Intelligent detection method and system for MPO connector pins

Abstract

The invention provides an intelligent detection method and system for MPO connector pins, and belongs to the technical field of visual detection. The method comprises the following steps: acquiring pin image data, accurately extracting geometric feature vectors of a tip and an axis, and respectively determining a tip offset vector and axis curvature distribution according to the geometric feature vectors; the method comprises the following steps: constructing an incidence matrix of an offset vector and curvature distribution, quantifying the correlation of the offset vector and the curvature distribution, and discriminating the offset caused by the natural bending of a pin according to a correlation threshold and direction consistency judgment; then pin depth point cloud is obtained, error sources such as ambient light interference and measuring equipment are positioned, a depth correction value is determined in a targeted mode, accurate correction of depth errors is achieved, and axis fitting precision and tip tiny offset capture are both considered. Finally, accurate data support is provided for pin quality evaluation, and the signal transmission stability and reliability of the MPO connector in a data center and an optical fiber transmission system are effectively guaranteed.

Description

Technical Field

[0001] This application relates to the field of visual inspection technology, specifically to an intelligent inspection method and system for MPO connector pins. Background Technology

[0002] Currently, the rapid development of optical communication networks has made MPO connectors, as multi-fiber push-pull connectors, play a crucial role in data centers and fiber optic transmission systems. The precise arrangement of their internal pins directly determines the stability and reliability of signal transmission. Even minute deviations or abnormal pin orientations can lead to increased insertion loss or even connection failure, thus affecting the performance and reliability of the entire network. Current detection methods largely rely on two-dimensional image analysis or manual visual inspection. These methods often struggle to fully capture the true state of the pins in three-dimensional space when dealing with the spatial arrangement of pins. When a pin tip experiences a micrometer-level shift, existing image processing methods struggle to simultaneously ensure the accuracy of axis fitting and the stability of depth information, resulting in a contradiction in the final orientation judgment: pursuing axis fitting accuracy increases depth error, while controlling depth error makes it difficult to capture minute tip shifts.

[0003] The information provided in the background section of this application is only for enhancing the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] In view of this, this application provides an intelligent detection method and system for MPO connector pins, which can achieve accurate correction of depth error while taking into account the accuracy of axis fitting and the capture of small tip offset.

[0005] In a first aspect, embodiments of this application provide an intelligent detection method for MPO connector pins. The method includes: acquiring pin image data of the MPO connector, and determining a pin tip geometric feature vector and a pin axis geometric feature vector based on the pin image data; determining a pin tip offset vector based on the pin tip geometric feature vector, the pin tip offset vector including a tip offset direction and a tip offset amplitude value; determining an axis curvature distribution based on the pin axis geometric feature vector; determining a correlation matrix between axis curvature and tip offset based on the pin tip offset vector and the axis curvature distribution; determining whether the correlation coefficient in the correlation matrix exceeds a preset correlation coefficient threshold; if it exceeds the preset correlation coefficient threshold, determining whether the pin tip offset vector and the axis curvature distribution have directional consistency; if it has directional consistency, determining the current pin offset factor as natural pin curvature, acquiring a pin depth point cloud, determining the pin depth error source based on the pin depth point cloud, the pin depth error source including ambient light interference error and measurement equipment error; and determining a depth correction value based on the corresponding pin depth error source.

[0006] Secondly, embodiments of this application provide an intelligent detection system for MPO connector pins, the system comprising: an acquisition module, a first determination module, a second determination module, a third determination module, a judgment module, a fourth determination module, and a fifth determination module. The system comprises the following modules: an acquisition module for acquiring pin image data of the MPO connector and determining the geometric feature vectors of the pin tip and the pin axis based on the pin image data; a first determination module for determining the pin tip offset vector based on the pin tip geometric feature vector, the pin tip offset vector including the offset direction and offset magnitude; a second determination module for determining the axis curvature distribution based on the pin axis geometric feature vector; a third determination module for determining the correlation matrix between the axis curvature and the tip offset based on the pin tip offset vector and the axis curvature distribution set; a judgment module for judging whether the correlation coefficient in the correlation matrix exceeds a preset correlation coefficient threshold, and if it exceeds the preset correlation coefficient threshold, judging whether the pin tip offset vector and the axis curvature distribution have directional consistency; a fourth determination module for determining the current pin offset factor as the natural bending of the pin if it is judged to have directional consistency, acquiring the pin depth point cloud, and determining the source of pin depth error based on the pin depth point cloud, the source of pin depth error including ambient light interference error and measurement equipment error; and a fifth determination module for determining the depth correction value based on the corresponding source of pin depth error.

[0007] This application provides an intelligent detection method and system for MPO connector pins. By acquiring pin image data of the MPO connector and accurately extracting the geometric feature vectors of the pin tips and the pin axis, it overcomes the limitations of traditional two-dimensional image analysis or manual visual inspection in comprehensively capturing the true three-dimensional state of the pins. Based on the geometric feature vectors of the pin tips, it determines the tip offset direction and amplitude value included in the pin tip offset vector, achieving precise quantitative identification of micron-level tip offsets and solving the problem that traditional image processing methods cannot accurately capture such micro-offsets. By using the geometric feature vectors of the pin axis to determine the axis curvature distribution, it clearly presents the bending state of the pin axis in space, providing a key basis for tracing the cause of tip offset. By constructing a correlation matrix between the pin tip offset vector and the axis curvature distribution, the correlation between the two is quantified, making the originally isolated offset and bending information an organic whole, facilitating root-cause analysis of the offset. By judging whether the correlation coefficient in the correlation matrix exceeds a preset correlation coefficient threshold, if it exceeds... The process further verifies the consistency between the pin tip offset vector and the axis curvature distribution. If the verification shows consistency, the pin tip offset is caused by the natural bending of the pin, eliminating irrelevant interference factors. After determining that the offset is caused by the natural bending of the pin, the source of the pin depth error is located by acquiring the pin depth point cloud, such as ambient light interference error or measurement equipment error. This specifically solves the problems that are difficult to avoid in traditional detection, such as image feature blurring caused by pin surface reflection under complex lighting conditions and errors caused by the measurement equipment itself. Finally, the depth correction value is determined based on the clear source of the pin depth error, realizing accurate correction of the depth error and effectively controlling the depth error in the attitude reconstruction process. At the same time, it takes into account the accuracy of axis fitting and the capture of small tip offsets, resolving the contradiction that is difficult to balance in traditional methods. Ultimately, it provides accurate data support for the quality evaluation of MPO connector pins, ensuring the signal transmission stability and reliability of MPO connectors in data centers and fiber optic transmission systems. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating an exemplary embodiment of the intelligent detection method for MPO connector pins provided in this application.

[0010] Figure 2This is a flowchart illustrating an intelligent detection method for MPO connector pins provided in another exemplary embodiment of this application.

[0011] Figure 3 This is a flowchart illustrating an intelligent detection method for MPO connector pins provided in another exemplary embodiment of this application.

[0012] Figure 4 This is a flowchart illustrating an intelligent detection method for MPO connector pins provided in another exemplary embodiment of this application.

[0013] Figure 5 This is a flowchart illustrating an intelligent detection method for MPO connector pins provided in another exemplary embodiment of this application.

[0014] Figure 6 This is a flowchart illustrating an intelligent detection method for MPO connector pins provided in another exemplary embodiment of this application.

