Intelligent detection system and method for aviation conduit identification mark belt

Abstract

The invention discloses an aviation conduit identification label belt intelligent detection system and method, the system comprises a scanning mechanism assembly and a processor unit, a fixed scanning device can carry out information acquisition and verification on conduit labels before pasting, a mobile scanning device can identify the pasted labels, and the labels can be identified through the processor unit. And the image preprocessing and distortion correction module is used for receiving image data from the scanning mechanism assembly and performing image optimization through an image preprocessing and distortion correction algorithm. The invention provides the intelligent detection system and method for the aviation conduit identification mark band, which are high in detection efficiency and stable in identification precision and have full-process closed-loop management and data tracing capabilities. The system aims to solve the problems of low manual detection efficiency and high error rate, and overcomes the difficulty in adaptability of the existing automation technology in a complex industrial environment.

Description

Technical Field

[0001] This invention belongs to the field of image data processing technology, specifically relating to an intelligent detection system and method for identifying labels on aviation duct tubes. Background Technology

[0002] In the field of aircraft manufacturing and maintenance, the duct system, as the "blood vessels" of critical systems such as aircraft hydraulics, fuel, and gas supply, is directly related to flight safety in terms of the accuracy of its assembly and maintenance. Duct labels (identification tapes) are the core carriers of information such as duct model, flow direction, and system code, and are an important basis for guiding assembly, ensuring maintenance, and achieving quality traceability.

[0003] However, current catheter label verification mainly relies on manual visual inspection, which suffers from significant problems such as low efficiency, susceptibility to human fatigue and subjective factors, and a high risk of missed and false detections. The error rate of manual inspection is between 0.3% and 1.2%, and it is difficult to achieve data traceability and integration with upper-level management systems (such as MES / ERP), which has become a bottleneck restricting the improvement of automation, digitalization and intelligence in the catheter assembly process.

[0004] While some existing technologies attempt to automate inspection using general OCR or barcode recognition, they generally suffer from poor environmental adaptability (e.g., changes in lighting, glare, label bending), unstable recognition accuracy, and the inability to achieve closed-loop verification between label information and task sheet information. Especially after the catheter label is affixed, image distortion caused by curved affixing makes it difficult for general OCR technology to reliably and accurately identify the label. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an intelligent detection system and method for aviation duct identification tags that features high detection efficiency, stable identification accuracy, and full-process closed-loop management and data traceability capabilities. This system aims to solve the problems of low efficiency and high error rate in manual detection and overcome the adaptability challenges of existing automation technologies in complex industrial environments.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An intelligent detection system for aviation duct identification tags. It includes a scanning mechanism assembly and a processor unit; the scanning mechanism includes a fixed scanning device for collecting and verifying information on unattached labels before the catheter labels are affixed; It also includes: a mobile scanning device, comprising a handheld support, an embedded camera module, and a supplementary lighting module, for identification after the label is affixed to the guide tube; and the processor unit for receiving image data from the scanning mechanism components and optimizing the image through image preprocessing and distortion correction algorithms.

[0007] This invention also discloses an intelligent detection method based on the above-mentioned detection system, specifically... A smart detection method for aviation duct identification tags includes the following steps: S1: Label Information Pre-collection: The unattached catheter labels are initially identified using a fixed scanning device. Multispectral imaging technology is used to eliminate reflective interference, extract the characters and barcode information on the labels, and compare them with the planned data in the task work order system to complete the first verification. S2: Label pasting and real-time verification: Paste the label that matches the pre-collected information to the designated position of the duct. After the operation is completed, trigger the mobile scanning device to perform secondary recognition on the pasted label, correct the surface deformation through the distortion compensation model, and restore the flat image of the label. S3: Closed-loop comparison and result judgment: The processor unit performs bidirectional verification between the actual collected information after correction in step S2 and the task work order data to complete the second verification; and performs the third verification based on the database information. If the triple verification information is consistent, the test is considered passed. If any one of the verification information is different, an audible and visual alarm will be triggered immediately, and the conduit will be locked and put into the pending process to prevent misinstallation and misuse. S4: Data traceability and system integration: All test records are automatically uploaded to the cloud database, supporting full lifecycle information traceability.

