Fine geometry reconstruction modeling method for optical fiber package

Abstract

The invention discloses a fine geometric reconstruction modeling method for an optical fiber package, which specifically comprises the following steps: scanning and photographing a sample optical fiber package by an industrial CT device; reconstructing the internal structural features of the sample optical fiber package by combining the scanned image with an FDK three-dimensional reconstruction algorithm to obtain a high-resolution three-dimensional image of the sample optical fiber package; uniformly discretizing the three-dimensional image along a circumferential direction, and enhancing and binarizing the discretized image to obtain a binarized two-dimensional cross-sectional image; calculating the centroid of each connected domain in the two-dimensional cross-sectional image to obtain the centroid coordinates of the optical fiber core cross-section; according to the optical fiber connection reconstruction algorithm, identifying the associated centroid points in the optical fiber cross-section and connecting them in sequence to obtain a fine geometric reconstruction model of the optical fiber package; the modeling method of the invention realizes fine geometric reconstruction modeling of an area of ​​interest of the optical fiber package, and can accurately reconstruct the real winding structure of the optical fiber inside a large-sized complex optical fiber package.

Description

Technical Field

[0001] The invention belongs to the technical field of optical fiber geometric reconstruction modeling, and in particular relates to a method for fine geometric reconstruction modeling of an optical fiber package. Background Art

[0002] The optical fiber package refers to a spindle-shaped entity formed by tightly winding the optical fiber on the surface of a truncated cone-shaped core tube. During the winding and molding process of the optical fiber package, the fluctuation of the optical fiber diameter, winding tension and hysteresis angle, the uneven coating of the adhesive, and the different thermal expansion coefficients of various materials will lead to various structural defects inside the optical fiber package, which will greatly reduce the reliability of the optical fiber when it is released. Therefore, it is necessary to observe and reconstruct the internal structure of the optical fiber package.

[0003] Industrial CT is a computer tomography technology used in the industrial field. It is based on X-ray detection, and performs multiple continuous end scans around the sample, and reconstructs the scan results into a three-dimensional visual image. It is a non-destructive testing technology that does not damage the performance of the object being tested. Compared with conventional non-destructive testing methods such as visual measurement, X-ray detection, and ultrasonic flaw detection, it has the advantages of fast detection speed, high resolution, and is not limited by the geometric structure of the object being tested. It is widely used in non-destructive testing and quality assessment of industrial equipment and products.

[0004] The optical fiber package is scanned by industrial CT to obtain a high-resolution three-dimensional image of the optical fiber package. In order to perform fine geometric reconstruction modeling based on the three-dimensional image, a series of digital image processing such as two-dimensional discretization, image enhancement, noise removal, image segmentation, and feature extraction are required to extract useful size and position information. Digital image processing, also known as computer image processing, refers to the process of converting image signals into digital signals and processing them using computers. With the rapid development and popularization of computer and multimedia technology, digital image processing technology has received unprecedented and widespread attention in many application fields and has achieved significant achievements, such as aerospace, biomedical engineering, materials engineering, industrial manufacturing, and robotic vision.

[0005] Digital image processing technology has the advantages of high precision, rich processing content, varied methods, and high flexibility. However, since images store information through an array of pixels, and the precision of each pixel is generally 256 levels, the amount of data is huge, and the processing speed is limited by computers and mathematical devices, which hinders its further development and promotion. Image binarization technology can greatly reduce the amount of data in the image, highlight the target object, and facilitate further processing of the image. The binarization of an image is to set the grayscale value of the pixel on the image to 0 or 255, that is, to present the entire image with only black and white visual effects. An image generally includes a target object, background, and noise. To extract the target object from the image, the commonly used method is image segmentation. That is, a grayscale threshold T is set, and all pixels in the image are divided into two categories by T: a group of pixels with grayscale values ​​greater than T and a group of pixels with grayscale values ​​less than T. According to the method of determining the threshold, the binarization algorithm can be roughly divided into two categories: the global threshold method and the local threshold method. The global threshold method determines a global threshold based on a certain rule, and uses the threshold to binarize all pixels of the entire image, such as the maximum inter-class variance method, the iteration method, etc. The local threshold method takes a single pixel as a unit, and calculates the threshold of the pixel according to a certain rule based on the relevant information of the pixels in its neighboring area. Typical algorithms include the Niblack algorithm and the Sauvola algorithm.

