability of laser processing, and improving production efficiency and workpiece processing quality.

[0004] Therefore, the industry urgently needs a technical solution that integrates high-precision three-dimensional reconstruction, adaptive path planning and real-time closed-loop control to meet the stringent requirements of process precision and consistency for large-scale production. SUMMARY

[0005] The present application aims to solve the problems in the background art and proposes a kind of optical fiber fabric visual activation method.

[0006] The technical solution of the present application is a kind of optical fiber fabric visual activation method, which includes the following specific implementation steps:

[0007] S1, a kind of optical fiber fabric visual activation system is constructed, and the system includes:

[0008] A laser engraving device is used to generate a high-energy laser beam and engrave and activate the optical fiber fabric.

[0009] a visual detection device for collecting image sequences of the optical fiber fabric and performing image processing and three-dimensional reconstruction;

[0010] a three-axis mechanical transmission device for driving the laser head and the workpiece to achieve three-dimensional spatial positioning;

[0011] a central control system for coordinating control of the laser engraving device, the visual detection device, and the three-axis mechanical transmission device;

[0012] an auxiliary negative pressure fixing device for flatly laying the optical fiber fabric without wrinkles through vacuum adsorption;

[0013] S2, fixing the optical fiber fabric on the table top, collecting image sequences through the visual detection device, generating a three-dimensional model of the optical fiber fabric through image enhancement, structure semantic analysis, and three-dimensional reconstruction;

[0014] S3, planning an engraving path based on the three-dimensional model, activating the laser using a gradient power control strategy, and dynamically adjusting the Z-axis focal length;

[0015] S4, performing real-time defect detection and automatically engraving after the current scanning period ends.

[0016] Preferably, the image enhancement adopts a structure-guided multi-scale image enhancement network FIRE-Net, which sequentially performs:

[0017] Based on the gray value of each pixel of the original image, the global minimum and maximum gray values are used for linear mapping to the range of 0-255, the gray value of the image is normalized, and the overall gray offset caused by uneven lighting or device differences is eliminated;

[0018] Based on the gray value normalized image, a multi-scale Gaussian filter is used to extract and enhance the local brightness difference in different spatial ranges, and a combination model of weighted original gray value and multi-scale local mean difference value is used to amplify structural details, while an intensity coefficient is used to control to avoid noise introduction;

[0019] The structure tensor is constructed by calculating the pixel gradient vector, the eigenvalue is extracted and the minimum value is taken as the orthogonal component to represent the edge strength, and the neighborhood weighted filtering is performed combined with the double weight factors of gray similarity and structural strength to suppress noise and preserve the fiber edge and texture details;

[0020] Based on the estimated main direction of the optical fiber and the local edge direction difference of the pixel, a directional mask is constructed, and the original filtered image and the mask enhanced image are fused by weighting to directionally enhance the fiber direction texture features and suppress background interference.

[0021] Preferably, the structure semantic analysis adopts an optical fabric structure semantic analysis method, which sequentially performs:

[0022] The YOLOv5 target detection model optimized by pruning is used to process the input image, identify the main structural units in the optical fiber fabric, and output a set of bounding boxes, including class labels, center coordinates, width and height dimensions, and confidence scores.

[0023] The spatial correlation scores between the structural elements are calculated by the structural compatibility function, exponential decay of Euclidean distance, and reciprocal of the difference in direction angle, and a spatial relationship matrix is constructed.

[0024] A graph model is constructed based on the high correlation threshold connection between the vertex corresponding structural elements and the edges, and the functional semantics carried by the edge weights are used to realize the computable expression of the relationship between the structures.

[0025] Based on the topological graph, the optical fiber path is constructed by the direction consistency and continuous connection criteria, and the geometric continuity and direction consistency are optimized by the evaluation function to complete the path labeling and record the start and end coordinates, curvature, length direction, and hierarchical relationship.