[0008] Furthermore, in step S1, information collection is performed using the following method: S11: Place the label to be inspected on the flat scanning bracket and start the high-resolution industrial camera to acquire images; S12: Adjust the light source angle and intensity of the supplementary light module to adjust the lighting conditions and ensure the brightness uniformity of the acquired image; S13: The acquired image is transmitted to the central processing unit, where a filtering algorithm is used to remove noise, and a contrast enhancement algorithm is used to improve image clarity. S14: Binarize and perform edge detection and localization on the preprocessed image, segment the left functional description area and the right coded information area, and extract the label character area and barcode features; S15: Dual-region parallel recognition, calling a dedicated character recognition model to extract the functional description text on the left and the encoded information on the right in parallel; S16: Automatically compare the identification results with the corresponding conduit design information in the task work order system to verify the consistency of the content.

[0009] Furthermore, in step S13, if the collected label is a ring label, a cylindrical unfolding model based on monocular vision is used to correct geometric distortion and generate a planar image.

[0010] Furthermore, in step S14, the processed image is segmented and region located, including the following sub-steps: S141: Detection of Circular Label Area and Determination of Central Axis: For the preprocessed grayscale image, Hough transform or contour analysis is used to detect parallel straight lines or ellipses representing the edge of the duct in the image; and the central axis of the duct (cylinder) in the image is fitted based on these features. S142: Parameter fitting for cylindrical model: In the direction perpendicular to the aforementioned central axis, the upper and lower edge pixels of the ring label are extracted; based on the perspective projection model and the extracted edge points, the radius R of the cylinder and its position parameters in the image coordinate system are fitted using the nonlinear least squares method. S143: Establish coordinate mapping relationships (expansion and transformation): Define the coordinates of a cylindrical surface: Suppose that any point on the cylindrical surface is determined by two parameters—the angle θ along the circumference and the height h along the axis; For each target pixel (u,v) on the corrected planar image, it corresponds to (θ,h) on the cylindrical surface, where u=R*θ and v=h; Based on the camera imaging model, establish a mathematical mapping relationship between the points (θ,h) on the cylindrical surface and the pixels (x,y) in the original distorted image; S144: Pixel Resampling and Planar Image Generation Iterate through each pixel (u,v) of the corrected planar image and use the mapping relationship from the previous step to calculate its corresponding floating-point coordinates (x,y) in the original distorted image.

[0011] Using bilinear or bicubic interpolation algorithms, the grayscale or color value of a point is calculated based on the pixel values ​​around the original image (x,y), and then assigned to the pixel (u,v) of the planar image. The final output is a planar image with no perspective distortion or curved surfaces, and the label content is horizontally unfolded. Post-processing of the S145 unfolded image: The unfolded planar image undergoes secondary enhancement processing to optimize the input quality for subsequent OCR recognition.

[0012] Furthermore, in step S2, the label pasting information verification is performed using the following method: S21: Bring the handheld support close to the affixed label and activate the embedded camera module to acquire images; S22: Adjust image brightness through dynamic illumination compensation algorithm to eliminate reflections and shadow interference; S23: The acquired image is transmitted to the central processing unit, and an image enhancement algorithm is used to correct the image distortion caused by curved surface pasting; S24: Based on the contour analysis algorithm, locate the overall label area, segment the left functional description area and the right coding information area, and extract the label character area and barcode features; S25: Dual-region parallel recognition, calling a dedicated character recognition model to extract the functional description text on the left and the encoded information on the right in parallel; S26: Automatically compare the identification results with the corresponding conduit design information in the task work order system to verify the consistency of the content.

[0013] Furthermore, the image enhancement algorithm correction in step S23 specifically includes: S231: Distortion Model Establishment: Surface distortion: The label is affixed to the cylindrical surface of the conduit with radius (R); Perspective distortion: During handheld scanning, let the parametric coordinates of a point (P) on a cylindrical surface in the world coordinate system be ((R,θ,h)). After perspective projection by the camera, it forms a pixel in the distorted image. .

[0014] S232: Correction Model and Core Mapping Formula: Correction via mapping function distorted image points Mapping to image points on the correction plane : in: (R) is the radius of the conduit cylinder; It is the unit normal vector representing the spatial orientation (axial direction) of the duct, used to characterize perspective distortion; It is the camera's intrinsic parameter matrix; The specific mapping can be achieved by solving the following inverse mapping: for each point on the calibration map... The image is then back-projected onto the surface of the cylindrical model, and then reprojected onto the original image to obtain its color values. The core relationship involves three-dimensional spatial coordinates. Transformation with image coordinates: in, As a scale factor, It is determined by the duct posture The determined rotation and translation matrix, (u,v), is the coordinate of the correction plane.