[0006] At present, the analysis of optical fiber cable package structure is mainly carried out through three-dimensional modeling and then finite element modeling and simulation. This method is based on observing the surface structure of the cable package, and then applying its surface structure to the inside of the cable package to draw an ideal cable package model. The internal structure of the cable package model obtained in this way is unknown to what extent it conforms to the actual situation, and it is impossible to simulate various defects that may exist inside the cable package. Therefore, it is necessary to introduce a three-dimensional representation method for the internal spatial structure of the actual cable package. Summary of the invention

[0007] The purpose of the present invention is to provide a method for fine geometric reconstruction modeling of an optical fiber package, which solves the problem that the prior art cannot simulate various defects that may exist inside a large-sized optical fiber package.

[0008] The technical solution adopted by the present invention is a method for fine geometric reconstruction and modeling of optical fiber line packages, specifically:

[0009] Step 1: Scan and photograph the sample optical fiber package using industrial CT equipment;

[0010] Step 2: Reconstruct the internal structure features of the sample optical fiber package by combining the scanned image with the FDK 3D reconstruction algorithm to obtain a high-resolution 3D image of the sample optical fiber package;

[0011] Step 3, the three-dimensional image is evenly discretized along the circumferential direction, and the discretized image is enhanced and binarized to obtain a binarized two-dimensional cross-sectional image;

[0012] Step 4: Calculate the centroid of each connected domain in the two-dimensional cross-sectional image to obtain the centroid coordinates of the optical fiber core cross-section; according to the optical fiber connection reconstruction algorithm, identify the associated centroid points in the optical fiber cross-section and connect them in order to obtain a fine geometric reconstruction model of the optical fiber package.

[0013] The present invention is also characterized in that:

[0014] The specific process of step 1 is: use an industrial CT device with a micro-focus ray source to scan and photograph the sample optical fiber package in an offset scanning manner.

[0015] The specific process of step 2 is:

[0016] A ray source is used to emit X-rays to penetrate the fiber optic cable package of the sample under test, and the attenuated photons are received by a flat-panel detector. O-xyz is the coordinate system of the fiber optic cable package of the sample under test, O-ab is the projection coordinate system of the flat-panel detector, S is the ray source, and O represents the position of the rotation center. A virtual detector is made parallel to the flat-panel detector and passes through point O. SO' represents the central ray of the ray source, SN represents a ray passing through point (x, y, z), Sk' is the projection of ray SN on the xOy plane, and the rotation angle of the ray source S relative to point O is taken as β, k is the cone angle, γ is the fan angle, and R represents the distance from the ray source to the rotation center.

[0017] The definition of the weighted process is that the X-rays at different angles are projected on the central ray by the vector formed by the ray source as the initial starting point and the flat panel detector as the end point. After weighting, the two-dimensional projection data is obtained. The weight is used to correct the two-dimensional projection data. The correction formula is:

[0018]

[0019] Where P(β,a,b) is the two-dimensional projection data under the rotation angle β, a and b represent the coordinates of the two-dimensional image obtained by projecting the flat panel detector in the O-xyz coordinate system, and P w (β, a, b) is the two-dimensional projection data obtained after correction;

[0020] The time domain filtering is performed on the corrected two-dimensional projection data, and the expression is:

[0021]

[0022] In the formula, represents the filtered two-dimensional projection data, H(a) represents the convolution kernel, Represents the convolution operation;

[0023] The filtered two-dimensional projection data is back-projected along the ray direction, and the back-projected data at all rotation angles passing through the voxel are accumulated and summed to obtain the reconstructed data f(x, y, z):

[0024]

[0025] Where N is the sampling times of the projection data;

[0026] The reconstructed data f(x, y, z) is input into the 3D reconstruction software VoxelStudio Recon to obtain a high-resolution 3D image of the sample optical fiber package.

[0027] The specific process of step 3 is as follows:

[0028] Step 3.1, take the center of the bottom end face of the three-dimensional image of the optical fiber package as the origin, and the axis perpendicular to the bottom end face of the optical fiber package as the Z axis, establish a semi-polar coordinate system, and cut n two-dimensional cross-sectional images of the optical fiber package on the ρOZ plane of the semi-polar coordinate system along the circumferential direction of the three-dimensional image with g as the step length to obtain a two-dimensional cross-sectional image stack I 1 ~I n ;

[0029]

[0030] Step 3.2: 2D cross-sectional image I 1 ~I n Perform image enhancement processing to obtain an enhanced two-dimensional cross-sectional image I 1’ ~I n’ ;

[0031] Step 3.3: Use the local adaptive threshold binarization algorithm to enhance the two-dimensional cross-sectional image I 1 '~I n’ Perform binarization processing to obtain a binary two-dimensional cross-sectional image I 1” ~I n” .