[0026] Preferably, the three-dimensional reconstruction adopts an optical fiber fabric three-dimensional structure reconstruction method based on direction perception semantic fusion, which is sequentially executed:

[0027] The left and right images are synchronously collected by the binocular camera and a Gaussian pyramid is constructed, the disparity similarity is calculated by the weighted matching cost function of the gray difference and the gradient difference, the optimal disparity is selected to generate a preliminary disparity map after guided filtering aggregation of the neighborhood cost, and finally the continuous and smooth disparity map is output by median-bilateral filtering and edge perception correction after left-right consistency check, sub-pixel interpolation, and occlusion region filling optimization.

[0028] According to the camera calibration parameters, the three-dimensional point cloud set P is generated by the pixel disparity value d in the disparity map D(x, y) and the image coordinates (u, v) using the depth calculation and horizontal coordinate conversion formula:

[0029]

[0030] Where f is the camera focal length; B is the baseline distance of the left and right cameras; d is the pixel disparity value; (u, v) is the image coordinates, i.e. (x, y); (c x ,c y ) is the principal point coordinates; (x, y, z) is the corresponding three-dimensional coordinate point; and M is the number of effective pixels.

[0031] Preferably, in step S3:

[0032] Based on the optical fiber path information in the three-dimensional model, a scanning path is generated with a point spacing of 0.2 mm horizontally and a line spacing of 0.3 mm vertically.

[0033] Each point on the path contains spatial coordinates, direction information, and curvature information, which are used for subsequent Z-axis focusing and power control.

[0034] Preferably, in step S3:

[0035] The gradient power control strategy increases the laser power according to the position of the input end of the fiber path segment:

[0036] The initial position laser power is 10W, and is increased by 2% for every 1cm extension, compensating for the light guide attenuation effect and ensuring that the laser activation effect at the tail end of the path is not lower than that at the start end;

[0037] Meanwhile, the laser head position is dynamically adjusted in combination with the Z-axis height information, so that the laser focal point always falls on the cladding center position, and the engraving depth is controlled within the range of 20μm±2μm.

[0038] Preferably, in step S4:

[0039] The defect detection is based on two types of indexes: one is whether the gray value of the processing area is lower than the set activation threshold, and the other is whether the offset distance of the engraving contour from the target path exceeds the set limit value;

[0040] When any condition triggers the defect recognition, the system records the spatial coordinates of the point and adds it to the list of supplementary engraving, and after the scanning period ends, the laser head automatically returns to perform the supplementary engraving operation.

[0041] Preferably, the visual detection device comprises:

[0042] The image acquisition unit is a combination of a binocular industrial camera and a telecentric lens fixed above the object table, which acquires the image sequence of the fiber fabric and provides shadowless illumination using a ring LED array light source;

[0043] The image processing unit detects and models the target in the collected image sequence of the fiber fabric using a deep learning method.

[0044] Compared with the prior art, the above technical scheme of the present application has the following beneficial technical effects:

[0045] The present application provides a kind of fiber fabric visual activation method, by deep fusion machine vision, artificial intelligence and laser processing technology, in imaging quality, three-dimensional reconstruction accuracy and processing control system three dimensions has the following beneficial technical effects:

[0046] In the imaging level, the structure-guided multi-scale image enhancement network and the direction perception semantic topology analysis technology are innovatively combined to effectively overcome the feature blurring problem caused by the interference of fabric surface reflection and fiber overlap in traditional methods, and the microstructure is accurately identified and continuously modeled;

[0047] In the dimension of 3D reconstruction and processing control, multi-scale stereo matching and semantic topology mapping technology are integrated to break through the spatial limitations of 2D images and construct a 3D model with complete physical properties and functional semantics. At the same time, a gradient power dynamic compensation mechanism is introduced to adaptively offset the inherent light-guiding attenuation effect of optical fiber, ensure the uniformity of light emission intensity throughout the path, and completely solve the problem of uneven light emission in traditional fixed power engraving process.