[0015] Furthermore, in step S24, to adapt to different catheter sizes and arbitrary shooting angles, key parameters (R) and... are automatically estimated from a single image. Specifically, it includes the following steps: S241: Initial localization: The label region is obtained based on contour analysis; S242: Feature extraction: Detect the edge lines of the label within this region; S243: Parameter Solving: Treat these two edge lines as projections of two generatrices (or sections) on a cylinder onto the image. Using a nonlinear optimization algorithm, adjust the parameters (R) and... This ensures that after the 3D cylindrical model is projected onto the image plane, the position of its corresponding generatrix best fits the detected edge line. S244: Image resampling and generation: The adjusted parameters (R) and Through mapping function Accurate calculations are performed, and each pixel of the corrected image is filled using bilinear interpolation. The final output is a distortion-free planar image with the label content horizontally unfolded, and region segmentation is performed.

[0016] Further, in step S4, the data traceability system includes: S41: Store each test result in the database of the central processing unit, recording the test time, test personnel, test equipment, and test results; S42: Upload the detection results to the upper-level management system via network communication protocol; S43: Record abnormal events in the operation log, including the abnormality type, abnormal time, and handling measures; S44: Through data correlation analysis algorithm, the detection results are compared with historical records to generate a quality traceability report.

[0017] Furthermore, in step S3, if the test result meets the standard, a test qualification certificate is generated, which includes an electronic record of timestamp, location code and operator ID and is synchronized to the management system. The relevant information is stored in the blockchain network.

[0018] Compared with existing technologies, the advantages of this invention are as follows: First, this invention constructs a fully intelligent detection closed loop encompassing "pre-collection - label verification - closed-loop comparison - data traceability." Through the collaborative operation of fixed and mobile scanning devices, it achieves dual recognition before and after label application, and also implements a triple verification mechanism of label information self-verification, task sheet information matching, and database information verification, covering the entire lifecycle quality control of duct labels. This effectively avoids the subjectivity and randomness of manual inspection, reducing the detection error rate to below 0.05%, and significantly improving the accuracy and reliability of aviation duct label detection.

[0019] Secondly, addressing the challenge of image distortion caused by labels affixed to curved surfaces of aviation ducts, this invention innovatively proposes a cylindrical unfolding model and dynamic distortion compensation algorithm based on monocular vision. This model can adapt to different duct radii and spatial orientations under arbitrary shooting angles. By establishing a precise three-dimensional spatial coordinate and image coordinate transformation relationship, it achieves high-precision planar restoration of curved labels, providing high-quality image input for subsequent OCR recognition, with a character recognition accuracy rate of over 99.8%.

[0020] Furthermore, the system highly integrates technologies such as multispectral imaging, dedicated character recognition models, blockchain notarization, and data correlation analysis, which not only realizes full lifecycle information traceability from label pre-collection to final assembly, but also ensures the traceability and quality control of each product. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the mobile scanning mechanism of the intelligent detection system for aviation duct identification labels provided by the present invention.

[0023] Figure 2 This is a flowchart of the mobile scanning mechanism of the intelligent detection method for aviation duct identification labels provided by the present invention.

[0024] Figure 3 This is a flowchart illustrating the workflow of the fixed scanning mechanism in the intelligent detection method for aviation duct identification tags provided by this invention. Reference numerals: 1. Handheld support frame; 2. Embedded camera module. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0027] In the description of this invention, it should be noted that if terms such as "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," or "outer" are used to indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this invention is usually placed, they are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0028] Furthermore, the terms "first," "second," and "third" are used only for distinguishing descriptions and should not be interpreted as indicating or implying relative importance.

[0029] Furthermore, the use of terms such as "horizontal," "vertical," and "suspended" does not imply that the component must be absolutely horizontal or suspended, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal relative to "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0030] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0031] It should be noted that, where there is no conflict, the features in the embodiments of the present invention can be combined with each other.

