Digital speckle based fiber-wound gas cylinder defect detection method and system

Abstract

The application discloses a kind of based on digital speckle's fiber winding gas cylinder defect detection method and system, high contrast speckle pattern is made on the surface of gas cylinder;Controllable load is applied to gas cylinder, and the image sequence of the surface of gas cylinder before and after load is collected;Based on the image sequence of the surface of gas cylinder, the full-field displacement field data of the surface of gas cylinder is calculated by digital speckle correlation method;The full-field displacement field data of the surface of gas cylinder is corrected by surface coordinate transformation;According to the full-field displacement field data of the surface of gas cylinder after correction, the full-field strain distribution image set of the surface of gas cylinder is calculated using difference method;The full-field strain distribution image set of the surface of gas cylinder is processed by image enhancement;Based on the full-field strain distribution image set of the surface of gas cylinder after enhancement, extract quantitative feature parameters, and utilize machine learning model to automatically identify and classify gas cylinder defect.By surface coordinate transformation correction, small microdefects can be identified, detection efficiency is improved, and detection cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of gas cylinder inspection technology, and in particular to a method and system for detecting defects in fiber-wound gas cylinders based on digital speckle. Background Technology

[0002] Fiber-wound gas cylinders are lightweight and high-strength, making them suitable for high-pressure hydrogen storage in hydrogen fuel cell vehicles. These cylinders consist of an inner liner and an outer carbon fiber / glass fiber epoxy resin winding layer. However, during long-term service, these cylinders are prone to defects such as fiber breakage, interlayer debonding, and resin cracking. While initially small in scale, these defects become stress concentration points that rapidly expand under sustained pressure loads, significantly reducing the cylinder's pressure-bearing capacity and fatigue life, posing serious safety hazards.

[0003] Currently, non-destructive testing of fiber-entangled gas cylinders relies on ultrasonic testing and X-ray inspection. Ultrasonic testing requires point-by-point scanning, is highly dependent on operator experience, is slow, and lacks sensitivity to defects where delamination is parallel to the sound beam propagation direction; furthermore, the coupling agent may contaminate the cylinder surface. X-ray inspection, on the other hand, has limited ability to detect fiber breaks and micro-cracks, and suffers from problems such as expensive equipment, radiation safety risks, and the need for strict protective measures, making it unsuitable for large-scale rapid inspection of gas cylinders in production sites or in service. Summary of the Invention

[0004] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, this invention proposes a method and system for detecting defects in fiber-wound gas cylinders based on digital speckle technology. The method excites internal defects in the gas cylinder through an optimized loading method, captures the resulting surface micro-strain field anomalies using digital speckle technology, enhances the signal-to-noise ratio of the defect signal through image processing, and achieves intelligent identification of gas cylinder defects based on quantization features. This method can identify minute micro-defects, improve detection efficiency, and reduce detection costs.

[0005] On one hand, embodiments of the present invention provide a method for detecting defects in fiber-wound gas cylinders based on digital speckle, comprising:

[0006] Create a high-contrast speckle pattern on the outer surface of the gas cylinder;

[0007] A controllable load is applied to the gas cylinder, and a sequence of images of the gas cylinder surface before and after the load is applied is acquired;

[0008] Based on the image sequence of the gas cylinder surface, the full-field displacement field data of the gas cylinder surface is calculated by the digital speckle correlation method.

[0009] The full-field displacement data of the gas cylinder surface is corrected by surface coordinate transformation to eliminate curvature error;

[0010] Based on the corrected full-field displacement field data of the gas cylinder surface, the image set of full-field strain distribution on the gas cylinder surface is calculated using the difference method;

[0011] Image enhancement processing is performed on the full-field strain distribution image set on the surface of the gas cylinder, including dynamic range gain, image decomposition and detail enhancement, normalization and histogram equalization, in order to amplify the micro-strain signal caused by gas cylinder defects.

[0012] Based on the enhanced full-field strain distribution image set of the gas cylinder surface, quantitative feature parameters are extracted, and machine learning models are used to automatically identify and classify gas cylinder defects.

[0013] According to some embodiments of the present invention, the speckle pattern has speckle size of 3-8 pixels, distribution density greater than 200 points / cm², local density variation coefficient less than 0.1, speckle pattern is made with high-contrast matte coating, and average grayscale gradient of speckle pattern is greater than 50 to ensure the accuracy of digital speckle calculation.

[0014] According to some embodiments of the present invention, the controllable load is low-pressure gas loading or mechanical loading. The pressure value of low-pressure gas loading is 5% to 15% of the rated working pressure of the gas cylinder, and the loading rate is controlled within 0.1 MPa / s; or mechanical loading is used to induce 0.1% to 0.5% circumferential strain in the gas cylinder.

[0015] According to some embodiments of the present invention, the digital speckle algorithm adopts the inverse combination Gauss-Newton algorithm with sub-pixel accuracy, a sub-region size of 21×21 pixels, and 20 iterations.