[0048] At the end-to-end control level, a real-time defect detection and automatic repair closed-loop system is constructed. Through synchronous quality monitoring and dynamic correction during high-speed processing, the consistency of processes and product yield are significantly improved. The central control system achieves intelligent closed-loop control from visual perception and path planning to laser execution through multi-module collaboration, which greatly improves the automation level and process robustness of complex fiber optic fabric processing. Attached Figure Description

[0049] Figure 1 This is a flowchart of a method for visual activation of fiber optic fabric proposed in this invention;

[0050] Figure 2 This is a schematic diagram of the laser activation process proposed in this invention;

[0051] Reference numerals: 1. Optical fiber, 2. Light source controller, 3. Weft fiber, 4. Warp yarn, 5. Laser head, 6. Laser. Detailed Implementation

[0052] Example 1, as Figure 1 As shown, the present invention proposes a method for visual activation of fiber optic fabrics, which includes the following specific implementation steps:

[0053] S1. Construct a fiber optic fabric visual activation system that integrates a laser engraving device, a visual inspection device, a three-axis mechanical transmission device, a central control system, and an auxiliary negative pressure fixing device. Specifically, it includes:

[0054] Laser engraving device, comprising laser, optical system, and adjustment components;

[0055] Lasers are used to generate high-energy laser beams for engraving or cutting materials.

[0056] An optical system is used to adjust the path and focus of the laser beam to ensure engraving accuracy;

[0057] Adjustment components are used to flexibly adjust the position of the laser head to adapt to different processing needs;

[0058] The visual inspection device includes: an image acquisition unit and an image processing unit;

[0059] An image acquisition unit, which is a combination of a binocular industrial camera and a telecentric lens fixed above the workbench, acquires image sequences of the optical fiber fabric and provides shadowless illumination using a ring-shaped LED array light source;

[0060] An image processing unit detects and models the target in the acquired image sequences of the optical fiber fabric using a deep learning method;

[0061] A three-axis mechanical transmission device, which includes a conveying belt and clamping mechanism, a servo motor and driver, and a photoelectric sensor;

[0062] The conveying belt and clamping mechanism are used for vacuum adsorption and fixation of the workpiece;

[0063] The servo motor is used to control the movement of the conveying belt to achieve precise positioning;

[0064] The photoelectric sensor is used to detect whether the product is in place to trigger the visual detection and laser engraving process;

[0065] A central control system, which includes an industrial computer, connects the visual detection device through a gigabit Ethernet and integrates a motion control card to achieve precise control of the laser trajectory;

[0066] An auxiliary negative pressure fixation device, which uses a micro-pore array inside the workbench, connects a vacuum pump system with a working negative pressure ≥-60kPa, and makes the optical fiber fabric flat without wrinkles at the microstructure scale; and a standard optical calibration board is used for binocular camera calibration to establish a bidirectional mapping relationship between image pixel coordinates and world coordinates; and a spatial coincidence adjustment mechanism is introduced to ensure that image analysis and laser processing instructions are consistent in three-dimensional space through precise alignment of the laser focusing plane and the visual detection plane (error controlled within 0.05mm);

[0067] S2, the optical fiber fabric to be processed is laid flat on the workbench, the vacuum adsorption device is started to ensure that the surface is flat and wrinkle-free; the three-axis mechanical transmission device is adjusted to fix the position of the optical fiber fabric to prevent displacement during processing; then the visual detection device is started, and the built-in binocular industrial camera is used to acquire left and right view images of the optical fiber fabric surface in a synchronous triggering manner. During the image acquisition process, the telecentric lens provides distortion-free imaging, and the ring-shaped LED light source realizes uniform illumination, thereby obtaining image sequences, specifically:

[0068] The image sequences are acquired, and gray scale normalization and contrast enhancement processing are performed based on a structure-guided multi-scale optical fiber image enhancement network (FIRE-Net, Fiber-oriented Reinforcement&Enhancement Network); then high-frequency texture features are extracted through an adaptive filter to strengthen the optical fiber edge information;

