Tail fiber surface coating quality visual detection method

Abstract

The invention discloses a tail fiber surface coating quality visual detection method, which comprises the following steps of: acquiring a tail fiber surface reflection image through a multispectral imaging device, recording an identification code and detection position information, and pre-processing to generate a standard detection image set; and extracting coating layer features including textures, defect forms and reflection spectrum intensity by using a deep convolutional neural network, matching the coating layer features with a standard feature library to judge the quality of the coating layer, and finally outputting a detection report and marking unqualified tail fibers. According to the method, the defect contrast ratio is enhanced by utilizing multispectral imaging, multi-dimensional feature extraction and accurate detection are realized by combining the deep convolutional neural network, and the tail fiber coating quality detection efficiency and accuracy can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition detection, and in particular to a method for visually detecting the coating quality of a pigtail surface. Background Art

[0002] In the field of fiber optic communications, pigtails are important components for connecting optical devices, and the quality of their surface coating directly affects the mechanical properties, transmission stability, and service life of the optical fiber. Traditional pigtail surface coating quality inspections mostly rely on manual visual inspections or simple optical microscope observations. This is not only inefficient, but also greatly affected by human factors, making it difficult to detect subtle defects at the micron level, such as tiny bubbles and shallow scratches. With the rapid development of fiber optic communication technology, the quality requirements for pigtails are increasing. There is an urgent need for an efficient, accurate, and automated method for pigtail surface coating quality inspection to meet the needs of large-scale production and high-quality products. At the same time, existing machine vision inspection methods have problems in pigtail coating quality inspection, such as incomplete extraction of complex defect features and insufficient detection accuracy, and are unable to meet the high-precision requirements of pigtail surface coating quality inspection. Summary of the Invention

[0003] In view of the above problems, the present invention provides a method for visually inspecting the coating quality of a pigtail surface.

[0004] The purpose of the present invention is achieved by adopting the following technical solutions:

[0005] A method for visually inspecting the surface coating quality of a pigtail fiber comprises the following steps:

[0006] S1. Collect the reflection image of the pigtail surface through the multispectral imaging device, and synchronously record and store the identification code and detection position information of the pigtail;

[0007] S2. Preprocessing the reflected image to generate a standard detection image set;

[0008] S3, extracting coating layer features of the standard detection image set through a deep convolutional neural network, wherein the coating layer features include a first feature, a second feature, and a third feature,

[0009] The first feature is a texture distribution parameter extracted by a multi-scale convolution kernel based on the texture uniformity of the coating surface;

[0010] The second feature is a geometric feature based on the defect morphology, which captures the boundary topology of bubbles and scratches through void convolution;

[0011] The third feature is to analyze the consistency of the coating thickness through the residual network based on the reflection spectrum intensity;

[0012] S4. Match the first feature, the second feature, and the third feature with a standard coating layer feature library to determine the coating layer quality:

[0013] Determine whether the coating uniformity is qualified based on the difference between the first feature and the standard texture;

[0014] Determine the defect type and location coordinates based on the matching result between the second feature and the defect template;

[0015] Calculate the thickness variation coefficient based on the light intensity distribution of the third characteristic and determine the thickness tolerance area;

[0016] S5. Output a test report containing identification code, defect type, location coordinates and severity, and trigger the sorting equipment to mark unqualified pigtails.

[0017] As a preferred embodiment, the multispectral imaging device includes a near-infrared light source in an annular array and a high-resolution CCD camera, and the wavelength of the light source is 850nm-1550nm, which is used to enhance the imaging contrast of micron-level defects on the coating surface.

[0018] As a preferred method, the first preprocessing is to screen out valid images containing a complete coating layer area and having a resolution higher than a preset threshold, and use them as a standard detection image set;

[0019] The second preprocessing is to remove image noise through adaptive filtering and enhance the contrast between the coating layer and the background based on histogram stretching;

[0020] The third preprocessing step is to divide the image into multiple sub-regions according to a preset grid and extract the texture gradient map of each sub-region.

[0021] As a preferred embodiment, in the second preprocessing, the adaptive filtering adopts a non-local mean filtering algorithm, and its window size is dynamically adjusted according to the image noise level, and the noise level is calculated by wavelet transform high-frequency component energy.

[0022] As a preferred embodiment, in the third pre-processing, the division spacing of the preset grid is adaptively adjusted according to the diameter of the pigtail to ensure that each sub-area covers at least 3 complete coating layer corrugation periods.