[0032] The specific process of step 3.2 is as follows:

[0033] A grayscale distribution histogram is drawn for the two-dimensional cross-sectional images I1~In. From the grayscale distribution histogram, the pixel grayscale levels of the two-dimensional cross-sectional image of the optical fiber package are distributed between [min, max], and the maximum grayscale value of all pixels is MAX. The grayscale range [min, max] of the two-dimensional cross-sectional image of the optical fiber package is linearly extended to [0, 255]. For the grayscale value T of any point in the image, B (x, y) performs linear grayscale transformation, and the transformation formula is:

[0034]

[0035] After transformation, the gray value of any point is T A (x, y), after transformation, the enhanced image I is obtained 1’ ~I n .

[0036] The specific process of step 3.3 is as follows: by calculating the number of connected domains n in the same area when different neighborhood radii are selected r The relative difference Δ with the actual number of fiber cross sections N is compared to obtain the neighborhood radius ρ required for the local adaptive threshold binarization method that can achieve the best binarization effect. The calculation formula for Δ is:

[0037]

[0038] Calculate the grayscale mean m(x,y) and standard deviation s(x,y) in the ρ neighborhood of any pixel:

[0039]

[0040]

[0041] Among them, h(i,j) is the gray value of any pixel in the ρ neighborhood of point (x,y);

[0042] Calculate the binarization threshold of each pixel in the neighborhood of ρ The calculation formula is:

[0043]

[0044] In the formula, ε is the denoising coefficient, which is pre-set by experience;

[0045] If the gray value of each pixel is greater than the binarization threshold Set it to white, otherwise set it to white, and get the binary two-dimensional cross-sectional image I 1” ~I n” .

[0046] The specific process of calculating the centroid of each connected domain in the two-dimensional cross-sectional image in step 4 is as follows: Use the centroid calculation tool regionprops in MATLAB to obtain the centroid positions and two-dimensional centroid coordinates (x 0 ,y 0 ), i = 1, 2, ..., n, the two-dimensional centroid coordinates (x 0 ,y 0 ) is converted into three-dimensional cylindrical coordinates, that is, the centroid coordinates of the optical fiber core cross section. The conversion formula is:

[0047]

[0048] Wherein, ρ, Z, θ are the three-dimensional coordinates corresponding to the cylindrical coordinate system, and n is the number of two-dimensional cross-sectional images obtained by intercepting the three-dimensional image along the circumferential direction.

[0049] According to the fiber connection reconstruction algorithm, the associated centroid points in the fiber cross section are identified and connected in sequence. The specific process is as follows:

[0050] Assume that the fiber diameter is d, the core diameter is D, and manually select the starting point p (l,m,q) , the cylindrical coordinates are (ρ 0 ,θ 0 ,Z 0 ), represents the centroid of the fiber core on the qth section of the mth turn of the lth layer of optical fiber, where l, m, and q are all positive integers.

[0051] Find the starting point p in the q+1th section (l,m,q) The neighborhood of radius δ;

[0052] Define the centroid points of the fiber core cross sections associated on two adjacent cross sections as the associated centroid points, and calculate the maximum distance between the associated centroid points of the two adjacent cross sections when the fiber is unwound as χ', and the calculation formula is:

[0053]

[0054] Calculate the distance between each point in the δ neighborhood and the starting point p (l,m,q) distance;

[0055]

[0056] If the calculated minimum distance χ min <χ', then the point with the minimum distance is the associated centroid point p in the q+1 section (l,m,q+1) , connect point p (l,m,q) With p (l,m,q+1) ;

[0057] If the minimum distance χ is calculated min ≥χ', then let the associated centroid point p in the k+1 section (l,m,q+1) The cylindrical coordinates are (ρ 0 ,θ 0 +1, Z 0 ), connect point p (l,m,q) With p (l,m,q+1) ;

[0058] The associated centroid points of all cross sections are identified in sequence and connected, thus completing the connection of one turn of optical fiber;

[0059] when When point p (l,m,q) The three-dimensional cylindrical coordinates are set as the starting point p of the next turn of optical fiber connection(l,m,q+1) , and then traverse all the sections in turn to find the associated points on each section to complete the connection of the next turn of optical fiber; until the reconstruction of the area to be reconstructed is completed.

[0060] The beneficial effects of the present invention are:

[0061] The fine geometric reconstruction modeling method of the optical fiber package of the present invention uses a high-resolution three-dimensional image taken by a high-resolution industrial CT device for reconstruction, and observes the fine winding structure inside a large-scale complex optical fiber package for the first time;

[0062] The present invention discretizes the three-dimensional image into a two-dimensional cross-sectional image, and then uses digital image processing techniques such as linear image enhancement and local adaptive threshold binarization to accurately obtain the three-dimensional coordinates of the cross-sectional centroid of the optical fiber core at each layer, each angle, and each turn, thereby obtaining the data basis for reconstruction and modeling.