[0069] Based on the semantic analysis method of optical weaving structure, the light depth network is used to identify the structure unit of optical fiber fabric, the topological graph model is constructed by fusing the directionality and semantic label, the path tracking and function labeling are realized by combining the structure continuity and the consistency of the principal axis direction, and the subsequent precise control and intelligent adjustment of laser process are supported;

[0070] Based on the semantic fusion of direction perception, a three-dimensional structure reconstruction method of optical fiber fabric is provided, which fuses multi-scale stereo matching, three-dimensional point cloud depth projection and structure guided semantic topological mapping, and constructs a high-precision, structure-continuous and semantic-consistent three-dimensional model of optical fiber fabric, as shown in Figure 2 The spatial relationship and relative height deviation of optical fiber 1, warp yarn 4 and weft fiber 3 are labeled.

[0071] Accordingly, the geometric reconstruction with sub-millimeter (0.1 mm level) accuracy is realized, which provides data basis for subsequent path planning and laser compensation.

[0072] S3, the control system plans the laser engraving path according to the three-dimensional model: the continuous laser points are generated with a point spacing of 0.2 mm in the transverse direction, and the parallel scanning lines are formed with a line spacing of 0.3 mm in the longitudinal direction.

[0073] When starting the laser engraving program, the laser head 5 moves along the path under the servo drive, and the Z-axis position is dynamically adjusted to keep the focal length constant according to the real-time height data.

[0074] The gradient power control strategy is adopted: from the starting position of the light source input end (connected to the light source controller 2), the initial value of the laser power is set to 10W, and the power is increased by 2% every 1cm along the extension direction of the optical fiber (for example, the power is increased to 11W at a distance of 5cm from the input end), and the light attenuation effect of the optical fiber 1 is accurately compensated to ensure that the end light intensity and the initial difference is less than 5%;

[0075] The laser 6 acts on the surface of the optical fiber cladding with a pulse width of 20ns and a frequency of 50kHz, and the engraving depth is strictly controlled at 50% of the cladding thickness (typical value 20μm);

[0076] S4, during the laser scanning process, the vision detection device continuously collects images of the processing area, and performs real-time gray histogram analysis, and determines that the defect is not activated when the gray value of the detection area is lower than the set threshold; at the same time, the connected domain positioning algorithm is used to identify abnormal points with engraving position deviation higher than the set threshold;

[0077] After finding the defect, the control system immediately marks the coordinates and automatically re-engraves after the current scanning period ends, and all process deviation data are stored in the database for optimizing subsequent processing parameters.

[0078] Embodiment two, the application proposes a kind of optical fiber fabric visual activation method, it further includes a kind of structural guidance type multi-scale fiber image enhancement network (FIRE-Net, Fiber-oriented Reinforcement&Enhancement Network) based on, specifically:

[0079] A1, by mapping the original image gray value to the uniform range, eliminate the overall gray offset caused by uneven illumination or acquisition equipment difference, ensure that image gray level is balanced:

[0080] Wherein, I x,y It represents the gray value of original image at pixel position (x,y);I min And I max It represents the minimum and maximum gray value of all pixels in original image respectively;I' x,y It represents the normalized image gray value;

[0081] A2, based on gray normalization image, the local brightness difference in different spatial range is extracted and enhanced by combining multi-scale Gaussian filter, and the structure details of image are amplified, so that fiber texture and edge are more clear and prominent:

[0082] Definition local contrast enhancement model of image:

[0083]

[0084] Wherein, I" x,y It represents the enhanced pixel gray value, including original gray and multi-scale local contrast enhancement information;λ It represents contrast enhancement intensity coefficient, control enhancement effect size, the embodiment takes 0.5~0.8, prevent excessive enhancement and introduce noise;S It represents scale set, corresponding different size Gaussian convolution kernel {3,5,7}, reflect local texture different scale features;G s (x,y) It represents two-dimensional Gaussian filter kernel of scale s, center in (x,y);μ s It represents the local mean of image I' under scale s, obtained by Gaussian smoothing;exp (·) It represents natural exponential function;σ s It represents the standard deviation of Gaussian function;