[0023] As a preferred embodiment, the deep convolutional neural network adopts a dual-branch structure, including:

[0024] The first branch: consists of the Inception module, which is used to extract multi-scale texture features;

[0025] The second branch consists of a U-Net encoder and is used for pixel-level segmentation of defect areas.

[0026] The feature maps of the two branches are fused through the channel attention mechanism and then input into the fully connected layer for classification.

[0027] As a preferred method, the thickness variation coefficient is calculated as follows:

[0028] The light intensity mean μ and standard deviation σ of each sub-area are calculated, and the thickness variation coefficient CV is defined as σ / μ×100%. When CV>5%, it is determined that the thickness is abnormal.

[0029] As a preferred approach, the severity of the defects is divided into four levels:

[0030] Minor: Single defect area <0.01mm 2 and does not penetrate the coating;

[0031] Moderate: defect area∈[0.01,0.1)mm 2 or there are 2 cross scratches;

[0032] Serious: defect area ≥ 0.1mm 2 Or it may cause the optical fiber core to be exposed;

[0033] Critical: Thickness variation coefficient CV∈[5%,8%) and no macro defects.

[0034] The beneficial effects of the present invention are:

[0035] The present invention adopts multispectral imaging equipment. Its annular array of near-infrared light sources and high-resolution CCD cameras significantly enhance the imaging contrast of micron-level defects on the coating surface at a wavelength of 850nm-1550nm, and can capture subtle defects that are difficult to detect with traditional methods. Combined with the extraction and analysis of multi-dimensional features by deep convolutional neural networks, the accuracy and efficiency of detection are greatly improved.

[0036] The present invention uses a deep convolutional neural network to extract the first feature (texture uniformity), second feature (defect morphological geometric feature) and third feature (thickness consistency characterized by reflection spectrum intensity) of the coating layer, and evaluates the quality of the coating layer from multiple angles. Compared with single-dimensional detection, it can more comprehensively and accurately judge the quality status of the coating layer and effectively avoid missed detection and false detection.

[0037] The present invention performs multi-stage preprocessing on the collected images, screens valid images, uses adaptive filtering to dynamically remove noise, and enhances contrast through histogram stretching. At the same time, it adaptively adjusts the grid segmentation spacing according to the fiber pigtail diameter to ensure that each sub-area can fully reflect the coating layer characteristics, thereby improving image quality and the reliability of subsequent feature extraction.

[0038] By matching the extracted features with a standard coating feature library, the present invention can automatically determine the coating uniformity, identify the defect type and location coordinates, and identify areas with out-of-tolerance thickness. It can also output a detailed inspection report based on the set four-level defect severity, triggering sorting equipment to mark unqualified pigtails, thereby realizing the automation of detection, judgment, and sorting, reducing labor costs, and improving production efficiency and product quality consistency. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative effort.

[0040] Figure 1 It is a flow chart of Example 1 of the present invention. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0042] The present disclosure provides a method for visually inspecting the coating quality of a pigtail surface. Figure 1 As shown, the following steps are included:

[0043] S1. Collect the reflection image of the pigtail surface through the multispectral imaging device, and synchronously record and store the identification code and detection position information of the pigtail;

[0044] S2. Preprocessing the reflected image to generate a standard detection image set;

[0045] S3, extracting coating layer features of the standard detection image set through a deep convolutional neural network, wherein the coating layer features include a first feature, a second feature, and a third feature,

[0046] The first feature is a texture distribution parameter extracted by a multi-scale convolution kernel based on the texture uniformity of the coating surface;

[0047] The second feature is a geometric feature based on the defect morphology, which captures the boundary topology of bubbles and scratches through void convolution;

[0048] The third feature is to analyze the consistency of the coating thickness through the residual network based on the reflection spectrum intensity;

[0049] S4. Match the first feature, the second feature, and the third feature with a standard coating layer feature library to determine the coating layer quality:

[0050] Determine whether the coating uniformity is qualified based on the difference between the first feature and the standard texture;

[0051] Determine the defect type and location coordinates based on the matching result between the second feature and the defect template;

[0052] Calculate the thickness variation coefficient based on the light intensity distribution of the third characteristic and determine the thickness tolerance area;

[0053] S5. Output a test report containing identification code, defect type, location coordinates and severity, and trigger the sorting equipment to mark unqualified pigtails.

[0054] As a preferred embodiment, the multispectral imaging device includes a near-infrared light source in an annular array and a high-resolution CCD camera, and the wavelength of the light source is 850nm-1550nm, which is used to enhance the imaging contrast of micron-level defects on the coating surface.