[0015] Figure 7 This is a flowchart illustrating an intelligent detection method for MPO connector pins provided in another exemplary embodiment of this application.

[0016] Figure 8 This is a flowchart illustrating an intelligent detection method for MPO connector pins provided in another exemplary embodiment of this application.

[0017] Figure 9 This is a flowchart illustrating an intelligent detection method for MPO connector pins provided in another exemplary embodiment of this application.

[0018] Figure 10 This is a flowchart illustrating an intelligent detection method for MPO connector pins provided in another exemplary embodiment of this application. Detailed Implementation

[0019] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this application will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this application.

[0020] The terms “a,” “one,” and “the” are used to indicate the existence of one or more elements / components / etc.; the terms “including” and “having” are used to indicate an open-ended inclusion and that other elements / components / etc. may exist in addition to those listed. The terms “first” and “second” are used only as markers and are not a limitation on the number of objects.

[0021] Currently, the rapid development of optical communication networks has made MPO connectors, as multi-fiber push-pull connectors, play a crucial role in data centers and fiber optic transmission systems. The precise arrangement of their internal pins directly determines the stability and reliability of signal transmission. Even minor pin deviations or abnormal orientations can lead to increased insertion loss or even connection failure, thus affecting the performance and reliability of the entire network. Current visual inspection methods largely rely on two-dimensional image analysis or manual visual inspection. These methods often struggle to fully capture the true state of pins in three-dimensional space when dealing with pin spatial arrangement. For example, under complex lighting conditions, pin surface reflections can blur image features, making visual inspection results susceptible to interference and misjudgments. Furthermore, as slender structures, the slight bending or tilting of the pin's axis in space directly affects the accuracy of its tip position, and tip offset further amplifies depth errors during orientation reconstruction. This accumulation of errors results in an uneven surface on the reconstructed 3D pin model, making it difficult to accurately reflect the actual spatial orientation. Especially during batch inspection, when the pin tip deviates by a micrometer, the existing visual inspection image processing methods have difficulty simultaneously ensuring the accuracy of axis fitting and the stability of depth information, resulting in a contradiction in the final attitude judgment: when pursuing the accuracy of axis fitting, the depth error increases, while when controlling the depth error, it is difficult to capture the tiny tip deviance.

[0022] Therefore, how to simultaneously achieve high precision in pin axis spatial fitting, effective control of depth error in attitude reconstruction, and accurate capture of minute tip offsets in image data-based detection has become a technical problem that urgently needs to be solved in an intelligent detection method and system for MPO connector pins.

[0023] This application provides an intelligent detection method for MPO connector pins, such as... Figure 1 The method shown is an intelligent detection method for MPO connector pins. This method may include the following steps:

[0024] Step S110: Obtain pin image data of the MPO connector, and determine the geometric feature vector of the pin tip and the geometric feature vector of the pin axis based on the pin image data;

[0025] Step S120: Determine the pin tip offset vector based on the geometric feature vector of the pin tip. The pin tip offset vector includes the offset direction and the offset magnitude.

[0026] Step S130: Determine the axis curvature distribution based on the geometric feature vector of the pin axis;

[0027] Step S140: Determine the correlation matrix between axis bending and tip offset based on the pin tip offset vector and axis curvature distribution;

[0028] Step S150: Determine whether the correlation coefficient in the correlation matrix exceeds the preset correlation coefficient threshold. If it exceeds the preset correlation coefficient threshold, determine whether the pin tip offset vector and the axis curvature distribution have directional consistency.

[0029] Step S160: If it is determined that the direction is consistent, the current pin offset factor is determined to be the natural bending of the pin, and the pin depth point cloud is obtained. The source of pin depth error is determined based on the pin depth point cloud. The source of pin depth error includes ambient light interference error and measurement equipment error.

[0030] Step S170: Determine the depth correction value based on the corresponding pin depth error source.

[0031] According to the intelligent detection method for MPO connector pins provided in this application, this method can overcome the limitations of traditional two-dimensional image analysis and manual inspection in capturing the true spatial state of pins by extracting the three-dimensional geometric features of the pin tip and pin axis, and achieve precise quantitative identification of micron-level tip offset. By constructing the correlation matrix between axis curvature and tip offset, the intrinsic relationship between tip offset and axis curvature is clarified, and tip offset caused by natural pin curvature is accurately identified. By establishing a pin depth point cloud to locate the source of depth error, such as ambient light interference or measurement equipment error, and determining the corresponding depth correction value, the problems that traditional detection methods cannot avoid, such as image blurring under complex lighting and equipment errors, are effectively solved. This method can balance the high precision of pin axis spatial fitting, depth error control of attitude reconstruction, and capture of minute tip offsets, resolving the contradiction that traditional methods cannot balance these three aspects. It provides reliable data support for pin quality assessment, improves the accuracy and stability of MPO connector pin detection, and ensures the reliability of signal transmission in optical communication networks.

[0032] The steps of the intelligent detection method for MPO connector pins provided in this application are described in detail below:

[0033] In one embodiment of this application, step S110, which involves acquiring pin image data of the MPO connector and determining the pin tip geometry and pin axis geometry based on the pin image data, further includes the following steps: Figure 2 As shown, the specific content is as follows:

[0034] Step S210: Perform background segmentation processing on the pin image data to determine the pin tip image data and pin axis image data;

[0035] Step S220: Determine the three-dimensional point cloud of the needle tip based on the needle tip image data, and determine the three-dimensional point cloud of the needle axis based on the needle axis image data;

[0036] Step S230: Based on the three-dimensional point cloud of the pin tip and the three-dimensional point cloud of the pin axis, an edge detection algorithm is used to extract features to determine the geometric feature vector of the pin tip and the geometric feature vector of the pin axis.

[0037] Specifically, when acquiring pin image data, an industrial-grade high-resolution camera can be used to capture pin images. Under a ring-shaped LED light source, the MPO connector pins are photographed from three different perspectives: front, 45-degree side, and 70-degree side. Five images are acquired from each perspective to ensure data redundancy, for a total of 15 raw images, avoiding the omission of pin spatial morphology information from a single perspective. Simultaneously, structured light triangulation is used to acquire pin depth data. A 650nm wavelength line laser is projected, with the angle between the camera and the laser set to 30 degrees. Subpixel-level parallax is calculated using the gray-scale centroid method to obtain point cloud data with a depth resolution of 0.8μm, completely recording the three-dimensional spatial depth information of the pin tip.