[0085] A3, by combining the weighted filtering of pixel similarity and local structure information, the key fiber edge and texture details are kept while smoothing the noise in image, avoid the edge blur caused by traditional filtering, specifically:

[0086] Gradient vector (I x ,I y ) at image pixel point (i,j) is calculated, and corresponding structure tensor is constructed:

[0087]

[0088] two eigenvalues of the tensor are calculated and

[0089] then the orthogonal component of the structure tensor is

[0090] The filtered pixel gray value is output:

[0091]

[0092] wherein, represents the filtered pixel gray value; Ω represents a neighborhood window centered at (x, y); represents a weight, which comprehensively considers the gray similarity and the local structure strength; σ represents a gray similarity parameter, which controls the influence of the gray difference on the weight, and in the embodiment, the value is empirically taken as 15-30, and the smaller the value is, the more sensitive it is; represents the orthogonal component of the structure tensor, which reflects the edge strength of the point; β represents a structure strength adjustment coefficient, which adjusts the edge retention degree; and Z represents a normalization constant, which ensures that the weights are added up to 1;

[0093] A4. According to the main direction of the optical fiber obtained by pre-analysis, the filtered image is directionally enhanced by using a direction similarity mask, the texture features along the optical fiber are highlighted, and the non-target background interference is suppressed:

[0094]

[0095] M θ (x, y) = cos 2 (θ-φ x,y );

[0096] wherein, represents the pixel value of the final output image after fusion; α represents a fusion coefficient, which balances the contributions of the original filtered image and the direction mask enhanced image, and in the embodiment, the value is taken as 0.6-0.9, so as to ensure that the original information is retained and the direction features are highlighted; M θ (x, y) represents a directional mask, which is generated by using the included angle between the local edge direction and the estimated main direction of the optical fiber; θ represents the main direction angle of the optical fiber, which is an estimated value of the overall trend of the optical fiber, and is extracted based on the frequency domain analysis or Fourier transform; φ x,y represents the local edge direction angle extracted at the pixel point (x, y) by using an edge detection algorithm such as a Sobel operator.

[0097] Embodiment three, the present application proposes a kind of optical fiber fabric visual activation method, still include a kind of based on light weaving structure semantic analysis method, its specific implementation process is as follows:

[0098] B1, input image Adopt the target detection network model of YOLOv5 core structure based on cutting optimization, realize the identification and frame selection of main structural unit in optical fiber fabric, and the neural network output is boundary box set:

[0099] Wherein, c i It represents the class label of the i-th boundary box (such as optical fiber fiber, warp yarn warp, weft yarn weft, node node);(x i ,y i ) represents the center coordinates of boundary box;(w i ,h i ) represents the width and height of boundary box;s i It represents confidence score, reflects prediction accuracy;N represents the number of identified structure elements;

[0100] B2, after obtaining structure element set, judge the geometric and functional association between each structure, construct spatial relationship matrix R, to support topological graph generation:

[0101] Definition of the correlation degree R ij Between structure pair:

[0102] Wherein, R ij It represents the spatial correlation degree score between the i-th and j-th structure;δ(c i ,c j ) represents structure compatible indication function;p i It represents the i-th structure center coordinate vector, p i =(x i ,y i );||p i -p j ||2 represents the Euclidean distance between the i-th and j-th structure;σ d It represents distance adjustment factor, controls the attenuation degree of long-distance connection (this embodiment is set according to experience 5-20 pixels);θ i , θ j Respectively, the main axis direction angle of the i-th and j-th structure boundary box;

[0103] B3, construct graph model G=(V,E), represent the topological relationship of optical fiber structure:

[0104] Vertex set V={v i} corresponds to each identified structure element b i, the attributes include category, coordinate, direction, size;

[0105] Edge set E = {e ij} represents that the connection is established when the high correlation degree (R ij > 0.7) exists between the structural units;

