A ZYNQ-based polarization-maintaining optical fiber fusion splicer axis alignment method and system

Abstract

The present invention relates to the technical field of optical fiber alignment of a polarization-maintaining optical fiber fusion splicer, and relates to a ZYNQ-based optical fiber alignment method and system. First, two optical fiber end face images are collected, and the collected images are denoised. The denoised images are threshold-segmented using an OTSU adaptive threshold segmentation algorithm, and binarized. The binarized images are processed using a connected domain processing algorithm to obtain coordinate information of optical fiber end faces belonging to the same connected domain. The coordinate information is processed using a formula to obtain center coordinates, and compensation is performed. The PS side determines whether the current position coordinates are at a predetermined position coordinate, and rotates the optical fiber to the predetermined position coordinate. The stress area is determined by detecting the end face image, and an optimal rotation angle is obtained. Accelerated processing is performed on an FPGA, thereby achieving rapid axis alignment and improving axis alignment accuracy. This solves the problem of how to achieve precise alignment of a birefringence axis before polarization-maintaining optical fiber fusion splicing in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of optical fiber alignment of a polarization-maintaining optical fiber fusion splicer, and in particular to an alignment method and system for a polarization-maintaining optical fiber fusion splicer based on ZYNQ. Background Art

[0002] A polarization-maintaining fiber fusion splicer is a device used for optical fiber connections. It precisely aligns and fuses two optical fibers together, establishing a connection. In optical fiber communications, since optical fibers transmit light signals, precise alignment of the fiber end faces is crucial during connection to avoid signal loss and malfunction. Traditional optical fiber connection methods, such as mechanical and adhesive splicing, require manual operation and are prone to errors and instability. However, a polarization-maintaining fusion splicer achieves high-quality, efficient fiber connections through automated, precise alignment and splicing. With the continuous advancement of optical fiber communication technology, polarization-maintaining fusion splicers are constantly being upgraded and improved to meet the growing demand for optical fiber connections.

[0003] The mainstream polarization-maintaining fiber fusion splicers currently in use in China primarily use two polarization-maintaining fiber alignment methods: the POL method, invented by Ericsson of Sweden, and the PAS method, invented by Fujikura of Japan. The POL method uses a side-illumination light source to plot an intensity vs. rotation angle curve on an observation screen for comparison. The main difference between the two alignment methods lies in the position at which the fiber image is observed. The POL observation screen is located at the focal point of the fiber's cylindrical lens effect ("Polarization of observation bylens-effect tracing"). The PAS observation screen is positioned more forward, at the midpoint between the fiber core and the focal point. This is known as the Profile Alignment System, also known as direct center image monitoring.

[0004] Currently, optical fiber alignment is primarily achieved through light intensity distribution detection. This method obtains the optical fiber's light intensity curve and determines the rotation angle when the peak is obtained. This method suffers from peak fuzziness, large errors, and low precision. To address the above shortcomings of special polarization-maintaining fiber alignment solutions, this paper addresses the existing problem of achieving precise alignment of the birefringence axis before polarization-maintaining fiber splicing. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies in the prior art and to propose an axis alignment method and system for a polarization-maintaining optical fiber fusion splicer based on ZYNQ.

[0006] In order to achieve the above object, the present invention is achieved through the following technical solutions:

[0007] In a first aspect, the present invention provides a method for aligning a polarization-maintaining optical fiber fusion splicer based on ZYNQ, the method comprising the following steps:

[0008] Step S1: collecting two fiber end face images;

[0009] Step S2: performing denoising on the collected image;

[0010] Step S3: Use the OTSU adaptive threshold segmentation algorithm to perform threshold segmentation on the image denoised in step S2 to obtain the optimal threshold h;

[0011] Step S4: using an edge operator to perform binarization processing on the image obtained in step S3 to obtain a binary image with a black and white effect;

[0012] Step S5: performing a connected domain processing algorithm on the binary image with a black and white effect to obtain coordinate information of the optical fiber end faces belonging to the same connected domain;

[0013] Step S6: Process the coordinate information obtained in step S5 using the coordinate formula to obtain the center coordinates, i.e., the current position coordinates;