[0038] Specifically, the effective regions corresponding to the pin tip and pin axis can be accurately separated from the pin image data, eliminating background, connector housing, and other irrelevant interference factors, providing a clean image data source for subsequent 3D point cloud construction and feature extraction. By fusing pin images and pin depth data, the 2D tip and axis image data are transformed into 3D point cloud data, fully preserving the spatial position and morphological information of the pin tip and axis. Edge detection is performed on the 3D point cloud of the pin tip and the 3D point cloud of the pin axis respectively. An improved Canny edge detection algorithm (adapted to 3D point cloud data) can be used, with a low threshold of 50 and a high threshold of 150. By calculating the angle between the normal vector of each point in the point cloud and the normal vector of its neighboring points, edge points with abrupt changes in normal vector are selected. The edge points are then connected and smoothed to obtain continuous tip edge curves and axis edge curves. Based on the tip edge curves, a cone model is fitted using the least squares method. Through iterative optimization, the key parameters of the cone are obtained, including the radius of curvature of the tip vertex, the cone angle, and the end face tilt angle, ultimately forming a geometric feature vector of the pin tip containing the above parameters. Based on the axis edge curve and the three-dimensional point cloud of the axis, the least squares method is used to fit a spatial straight line as the axis baseline. The distance deviation of each point in the three-dimensional point cloud of the axis from the baseline is calculated, and the curvature distribution characteristic data of the axis is obtained (by dividing the axis into several equal-length segments and calculating the local curvature value of each segment). At the same time, the spatial orientation parameters of the axis (such as the angle between the axis and the X-axis and Y-axis of the reference coordinate system) are extracted to form the geometric feature vector of the pin axis, which fully characterizes the spatial curvature and tilt state of the axis.

[0039] For example, for pin inspection of a certain MPO connector, a frontal view with an original pin image of 4096×3000 pixels is first selected. The pin tip image data (keeping only the tip region) and the pin axis image data (keeping only the axis region) are then separated. The separated pin tip image data is registered with a global depth map (depth resolution 0.8μm) acquired by structured light triangulation. The depth values ​​corresponding to 7200 pixels within the tip mask are extracted and converted into three-dimensional world coordinates (X∈[12.3,15.6]mm, Y∈[8.5,10.2]mm, Z∈[2.1,2.8]mm) using a calibrated camera intrinsic parameter matrix, resulting in an average density of 120 points / μm. The three-dimensional point cloud of the pin tip, with m², no repeating points, and no outliers, was processed simultaneously with the pin axis image data. 3100 uniformly distributed three-dimensional coordinate points with a 10μm interval were extracted to form a three-dimensional point cloud of the pin axis covering the entire 100mm length. Finally, the improved Canny edge detection algorithm (low threshold 50, high threshold 150) was used to filter edge points and smooth the three-dimensional point cloud of the pin tip and the pin axis, respectively, to obtain continuous tip edge curves and axis edge curves. Based on the tip edge curves, a cone model was fitted using the least squares method to extract the geometric feature vector of the pin tip with a vertex curvature radius of 1.6μm, a cone angle of 8.3 degrees, and an end face tilt angle of 0.22 degrees.

[0040] For example, a spatial baseline can be fitted based on the edge curve of the axis and the three-dimensional point cloud of the axis to divide the pin axis into 10 segments (10mm each). After the division, the curvature value of each segment is calculated as [0.01, 0.02, 0.015, 0.025, 0.03, 0.02, 0.018, 0.022, 0.019, 0.021] cm⁻¹. At the same time, the spatial orientation parameters of the axis with an angle of 3.2 degrees with the X-axis and an angle of 1.8 degrees with the Y-axis are extracted to form a complete geometric feature vector of the pin axis.

[0041] In the above method, the spatial position and morphological information of the pin tip and pin axis are fully preserved by constructing a 3D point cloud, breaking the limitation that 2D images can only reflect planar information. By adapting the Canny algorithm and least squares modeling to 3D point clouds, key geometric parameters such as the radius of curvature of the tip vertex, the cone angle, and the curvature distribution of the axis are accurately extracted, ensuring the accuracy of geometric features. This forms a complete transformation link from 2D images to 3D features, which not only achieves comprehensive capture of the 3D spatial morphology of the pin tip and axis, but also provides reliable and high-precision core data support for subsequent calculation of the tip offset vector and analysis of the axis bending trend. This reduces the risk of visual detection misjudgment caused by incomplete or insufficient data from the source, laying the foundation for intelligent detection of MPO connector pins.

[0042] In one embodiment of this application, step S120 involves determining a pin tip offset vector based on the pin tip geometric feature vector. The pin tip offset vector includes the pin tip offset direction and the pin tip offset magnitude. The step also includes the following steps: Figure 3 As shown, the specific content is as follows:

[0043] Step S310: Determine the position of the tip vertex based on the geometric feature vector of the pin tip;

[0044] Step S320: Determine the tip offset angle based on the tip vertex position and the preset tip position;

[0045] Step S330: Determine the tip offset direction and tip offset amplitude value based on the tip offset angle;

[0046] Step S340: Determine the pin tip offset vector based on the tip offset direction and tip offset amplitude value.

[0047] Specifically, for example, in the detection of a pin of an MPO connector, based on the extracted geometric feature vector of the pin tip (including a cone angle of 8.4 degrees, a vertex radius of curvature of 1.7 μm, and an end face tilt angle of 0.28 degrees), the corresponding tip sub-image is subjected to Gaussian filtering with σ=1.5 for noise reduction. Then, the subpixel-level tip contour point set is extracted using the Canny algorithm (low threshold 50, high threshold 150). After iterative optimization using the cone model, the tip vertex position is determined to be (25.324 μm, 18.651 μm, 3.210 μm). Subsequently, using the preset tip position (25.300 μm, 18.600 μm, 3.210 μm) as a reference, the centroid deviation is calculated as ΔX=24.0 nm and ΔY=51.0 nm. After obtaining the principal direction vector through principal component analysis, the tip offset angle θ=64.9° is calculated using the formula θ=atan2(51.0 nm, 24.0 nm). Then... The surface texture of the pin tip was analyzed using the gray-level co-occurrence matrix, yielding features of 0.83 contrast, 0.91 correlation, and 0.65 energy. Combined with the texture orientation field extracted by the Gabor filter, it was found that the initial offset direction deviated from the texture trend by 7.2°. The offset direction was then corrected to 72.1°. Simultaneously, the initial geometric amplitude value L1 = 56.3nm and the texture equivalent offset amplitude L2 = 48.5nm were calculated. After weighted fusion (0.6×56.3+0.4×48.5), the final offset amplitude value of 53.2nm was obtained. Finally, the corrected offset direction of 72.1° was converted into a unit vector (0.305, 0.952, 0), which was multiplied by the amplitude value of 53.2nm to construct the pin tip offset vector V = (53.2×0.305nm, 53.2×0.952nm, 0) = (16.2nm, 50.7nm, 0). This offset vector represents the actual offset state of the pin tip.

[0048] In one embodiment of this application, step S130, determining the axis curvature distribution based on the geometric feature vector of the pin axis, further includes the following steps: Figure 4 As shown, the specific content is as follows:

[0049] Step S410: Divide the pin axis into multiple segments of the same length, and determine the three-dimensional coordinate values ​​of both ends of each segment based on the geometric feature vector of the pin axis.

[0050] Step S420: Determine the local curvature value of each shaft segment sample based on the three-dimensional coordinate values ​​at both ends of each shaft segment sample;

[0051] Step S430: Determine the axis curvature distribution based on the local curvature values ​​of each axis segment sample.

[0052] Specifically, the continuous pin axis is discretized, and standardized pin segment samples are obtained through uniform segmentation. Then, precise three-dimensional coordinates at both ends of each pin segment are extracted from the geometric feature vector of the pin axis. Based on these three-dimensional coordinates, the curvature of each pin segment is quantified through geometric calculations to obtain local curvature values, thus accurately reflecting the curvature state at different locations on the axis. By integrating the local curvature values ​​of all pin segment samples, curvature distribution data that comprehensively characterizes the overall curvature state of the pin axis is formed, providing an intuitive and quantitative basis for subsequent analysis of the correlation between the axis's curvature trend and tip offset.