[0106] Each edge is attached with an attribute weight ω ij = R ij , which is used to measure the connection strength;

[0107] Accordingly, the two-dimensional distribution information of the optical fiber fabric is converted into a graph structure expression, and the edge weight constraint is combined with the physical function relationship, so that the connection between structures has 'function semantics', and has dynamic interpretation and planning guidance;

[0108] B4, based on the topological graph G, path tracking and structure labeling are performed, for each complete optical fiber path, a tracking algorithm based on direction consistency and continuous connection is used to construct the path;

[0109] The path tracking evaluation function is defined as:

[0110] Wherein, S path represents the optical fiber path score, representing the overall geometric continuity and structural consistency of a path, the larger the value, the better; P represents the structure pair set on the path, the continuous node pair set obtained according to the graph search algorithm (such as DFS); λ1 and λ2 represent the weighting coefficients, which control whether to pay more attention to spatial distance or direction continuity when selecting a path, and satisfy λ1+λ2=1; cos(θ i -θ j ) represents the structure direction continuity factor, which is close to 1 when the two structure directions are consistent, otherwise it tends to 0, and is used to exclude the mutant path;

[0111] After the path construction is completed, each path is labeled as an optical fiber main chain, a warp yarn chain and a weft yarn chain, and the following attributes are attached:

[0112] Start and end point coordinates (used for laser starting point positioning);

[0113] Local curvature change (used for Z-axis trajectory adjustment);

[0114] Length and direction of each segment (used for laser energy gradient control);

[0115] Path ID and hierarchical relationship (used for cross-structure management).

[0116] In the fourth embodiment, the optical fiber fabric visual activation method proposed by the application further includes an optical fiber fabric three-dimensional structure reconstruction method based on direction perception semantic fusion, and the specific implementation process is as follows:

[0117] C1, synchronously acquire left and right image pairs <I L (x,y), I R (x,y)>, construct Gaussian pyramid for each image (s is scale level), for each pixel (x,y), calculate matching cost within disparity range d∈[d min ,d max ]:

[0118]

[0119] where I L (x,y) and I R (x,y) represent image gray value of left view and right view respectively; d represents disparity value (pixel horizontal offset); C(x,y,d) represents matching cost, similarity score corresponding to candidate disparity, the smaller the value, the better the matching; α1, α2 represent weight coefficients in matching cost function, satisfying α1+α2=1; and represent gradient value of left image and right image respectively, calculated by Sobel operator;

[0120] Aggregate cost of each candidate disparity within its pixel neighborhood, use guided filtering method for weighted average:

[0121] where C (x,y,d) represents aggregated matching cost; Z represents normalization factor, Z=∑ω ij ; ω ij represents cost aggregation weight; N(x,y) represents pixel neighborhood set, i.e. fixed window centered on pixel (x,y);

[0122] Select disparity d with minimum cost value for each pixel, get preliminary disparity map:

[0123]

[0124] where d * (x,y) represents optimal disparity value, best disparity value selected under the condition of minimum cost;

[0125] Then perform the following optimization:

[0126] Left-right consistency check: compare left image disparity map with right image calculation result in reverse, eliminate inconsistent points;

[0127] Sub-pixel interpolation: perform parabolic fitting on matching cost curve, improve disparity accuracy to sub-pixel level;

[0128] Occlusion region filling: using guided filling method (such as nearest edge copy) to repair the disparity hole of the occluded region;

[0129] Then use the combination of median filtering and bilateral filtering to smooth the disparity map and perform edge-aware correction strategy on the abnormal jump region, ensuring the continuity and smoothness of the disparity map and the structure preservation;

[0130] Output disparity map: D(x,y)=d * (x,y);

[0131] Where D(x,y) represents the final output disparity map;

[0132] C2, according to the camera calibration parameters, convert the disparity map D(x,y) to a three-dimensional point cloud in the world coordinate:

[0133]

[0134] According to this, generate a point cloud set P: (M is the number of valid pixel points);