[0032] like Figures 1-2 As shown, an intelligent detection system for aviation conduit identification tags is provided. It includes a scanning mechanism assembly and a processor unit; the scanning mechanism includes a fixed scanning device for collecting and verifying information on unattached labels before the catheter labels are affixed; It also includes: a mobile scanning device, comprising a handheld support frame 1, an embedded camera module 2, and a supplementary lighting module, for identification after the label is affixed to the guide tube; and the processor unit for receiving image data from the scanning mechanism components and optimizing the image through image preprocessing and distortion correction algorithms.

[0033] This intelligent detection system adds a mobile scanning device to the traditional fixed scanning device. The mobile scanning device identifies the pasted labels, corrects distortion, and optimizes the image. This effectively avoids the problem of high error rates caused by traditional general-purpose OCR systems failing to directly apply simple edge detection algorithms suitable for planar scanning to the positioning of curved labels in complex environments. This invention systematically solves the positioning problem throughout the entire process by designing two positioning mechanisms, optimized at different stages.

[0034] Based on this system design, the present invention also discloses an intelligent detection method based on the above detection system, specifically: A smart detection method for aviation duct identification tags includes the following steps: S1: Label Information Pre-collection: The unattached catheter labels are initially identified using a fixed scanning device. Multispectral imaging technology is used to eliminate reflective interference, extract the characters and barcode information on the labels, and compare them with the planned data in the task work order system to complete the first verification. S2: Label pasting and real-time verification: Paste the label that matches the pre-collected information to the designated position of the duct. After the operation is completed, trigger the mobile scanning device to perform secondary recognition on the pasted label, correct the surface deformation through the distortion compensation model, and restore the flat image of the label. S3: Closed-loop comparison and result judgment: The processor unit performs bidirectional verification between the actual collected information after correction in step S2 and the task work order data to complete the second verification; and performs the third verification based on the database information. If the triple verification information is consistent, the test is considered passed. If any one of the verification information is different, an audible and visual alarm will be triggered immediately, and the conduit will be locked and put into the pending process to prevent misinstallation and misuse. S4: Data traceability and system integration: All test records are automatically uploaded to the cloud database, supporting full lifecycle information traceability.

[0035] In step S1, information collection is performed using the following method: S11: Place the label to be inspected on the flat scanning bracket and start the high-resolution industrial camera to acquire images; S12: Adjust the light source angle and intensity of the supplementary light module to adjust the lighting conditions and ensure the brightness uniformity of the acquired image; S13: The acquired image is transmitted to the central processing unit, where a filtering algorithm is used to remove noise, and a contrast enhancement algorithm is used to improve image clarity. S14: Binarize and perform edge detection and localization on the preprocessed image, segment the left functional description area and the right coded information area, and extract the label character area and barcode features; S15: Dual-region parallel recognition, calling a dedicated character recognition model to extract the functional description text on the left and the encoded information on the right in parallel; S16: Automatically compare the identification results with the corresponding conduit design information in the task work order system to verify the consistency of the content.

[0036] In step S13, if the collected label is a ring label, a cylindrical unfolding model based on monocular vision is used to correct geometric distortion and generate a planar image.

[0037] In step S14, the processed image is segmented and region located, including the following sub-steps: S141: Detection of Circular Label Area and Determination of Central Axis: For the preprocessed grayscale image, Hough transform or contour analysis is used to detect parallel straight lines or ellipses representing the edge of the duct in the image; and the central axis of the duct (cylinder) in the image is fitted based on these features. S142: Parameter fitting for cylindrical model: In the direction perpendicular to the aforementioned central axis, the upper and lower edge pixels of the ring label are extracted; based on the perspective projection model and the extracted edge points, the radius R of the cylinder and its position parameters in the image coordinate system are fitted using the nonlinear least squares method. S143: Establish coordinate mapping relationships (expansion and transformation): Define the coordinates of a cylindrical surface: Suppose that any point on the cylindrical surface is determined by two parameters—the angle θ along the circumference and the height h along the axis; For each target pixel (u,v) on the corrected planar image, it corresponds to (θ,h) on the cylindrical surface, where u=R*θ and v=h; Based on the camera imaging model, establish a mathematical mapping relationship between the points (θ,h) on the cylindrical surface and the pixels (x,y) in the original distorted image; S144: Pixel Resampling and Planar Image Generation Iterate through each pixel (u,v) of the corrected planar image and use the mapping relationship from the previous step to calculate its corresponding floating-point coordinates (x,y) in the original distorted image.