[0053] For example, when inspecting the pins of an MPO connector, the geometric feature vector of the pin axis is first analyzed, and the three-dimensional coordinates of the starting point of the axis are extracted as (50.0μm, 30.0μm, 5.0μm) and the three-dimensional coordinates of the ending point are (150.0μm, 30.0μm, 5.0μm). The total length of the axis is calculated as L = 100.0μm. According to the detection accuracy requirements, the axis is divided into 10 axis segment samples (N=10), and the length of each axis segment sample is l = 100.0μm / 10 = 10.0μm. They are numbered from 1 to 10 according to the axis direction. By analyzing the three-dimensional point cloud data in the geometric feature vector of the axis and combining it with a linear interpolation algorithm, the three-dimensional coordinates of the two ends of each axis segment sample are determined. For example, the coordinates of the two ends of axis segment sample No. 1 are (50.0μm, 30.0μm, 5.0μm) and (60.0μm, 30.1μm, 5.0μm), and the coordinates of the two ends of axis segment sample No. 2 are (60.0μm, 30.1μm, 5.0μm) and (70.0μm, 30.3μm, 5.0μm). The coordinates of the remaining axis segment samples are extracted sequentially according to the same rule, all accurate to 0.1μm.

[0054] Next, the local curvature value of each shaft segment sample is calculated. Taking shaft segment sample No. 3 as an example, its coordinates at both ends are (70.0μm, 30.3μm, 5.0μm) and (80.0μm, 30.6μm, 5.0μm). First, the chord length S3 = 10.0045μm is calculated. Then, the vertical distance d3 = 0.15μm from the midpoint of the shaft segment (75.0μm, 30.45μm, 5.0μm) to the axis fitting baseline is calculated. Then, the local curvature radius R3 = 83.4μm, and the local curvature value K3 = 1 / 83.4 = 0.012μm⁻¹ is obtained.

[0055] Following the same method described above, the local curvature values ​​of 10 segments in the pin axis were calculated sequentially as follows: [0.01, 0.011, 0.012, 0.015, 0.018, 0.016, 0.013, 0.011, 0.01, 0.009] μm⁻¹. Finally, these local curvature values ​​were arranged in segment number order, and a curvature distribution curve was constructed with the segment number as the x-axis and the local curvature value as the y-axis. Simultaneously, a distribution array containing the coordinates of each segment and its corresponding curvature value was generated. This curvature distribution accurately reflects the actual bending trend of the pin axis. For example, the middle region (segments numbered 4-6) may have the greatest curvature, gradually flattening out towards both ends.

[0056] In one embodiment of this application, step S140, which determines the correlation matrix between axial curvature and tip offset based on the pin tip offset vector and the axial curvature distribution, further includes the following steps: Figure 5 As shown, the specific content is as follows:

[0057] Step S510: Using a rotation matrix algorithm, the pin tip offset vector is mapped to the local coordinate system of the pin axis to determine the local tip offset vector;

[0058] Step S520: Using the Pearson correlation coefficient algorithm, calculate the correlation coefficient between each component of the local tip offset vector and the curvature of each axis segment sample of the axis curvature distribution;

[0059] Step S530: Determine the correlation matrix between axis bending and tip offset based on the correlation coefficient.

[0060] Specifically, to eliminate the deviation between the global coordinate system and the spatial orientation of the needle axis, the global needle tip offset vector is precisely mapped to a local coordinate system that fits the needle axis through coordinate system transformation, ensuring that the direction of the offset vector and the bending trend of the axis are in the same reference system. Next, the linear correlation between each component of the local tip offset vector and the curvature of each axis segment sample is quantified. The correlation coefficient visually presents the strength of the correlation between offset and bending at different locations, providing a quantitative basis for constructing the correlation matrix. Finally, the dispersed correlation coefficients are integrated into a structured correlation matrix according to a fixed dimension, making the correlation between axis bending (curvature of each axis segment) and tip offset (local offset vector components) visible and standardized, providing direct data support for subsequent consistency judgment of bending direction and offset direction.

[0061] For example, in the detection of pins of a certain MPO connector, the pin tip offset vector V in the global coordinate system has been obtained. global =(20.5nm, 15.3nm, 3.2nm), the local coordinate system of the pin axis is constructed based on the direction vector of the axis fitting baseline, and the rotation matrix R is known (the axis makes an angle of 45° with the global X-axis and 30° with the Y-axis). Then the rotation matrix R is:

[0062]

[0063] V global After transposing and multiplying by R, we obtain the local tip offset vector, whose components in the x, y, and z directions are V, respectively. x_local V y_local V z_local Given that the pin axis is uniformly divided into 8 equal segments along the central direction, the curvature distribution of each segment is [0.012μm⁻¹, 0.015μm⁻¹, 0.018μm⁻¹, 0.021μm⁻¹, 0.024μm⁻¹, 0.020μm⁻¹, 0.016μm⁻¹, 0.013μm⁻¹]. After standardizing the offset components and curvature values, the Pearson correlation coefficient is calculated. For example, the component V of the pin tip offset vector in the x-direction. x_local The correlation coefficients with the curvature of the eight axis segments are [0.65, 0.72, 0.78, 0.83, 0.87, 0.81, 0.74, 0.68], respectively. The component V of the pin tip offset vector in the y-direction... y_local The corresponding correlation coefficients are [0.23, 0.28, 0.32, 0.37, 0.41, 0.36, 0.30, 0.25], and the component V of the pin tip offset vector in the z-direction is... z_localThe corresponding correlation coefficients are [-0.18, -0.14, -0.10, -0.06, -0.03, -0.07, -0.11, -0.15]. Based on these correlation coefficients, a 3×8 correlation matrix M between axis curvature and tip offset is constructed as follows:

[0064]

[0065] In one embodiment of this application, in step S150, it is determined whether the correlation coefficient in the correlation matrix exceeds a preset correlation coefficient threshold. If it is determined that the correlation coefficient exceeds the preset correlation coefficient threshold, it is determined whether the pin tip offset vector and the axis curvature distribution have directional consistency. Specifically, combinations with strong linear correlation between pin tip offset and axis curvature can be screened first, and then the directional matching between the two can be further verified to accurately determine the core cause of the tip offset (whether it is caused by the natural curvature of the axis). The preset correlation coefficient threshold can be determined based on the detection accuracy requirements of MPO connector pins, industry standards, and a large amount of experimental data statistics. Since the Pearson correlation coefficient ranges from [-1, 1], combined with the physical correlation characteristics between pin axis curvature and tip offset, the preset threshold is set to 0.7 (absolute value), that is, when the absolute value of the correlation coefficient is ≥0.7, it is determined that there is a strong linear correlation between the two. Each element (i.e., the correlation coefficient between each offset component and each axis curvature) in the correlation matrix between axis curvature and tip offset (dimension is 3×N, where 3 is the X, Y, Z components of the local tip offset vector, and N is the number of axis segment samples) is iterated one by one. The absolute value of each element is compared with a preset correlation coefficient threshold (0.7), and the position information (including the corresponding offset component and axis segment sample number) of all elements with an absolute correlation coefficient ≥ 0.7 is recorded to form a strongly correlated combination set. From the strongly correlated combination set, the core data corresponding to each combination is extracted: one is the local tip offset vector component corresponding to the combination (such as the X-direction offset component V). x_local The second aspect is the curvature direction data of the shaft segment sample corresponding to the combination (the bending direction vector calculated from the three-dimensional coordinates of both ends of the shaft segment, with a dimension of 3×1, representing the spatial direction of the shaft segment's curvature). A vector cosine similarity algorithm is used to calculate the directional consistency between the offset component and the bending direction vector. Considering the actual needs of pin detection, the preset threshold for directional consistency is set to 0.8 (cosθ≥0.8 is considered directionally consistent). The calculated cosine similarity cosθ is compared with the preset threshold for directional consistency. If cosθ≥0.8, it is determined that the offset vector component of the pin tip corresponding to this strongly correlated combination has directional consistency with the axis curvature distribution, meaning the tip offset is caused by the natural curvature of the shaft segment.