[0135] Where f is the camera focal length; B is the baseline distance between left and right cameras; d is the pixel disparity value; (u,v) is the image coordinate, i.e. (x,y); (c x ,c y ) is the principal point coordinate (the position of the camera optical center in the image coordinate system); (x,y,z) is the corresponding three-dimensional coordinate point;

[0136] C3, introduce a direction-aware double filtering algorithm for point cloud filtering and structure surface reconstruction, specifically:

[0137] Use statistical outlier removal (Statistical Outlier Removal) to judge abnormal points and remove them according to the average distance of points in the neighborhood to the center point;

[0138] Combined with the direction of the main shaft of the structure, a direction-guided bilateral filter is designed:

[0139]

[0140] Where P i and P j represent the three-dimensional point coordinate vectors of points i and j respectively; P i filt represents the new position of point i after filtering; N(i) represents the neighborhood set of point i; θ i , θ j represent the structure direction angles corresponding to points i and j respectively; σ s represents the spatial Gaussian kernel parameter; σ θ represents the direction similarity control parameter; Zi denotes a normalization factor to prevent the weighted sum from being distorted;

[0141] C4, the generated semantic topology graph G is fused to map the semantic attributes of each path node to the three-dimensional point cloud surface, and the mapping method is as follows:

[0142]

[0143] wherein, v i denotes the three-dimensional position of the structure node in the topology graph; ε denotes the mapping tolerance threshold; Label(·) denotes the semantic label of the point or node.

[0144] The embodiments of the present application are described in detail above in combination with the drawings, but the present application is not limited thereto, and various changes can be made within the knowledge of those skilled in the art without departing from the purpose of the present application.

Claims

1. A method for visual activation of fiber optic fabric, characterized in that, The specific implementation steps include the following: S1. Construct a fiber optic fabric visual activation system, the system comprising: Laser engraving device, used to generate high-energy laser beams and engrave and activate fiber optic fabrics; A visual inspection device is used to acquire image sequences of fiber optic fabrics and perform image processing and 3D reconstruction. A three-axis mechanical transmission device is used to drive the laser head and workpiece to achieve three-dimensional spatial positioning. The central control system is used to coordinate and control the laser engraving device, the vision inspection device, and the three-axis mechanical transmission device. An auxiliary negative pressure fixing device is used to lay the optical fiber fabric flat without wrinkles by vacuum adsorption. S2. Fix the fiber optic fabric to the platform, collect image sequences through a vision inspection device, and generate a three-dimensional model of the fiber optic fabric through image enhancement, structural semantic analysis and three-dimensional reconstruction. S3. Based on the 3D model, plan the laser engraving path, use the gradient power control strategy for laser activation, and dynamically adjust the Z-axis focal length. S4. Perform real-time defect detection and automatically re-carve after the current scanning cycle ends.

2. The method for visual activation of fiber optic fabric according to claim 1, characterized in that, Image enhancement is performed using the structure-guided multi-scale image enhancement network FIRE-Net, following these steps: Based on the gray values ​​of each pixel in the original image, the image is linearly mapped to the range of 0-255 using the global minimum and maximum gray values. This gray-level normalization eliminates the overall gray-level shift caused by uneven lighting or device differences. Based on grayscale normalized images, a multi-scale Gaussian filter is used to extract and enhance the local brightness differences in different spatial ranges. The structural details are amplified by a combination model of weighted original grayscale and multi-scale local mean difference, while noise is avoided by controlling the intensity coefficient. By constructing a structure tensor through pixel gradient vector calculation, extracting feature values ​​and taking the minimum value as an orthogonal component to represent edge strength, and combining gray-level similarity and structure strength as dual weighting factors for neighborhood weighted filtering, noise is suppressed while preserving fiber edge and texture details. A directional mask is constructed based on the pre-estimated difference between the main direction of the optical fiber and the local edge direction of the pixel. The original filtered image and the mask-enhanced image are then fused by weighted fusion to directionally enhance the texture features of the optical fiber orientation and suppress background interference.