[0014] Step S7: Assume that the centers of the stress areas of the two optical fiber end faces are on the same horizontal line, that is, the pre-position coordinates are at the same height, and the current position coordinates are (x ave ,y ave ), using a two-step rotation method, when the current position coordinate value is far from the predetermined position coordinate value, the rotation speed of the polarization-maintaining fiber fusion splicer is set to high speed, and coarse adjustment is performed. The difference between the current position coordinate and the predetermined position coordinate is continuously calculated through the PL module and the PS module. When the difference is less than or equal to the coarse adjustment set threshold, the rotation speed of the polarization-maintaining fiber fusion splicer is reduced, and fine adjustment is performed. The fuzzy PID algorithm is used to continuously reduce the difference between the current position coordinate and the predetermined position coordinate. When the difference is less than or equal to the fine adjustment set threshold, that is, the current position coordinate reaches the predetermined position coordinate, the operation is stopped;

[0015] Step S8: Calculate the difference between the curve fitted by measuring the current position coordinates and the actual position coordinates, compensate the current position coordinates, obtain the compensated center coordinates, adjust the rotation of the polarization-maintaining fiber fusion splicer, and complete the axis alignment.

[0016] Preferably, in step S1, a grayscale MT9V034 camera is used in conjunction with a high-magnification lens to capture fiber endface images. The image resolution is set to 640*480, and the fiber endface image is stored as an 8-bit grayscale image in RGB888 format. Images captured using this camera have fast exposure speeds, minimal distortion, and a high frame rate, reducing errors caused by slow image acquisition speeds and improving system response accuracy. The PL VDMA in the ZYNQ is used to transfer the image to DDR3 memory, where the camera image is read out using VDMA.

[0017] Preferably, in step S2, a 3*3 convolution operator or a 5*5 convolution operator is used to scan and denoise the collected image, and a threshold is set. When the threshold is less than the threshold, denoising is performed again until the threshold is reached. Before image binarization, sharp noise in the image must be eliminated and image smoothing, blurring, and other functional effects must be achieved. Compared with performing denoising after binarization, this method effectively avoids the inaccurate threshold distinction of the adaptive threshold segmentation algorithm, resulting in large errors in image judgment position, which ultimately affects the center position calibration.

[0018] Preferably, the OTSU adaptive threshold segmentation algorithm calculates the grayscale values ​​of all pixels in the image and selects a value that maximizes the variance between the grayscale values ​​of all pixels and the selected grayscale value. This value is the optimal threshold h. Compared to fixed threshold segmentation algorithms, OTSU can obtain the optimal threshold value, which is calculated by maximizing the between-class variance. Therefore, it has stronger anti-interference performance and better threshold segmentation results.

[0019] Preferably, the image obtained in step S3 is binarized using an edge operator as described in step S4 to obtain a binary image with a black and white effect, specifically: if the pixel value of the currently scanned pixel is x, if x>h, then 0 is output, otherwise, 255 is output, thereby obtaining a frame of black and white image containing the image information of the optical fiber end face, and an erosion and dilation operation is performed on it to eliminate isolated noise points, and the connected areas are white and the background is black.

[0020] Preferably, the connected domain processing algorithm is performed on the binary image with black and white effect in step S5 to obtain coordinate information belonging to the same domain, including the x-axis and the y-axis, specifically: a connected domain information table is set, if the pixel coordinate data of a point in the 8-neighborhood domain is S11, S12, S13, S21, S22, S23; two RAMs are used to store the connected domain information table, and a dual-port RAM is used to store the coordinate information (x1, y1), (x2, x2) (x3, y3) ... (xn, yn) in the connected body scanned by the connected domain algorithm;

[0021] When scanning the image line by line from left to right, the mark on the point being scanned is related to the four points in its previous neighborhood. This point mark includes three cases:

[0022] (1) The four points in the previous neighborhood are all white, indicating that this is the starting point of a new connected domain. S23 is given a new label and the new connected domain information is updated in the connected domain information table.

[0023] (2) Two of the four points have different labels. This can only happen in two cases: S13 and S21 have different labels, or S13 and S11 have different labels. In this case, it means that a new point has been created to connect the previously disconnected area. The label of S21 or S11 is assigned to S22, and the label of S22 is also assigned to S23.

[0024] (3) Among the 4 points, 1, 2, 3 or 4 are marked, but their labels are the same, so the label of S11 is assigned to S22.