[0063] The present invention uses an optical fiber connection reconstruction algorithm based on distance calculation to determine the associated centroid points, which can achieve fine geometric reconstruction modeling of the area of ​​interest of the optical fiber package and accurately reconstruct the real winding structure of the optical fiber inside the large-sized and complex optical fiber package. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a projection diagram of X-ray penetrating the optical fiber package of the sample under inspection in the present invention;

[0065] Figure 2 is a high-resolution three-dimensional image of the sample optical fiber package obtained in Example 2 of the present invention;

[0066] Figure 3 is a schematic diagram of establishing a semi-polar coordinate system in Example 2 of the present invention;

[0067] Figure 4 is a schematic diagram of a two-dimensional cross-sectional image stack obtained in Example 2 of the present invention;

[0068] Figure 5 is the enhanced image obtained in Example 2 of the present invention;

[0069] Figure 6 is the binarized two-dimensional cross-sectional image obtained in Example 2 of the present invention;

[0070] Figure 7 is a schematic diagram of the centroid positions and two-dimensional centroid coordinates of all connected domains in the binary image for obtaining a two-dimensional cross-sectional image in Example 2 of the present invention;

[0071] Figure 8 is a schematic diagram of the centroid coordinates of the cross section of the optical fiber core obtained in Example 2 of the present invention;

[0072] Fig. 9It is a flow chart of the optical fiber connection reconstruction algorithm in the present invention;

[0073] Fig.10 It is a schematic diagram of the fine geometric reconstruction model of the optical fiber package finally obtained in Example 3 of the present invention. DETAILED DESCRIPTION

[0074] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0075] Example 1

[0076] The optical fiber package fine geometry reconstruction modeling method of the present invention comprises the following steps: scanning and photographing a sample optical fiber package by an industrial CT device; reconstructing the internal structure characteristics of the sample optical fiber package by combining the scanned image with an FDK three-dimensional reconstruction algorithm to obtain a high-resolution three-dimensional image of the sample optical fiber package; uniformly discretizing the three-dimensional image along a circumferential direction, and enhancing and binarizing the discretized image to obtain a binarized two-dimensional cross-sectional image; calculating the centroid of each connected domain in the two-dimensional cross-sectional image to obtain the centroid coordinates of the cross section of the optical fiber core; and identifying the associated centroid points in the optical fiber cross section according to the optical fiber connection reconstruction algorithm and connecting them in sequence to obtain a fine geometry reconstruction model of the optical fiber package. The present invention uses high-resolution three-dimensional images taken by high-resolution industrial CT equipment for reconstruction, and observes the fine winding structure inside a large-sized complex optical fiber package for the first time; by discretizing the three-dimensional image into a two-dimensional cross-sectional image, and then using digital image processing techniques such as linear image enhancement and local adaptive threshold binarization, the three-dimensional coordinates of the centroid of the optical fiber core cross section at each layer, each angle, and each turn are accurately obtained, and the data basis for reconstruction modeling is obtained; using an optical fiber connection reconstruction algorithm based on distance calculation to determine the associated centroid points, the fine geometric reconstruction modeling of the optical fiber package's area of ​​interest is finally achieved, and the real optical fiber winding structure inside a large-sized complex optical fiber package can be accurately reconstructed.

[0077] Example 2

[0078] In this embodiment, the optical fiber package fine geometry reconstruction modeling method is specifically: using an industrial CT device with a micro-focus ray source to scan and photograph the sample optical fiber package in an offset scanning manner;

[0079] Industrial CT equipment is mainly composed of subsystems such as micro-focus radiation source, collimator system, detector system, multi-degree-of-freedom mechanical turntable system, data acquisition and transmission system, image reconstruction workstation and radiation safety protection system.

[0080] Since the sample fiber package is large in size, it is scanned multiple times from top to bottom using offset scanning. The system uses a micro-focus ray source with a focal size of only 6 microns, and its penumbra is negligible, so that the distance from the sample to the focal point can be minimized, the imaging range can be reduced, and the imaging resolution can be improved. Finally, the image formed by the geometric projection is optically magnified by coupling the detector through the objective lens, thereby achieving high-resolution imaging of large-size samples.