3. The method for visual activation of fiber optic fabric according to claim 2, characterized in that, Structural semantic analysis employs a method based on optical weave structure semantic analysis, performed sequentially as follows: The YOLOv5 target detection model with cropping optimization is used to process the input image, identify the main structural units in the fiber optic fabric and output a set of bounding boxes, including category labels, center coordinates, width and height dimensions and confidence scores; Spatial correlation scores between structural elements are calculated using structural compatibility functions, exponential decay of Euclidean distance, and the reciprocal of the direction angle difference, and a spatial relationship matrix is ​​constructed. Construct a graph model where vertices correspond to structural elements and edges are connected based on a high correlation threshold, and use edge weights to carry functional semantics to realize a computable expression of the relationships between structures; Based on the topology map, fiber optic paths are constructed using directional consistency and continuous connection criteria. Geometric continuity and directional consistency are optimized through evaluation functions. Path labeling is completed, and start and end coordinates, curvature, length direction, and hierarchical relationships are recorded.

4. The method for visual activation of fiber optic fabric according to claim 3, characterized in that, The 3D reconstruction employs a fiber optic fabric 3D structure reconstruction method based on direction-aware semantic fusion, which is executed sequentially as follows: By simultaneously acquiring left and right images with a binocular camera and constructing a Gaussian pyramid, the disparity similarity is calculated by combining the weighted matching cost function of gray-level difference and gradient difference. After guided filtering and aggregating neighborhood costs, the optimal disparity is selected to generate a preliminary disparity map. Then, the accuracy is optimized by left-right consistency check, sub-pixel interpolation and occlusion area filling. Finally, a continuous and smooth disparity map is output by median-bilateral filtering and edge perception correction. Based on the camera calibration parameters, a 3D point cloud set P is generated using the pixel disparity values ​​d in the disparity map D(x,y) and the image coordinates (u,v), along with depth calculation and horizontal coordinate transformation formulas. ； ； ； ； Where f is the camera focal length; B is the baseline distance between the left and right cameras; d is the pixel parallax value; (u,v) are the image coordinates; (c x ,c y (x, y, z) are the main point coordinates; (x, y, z) are the corresponding 3D coordinates; M is the number of effective pixels.

5. The method for visual activation of fiber optic fabric according to claim 4, characterized in that, In step S3: Based on the fiber optic path information in the 3D model, a scanning path is generated horizontally with a point spacing of 0.2mm and vertically with a line spacing of 0.3mm. Each point on the path contains spatial coordinates, orientation information, and curvature information, which are used for subsequent Z-axis focusing and power control.

6. The method for visual activation of fiber optic fabric according to claim 5, characterized in that, In step S3: The gradient power control strategy increases the laser power in segments based on the distance of the fiber path from the input end: The initial laser power is 10W, increasing by 2% for every 1cm extension to compensate for the optical guide attenuation effect and ensure that the laser activation effect at the end of the path is not lower than that at the beginning. Simultaneously, the laser head position is dynamically adjusted based on the Z-axis height information, ensuring that the laser focus always falls on the center of the cladding, and the engraving depth is controlled within the range of 20μm±2μm.

7. The method for visual activation of fiber optic fabric according to claim 6, characterized in that, In step S4: Defect detection is based on two types of indicators: first, whether the gray value of the processing area is lower than the set activation threshold; second, whether the offset distance between the engraved outline and the target path exceeds the set limit. When any condition triggers defect identification, the system records the spatial coordinates of the point and adds it to the recarving list. After the scanning cycle ends, the laser head automatically returns to perform the recarving operation.

8. The method for visual activation of fiber optic fabric according to claim 1, characterized in that, The visual inspection device includes: The image acquisition unit is a combination of a binocular industrial camera and a telecentric lens fixed above the platform. It acquires image sequences of fiber optic fabric and provides shadowless illumination using a ring LED array light source. The image processing unit uses deep learning methods to detect and model targets in the acquired optical fiber fabric image sequence.

Images