[0038] Using bilinear or bicubic interpolation algorithms, the grayscale or color value of a point is calculated based on the pixel values ​​around the original image (x,y), and then assigned to the pixel (u,v) of the planar image. The final output is a planar image with no perspective distortion or curved surfaces, and the label content is horizontally unfolded. Post-processing of the S145 unfolded image: The unfolded planar image undergoes secondary enhancement processing to optimize the input quality for subsequent OCR recognition.

[0039] In short, in fixed scanning, there are two scenarios for the labels. One is that the label is planar before being pasted. In this case, an edge detection algorithm is used to locate the entire label area. Under controlled, uniform lighting conditions, leveraging high contrast, edge detection and Hough linear transform are used to quickly and accurately locate the rectangular boundaries of the planar label. This method assumes that the label edges are straight lines and the image is distortion-free, making it suitable for high-throughput, high-precision processing in the label pre-acquisition stage. The other scenario is that the label is already a ring-shaped label before being pasted. In this case, steps S141-S145 are required to unfold the ring-shaped label into a planar shape.

[0040] Finally, the extracted plane is divided into left and right sides. Generally, the left side contains Chinese characters and the right side contains codes. The characters and codes are recognized separately, and the recognition results are compared with the preset settings. If they both meet the preset settings, the results are stored. If they do not match, an error warning is displayed.

[0041] For example, consider the ring-shaped label affixed to a Φ24mm diameter aviation aluminum alloy conduit. The label is made of laminated PET, 12mm wide, and is affixed around the conduit, with characters and a barcode printed on its surface.

[0042] Image acquisition: A high-resolution industrial camera (20 megapixels) with a fixed scanning device is used to vertically photograph the ring-shaped label to obtain the original color image.

[0043] Preprocessing and feature extraction: The image is converted to grayscale and then Gaussian filtered to remove noise.

[0044] Strong edges in the image were obtained using the Canny edge detection operator. It can be clearly seen that the top and bottom edges of the label appear as two curved arcs, and the ends of the conduit show partially circular outlines.

[0045] Model fitting: Using the detected set of upper and lower edge points of the label, combined with the circular contour cues of the duct end face, a cylinder fitting program based on the RANSAC (Random Sample Consensus) algorithm was run to robustly estimate the cylinder radius as R=12.1mm (which is basically consistent with the physical size), and the equation of the cylinder axis in the image was determined.

[0046] Perform the unfolding transformation: Set the resolution of the unfolded planar image to 600 pixels (width) x 150 pixels (height), with the width corresponding to the label perimeter (approximately 2πR≈76mm) and the height corresponding to the label width.

[0047] Based on the fitted cylinder parameters and camera intrinsics (which have been pre-calibrated), calculate the mapping position of each pixel in the unfolded image in the original image.

[0048] Pixel values ​​are sampled using bilinear interpolation to generate the final planar corrected image.

[0049] Effect verification: The corrected image is then input into the system's dedicated character recognition model. The uncorrected original image, due to severe character distortion, has an OCR recognition rate of only ~78%.

[0050] After processing by this geometric distortion correction algorithm, the characters are restored to a standard horizontal arrangement, and the recognition rate is significantly improved to >99.5%, effectively solving the problem of recognizing circular labels.

[0051] The label's overall area is located based on an edge detection algorithm, and the left functional description area and the right encoded information area are segmented.

[0052] In step S2, the label pasting information verification is performed using the following method: S21: Bring the handheld support close to the affixed label and activate the embedded camera module to acquire images; S22: Adjust image brightness through dynamic illumination compensation algorithm to eliminate reflections and shadow interference; S23: The acquired image is transmitted to the central processing unit, and an image enhancement algorithm is used to correct the image distortion caused by curved surface pasting; S24: Based on the contour analysis algorithm, locate the overall label area, segment the left functional description area and the right coding information area, and extract the label character area and barcode features; S25: Dual-region parallel recognition, calling a dedicated character recognition model to extract the functional description text on the left and the encoded information on the right in parallel; S26: Automatically compare the identification results with the corresponding conduit design information in the task work order system to verify the consistency of the content.