[0081] When X-rays penetrate an object, they will attenuate. The detector receives the attenuated photons and reconstructs the internal structure of the object through a mathematical 3D reconstruction algorithm, thereby obtaining a 3D image of the object. The FDK algorithm is a typical analytical algorithm. It is a 3D extension of the 2D fan beam reconstruction algorithm. The projection diagram is shown in the figure. Figure 1 As shown; the specific process is: use a ray source to emit X-rays to penetrate the optical fiber package of the sample to be inspected, and receive the attenuated photons through the flat-panel detector, with O-xyz as the coordinate system of the optical fiber package of the sample to be inspected, O-ab as the projection coordinate system of the flat-panel detector, S as the ray source, and O as the position of the rotation center; make a virtual detector parallel to the flat-panel detector and passing through point O, SO' represents the central ray of the ray source, SN represents a ray passing through the point (x, y, z), Sk' is the projection of ray SN on the xOy plane, the rotation angle of the ray source S relative to point O is taken as β, k is the cone angle, γ is the fan angle, and R represents the distance from the ray source to the rotation center;

[0082] The definition of the weighted process is that the X-rays at different angles are projected on the central ray by the vector formed by the ray source as the initial starting point and the flat panel detector as the end point. After weighting, the two-dimensional projection data is obtained. The weight is used to correct the two-dimensional projection data. The correction formula is:

[0083]

[0084] Where P(β,a,b) is the two-dimensional projection data under the rotation angle β, a and b represent the coordinates of the two-dimensional image obtained by projecting the flat panel detector in the O-xyz coordinate system, and P w (β, a, b) is the two-dimensional projection data obtained after correction;

[0085] The time domain filtering is performed on the corrected two-dimensional projection data, and the expression is:

[0086]

[0087] In the formula, represents the filtered two-dimensional projection data, H(a) represents the convolution kernel, Represents the convolution operation;

[0088] The filtered two-dimensional projection data is back-projected along the ray direction, and the back-projected data at all rotation angles passing through the voxel are accumulated and summed to obtain the reconstructed data f(x, y, z):

[0089]

[0090] Where N is the sampling times of the projection data;

[0091] Input the reconstruction data f(x, y, z) into the 3D reconstruction software VoxelStudio Recon and obtain Figure 2 A high-resolution 3D image of a sample fiber optic cable package is shown.

[0092] like Figure 3 As shown, with the center of the bottom end face of the optical fiber package as the origin and the axis perpendicular to the bottom end face of the optical fiber package as the Z axis, a semi-polar coordinate system is established. The three-dimensional image is cut along the circumferential direction with g as the step length on the ρOZ plane of the semi-polar coordinate system to obtain n two-dimensional cross-sectional images of the optical fiber package, and a two-dimensional cross-sectional image stack I is obtained. 1 ~I n like Figure 4 As shown;

[0093]

[0094] A grayscale distribution histogram is drawn for the two-dimensional cross-sectional images I1~In. From the grayscale distribution histogram, the pixel grayscale levels of the two-dimensional cross-sectional image of the optical fiber package are distributed between [min, max], and the maximum grayscale value of all pixels is MAX. The grayscale range [min, max] of the two-dimensional cross-sectional image of the optical fiber package is linearly extended to [0, 255]. For the grayscale value T of any point in the image, B (x, y) performs linear grayscale transformation, and the transformation formula is:

[0095]

[0096] After transformation, the gray value of any point is T A (x, y), after transformation, the enhanced image I is obtained 1’ ~I n like Figure 5 shown.

[0097] The local adaptive binarization algorithm contains two main parameters, namely the neighborhood radius ρ and the denoising coefficient ε. First, using the imdistline image length measurement tool, nine pairs of adjacent fiber core sections are evenly selected from the inside to the outside and from the top to the bottom in the two-dimensional cross-sectional image of the optical fiber package, and the mean distance D between the centroids of adjacent fiber sections is measured. m , then the selection range of the neighborhood radius of the local adaptive binarization algorithm is determined to be [Dm ,2D m ] Then, the number of fiber sections N actually contained in the local image is counted. m At the beginning, the step size is 1, and the number of connected domains n in the same area is calculated in turn when different neighborhood radii are selected r The relative difference Δ with the actual number of fiber cross sections N is compared to obtain the neighborhood radius ρ required by the local adaptive threshold binarization algorithm that can achieve the best binarization effect. The calculation formula for Δ is:

[0098]

[0099] Calculate the grayscale mean m(x,y) and standard deviation s(x,y) in the ρ neighborhood of any pixel:

[0100]

[0101]

[0102] Among them, h(i,j) is the gray value of any pixel in the ρ neighborhood of point (x,y);

[0103] Calculate the binarization threshold of each pixel in the neighborhood of ρ The calculation formula is:

[0104]

[0105] In the formula, ε is the denoising coefficient, which is pre-set by experience;

[0106] If the gray value of each pixel is greater than the binarization threshold Set it to white, otherwise set it to white, and get the binary two-dimensional cross-sectional image I 1” ~I n” like Figure 6 shown.