[0066] In one embodiment of this application, in step S160, if it is determined that there is directional consistency, the current pin offset factor is determined to be the natural bending of the pin, and a pin depth point cloud is obtained. The source of pin depth error is determined based on the pin depth point cloud. The source of pin depth error includes ambient light interference error and measurement equipment error. The method also includes the following steps: Figure 6 As shown, the specific content is as follows:

[0067] Step S610: Divide the pin depth point cloud into multiple local point cloud blocks, and use principal component analysis algorithm to determine the normal vector of each local point cloud block;

[0068] Step S620: Determine whether the angle between the normal vectors of adjacent local point cloud blocks exceeds a preset angle threshold. If it is determined that the angle exceeds the preset angle threshold, then the current pin depth error source is determined to be ambient light interference error.

[0069] Specifically, the pin depth point cloud is uniformly divided into multiple local point cloud blocks along three-dimensional space. Each local point cloud block is 5μm × 5μm × 5μm in size, and all local point cloud blocks are numbered sequentially according to their spatial location to ensure coverage of the entire pin depth region without overlap. The PCA algorithm is executed individually for each local point cloud block. The mean vector of all points within the block is calculated to eliminate the effects of translation. Then, the covariance matrix of the point cloud is constructed, and the three eigenvalues ​​(λ1≥λ2≥λ3) and corresponding eigenvectors (v1, v2, v3) of the covariance matrix are solved through eigenvalue decomposition. The magnitude of the eigenvalues ​​reflects the dispersion of the point cloud distribution along the corresponding eigenvector direction. The eigenvector v3 corresponding to the smallest eigenvalue λ3 is the normal vector of the local point cloud block, as its direction is perpendicular to the local surface of the block, representing the spatial orientation of the local surface. For all adjacent local point cloud blocks (including adjacent blocks in the X, Y, and Z directions), the angle between their normal vectors is calculated. The angle calculation uses the vector dot product formula. The calculated angle between adjacent normal vectors is compared one by one with the preset angle threshold (15°). If the angle between the normal vectors of a group of adjacent local point cloud blocks is ≥15°, the source of the pin depth error in the region is determined to be ambient light interference error, and the spatial location information of the region (corresponding to the local point cloud block number) is recorded. If the angle is <15°, the depth point cloud in the region is determined to be continuous, and it is not considered as ambient light interference error for the time being.

[0070] For example, in the detection of pins of an MPO connector, a pin depth point cloud with a depth resolution of 0.8 μm has been acquired. After preprocessing to remove outliers, it is divided into 100 local point cloud blocks (numbered 1-100) using a uniform grid partitioning method. Each local point cloud block has a size of 5 μm × 5 μm × 5 μm. The PCA algorithm is performed on the local point cloud block numbered 25: after calculating the mean vector of the 89 points within the block, a covariance matrix is ​​constructed and eigenvalue decomposition is performed, yielding eigenvalues ​​λ1 = 0.085, λ2 = 0.023, and λ3 = 0.004, with corresponding eigenvectors v1 = (0.92, 0.35, 0.18), v2 = (0.31, -0.89, 0.32), and v3 = (0.25, 0.28, -0.93). The v3 corresponding to the smallest eigenvalue λ3 is taken as the normal vector of this point cloud block, and after standardization, it becomes n. 25 =(0.25,0.28,-0.93). For the local point cloud patch numbered 26, which is adjacent to number 25, the normalized normal vector n is also obtained using the PCA algorithm. 26 =(0.87,-0.42,0.24). The angle between the two is calculated to be 97.1°. Comparing this angle (97.1°) with the preset angle threshold (15°), 97.1°≥15°, therefore, the depth region corresponding to local point cloud blocks numbered 25 and 26 is determined to be caused by ambient light interference error. Continuing to traverse other adjacent local point cloud blocks, it is found that the angles between the normal vectors of blocks numbered 43 and 44, and 67 and 68 are 23.5° and 18.7° respectively, both exceeding the preset angle threshold of 15°, and therefore are all determined to be areas of ambient light interference error.

[0071] In the above method, the overall point cloud is decomposed into local blocks using a uniform grid partitioning method. Principal component analysis (PCA) is then used to accurately extract the normal vector of each block. The normal vector effectively characterizes the spatial orientation of the local point cloud, providing a quantitative geometric feature basis for error judgment. By comparing the angle between adjacent normal vectors with a preset threshold, regions of depth point cloud "jumps" caused by ambient light interference can be intuitively identified, accurately distinguishing between ambient light interference and normal geometric changes, thus avoiding misjudgment. This provides a clear error area and cause basis for subsequent depth value correction, effectively reducing the impact of ambient light interference on pin spatial attitude reconstruction, and further improving the reliability and accuracy of MPO connector pin intelligent detection.

[0072] In one embodiment of this application, after determining that the source of the current pin depth error is ambient light interference error, step S160 further includes the following steps, such as... Figure 7 As shown, the specific content is as follows:

[0073] Step S710: Mark local point cloud blocks that exceed the preset angle threshold as abnormal point cloud blocks, and mark local point cloud blocks that do not exceed the preset angle threshold as normal point cloud blocks;

[0074] Step S720: Obtain the depth value of the abnormal point cloud block and the depth value of the normal point cloud block adjacent to the abnormal point cloud block respectively;

[0075] Step S730: Subtract the depth value of the abnormal point cloud block from the depth value of the normal point cloud block adjacent to the abnormal point cloud block, and use the result of the subtraction operation as the depth correction value for the corresponding ambient light interference error.

[0076] Specifically, for example, in the depth detection of a certain MPO connector pin, a preset threshold of the angle between the normal vectors of adjacent local point cloud blocks is set to 15°. After traversing all 5μm×5μm×5μm local point cloud blocks, local point cloud blocks numbered 25 and 26 are marked as abnormal point cloud blocks (their angles with the normal vectors of adjacent blocks are 97.1° and 89.3°, respectively, both exceeding the preset angle threshold). Adjacent local point cloud blocks numbered 24, 27, 35, and 36 are marked as normal point cloud blocks (all adjacent angles are less than 15°). Subsequently, the average depth of abnormal point cloud block 25 is obtained as 2.51μm, and the average depth of abnormal point cloud block 26 is obtained as 2.53μm. At the same time, the average depths of the corresponding normal point cloud blocks 24, 35, and 44 are extracted as 2.32μm, 2.33μm, and 2.31μm, respectively. The average depths of the normal point cloud blocks 27, 36, and 45 corresponding to point cloud block 26 are 2.35μm, 2.34μm, and 2.33μm, respectively. Next, the reference depth value of the abnormal point cloud block 25 is calculated as (2.32+2.33+2.31) / 3=2.32μm, and the depth correction value is 2.32-2.51=-0.19μm. The reference depth value of the abnormal point cloud block 26 is (2.35+2.34+2.33) / 3=2.34μm, and the depth correction value is 2.34-2.53=-0.19μm. After subtracting 0.19μm from the depth values ​​of the two abnormal point cloud blocks, their depth values ​​become 2.32μm and 2.34μm, respectively. The depth values ​​are continuous and smooth with the depth values ​​of the adjacent normal point cloud blocks. The angle between the recalculated adjacent normal vectors is less than 15°, which is consistent with the true depth trend of the pin.