[0025] Preferably, the coordinate formula in step S6 is as follows:

[0026]

[0027] y=ax 2 +bx+c

[0028] a=(||g(A)||+||g(C)||-2||g(B)||) / 2

[0029] b=(||g(C)||-||g(A)||) / 2

[0030] c=||g(B)||

[0031] Where g(A) is the derivative of f(x-1), g(C) is the derivative of f(x+1), and g(B) is the derivative of f(x).

[0032] The outline of the stress zone can be obtained from the above formula, and the center coordinates can be obtained from the following formula;

[0033] x=x1+x2+x3......xn

[0034] y=y1+y2+y3......yn.

[0035] Preferably, the PID formula for the rotation alignment in step S8 is as follows:

[0036]

[0037] Where, e(t) is the difference between the current fiber position and the predetermined position; u(t) is the current output control signal; K p is the proportional coefficient; T t Integration time constant; T D is the differential time constant; t is the time interval from the start of adjustment to the output of the current control quantity.

[0038] In a second aspect, the present invention provides a ZYNQ-based polarization-maintaining fiber fusion splicer alignment system, the system comprising a camera for capturing fiber end faces, a ZYNQ mainboard, and the ZYNQ mainboard comprising a PS module and a PL module;

[0039] The PL module integrates an optical fiber image acquisition module and an image processing module; the image processing module mainly performs denoising, OTSU adaptive threshold segmentation algorithm processing, binarization processing and connected domain algorithm processing on the collected light image;

[0040] The PS module and the PL module cooperate to process the center coordinates obtained by the connected domain algorithm. The polarization-maintaining fiber fusion splicer adopts two-step rotation to realize axis alignment, measures the error between the current position coordinate and the actual position coordinate, and compensates for it to complete the axis alignment.

[0041] The present invention has the following beneficial effects: (1) The present invention proposes to determine the stress area by detecting the end face image, obtain the optimal rotation angle, and implement accelerated processing on the FPGA, thereby achieving the purpose of rapid axis determination and improving the axis determination accuracy. In addition, the end face of the polarization-maintaining optical fiber generally includes panda type, bow tie type, elliptical cladding type, etc. Currently, general image algorithms can only target a single end face type, while this solution can identify polarization-maintaining optical fibers with multiple different end face types.

[0042] (2) The method and system described in the present invention can apply image processing methods to polarization-maintaining optical fiber fusion splicers for the first time. The image processing methods can more accurately obtain the azimuth angle of the rotation core, and the image processing algorithm can be run on the FPGA. Due to the parallel processing characteristics of the FPGA, the response time can be greatly accelerated. Compared with the PAS method and other methods used to obtain the rotation angle when the general polarization-maintaining optical fusion splicer is aligning the axis, this method can obtain a more accurate rotation angle by using the image processing method mentioned in this patent as long as the image of the optical fiber end face is collected during the aligning process, and the speed will be faster and the space occupied will be smaller. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Schematic diagram of the process of the present invention;

[0044] Figure 2 Schematic diagram of a pixel connected domain of the present invention;

[0045] Figure 3 is the convolution operator template of the present invention;

[0046] Figure 4 Schematic diagram of complete connectivity and incomplete connectivity;

[0047] Figure 5 Schematic diagram of edge oscillation;

[0048] Figure 6 Schematic diagram of the center coordinates of different optical fiber end face identification locations of the present invention;

[0049] Figure 7 Schematic diagram of monitoring the stress area on the optical fiber end face using the connected domain method;

[0050] Figure 8 This is a schematic diagram of the first case among the four-point judgment cases in the field;

[0051] Figure 9 This is a schematic diagram of the second case in the judgment of the four points in the field;

[0052] Figure 10 This is a schematic diagram of the third case in the judgment of 4 points in the field. DETAILED DESCRIPTION

[0053] The present invention will be further described below with reference to the examples, but they are not intended to limit the present invention.

[0054] A ZYNQ-based polarization-maintaining fiber fusion splicer alignment system, comprising a camera for capturing fiber end faces and a ZYNQ mainboard, wherein the ZYNQ mainboard comprises a PS module and a PL module;

[0055] The PL module integrates an optical fiber image acquisition module and an image processing module; the image processing module mainly performs denoising, OTSU adaptive threshold segmentation algorithm, binarization and connected domain algorithm processing on the collected light image;

[0056] The PS module and the PL module cooperate to process the center coordinates obtained by the connected domain algorithm. The polarization-maintaining fiber fusion splicer adopts two-step rotation to realize axis alignment, measures the error between the current position coordinate and the actual position coordinate, and compensates for it to complete the axis alignment.