[0107] The centroids of each connected domain in the two-dimensional cross-sectional image are calculated to obtain the centroid coordinates of the fiber core cross-section. According to the fiber connection reconstruction algorithm, the associated centroid points in the fiber cross-section are identified and connected in order to obtain the fine geometric reconstruction model of the fiber package.

[0108] In this embodiment, the three-dimensional image is discretized in two dimensions along the circumferential direction to obtain a stack of two-dimensional cross-sectional images of the optical fiber package. Then, based on the grayscale distribution histogram of the two-dimensional cross-sectional image, the grayscale interval of the main area of ​​the optical fiber is determined, and the interval is linearly stretched to achieve image enhancement and better distinguish the background, the optical fiber coating and the optical fiber core. Through image measurement and program calculation, the optimal neighborhood radius of the local adaptive threshold binarization algorithm is obtained, and each two-dimensional cross-sectional image is binarized, and the centroid coordinates of the optical fiber core cross section in each image are calculated. Finally, by calculating the two-dimensional distance between each point in the associated local area in the adjacent cross section and the previous point, the associated centroid points in the adjacent cross sections are determined and identified and connected in sequence, so as to achieve fine geometric reconstruction modeling of the area of ​​interest of the optical fiber package.

[0109] Example 3

[0110] On the basis of Example 1 and Example 2, in this embodiment, the centroid of each connected domain in the two-dimensional cross-sectional image is calculated to obtain the centroid coordinates of the optical fiber core cross section; according to the optical fiber connection reconstruction algorithm, the associated centroid points in the optical fiber cross section are identified and connected in order to obtain the fine geometric reconstruction model of the optical fiber package. The specific process is as follows: Figure 7 As shown in the figure, the centroid calculation tool regionprops in MATLAB is used to obtain the centroid positions and two-dimensional centroid coordinates (x 0 ,y 0 ), i=1,2,…,n, such as Figure 8 As shown, the two-dimensional centroid coordinates (x 0 ,y 0 ) is converted into three-dimensional cylindrical coordinates, that is, the centroid coordinates of the optical fiber core cross section. The conversion formula is:

[0111]

[0112] Wherein, ρ, Z, θ are the three-dimensional coordinates corresponding to the cylindrical coordinate system, and n is the number of two-dimensional cross-sectional images obtained by intercepting the three-dimensional image along the circumferential direction.

[0113] The optical fiber connection reconstruction algorithm process is as follows Fig. 9 As shown, assume that the fiber diameter is d, the core diameter is D, and the starting point p is manually selected. (l,m,q) , the cylindrical coordinates are (ρ 0 ,θ 0 ,Z 0 ), represents the centroid of the fiber core on the qth section of the mth turn of the lth layer of optical fiber, where l, m, and q are all positive integers.

[0114] Find the starting point p in the q+1th section (l,m,q) The neighborhood of radius δ;

[0115] Define the centroid points of the fiber core cross sections associated on two adjacent cross sections as the associated centroid points, and calculate the maximum distance between the associated centroid points of the two adjacent cross sections when the fiber is unwound as χ', and the calculation formula is:

[0116]

[0117] Calculate the distance between each point in the δ neighborhood and the starting point p (l,m,q) distance;

[0118]

[0119] If the minimum distance χ is calculated min <χ', then the point with the minimum distance is the associated centroid point p in the q+1 section (l,m,q+1) , connect point p (l,m,q) With p (l,m,q+1) ;

[0120] If the minimum distance χ is calculated min ≥χ', then let the associated centroid point p in the k+1 section (l,m,q+1) The cylindrical coordinates are (ρ 0 ,θ 0 +1, Z 0 ), connect point p (l,m,q) With p (l,m,q+1) ;

[0121] The associated centroid points of all cross sections are identified in sequence and connected, thus completing the connection of one turn of optical fiber;

[0122] when When point p (l,m,q) The three-dimensional cylindrical coordinates are set as the starting point p of the next turn of optical fiber connection (l,m,q+1) The coordinates of the fiber optic cable are then traversed through all cross sections to find the associated points on each cross section, and the connection of the next turn of optical fiber is completed; until the reconstruction of the area to be reconstructed is completed, such as Fig.10 shown; according to Fig.10 The winding shape of each turn of optical fiber, the position change across turns, and the structure of the unwinding section across layers can be directly observed, which helps us predict the reliability of optical fiber packages and optimize the manufacturing process of optical fiber packages.