[0077] In one embodiment of this application, step S160, determining the source of pin depth error based on the pin depth point cloud, further includes the following steps: Figure 8 As shown, the specific content is as follows:

[0078] Step S810: Based on the pin depth point cloud, the optimal planar model of each local point cloud block is selected using the random sample consensus algorithm;

[0079] Step S820: Determine whether the external point ratio of the optimal planar model exceeds the preset ratio threshold. If it is determined that it exceeds the preset ratio threshold, then the source of the current pin depth error is determined to be the measurement equipment error.

[0080] Specifically, based on the depth resolution (0.8μm) and detection accuracy requirements of the pin depth point cloud, the core parameters of the RANSAC algorithm can be set as follows: the number of random samplings is set to 1000 times to ensure that enough sample combinations are traversed to find the optimal model; the distance threshold is set to 1μm, that is, points in the point cloud data with a distance ≤1μm from the fitting plane are judged as inliers (points that conform to the planar model), and points with a distance >1μm are judged as outliers (points that deviate from the planar model); the inlier ratio threshold is set to 60%, that is, a model is considered valid only when the proportion of inliers in the fitted model is ≥60%. For each local point cloud block, the RANSAC algorithm is executed: three non-collinear points are randomly selected from the point cloud dataset to fit an initial planar model; then, the distances from all other points in the block to the initial planar model are calculated, and the number and proportion of interior points are counted; the above sampling, fitting, and statistical process is repeated until the set number of sampling times ends; finally, from all valid models (interior point proportion ≥ 60%), the planar model with the most interior points and the highest interior point proportion is selected as the optimal planar model for the local point cloud block. If all models do not meet the interior point proportion threshold, the block is marked as an abnormal point cloud block. The calculated exterior point proportion is compared with the preset proportion threshold (15%) one by one: if the exterior point proportion of a local point cloud block is ≥ 15%, the pin depth error in this area is determined to be due to measurement equipment error, and the spatial location information of this area (corresponding to the local point cloud block number) is recorded; if the exterior point proportion is < 15%, the point cloud dispersion in this area is determined to be normal, with no obvious measurement equipment error.

[0081] For example, in the detection of pins of an MPO connector, local point cloud block number 38 includes 90 three-dimensional coordinate points. Based on the RANSAC algorithm, an optimal planar model was fitted: the sampling count was set to 1000 times, the distance threshold to 1 μm, and the interior point ratio threshold to 60%. After iterative fitting, the planar model with the most interior points (70 interior points and 20 exterior points) was selected. The exterior point ratio of this optimal planar model was calculated as: exterior point ratio = 20 / 90 × 100% ≈ 22.2%. Comparing this exterior point ratio (22.2%) with the preset ratio threshold (15%), 22.2% ≥ 15%. Therefore, the depth error of the pin depth region corresponding to local point cloud block number 38 is determined to be due to measurement equipment error. Continuing to traverse other local point cloud blocks, it was found that the external point ratios of the optimal planar model of local point cloud blocks numbered 12, 55, and 89 were 17.5%, 19.3%, and 16.8%, respectively, all exceeding 15%, and were all determined to be in the measurement equipment error area; the external point ratios of the remaining local point cloud blocks were all within 15%, and were not considered as measurement equipment errors for the time being.

[0082] In the above method, by employing the random sample consensus algorithm, leveraging its noise and outlier resistance capabilities, it can accurately fit the optimal planar model representing the true morphology of the pin surface from point cloud data containing interference. This avoids the influence of noise on model fitting and provides a reliable quantitative benchmark for error determination. By comparing the proportion of outliers with a preset threshold, the dispersion of the point cloud data can be objectively quantified, accurately distinguishing between dispersion caused by measurement equipment errors (outlier proportion exceeding the standard) and slight dispersion caused by normal geometric changes, thus avoiding misjudgment of the error source. In this way, precise location of measurement equipment errors can be achieved without additional auxiliary equipment, clarifying the error area and cause, providing a targeted basis for subsequent depth value correction, effectively reducing the interference of measurement equipment errors on pin spatial attitude reconstruction, and further improving the accuracy and reliability of MPO connector pin intelligent detection.

[0083] In one embodiment of this application, step S170, determining the depth correction value based on the corresponding pin depth error source, further includes the following steps: Figure 9 As shown, the specific content is as follows:

[0084] Step S910: When the source of the corresponding pin depth error is the measurement equipment error, obtain the depth measurement sequence of the measurement equipment, and fit a linear model using the least squares method based on the depth measurement sequence;

[0085] Step S920: Determine the depth correction value of the corresponding measuring equipment error based on the linear model.

[0086] Specifically, for the local depth data area of ​​the pins that has been determined to be due to measurement equipment error, the measurement equipment (such as a structured light depth camera) is controlled to perform continuous and repeated depth measurements on this area. The number of measurements N is set to 10-20 times, and the equipment parameters, shooting angle, and lighting conditions are kept consistent for each measurement to avoid additional interference. The depth values ​​obtained from the N measurements are arranged in the measurement order to form a depth measurement sequence D=[d1,d2,...,d N ], where d i Let be the depth value of the i-th measurement, in micrometers (μm), with an accuracy consistent with the device's depth resolution (e.g., 0.8 μm). Assume the systematic error of the measuring device exhibits a linear trend, and use the least squares method to solve for the linear model parameters to construct the final linear model. For each measurement value in the depth measurement sequence D (corresponding to a measurement number), substitute the measurement value into the linear model to calculate the model prediction value corresponding to that measurement number. This prediction value represents the ideal pin depth at that measurement number after eliminating systematic errors. Then, use the ideal value predicted by the model to offset the systematic errors in the original measurement value; the depth correction value is the model prediction value minus the actual depth measurement value.

[0087] For example, the depth measurement sequence D = [2350μm, 2370μm, 2360μm, 2390μm, 2380μm, 2400μm, 2410μm, 2390μm, 2420μm, 2430μm]. The linear model is y = 8x + 2346. Calculate the model prediction value for each measurement number in the depth measurement sequence D, for example, the model prediction value y when x = 1. 1p ᵣ e d = 8 × 1 + 2346 = 2354 μm. When x = 2, y 2p ᵣ e d = 8 × 2 + 2346 = 2362 μm, and so on, to obtain 10 model prediction values. Finally, the depth correction values ​​are calculated. For example, when x = 1, the depth correction value Δd1 = 2354 - 2350 = 4 μm; when x = 2, the depth correction value Δd2 = 2362 - 2370 = -8 μm, until all depth correction values ​​are obtained.