[0053] The image enhancement algorithm correction in step S23 specifically includes: S231: Distortion Model Establishment: Surface distortion: The label is affixed to the cylindrical surface of the conduit with radius (R); Perspective distortion: During handheld scanning, let the parametric coordinates of a point (P) on a cylindrical surface in the world coordinate system be ((R,θ,h)). After perspective projection by the camera, it forms a pixel in the distorted image. .

[0054] S232: Correction Model and Core Mapping Formula: Correction via mapping function distorted image points Mapping to image points on the correction plane : in: (R) is the radius of the conduit cylinder; It is the unit normal vector representing the spatial orientation (axial direction) of the duct, used to characterize perspective distortion; It is the camera's intrinsic parameter matrix; The specific mapping can be achieved by solving the following inverse mapping: for each point on the calibration map... The image is then back-projected onto the surface of the cylindrical model, and then reprojected onto the original image to obtain its color values. The core relationship involves three-dimensional spatial coordinates. Transformation with image coordinates: in, As a scale factor, It is determined by the duct posture The determined rotation and translation matrix, (u,v), is the coordinate of the correction plane.

[0055] In step S24, to adapt to different catheter sizes and arbitrary shooting angles, key parameters (R) and [other parameters] are automatically estimated from a single image. Specifically, it includes the following steps: S241: Initial localization: The label region is obtained based on contour analysis; S242: Feature extraction: Detect the edge lines of the label within this region; S243: Parameter Solving: Treat these two edge lines as projections of two generatrices (or sections) on a cylinder onto the image. Using a nonlinear optimization algorithm, adjust the parameters (R) and... This ensures that after the 3D cylindrical model is projected onto the image plane, the position of its corresponding generatrix best fits the detected edge line. S244: Image resampling and generation: The adjusted parameters (R) and Through mapping function Accurate calculations are performed, and each pixel of the corrected image is filled using bilinear interpolation. The final output is a distortion-free planar image with the label content horizontally unfolded, and region segmentation is performed.

[0056] In mobile scanning, a contour analysis algorithm is used to locate the entire label area. Under complex lighting and background interference, dynamic lighting compensation is first used to improve image quality, followed by the extraction of all contours in the image. Contours are filtered based on prior geometric features of the known label (such as area range, aspect ratio range, and contour convexity). Polygon approximation is performed on candidate contours to obtain their circumscribed polygons. This method does not rely on complete straight edges and can effectively cope with label shape distortion and edge breaks caused by curved surface pasting and shooting angles, ensuring robust positioning.

[0057] Information is extracted from the final generated image and compared with the database to determine if it matches the database information. If it matches, the recognition result is stored; otherwise, an error warning is issued.

[0058] For example: Handheld scanning of a ring-shaped label already affixed to a hydraulic conduit with a diameter of Φ28mm (radius R=14mm). The camera is positioned at approximately a 45-degree angle to the conduit's axis, resulting in significant perspective and surface distortion in the image. Algorithm execution flow: Input: Image after S22 dynamic illumination compensation.

[0059] Parameter estimation: Contour analysis locates the approximate area of ​​the label.

[0060] Within this region, two distinct curved edges were detected using the Canny operator and the Hough line transform.

[0061] Using these two curves as observation data, initialize (R=10mm), For the vertical direction of the image, an optimization library is used for iterative solution. After approximately 5 iterations, the estimated value converges. , (In the camera coordinate system) it accurately reflects the spatial tilt angle of the duct.

[0062] Image correction: using the estimated and A 400×120 pixel corrected image is generated according to the above mapping model.

[0063] Quantifying the effects: Directly performing OCR recognition on the original distorted image yields an accuracy rate of only 65%.

[0064] After correction using this distortion compensation algorithm, the accuracy rate of OCR recognition is improved to 98.7%.

[0065] Robustness demonstration: Experiments show that the algorithm can stably estimate parameters and achieve effective correction within the range of duct diameter 10mm-50mm and camera tilt angle 0°-70°, and has a certain fault tolerance for partially occluded and weakly textured areas.

[0066] In step S3, if the test result meets the standard, a test qualification certificate is generated, which includes an electronic record of timestamp, location code and operator ID and is synchronized to the management system. The relevant information is stored in the blockchain network.