[0055] As a preferred method, the first preprocessing is to screen out valid images containing a complete coating layer area and having a resolution higher than a preset threshold, and use them as a standard detection image set;

[0056] The second preprocessing is to remove image noise through adaptive filtering and enhance the contrast between the coating layer and the background based on histogram stretching;

[0057] The third preprocessing step is to divide the image into multiple sub-regions according to a preset grid and extract the texture gradient map of each sub-region.

[0058] As a preferred embodiment, in the second preprocessing, the adaptive filtering adopts a non-local mean filtering algorithm, and its window size is dynamically adjusted according to the image noise level, and the noise level is calculated by wavelet transform high-frequency component energy.

[0059] As a preferred embodiment, in the third pre-processing, the division spacing of the preset grid is adaptively adjusted according to the diameter of the pigtail to ensure that each sub-area covers at least 3 complete coating layer corrugation periods.

[0060] As a preferred embodiment, the deep convolutional neural network adopts a dual-branch structure, including:

[0061] The first branch: consists of the Inception module, which is used to extract multi-scale texture features;

[0062] The second branch consists of a U-Net encoder and is used for pixel-level segmentation of defect areas.

[0063] The feature maps of the two branches are fused through the channel attention mechanism and then input into the fully connected layer for classification.

[0064] As a preferred method, the thickness variation coefficient is calculated as follows:

[0065] The light intensity mean μ and standard deviation σ of each sub-area are calculated, and the thickness variation coefficient CV is defined as σ / μ×100%. When CV>5%, it is determined that the thickness is abnormal.

[0066] As a preferred approach, the severity of the defects is divided into four levels:

[0067] Minor: Single defect area <0.01mm 2 and does not penetrate the coating;

[0068] Moderate: defect area∈[0.01,0.1)mm 2 or there are 2 cross scratches;

[0069] Serious: defect area ≥ 0.1mm 2 Or it may cause the optical fiber core to be exposed;

[0070] Critical: Thickness variation coefficient CV∈[5%,8%) and no macro defects.

[0071] Specifically, in the disclosed embodiment, the multispectral imaging device utilizes a near-infrared light source in a circular array and a high-resolution CCD camera, with the wavelength range of the near-infrared light source set to 850nm-1550nm. This device is installed at the inspection station of the pigtail production line, ensuring that the pigtails can enter the imaging device's field of view completely and clearly during transmission. The imaging device also establishes a communication connection with the data storage system and the detection control system, ensuring that the collected images, pigtail identification codes, and detection position information can be accurately transmitted and stored in real time.

[0072] When the fiber pigtail is transported along the production line to the inspection station, the multispectral imaging equipment begins operating. A high-resolution CCD camera captures the reflected image of the pigtail surface. A barcode scanner or other identification device simultaneously records the pigtail's identification code. The equipment-mounted position sensor or an image positioning algorithm determines and stores the inspection location information. The captured image and recorded information are then synchronously transmitted to a data storage system for subsequent processing and traceability.

[0073] The first preprocessing: The collected reflection images are screened, and a resolution threshold is set according to the integrity of the coating area in the image and the image resolution. Images containing complete coating areas and with resolutions higher than the preset threshold are screened out as valid images, and images that do not meet the requirements are eliminated.

[0074] Second preprocessing: De-noising the valid image using a non-local means filtering algorithm. First, the energy of the image's high-frequency components is calculated using a wavelet transform to assess the image's noise level. The non-local means filtering algorithm's window size is then dynamically adjusted based on the noise level to remove noise from the image. A histogram stretching algorithm is then used to enhance the contrast between the coating and the background, making the coating's details more prominent.

[0075] Third pre-processing: Based on the diameter of the pigtail, the preset grid spacing is adaptively adjusted to ensure that each sub-region covers at least three complete coating corrugation cycles. After the image is divided into multiple sub-regions according to the preset grid, a texture gradient map is extracted for each sub-region to highlight the texture variation characteristics of the coating surface.

[0076] The deep convolutional neural network adopts a two-branch structure. The first branch, consisting of an Inception module, feeds a set of standard inspection images into the first branch. Using multiple convolution kernels of different scales, the Inception module extracts multi-scale texture distribution parameters based on texture uniformity of the coating surface, which is the first feature. The second branch, consisting of a U-Net encoder, feeds the image into the second branch and uses dilated convolution to capture the geometric features of defects such as bubbles and scratches, achieving pixel-level segmentation of defect areas and obtaining the second feature. Simultaneously, a residual network analyzes the image's reflected spectral intensity, calculating the mean and standard deviation of the light intensity in each subregion to obtain a coating thickness consistency feature based on the reflected spectral intensity, which is the third feature. The feature maps extracted by the two branches are fused using a channel attention mechanism and then fed into a fully connected layer for classification.