[0123] Through the above-mentioned manner, the fine geometric reconstruction modeling method of the optical fiber package of the present invention performs multiple offset scanning and shooting of the sample optical fiber package through an industrial CT device with a micro-focus ray source and an objective lens coupled detector to obtain a high-resolution three-dimensional image. The three-dimensional image is then discretized in two dimensions along the circumferential direction to obtain a stack of two-dimensional cross-sectional images of the optical fiber package. Afterwards, the grayscale interval of the main area of ​​the optical fiber is determined according to the grayscale distribution histogram of the two-dimensional cross-sectional image, and the interval is linearly stretched to achieve image enhancement and better distinguish the background, the optical fiber coating and the optical fiber core. Again, through image measurement and program calculation, the optimal neighborhood radius of the local adaptive threshold binarization algorithm is obtained, each two-dimensional cross-sectional image is binarized, and the centroid coordinates of the optical fiber core cross section in each image are calculated. Finally, by calculating the two-dimensional distance between each point in the associated local area in the adjacent cross section and the previous point, the associated centroid points in the adjacent cross sections are determined and identified and connected in sequence, so as to achieve fine geometric reconstruction modeling of the area of ​​interest of the optical fiber package.

Claims

1. A method for fine geometric reconstruction and modeling of optical fiber line packages, characterized in that: Specifically: Step 1: Scan and photograph the sample optical fiber package using industrial CT equipment; Step 2: Reconstruct the internal structure features of the sample optical fiber package by combining the scanned image with the FDK 3D reconstruction algorithm to obtain a high-resolution 3D image of the sample optical fiber package; The specific process of step 2 is as follows: A ray source is used to emit X-rays to penetrate the fiber optic cable package of the sample under test, and the attenuated photons are received by a flat-panel detector. O-xyz is the coordinate system of the fiber optic cable package of the sample under test, O-ab is the projection coordinate system of the flat-panel detector, S is the ray source, and O represents the position of the rotation center. A virtual detector is made parallel to the flat-panel detector and passes through point O. SO' represents the central ray of the ray source, SN represents a ray passing through point (x, y, z), Sk' is the projection of ray SN on the xOy plane, the rotation angle of the ray source S relative to point O is taken as β, k is the cone angle, γ is the fan angle, and R represents the distance from the ray source to the rotation center. The definition of the weighted process is that the X-rays at different angles are projected on the central ray by the vector formed by the ray source as the initial starting point and the flat panel detector as the end point. After weighting, the two-dimensional projection data is obtained. The weight is used to correct the two-dimensional projection data. The correction formula is: Where P(β,a,b) is the rotation angle β The two-dimensional projection data under a , b represents the coordinates of the two-dimensional image obtained by projecting the flat panel detector in the O-xyz coordinate system, and Pw(β,a,b) is the two-dimensional projection data obtained after correction; The time domain filtering is performed on the corrected two-dimensional projection data, and the expression is: In the formula, represents the filtered two-dimensional projection data, H(a) represents the convolution kernel, and ⊗ represents the convolution operation; The filtered two-dimensional projection data is back-projected along the ray direction, and the back-projected data at all rotation angles passing through the voxel are accumulated and summed to obtain the reconstructed data f(x, y, z): In the formula, N is the sampling times of the projection data; Input the reconstructed data f(x, y, z) into the 3D reconstruction software VoxelStudio Recon to obtain a high-resolution 3D image of the sample optical fiber package; Step 3, the three-dimensional image is evenly discretized along the circumferential direction, and the discretized image is enhanced and binarized to obtain a binarized two-dimensional cross-sectional image; The specific process of step 3 is as follows: Step 3.1, take the center of the bottom end face of the three-dimensional image of the optical fiber package as the origin, and the axis perpendicular to the bottom end face of the optical fiber package as the Z axis, establish a semi-polar coordinate system, and cut n two-dimensional cross-sectional images of the optical fiber package on the ρOZ plane of the semi-polar coordinate system along the circumferential direction of the three-dimensional image with g as the step length, and obtain a two-dimensional cross-sectional image stack I1~I n ; ； Step 3.2: For the two-dimensional cross-sectional images I1~I n Perform image enhancement processing to obtain an enhanced two-dimensional cross-sectional image I 1’ ~I n’ ; Step 3.3: Use the local adaptive threshold binarization algorithm to enhance the two-dimensional cross-sectional image I 1’ ~I n’ Perform binarization processing to obtain a binary two-dimensional cross-sectional image I 1” ~I n”； Step 4: Calculate the centroid of each connected domain in the two-dimensional cross-sectional image to obtain the centroid coordinates of the optical fiber core cross section; according to the optical fiber connection reconstruction algorithm, identify the associated centroid points in the optical fiber cross section and connect them in order to obtain a fine geometric reconstruction model of the optical fiber package.