[0067] In step S4, the data traceability system includes: S41: Store each test result in the database of the central processing unit, recording the test time, test personnel, test equipment, and test results; S42: Upload the detection results to the upper-level management system via network communication protocol; S43: Record abnormal events in the operation log, including the abnormality type, abnormal time, and handling measures; S44: Through data correlation analysis algorithm, the detection results are compared with historical records to generate a quality traceability report.

[0068] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. An intelligent detection system for aviation duct identification labels, comprising a scanning mechanism assembly and a processor unit; the scanning mechanism includes a fixed scanning device for collecting and verifying information on unattached labels before the duct labels are affixed, characterized in that: It also includes: a mobile scanning device, comprising a handheld support frame (1), an embedded camera module (2) and a supplementary lighting module, for identification after the label is attached to the conduit; the processor unit is used to receive image data from the scanning mechanism components and optimize the image through image preprocessing and distortion correction algorithms.

2. A method for intelligent detection of aviation duct identification tags, employing the aforementioned intelligent detection system, characterized in that, Includes the following steps: S1: Label Information Pre-collection: The unattached catheter labels are initially identified using a fixed scanning device. Multispectral imaging technology is used to eliminate reflective interference, extract the characters and barcode information on the labels, and compare them with the planned data in the task work order system to complete the first verification. S2: Label pasting and real-time verification: Paste the label that matches the pre-collected information to the designated position of the duct. After the operation is completed, trigger the mobile scanning device to perform secondary recognition on the pasted label, correct the surface deformation through the distortion compensation model, and restore the flat image of the label. S3: Closed-loop comparison and result judgment: The processor unit performs bidirectional verification between the actual collected information after correction in step S2 and the task work order data to complete the second verification; and performs the third verification based on the database information. If the triple verification information is consistent, the test is considered passed. If any one of the verification information is different, an audible and visual alarm will be triggered immediately, and the conduit will be locked and put into the pending process to prevent misinstallation and misuse. S4: Data traceability and system integration: All test records are automatically uploaded to the cloud database, supporting full lifecycle information traceability.

3. The intelligent detection method for aviation duct identification tags according to claim 2, wherein in step S1, information collection is performed by the following method: S11: Place the label to be inspected on the flat scanning bracket and start the high-resolution industrial camera to acquire images; S12: Adjust the light source angle and intensity of the supplementary light module to adjust the lighting conditions and ensure the brightness uniformity of the acquired image; S13: The acquired image is transmitted to the central processing unit, where a filtering algorithm is used to remove noise, and a contrast enhancement algorithm is used to improve image clarity. S14: Binarize and perform edge detection and localization on the preprocessed image, segment the left functional description area and the right coded information area, and extract the label character area and barcode features; S15: Dual-region parallel recognition, calling a dedicated character recognition model to extract the functional description text on the left and the encoded information on the right in parallel; S16: Automatically compare the identification results with the corresponding conduit design information in the task work order system to verify the consistency of the content.

4. In the intelligent detection method for aviation duct identification tags according to claim 3, in step S13, if the collected tag is a ring tag, a cylindrical unfolding model based on monocular vision is used for geometric distortion correction to generate a planar image.

5. The intelligent detection method for aviation duct identification tags according to claim 3, wherein step S14, segmenting and locating the processed image, includes the following sub-steps: S141: Detection of Circular Label Area and Determination of Central Axis: For the preprocessed grayscale image, Hough transform or contour analysis is used to detect parallel straight lines or ellipses representing the edge of the duct in the image; and the central axis of the duct (cylinder) in the image is fitted based on these features. S142: Parameter fitting for cylindrical model: In the direction perpendicular to the aforementioned central axis, the upper and lower edge pixels of the ring label are extracted; based on the perspective projection model and the extracted edge points, the radius R of the cylinder and its position parameters in the image coordinate system are fitted using the nonlinear least squares method. S143: Establish coordinate mapping relationships (expansion and transformation): Define the coordinates of a cylindrical surface: Suppose that any point on the cylindrical surface is determined by two parameters—the angle θ along the circumference and the height h along the axis; For each target pixel (u,v) on the corrected planar image, it corresponds to (θ,h) on the cylindrical surface, where u=R*θ and v=h; Based on the camera imaging model, establish a mathematical mapping relationship between the points (θ,h) on the cylindrical surface and the pixels (x,y) in the original distorted image; S144: Pixel Resampling and Planar Image Generation Iterate through each pixel (u,v) of the corrected planar image and use the mapping relationship from the previous step to calculate its corresponding floating-point coordinates (x,y) in the original distorted image; Using bilinear or bicubic interpolation algorithms, the grayscale or color value of a point is calculated based on the pixel values ​​around the original image (x,y), and then assigned to the pixel (u,v) of the planar image. The final output is a planar image with no perspective distortion or curved surfaces, and the label content is horizontally unfolded. Post-processing of the S145 unfolded image: The unfolded planar image undergoes secondary enhancement processing to optimize the input quality for subsequent OCR recognition.