[0077] The extracted first feature is compared with the relevant parameters of the standard texture in the standard coating feature library, the difference between the two is calculated, and the difference threshold is set. When the difference exceeds the threshold, the coating uniformity is judged to be unqualified; the second feature is matched with the pre-established defect template, the defect type is determined according to the matching result, and the position coordinates of the defect are obtained; according to the light intensity distribution of each sub-area in the third feature, the thickness variation coefficient CV = σ / μ×100% is calculated, and when CV>5%, it is judged to be an abnormal thickness area. At the same time, according to the defect area, whether it penetrates the coating, whether it causes the exposure of the optical fiber core, etc., combined with the set four-level defect severity standard (minor: single defect area <0.01mm 2 And does not penetrate the coating layer; Moderate: defect area∈[0.01,0.1)mm 2 Or there are two cross scratches; severe: defect area ≥ 0.1mm 2 Or it may cause the fiber core to be exposed; critical: thickness variation coefficient CV∈[5%,8%) and no macro defects), determine the severity of the defect.

[0078] Based on the quality assessment results, the inspection and control system generates an inspection report containing the pigtail identification code, defect type, location coordinates, and severity. The report is stored in the data management system for subsequent query and statistical analysis. Simultaneously, instructions are sent to the sorting equipment, triggering it to mark the unqualified pigtails, such as by inkjet or labeling, to clearly mark them for subsequent sorting.

[0079] Through the above specific implementation methods, efficient and accurate detection of the surface coating quality of the pigtail can be achieved, meeting the quality detection requirements in the large-scale production process of the pigtail.

[0080] The disclosed embodiment uses a multispectral imaging device. Its annular array of near-infrared light sources and high-resolution CCD camera significantly enhance the imaging contrast of micron-level defects on the coating surface at a wavelength of 850nm-1550nm, and can capture subtle defects that are difficult to detect with traditional methods. Combined with the extraction and analysis of multi-dimensional features by deep convolutional neural networks, the accuracy and efficiency of detection are greatly improved.

[0081] The disclosed embodiment uses a deep convolutional neural network to extract the first feature (texture uniformity), the second feature (defect morphology geometric feature) and the third feature (thickness consistency characterized by the intensity of the reflected spectrum) of the coating layer, and evaluates the quality of the coating layer from multiple angles. Compared with single-dimensional detection, it can more comprehensively and accurately judge the quality status of the coating layer and effectively avoid missed detection and false detection.

[0082] The disclosed embodiment performs multi-stage preprocessing on the captured images, screens valid images, uses adaptive filtering to dynamically remove noise, and enhances contrast through histogram stretching. At the same time, it adaptively adjusts the grid segmentation spacing according to the fiber pigtail diameter to ensure that each sub-area can fully reflect the coating layer characteristics, thereby improving image quality and the reliability of subsequent feature extraction.

[0083] By matching the extracted features with a standard coating feature library, the disclosed embodiment can automatically determine the coating uniformity, identify the defect type and location coordinates, and identify areas with out-of-tolerance thickness. It can also output a detailed inspection report based on the set four-level defect severity, triggering sorting equipment to mark unqualified pigtails, thereby realizing the automation of inspection, judgment, and sorting, reducing labor costs, and improving production efficiency and product quality consistency.

[0084] The above description and the accompanying drawings fully illustrate the embodiments of the present disclosure so that those skilled in the art can practice them. Other embodiments may include structural, logical, electrical, process and other changes. The embodiments represent only possible variations. Unless explicitly required, individual components and functions are optional, and the order of operations may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the words used in this application are only used to describe the embodiments and are not used to limit the claims. As used in the description of the embodiments and claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to also include plural forms. Similarly, the term "and / or" as used in this application refers to any and all possible combinations of one or more associated listings. In addition, when used in this application, the term "comprise" and its variations "comprises" and / or comprising refer to the presence of stated features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups of these. In the absence of further restrictions, an element defined by the sentence "comprising a..." does not exclude the presence of other identical elements in the process, method or device that includes the element. In this article, each embodiment may focus on the differences from other embodiments, and the same and similar parts between the various embodiments can be referenced to each other. For the methods, products, etc. disclosed in the embodiments, if they correspond to the method part disclosed in the embodiments, then the relevant parts can be found in the description of the method part.