2. The optical fiber package fine geometry reconstruction modeling method according to claim 1, characterized in that: The specific process of step 1 is: use an industrial CT device with a micro-focus ray source to scan and photograph the sample optical fiber package in an offset scanning manner.

3. The optical fiber package fine geometry reconstruction modeling method according to claim 1, characterized in that: The specific process of step 3.2 is as follows: A grayscale distribution histogram is drawn for the two-dimensional cross-sectional images I1~In. From the grayscale distribution histogram, the pixel grayscale levels of the two-dimensional cross-sectional image of the optical fiber package are distributed between [min, max], and the maximum grayscale value of all pixels is MAX. The grayscale range [min, max] of the two-dimensional cross-sectional image of the optical fiber package is linearly extended to [0, 255]. For the grayscale value T of any point in the image, B (x, y) performs linear grayscale transformation, and the transformation formula is: After transformation, the gray value of any point is T A (x, y), after transformation, the enhanced image I is obtained 1’ ~I n .

4. The optical fiber package fine geometry reconstruction modeling method according to claim 1, characterized in that: The specific process of step 3.3 is as follows: by calculating the number of connected domains n in the same area when different neighborhood radii are selected r The relative difference Δ with the actual number of fiber cross sections N is compared to obtain the neighborhood radius ρ required for the local adaptive threshold binarization method that can achieve the best binarization effect. The calculation formula for Δ is: Calculate any pixel ρ Neighborhood grayscale mean m ( x , y ) and standard deviation s ( x , y ): in, h ( i , j ) is the point ( x , y )of ρ The gray value of any pixel in the neighborhood; calculate ρ The binarization threshold of each pixel in the neighborhood , the calculation formula is: In the formula, ε is the denoising coefficient, which is pre-set by experience; If the gray value of each pixel is greater than the binarization threshold , it is set to white, otherwise it is set to white, and a binary two-dimensional cross-sectional image I is obtained. 1” ~I n” .

5. The optical fiber package fine geometry reconstruction modeling method according to claim 4, characterized in that: The specific process of calculating the centroid of each connected domain in the two-dimensional cross-sectional image in step 4 is as follows: matlab Centroid Calculation Tool regionprops , get the i The centroid positions and two-dimensional centroid coordinates of all connected domains in the binary image of a two-dimensional cross-sectional image , i=1,2,…,n, the two-dimensional centroid coordinates Converted into three-dimensional cylindrical coordinates, that is, the centroid coordinates of the optical fiber core cross section, the conversion formula is: ； In the formula, ρ , Z , θ That is, the three-dimensional coordinates corresponding to the cylindrical coordinate system, n It is the number of two-dimensional cross-sectional images obtained by cutting the three-dimensional image along the circumferential direction.

6. The optical fiber package fine geometry reconstruction modeling method according to claim 4, characterized in that: The specific process of identifying the associated centroid points in the optical fiber cross section and connecting them in order according to the optical fiber connection reconstruction algorithm is as follows: Assume the fiber diameter is d The core diameter is D , manually select the starting point p (l,m,q) , the cylindrical coordinates are ( ρ 0, θ 0, Z 0), indicating the l Tier m Optical fiber turn q The optical fiber core centroid point on the cross section is l , m , q are all positive integers, q =1,2,……, ; In the q +1 section to find the starting point p (l,m,q) The radius is δ Neighborhood of; Define the centroid points of the fiber core cross sections associated on two adjacent cross sections as the associated centroid points, and calculate the maximum distance between the associated centroid points of two adjacent cross sections when the fiber is unwound: , the calculation formula is: ； calculate δ Points in the neighborhood and the starting point p (l,m,q) distance; If the minimum distance calculated , then the minimum distance point is q +1 Associated centroid point in cross section p (l,m,q+1) , connect the points p (l,m,q) and p (l,m,q+1) ; If the minimum distance calculated , then suppose k +1 Associated centroid point in cross section p (l,m,q+1) The cylindrical coordinates are ( ρ 0, θ 0+1, Z 0), connection point p (l,m,q) and p (l,m,q+1) ; The associated centroid points of all cross sections are identified in sequence and connected, thus completing the connection of one turn of optical fiber; when When p (l,m,q) The three-dimensional cylindrical coordinates are set as the starting point of the next turn of optical fiber connection p (l,m,q+1) , and then traverse all the sections in turn to find the associated points on each section to complete the connection of the next turn of optical fiber; until the reconstruction of the area to be reconstructed is completed.

Images