6. The intelligent detection method for aviation duct identification labels according to claim 2, wherein in step S2, the label affixing information verification is performed by the following method: S21: Bring the handheld support close to the affixed label and activate the embedded camera module to acquire images; S22: Adjust image brightness through dynamic illumination compensation algorithm to eliminate reflections and shadow interference; S23: The acquired image is transmitted to the central processing unit, and an image enhancement algorithm is used to correct the image distortion caused by curved surface pasting; S24: Based on the contour analysis algorithm, locate the overall label area, segment the left functional description area and the right coding information area, and extract the label character area and barcode features; S25: Dual-region parallel recognition, calling a dedicated character recognition model to extract the functional description text on the left and the encoded information on the right in parallel; S26: Automatically compare the identification results with the corresponding conduit design information in the task work order system to verify the consistency of the content.

7. The intelligent detection method for aviation duct identification tags according to claim 6, wherein the image enhancement algorithm correction in step S23 specifically includes: S231: Distortion Model Establishment: Surface distortion: The label is affixed to the cylindrical surface of the conduit with radius (R); Perspective distortion: During handheld scanning, let the parametric coordinates of a point (P) on a cylindrical surface in the world coordinate system be ((R,θ,h)). After perspective projection by the camera, it forms a pixel in the distorted image. ; S232: Correction Model and Core Mapping Formula: Correction via mapping function distorted image points Mapping to image points on the correction plane : in: (R) is the radius of the conduit cylinder; It is the unit normal vector representing the spatial orientation (axial direction) of the duct, used to characterize perspective distortion; It is the camera's intrinsic parameter matrix; The specific mapping can be achieved by solving the following inverse mapping: for each point on the calibration map... The image is then back-projected onto the surface of the cylindrical model, and then reprojected onto the original image to obtain its color values. The core relationship involves three-dimensional spatial coordinates. Transformation with image coordinates: in, As a scale factor, It is determined by the duct posture The determined rotation and translation matrix, (u,v), is the coordinate of the correction plane.

8. The intelligent detection method for aviation duct identification tags according to claim 6, wherein in step S24, to adapt to different duct sizes and arbitrary shooting angles, key parameters (R) and [other parameters] are automatically estimated from a single image. Specifically, it includes the following steps: S241: Initial localization: The label region is obtained based on contour analysis; S242: Feature extraction: Detect the edge lines of the label within this region; S243: Parameter Solving: Treat these two edge lines as projections of two generatrices (or sections) on a cylinder onto the image. Using a nonlinear optimization algorithm, adjust the parameters (R) and... This ensures that after the 3D cylindrical model is projected onto the image plane, the position of its corresponding generatrix best fits the detected edge line. S244: Image resampling and generation: The adjusted parameters (R) and Through mapping function Accurate calculations are performed, and each pixel of the corrected image is filled using bilinear interpolation. The final output is a distortion-free planar image with the label content horizontally unfolded, and region segmentation is performed.

9. The intelligent detection method for aviation duct identification tags according to claim 2, wherein in step S4, the data traceability system comprises: S41: Store each test result in the database of the central processing unit, recording the test time, test personnel, test equipment, and test results; S42: Upload the detection results to the upper-level management system via network communication protocol; S43: Record abnormal events in the operation log, including the abnormality type, abnormal time, and handling measures; S44: Through data correlation analysis algorithm, the detection results are compared with historical records to generate a quality traceability report.

10. In the intelligent detection method for aviation duct identification tags according to claim 2, in step S3, if the detection result meets the standard, a certificate of conformity is generated, which includes an electronic record of timestamp, location code and operator ID and is synchronized to the management system, and the relevant information is stored in the blockchain network.

Images