[0057] like Figure 1 As shown, a method for aligning the polarization-maintaining optical fiber fusion splicer based on ZYNQ, the method comprising the following steps:

[0058] Step S1: Capture two fiber end face images; use a grayscale MT9V034 camera with a high magnification lens tube to capture the fiber end face image. The fiber end face image is set to 640*480 resolution, and the fiber end face image is an 8-bit grayscale image, which is spliced ​​into RGB888 for storage.

[0059] Step S2: De-noise the collected image; use a 3*3 convolution operator or a 5*5 convolution operator to scan and de-noise the collected image, and set a threshold. When the value is less than the threshold, perform de-noising again until the threshold is reached. Figure 4 As shown, after multiple denoising, the key is to obtain a "completely connected" edge, that is, the area of ​​each edge point must have another edge point. Disconnected edge points will greatly affect the results; the operator template is as follows Figure 3 As shown in the figure, before the image is binarized, it is necessary to first eliminate the sharp noise of the image and achieve image smoothing, blurring and other functional effects. Compared with denoising after binarization, this method can effectively avoid the inaccurate threshold distinction of the adaptive threshold segmentation algorithm, resulting in large errors in image judgment position, which ultimately affects the center position calibration.

[0060] Step S3: Use the OTSU adaptive threshold segmentation algorithm to perform threshold segmentation on the image denoised in step S2. The OTSU adaptive threshold segmentation algorithm calculates the grayscale values ​​of all pixels in the image and selects a value that maximizes the variance between the grayscale values ​​of all pixels and the selected grayscale value. This value is the optimal threshold h, which achieves the best segmentation effect, resulting in an image with minimal noise and only connected edges. During binarization, the Sobel edge detection operator is used to detect the edges of the stress zone.

[0061] Step S4: Use a 3*3 edge detection operator to binarize the image obtained in step S3. Using the edge detection operator to perform convolution on the image row by row can obtain an image with a black and white effect and contour features. Only edge information exists, and a binary image with a black and white effect is obtained; specifically: if the pixel value of the currently scanned pixel is x, if x>h, then output 0, otherwise, output 255, thereby obtaining a frame of black and white image containing the image information of the optical fiber end face, and perform an erosion and expansion operation on it to eliminate isolated noise points. The connected areas are white and the background is black.

[0062] Step S5: performing a connected domain processing algorithm on the binary image with a black and white effect to obtain coordinate information of the optical fiber end faces belonging to the same connected domain, and obtaining the outline of the stress area; Figure 7 This is a schematic diagram of the connected domain method for monitoring the stress zone on the optical fiber end face. The stress zone is the black circle in the figure, but there are also other shapes, such as dumbbell shape, etc. During detection, the center coordinates of a detected contour are actually the center coordinates of this small circle. There are two stress zones on the optical fiber end face, namely two small circles.

[0063] Step S6: Use Verilog to write a connected domain processing algorithm, and use the coordinate formula to process the coordinate information obtained in step S5 to obtain the center coordinates, that is, the current position coordinates;

[0064] Specifically: Set the connected domain information table, as shown in Table 1, if the pixel coordinate data of a point in the 8-neighborhood is S11, S12, S13, S21, S22, S23, Figure 2 As shown,

[0065] Table 1 Connected domain information table

[0066] Points Label 1 1 1 2 1 -1

[0067] When scanning an image line by line from left to right, what label the scanned point should be marked with is actually only related to the points that have been scanned before it, that is, the 3 points above and 1 point on the left in its neighborhood. Therefore, what label a certain point should be marked with is only related to the four points in its previous neighborhood, that is, the left, upper left, upper and upper right points. Therefore, to scan the image with such a 2x3 operator, only one line of image needs to be cached. The next step is to judge the situation of the 4 points in the neighborhood to determine how the current point should be marked. There are 16 cases for a total of 4 points, but in reality there are only three cases, such as Figure 8-10 As shown:

[0068] (1) The four points in the previous neighborhood are all white, indicating that this is the starting point of a new connected domain. S23 is given a new label and the new connected domain information is updated in the connected domain information table.

[0069] (2) Two of the four points have different labels. This can only happen in two cases: S13 and S21 have different labels, or S13 and S11 have different labels. There will not be a case where three points have different labels, or four points have different labels. In this case, it means that a new point has been created to connect the previously disconnected area. The label of S21 or S11 is assigned to S22, and the label of S22 is also assigned to S23.