[0088] In the above method, the depth correction value calculated based on a linear model can specifically offset the systematic errors of the measuring equipment. A unified correction value balances batch testing efficiency and correction accuracy, while independent correction values ​​meet the requirements of high-precision testing. Accurate optimization of depth data can be achieved without relying on additional calibration equipment. This significantly reduces the mean square error of the corrected pin depth values, effectively improving the accuracy of the depth data. This provides reliable data support for subsequent pin spatial attitude reconstruction, contact deviation simulation analysis, and quality assessment, further enhancing the accuracy and stability of MPO connector pin intelligent detection.

[0089] In one embodiment of this application, after determining the depth correction value, step S170 further includes the following steps: Figure 10 As shown, the specific content is as follows:

[0090] Step S1100: Determine the pin depth deviation and pin posture deviation based on the depth correction value;

[0091] Step S1200: Determine the pin signal transmission loss based on the pin depth deviation and pin orientation deviation;

[0092] Step S1300: Determine whether the pin signal transmission loss exceeds the preset loss threshold. If it is determined that the pin exceeds the preset loss threshold, mark the current pin as a potential contact defect and determine the pin quality score based on the pin depth deviation, pin posture deviation and pin signal transmission loss.

[0093] Step S1400: Determine whether the quality score exceeds the preset score threshold. If the score does not exceed the preset score threshold, mark the current pin as unqualified.

[0094] Specifically, the standard depth value of an MPO connector pin is the ideal depth when the pin has no deviation and no error. The actual pin depth value after depth correction is obtained; the pin depth deviation is the difference between the adjusted actual depth value and the actual pin depth value. The unit is uniformly micrometers (μm), accurate to 0.01μm. This value reflects the degree of deviation of the depth direction from the standard state. Pin attitude deviation is mainly characterized by tilt angle deviation. The standard attitude parameter is the ideal tilt angle between the pin axis and the reference plane. The actual tilt angle in the corrected pin attitude distribution data is obtained; the pin attitude deviation is the difference between the actual tilt angle in the corrected pin attitude distribution data and the ideal tilt angle between the pin axis and the reference plane. Pin attitude deviation reflects the degree of deviation of the pin's spatial attitude from the standard state.

[0095] Specifically, the weights corresponding to pin depth deviation and pin posture deviation can be obtained separately. Based on these weights, a weighted calculation is performed on the pin depth deviation and pin posture deviation, and the result is used as the pin signal transmission loss. For example, if the pin depth deviation is 0.05 mm and the pin posture deviation is 0.1 degrees, with a weight of 0.5 for the depth deviation and 0.8 for the posture deviation, then the signal transmission loss is Loss = 0.5 * 0.05 + 0.8 * 0.1 = 0.105 dB. If the preset loss threshold is 0.1 dB, then this value exceeds the preset loss threshold, indicating a risk to signal transmission stability, and the current pin is marked as potentially having poor contact.

[0096] Furthermore, the pin depth deviation and pin orientation deviation can be squared separately, summed, and then the square root can be taken. The result is multiplied by its corresponding weighting factor to obtain the pin offset quality scoring factor. The signal transmission loss is multiplied by its corresponding weighting factor to obtain the pin signal loss quality scoring factor. Therefore, the pin quality score equals the pin offset quality scoring factor minus the pin signal loss quality scoring factor. For example, if the signal transmission loss is 0.105 and the corresponding weighting factor is 50, then the pin signal loss quality scoring factor = 0.105 * 50 = 5.25. If the pin depth deviation is 0.05 and the pin orientation deviation is 0.1, the pin offset quality scoring factor = ... Therefore, the pin quality score = 100 - 5.25 - 3.36 = 91.39. If the preset scoring threshold is set to 95, and 91.39 < 95, then the current pin is marked as unqualified. For example, the evaluation results can be linked with production batch data. If the defect rate within a batch exceeds 5%, a production line parameter adjustment command is triggered, such as adjusting the pin crimping pressure from 10N to 12N to optimize the quality of subsequent produced pins.

[0097] This application also provides an intelligent detection system for MPO connector pins, which may include an acquisition module, a first determination module, a second determination module, a third determination module, a judgment module, a fourth determination module, and a fifth determination module. The system comprises the following modules: an acquisition module for acquiring pin image data of the MPO connector and determining the geometric feature vectors of the pin tip and the pin axis based on the pin image data; a first determination module for determining the pin tip offset vector based on the pin tip geometric features, the pin tip offset vector including the offset direction and the offset magnitude; a second determination module for determining the axis curvature distribution based on the pin axis geometric feature vector; a third determination module for determining the correlation matrix between the axis curvature and the tip offset based on the pin tip offset vector and the axis curvature distribution set; a judgment module for judging whether the correlation coefficient in the correlation matrix exceeds a preset correlation coefficient threshold, and if it exceeds the preset correlation coefficient threshold, judging whether the pin tip offset vector and the axis curvature distribution have directional consistency; a fourth determination module for determining the current pin offset factor as the natural bending of the pin if it is judged to have directional consistency, acquiring the pin depth point cloud, and determining the source of pin depth error based on the pin depth point cloud, the source of pin depth error including ambient light interference error and measurement equipment error; and a fifth determination module for determining the depth correction value based on the corresponding source of pin depth error.

[0098] It should be noted that the embodiments of the intelligent detection system for MPO connector pins provided in this application can be used to execute the processing flow of the embodiments of the intelligent detection method for MPO connector pins in the above embodiments. Its functions will not be repeated here, but can be referred to the detailed description of the above method embodiments.

[0099] This application also provides an electronic device, which includes one or more processors and memory resources represented by memory for storing instructions executable by the processor, such as application programs. The application programs stored in the memory may include one or more modules, each corresponding to a set of instructions. Furthermore, the processor is configured to execute instructions to perform the aforementioned intelligent detection method for MPO connector pins.

[0100] The electronic device may also include a power supply component configured to perform power management of the electronic device, a wired or wireless network interface configured to connect the electronic device to a network, and an input / output (I / O) interface. The electronic device can be operated based on operating devices stored in memory, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or similar.

[0101] In one embodiment, a computer device, which may be a server, is also provided. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an ambient light for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores data. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an intelligent detection method for MPO connector pins.

[0102] In one embodiment, a computer device is provided, which may be a terminal. The computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides ambient light for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements an intelligent detection method for MPO connector pins. The display unit of the computer device is used to form a visually visible image and may be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0103] This application also provides a non-transitory computer-readable storage medium, which, when the instructions in the storage medium are executed by the processor of the electronic device, enables the electronic device to perform an intelligent detection method for MPO connector pins.

[0104] This application may take the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-readable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to: phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0105] It should be noted that although the steps of the intelligent detection method for MPO connector pins in this application are described in a specific order in the accompanying drawings, this does not require or imply that these steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps, such as omitting certain steps, combining multiple steps into one step, and / or breaking down one step into multiple steps, should all be considered part of this application.

[0106] It should be understood that this application is not limited to the detailed structure and arrangement of the modules of the intelligent detection system for MPO connector pins proposed in this specification. This application can have other embodiments and can be implemented and executed in various ways. The foregoing variations and modifications fall within the scope of this application. It should be understood that the application and its definition in this specification extend to all alternative combinations of two or more individual features mentioned or apparent in the text and / or drawings. All these different combinations constitute multiple alternative aspects of this application. The embodiments described in this specification illustrate the best known mode for implementing this application and will enable those skilled in the art to utilize this application.