[0085] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these effects are performed in hardware or software may depend on the specific application and design constraints of the technical solution. The technicians may use different methods for each specific application to implement the described effects, but such implementations should not be considered to exceed the scope of the embodiments of the present disclosure. The technicians will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0086] The flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to the embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment or part of the code, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical functions. In some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed in parallel, or they can sometimes be executed in the opposite order, which can depend on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in an order different from that disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed in parallel, or they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by dedicated hardware-based devices that perform the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

Claims

1. A method for visually inspecting the surface coating quality of a pigtail, characterized in that: The steps include: S1. Collect the reflection image of the pigtail surface through the multispectral imaging device, and synchronously record and store the identification code and detection position information of the pigtail; S2. Preprocessing the reflected image to generate a standard detection image set; S3, extracting coating layer features of the standard detection image set through a deep convolutional neural network, wherein the coating layer features include a first feature, a second feature, and a third feature, The first feature is a texture distribution parameter extracted by a multi-scale convolution kernel based on the texture uniformity of the coating surface; The second feature is a geometric feature based on the defect morphology, which captures the boundary topology of bubbles and scratches through void convolution; The third feature is to analyze the consistency of the coating thickness through the residual network based on the reflection spectrum intensity; S4. Match the first feature, the second feature, and the third feature with a standard coating layer feature library to determine the coating layer quality: Determine whether the coating uniformity is qualified based on the difference between the first feature and the standard texture; Determine the defect type and location coordinates based on the matching result between the second feature and the defect template; Calculate the thickness variation coefficient based on the light intensity distribution of the third characteristic and determine the thickness tolerance area; S5. Output a test report containing identification code, defect type, location coordinates and severity, and trigger the sorting equipment to mark unqualified pigtails.

2. The method for visually inspecting the surface coating quality of a pigtail according to claim 1, wherein: The multispectral imaging device includes a near-infrared light source in an annular array and a high-resolution CCD camera. The wavelength of the light source is 850nm-1550nm, and is used to enhance the imaging contrast of micron-level defects on the coating surface.

3. The method for visually inspecting the surface coating quality of a pigtail according to claim 1, wherein: First preprocessing, screening out valid images containing complete coating areas and with resolutions higher than a preset threshold, and using them as standard inspection image sets; The second preprocessing is to remove image noise through adaptive filtering and enhance the contrast between the coating layer and the background based on histogram stretching; The third preprocessing is to divide the image into multiple sub-regions according to a preset grid and extract the texture gradient map of each sub-region.

4. The method for visually inspecting the surface coating quality of a pigtail according to claim 3, wherein: In the second preprocessing, the adaptive filtering adopts a non-local mean filtering algorithm, and the window size is dynamically adjusted according to the image noise level. The noise level is calculated by wavelet transforming the high-frequency component energy.

5. The method for visually inspecting the surface coating quality of a pigtail according to claim 4, wherein: In the third pre-processing, the division spacing of the preset grid is adaptively adjusted according to the diameter of the pigtail, so that each sub-area covers at least 3 complete coating layer corrugation periods.

6. The method for visually inspecting the surface coating quality of a pigtail according to claim 1, wherein: The deep convolutional neural network adopts a dual-branch structure, including: The first branch: consists of the Inception module, which is used to extract multi-scale texture features; The second branch consists of a U-Net encoder and is used for pixel-level segmentation of defect areas. The feature maps of the two branches are fused through the channel attention mechanism and then input into the fully connected layer for classification.

7. The method for visually inspecting the surface coating quality of a pigtail according to claim 1, wherein: The calculation method of the thickness variation coefficient is: The light intensity mean μ and standard deviation σ of each sub-area are calculated, and the thickness variation coefficient CV is defined as σ / μ×100%. When CV>5%, it is determined that the thickness is abnormal.

8. The method for visually inspecting the surface coating quality of a pigtail according to claim 1, wherein: The severity of the defects is divided into four levels: Minor: Single defect area <0.01mm 2 and does not penetrate the coating; Moderate: defect area∈[0.01,0.1)mm 2 or there are 2 cross scratches; Serious: defect area ≥ 0.1mm 2 Or it may cause the optical fiber core to be exposed; Critical: Thickness variation coefficient CV∈[5%,8%) and no macro defects.

Images