[0070] (3) Among the 4 points, 1, 2, 3 or 4 are marked, but their labels are the same, so the label of S11 is assigned to S22.

[0071] Two RAMs are used to store the information table of the connected domain, and a dual-port RAM is used to store the coordinate information (x1, y1), (x2, x2) (x3, y3) ... (xn, yn) in the connected domain algorithm. The characteristics of this algorithm are (1) it can be implemented using a low-end FPGA without any external memory. It also does not require DDR for storage, and only a dozen RAM blocks are used, consuming less logic resources.

[0072] 2) High real-time performance with minimal and constant latency. Because this method utilizes parallel pipeline processing—that is, a single scan of the image completes the identification of all connected regions—the latency for identifying each connected region is constant and does not increase with the number of connected regions in the image. The latency is minimal, approximately the time it takes to scan a dozen or so lines of the image. This results in low memory consumption and minimal latency.

[0073] 3) It can provide various statistical information about connected regions, such as area, perimeter, and coordinates of the center point of the circumscribed rectangle. When identifying connected regions, multiple denoising steps are required to obtain an image free of excess noise. This ensures that the current image contains only two objects to be identified. This method provides more accurate center coordinates for this system. S1-S6 are all implemented on the FPGA within the Zynq, which improves processing speed.

[0074] The coordinate information data stored in the dual-port RAM is transferred to the DDR memory on the PS side through DMA. The dual-core Cortex-A9 processor in the PS is used to write C language code to calculate the sum of the x-coordinate and the y-coordinate through the following coordinate formula. This coordinate is the end face of the polarization-maintaining fiber. When extracting the edge of the image, it is necessary to prevent the edge oscillation problem. Figure 5 As shown, the edge oscillation problem is prevented by calculating the sub-pixel maximum point.

[0075]

[0076] y=ax 2 +bx+c

[0077] a=(||g(A)||+||g(C)||-2||g(B)||) / 2

[0078] b=(||g(C)||-||g(A)||) / 2

[0079] c=||g(B)||

[0080] Where g(A) is the derivative of f(x-1), g(C) is the derivative of f(x+1), g(B) is the derivative of f(x), and x is the x-coordinate of the currently calculated coordinate point. Each time you run these five formulas, you can get a set of x and y coordinates. Running them many times can get many sets of x and y coordinates.

[0081] The outline of the stress zone can be obtained from the above formula, and the center coordinates can be obtained from the following formula.

[0082] x=x1+x2+x3......xn

[0083] y=y1+y2+y3......yn;

[0084] The coordinates of the stress zone can be used to obtain the center coordinates of closed-loop connected figures of various shapes, such as Figure 6 shown.

[0085] Step S7: Assume that the centers of the stress areas of the two optical fiber end faces are on the same horizontal line, that is, the pre-position coordinates are at the same height, and the current position coordinates are (x ave ,y ave ), adopting a two-step rotation method, when the current position coordinate value is far away from the predetermined position coordinate value, the optical fiber is clamped by the optical fiber clamp of the polarization-maintaining optical fiber fusion splicer, and the rotation of the optical fiber clamp drives the rotation of the optical fiber, that is, the optical fiber end face is rotating, and the rotation speed of the polarization-maintaining optical fiber fusion splicer is set to high speed, and coarse adjustment is performed, and the difference between the current position coordinate and the predetermined position coordinate is continuously calculated by the PL module and the PS module. When the difference is less than or equal to the coarse adjustment set threshold, the rotation speed of the polarization-maintaining optical fiber fusion splicer is reduced, and fine adjustment is performed, and the fuzzy PID algorithm is used to continuously reduce the difference between the current position coordinate and the predetermined position coordinate. When the difference is less than or equal to the fine adjustment set threshold, that is, the current position coordinate reaches the predetermined position coordinate, the operation is stopped;

[0086] Step S8: Calculate the difference between the curve fitted by measuring the current position coordinates and the actual position coordinates. The actual coordinates are the pixel coordinates (accurate values) measured by professional instruments, but the ones calculated by the image algorithm have a certain error value compared with the actual real coordinates, so compensation is required. The entire alignment process is to continuously rotate the optical fiber. Each frame of the image can calculate a coordinate to compensate for the current position coordinates to obtain the compensated center coordinates; compare the actual coordinates with the coordinates calculated by the algorithm each time, calculate 100 times, and obtain the average difference. In the next calculation, add or subtract this difference; adjust the polarization-maintaining fiber fusion rotation to complete the axis alignment.