Claims

1. A method for intelligent detection of MPO connector pins, characterized in that, include: Acquire pin image data of the MPO connector, and determine the geometric feature vector of the pin tip and the geometric feature vector of the pin axis based on the pin image data; The pin tip offset vector is determined based on the geometric feature vector of the pin tip, and the pin tip offset vector includes the tip offset direction and the tip offset magnitude value; The axis curvature distribution is determined based on the geometric feature vector of the pin axis. The correlation matrix between axis curvature and tip offset is determined based on the pin tip offset vector and the axis curvature distribution; Determine whether the correlation coefficient in the correlation matrix exceeds a preset correlation coefficient threshold. If it is determined that the correlation coefficient exceeds the preset correlation coefficient threshold, then determine whether the pin tip offset vector and the axis curvature distribution have directional consistency. If the direction consistency is determined, the current pin offset factor is determined to be the natural bending of the pin, and the pin depth point cloud is obtained. The source of pin depth error is determined based on the pin depth point cloud. The source of pin depth error includes ambient light interference error and measurement equipment error. The depth correction value is determined based on the corresponding source of the pin depth error.

2. The intelligent detection method for MPO connector pins according to claim 1, characterized in that, The step of determining the geometric feature vector of the pin tip and the geometric feature vector of the pin axis based on the pin image data includes: The pin image data is subjected to background segmentation processing to determine the pin tip image data and the pin axis image data; The three-dimensional point cloud of the needle tip is determined based on the needle tip image data, and the three-dimensional point cloud of the needle axis is determined based on the needle axis image data. Based on the three-dimensional point cloud of the pin tip and the three-dimensional point cloud of the pin axis, an edge detection algorithm is used to extract features to determine the geometric feature vector of the pin tip and the geometric feature vector of the pin axis.

3. The intelligent detection method for MPO connector pins according to claim 1, characterized in that, The step of determining the pin tip offset vector based on the pin tip geometric feature vector includes: The position of the tip vertex is determined based on the geometric feature vector of the pin tip. The tip offset angle is determined based on the tip vertex position and the preset tip position; The tip offset direction and tip offset magnitude are determined based on the tip offset angle. The pin tip offset vector is determined based on the tip offset direction and the tip offset magnitude.

4. The intelligent detection method for MPO connector pins according to claim 1, characterized in that, Determining the axis curvature distribution based on the geometric feature vector of the pin axis includes: The pin axis is evenly divided into multiple segments of the same length, and the three-dimensional coordinate values ​​of both ends of each segment are determined according to the geometric feature vector of the pin axis. The local curvature value of each shaft segment sample is determined based on the three-dimensional coordinate values ​​at both ends of each shaft segment sample; The curvature distribution of the axis is determined based on the local curvature values ​​of each axis segment sample.

5. The intelligent detection method for MPO connector pins according to claim 4, characterized in that, The step of determining the correlation matrix between axis curvature and tip offset based on the pin tip offset vector and the axis curvature distribution includes: A rotation matrix algorithm is used to map the pin tip offset vector to the local coordinate system of the pin axis to determine the local tip offset vector; The Pearson correlation coefficient algorithm is used to calculate the correlation coefficient between each component of the local tip offset vector and the curvature of each of the axis segment samples of the axis curvature distribution; The correlation matrix between axis curvature and tip offset is determined based on the correlation coefficient.

6. The intelligent detection method for MPO connector pins according to claim 1, characterized in that, The sources of pin depth error are determined based on the pin depth point cloud, including: The pin depth point cloud is divided into multiple local point cloud blocks, and the normal vector of each local point cloud block is determined by principal component analysis algorithm. Determine whether the angle between the normal vectors of adjacent local point cloud blocks exceeds a preset angle threshold. If it is determined that the angle exceeds the preset angle threshold, then the current pin depth error source is determined to be ambient light interference error. After determining that the current pin depth error source is ambient light interference error, the method further includes: Local point cloud blocks that exceed the preset angle threshold are marked as abnormal point cloud blocks, and local point cloud blocks that do not exceed the preset angle threshold are marked as normal point cloud blocks; The depth values ​​of the abnormal point cloud block and the normal point cloud block adjacent to the abnormal point cloud block are obtained respectively. The depth value of the abnormal point cloud block is subtracted from the depth value of the normal point cloud block adjacent to the abnormal point cloud block, and the result of the subtraction is used as the depth correction value corresponding to the ambient light interference error.

7. The intelligent detection method for MPO connector pins according to claim 1, characterized in that, The sources of pin depth error are determined based on the pin depth point cloud, including: Based on the pin depth point cloud, the optimal planar model for each local point cloud block is selected using a random sample consensus algorithm. Determine whether the external point ratio of the optimal planar model exceeds a preset ratio threshold. If it is determined that it exceeds the preset ratio threshold, then the current pin depth error source is determined to be a measurement device error.

8. The intelligent detection method for MPO connector pins according to claim 7, characterized in that, After determining that the current pin depth error source is a measuring device error, the method further includes: When the source of the pin depth error is the measurement equipment error, the depth measurement sequence of the measurement equipment is obtained, and a linear model is fitted using the least squares method based on the depth measurement sequence. The depth correction value corresponding to the error of the measuring device is determined based on the linear model.

9. The intelligent detection method for MPO connector pins according to claim 1, characterized in that, After determining the depth correction value, the method further includes: The pin depth deviation and pin posture deviation are determined based on the depth correction value; The pin signal transmission loss is determined based on the pin depth deviation and the pin orientation deviation. Determine whether the pin signal transmission loss exceeds a preset loss threshold. If it exceeds the preset loss threshold, mark the current pin as a potential contact defect and determine the pin quality score based on the pin depth deviation, the pin posture deviation, and the pin signal transmission loss. Determine whether the quality score exceeds a preset score threshold. If it does not exceed the preset score threshold, mark the current pin as unqualified.

10. An intelligent detection system for MPO connector pins, characterized in that, include: The acquisition module is used to acquire pin image data of the MPO connector and determine the geometric feature vector of the pin tip and the geometric feature vector of the pin axis based on the pin image data. The first determining module is used to determine the pin tip offset vector based on the geometric feature vector of the pin tip, wherein the pin tip offset vector includes the tip offset direction and the tip offset magnitude value. The second determining module is used to determine the axis curvature distribution based on the geometric feature vector of the pin axis; The third determining module is used to determine the correlation matrix between axis curvature and tip offset based on the pin tip offset vector and the axis curvature distribution set; The judgment module is used to determine whether the correlation coefficient in the correlation matrix exceeds a preset correlation coefficient threshold. If it is determined that the correlation coefficient exceeds the preset correlation coefficient threshold, it is then determined whether the pin tip offset vector and the axis curvature distribution have directional consistency. The fourth determining module is used to determine the current pin offset factor as the natural bending of the pin if the direction consistency is determined, and to obtain the pin depth point cloud. The source of the pin depth error is determined based on the pin depth point cloud. The source of the pin depth error includes ambient light interference error and measurement equipment error. The fifth determining module is used to determine the depth correction value based on the corresponding source of the pin depth error.

Images