[0087] The PID formula for rotating core alignment is as follows:

[0088]

[0089] Where, e(t) is the difference between the current fiber position and the predetermined position; u(t) is the current output control signal; K p is the proportional coefficient; T t Integration time constant; T D is the differential time constant; t is the time interval from the start of adjustment to the output of the current control quantity.

[0090] Based on the curve fitted by measuring the current position coordinates, the least squares method is used to minimize the sum of squared errors to select the optimal curve fitting data points. Generally, there are the following steps:

[0091] 1. Determine the form of the fitting function. According to the coordinates of the measurement points, select a suitable fitting function and determine its form;

[0092] 2. Determine the error function. Common error functions include mean square error and mean absolute error.

[0093] 3. Construct an optimization model for the error function;

[0094] 4. Solve for the coefficients of the optimal fit function, differentiate the error function, and obtain the closed-form solution of the coefficients of the optimal fit function; based on the form of the fitting function and the definition of the error function, solve for the coefficients of the optimal fit function.

[0095] like Figure 6 Figure 2 shows a schematic diagram of the center coordinates of different fiber end face identification locations. The ultra-fine polarization-maintaining fiber involved in the present invention is a special fiber with a diameter as thin as 40 μm, two stress zones symmetrically distributed on either side of the fiber core, and a structure that can be shaped like a bow tie, panda, or bow tie. The direction of the line connecting the two stress zones is called the stress axis direction of the ultra-fine polarization-maintaining fiber.

[0096] The above shows and describes the basic principles, main features, and advantages of the present invention. However, the above is only a specific embodiment of the present invention, and the technical features of the present invention are not limited thereto. Any other implementation methods derived by any person skilled in the art without departing from the technical solution of the present invention should be included in the patent scope of the present invention.

Claims

1. A method for aligning the polarization-maintaining optical fiber fusion splicer based on ZYNQ, characterized in that: The method comprises the following steps: Step S1: collecting two fiber end face images; Step S2: performing denoising on the collected image; Step S3: Use the OTSU adaptive threshold segmentation algorithm to perform threshold segmentation on the image denoised in step S2 to obtain the optimal threshold h; Step S4: using an edge operator to perform binarization processing on the image obtained in step S3 to obtain a binary image with a black and white effect; Step S5: performing a connected domain processing algorithm on the binary image with a black and white effect to obtain coordinate information of the optical fiber end faces belonging to the same connected domain; Step S6: Process the coordinate information obtained in step S5 using the coordinate formula to obtain the center coordinates, i.e., the current position coordinates; Step S7: Assume that the centers of the stress areas of the two optical fiber end faces are on the same horizontal line, that is, the pre-position coordinates are at the same height, and the current position coordinates are (x ave ,y ave ), using a two-step rotation method, when the current position coordinate value is far from the predetermined position coordinate value, the rotation speed of the polarization-maintaining fiber fusion splicer is set to high speed, and coarse adjustment is performed. The difference between the current position coordinate and the predetermined position coordinate is continuously calculated through the PL module and the PS module. When the difference is less than or equal to the coarse adjustment set threshold, the rotation speed of the polarization-maintaining fiber fusion splicer is reduced, and fine adjustment is performed. The fuzzy PID algorithm is used to continuously reduce the difference between the current position coordinate and the predetermined position coordinate. When the difference is less than or equal to the fine adjustment set threshold, that is, the current position coordinate reaches the predetermined position coordinate, the operation is stopped; Step S8: Calculate the difference between the curve fitted by measuring the current position coordinates and the actual position coordinates, compensate the current position coordinates, obtain the compensated center coordinates, adjust the rotation of the polarization-maintaining fiber fusion splicer, and complete the axis alignment.

2. The ZYNQ-based polarization-maintaining optical fiber fusion splicer alignment method according to claim 1, characterized in that: In step S1, a grayscale MT9V034 camera is used in conjunction with a high-magnification lens barrel to capture the fiber end face image. The fiber end face image is set to a resolution of 640*480, and the fiber end face image is an 8-bit grayscale image, which is spliced ​​into RGB888 for storage.

3. The ZYNQ-based polarization-maintaining optical fiber fusion splicer alignment method according to claim 1, characterized in that: In step S2, a 3*3 convolution operator or a 5*5 convolution operator is used to perform scanning denoising on the collected image, and a threshold is set. When the value is less than the threshold, denoising is performed again until the threshold is reached.

4. The method for aligning the polarization-maintaining optical fiber fusion splicer based on ZYNQ according to claim 1, characterized in that: The OTSU adaptive threshold segmentation algorithm is to select a value by counting the grayscale values ​​of all pixels in the image so that the variance between the grayscale values ​​of all pixels and the selected grayscale value is the largest. Then this value is the optimal threshold h.

5. The ZYNQ-based polarization-maintaining optical fiber fusion splicer alignment method according to claim 1, characterized in that: Step S4 uses the edge operator to binarize the image obtained in step S3 to obtain a binary image with a black and white effect. Specifically, if the pixel value of the currently scanned pixel is x, if x>h, then output 0, otherwise, output 255, thereby obtaining a frame of black and white image containing the image information of the optical fiber end face, and perform an erosion and expansion operation on it to eliminate isolated noise points. The connected areas are white and the background is black.

6. The method for aligning the polarization-maintaining optical fiber fusion splicer based on ZYNQ according to claim 1, characterized in that: In step S5, the connected domain processing algorithm is performed on the binary image with black and white effect to obtain coordinate information belonging to the same domain, including the x-axis and the y-axis. Specifically, a connected domain information table is set. If the pixel coordinate data of a point in the 8-neighborhood domain is S11, S12, S13, S21, S22, S23; two RAMs are used to store the connected domain information table, and a dual-port RAM is used to store the coordinate information (x1, y1), (x2, x2) (x3, y3) ... (xn, yn) scanned in the connected body by the connected domain algorithm; When scanning the image line by line from left to right, the mark on the point being scanned is related to the four points in its previous neighborhood. This point mark includes three cases: (1) The four points in the previous neighborhood are all white, indicating that this is the starting point of a new connected domain. S23 is given a new label and the new connected domain information is updated in the connected domain information table. (2) Two of the four points have different labels. This can only happen in two cases: S13 and S21 have different labels, or S13 and S11 have different labels. In this case, it means that a new point has been created to connect the previously disconnected area. The label of S21 or S11 is assigned to S22, and the label of S22 is also assigned to S23. (3) Among the 4 points, 1, 2, 3 or 4 are marked, but their labels are the same, so the label of S11 is assigned to S22.

7. The method for aligning the polarization-maintaining optical fiber fusion splicer based on ZYNQ according to claim 1, characterized in that: The coordinate formula in step S6 is as follows: y=ax 2 +bx+c a=(||g(A)||+||g(C)||-2||g(B)||) / 2 b=(||g(C)||-||g(A)||) / 2 c=||g(B)|| Where g(A) is the derivative of f(x-1), g(C) is the derivative of f(x+1), and g(B) is the derivative of f(x). The outline of the stress zone can be obtained from the above formula, and the center coordinates can be obtained from the following formula; x=x1+x2+x3......xn y=y1+y2+y3......yn.

8. The ZYNQ-based polarization-maintaining optical fiber fusion splicer alignment method according to claim 1, characterized in that: The PID formula for the rotation centering in step S8 is as follows: Where, e(t) is the difference between the current fiber position and the predetermined position; u(t) is the current output control signal; K p is the proportional coefficient; T t Integration time constant; T D is the differential time constant; t is the time interval from the start of adjustment to the output of the current control quantity.

9. The system for the ZYNQ-based polarization-maintaining optical fiber fusion splicer alignment method according to any one of claims 1 to 8, characterized in that: The system includes a camera for collecting optical fiber end faces and a ZYNQ mainboard, wherein the ZYNQ mainboard includes a PS module and a PL module; The PL module integrates an optical fiber image acquisition module and an image processing module; the image processing module mainly performs denoising, OTSU adaptive threshold segmentation algorithm processing, binarization processing and connected domain algorithm processing on the collected light image; The PS module and the PL module cooperate to process the center coordinates obtained by the connected domain algorithm. The polarization-maintaining fiber fusion splicer adopts two-step rotation to realize axis alignment, measures the error between the current position coordinate and the actual position coordinate, and compensates for it to complete the axis alignment.

Images