[0016] According to some embodiments of the present invention, the surface coordinate transformation correction is based on the CAD model of the gas cylinder and stereo vision calibration data, dividing the gas cylinder surface into a triangular mesh, with each mesh node storing the local radius of curvature R, and the displacement field correction formula is as follows:

[0017]

[0018] In the formula, u is the displacement component of a point on the surface of the gas cylinder in the x-direction; Δu measured It is the x-direction displacement directly calculated by digital speckle; Δu corrected It is the corrected displacement in the x-direction; u / x is the rate of change of displacement along the x-direction; 2 z / x 2 The curvature term characterizes the local curvature of the gas cylinder surface in the x-direction. 2 z / x 2The larger the value, the more drastic the curvature change, and the greater the correction amount for a standard cylindrical surface. 2 z / x 2 It is proportional to 1 / R; z(x,y) is a function of the surface height of the gas cylinder, obtained through stereo vision calibration.

[0019] According to some embodiments of the present invention, the image enhancement processing of the full-field strain distribution image set on the surface of the gas cylinder includes:

[0020] The gain factor is adaptively adjusted based on the standard deviation of the strain values ​​across the entire field. The gain formula is α = k / σ, where σ is the standard deviation and k is a constant.

[0021] A guided filtering algorithm is used to decompose the full-field strain distribution map into a base layer and a detail layer, with a filtering radius parameter of 15-25 pixels.

[0022] An enhancement factor γ is applied to the image detail layer, with a value ranging from 2.0 to 5.0.

[0023] Image contrast is enhanced using histogram equalization algorithms.

[0024] According to some embodiments of the present invention, the quantized characteristic parameters include local strain concentration factor, gradient magnitude entropy, anomalous region area, and shape factor.

[0025] According to some embodiments of the present invention, the machine learning model is a support vector machine or a random forest classifier, and the training data includes fiber breakage, delamination, and resin cracking defect samples.

[0026] In another aspect, embodiments of the present invention provide a fiber-wound gas cylinder defect detection system based on digital speckle, used to implement the aforementioned fiber-wound gas cylinder defect detection method based on digital speckle. The system includes:

[0027] The speckle pattern fabrication module is used to prepare speckle patterns that meet the requirements on the surface of the gas cylinder;

[0028] Loading module for applying precisely controlled low-pressure inflation or mechanical loads to gas cylinders;

[0029] The image acquisition module is used to acquire image sequences of the gas cylinder surface before and after the load is applied;

[0030] The data processing module is used for displacement field calculation, strain field calculation, image enhancement, and defect identification.

[0031] According to some embodiments of the present invention, the loading module includes a precision gas source, a pressure regulating valve, a safety relief valve, and a pressure sensor for achieving precise low-pressure gas filling; the image acquisition module includes multiple industrial cameras and an illumination source, with the multiple industrial cameras forming a camera array, which is stereo-calibrated to achieve panoramic coverage of the gas cylinder; the data processing module also includes a result output unit for generating a three-dimensional defect distribution map and an inspection report.

[0032] The fiber-wound gas cylinder defect detection method and system based on digital speckle according to embodiments of the present invention have at least the following beneficial effects:

[0033] By combining load excitation with image enhancement, micron-level strain changes can be effectively detected, uncovering minute and deep defects that are difficult to detect with traditional ultrasound and X-rays, thus reducing the false negative rate. Employing a camera array for single-image or rapid scanning replaces point-scan ultrasonic testing, enabling rapid full-field measurement of gas cylinders. Full-process automation shortens inspection time and improves efficiency, making it suitable for rapid inspection on production lines. It provides full-field, color strain distribution cloud maps, clearly showing the location, size, and relative severity of defects. Combining quantitative features and machine learning reduces human error, ensuring objective and traceable results. Coordinate transformation algorithms address surface measurement errors, adapting to various gas cylinder specifications. It is safely used throughout the entire lifecycle of gas cylinders, from factory inspection and periodic calibration to end-of-life assessment, providing continuous data support for safe use. It can be applied to all stages of the lifecycle of fiber-wound gas cylinders, including manufacturing quality inspection, damage assessment after type testing, and in-service periodic inspection. By integrating optical measurement, precision control, image processing, and artificial intelligence technologies, an integrated solution has been formed, which lowers the operating threshold, requires less equipment investment and no consumables, has low testing costs, and improves the repeatability and consistency of gas cylinder testing.

[0034] In another aspect, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program, the processor executes the above-described method for detecting defects in fiber-wound gas cylinders based on digital speckle.

[0035] On the other hand, an embodiment of the present invention provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for detecting defects in fiber-wound gas cylinders based on digital speckle.

[0036] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0037] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0038] Figure 1 This is a flowchart of a fiber-wound gas cylinder defect detection method based on digital speckle according to an embodiment of the present invention;

[0039] Figure 2 This is a flowchart illustrating the image enhancement processing of the full-field strain distribution image set on the surface of the gas cylinder in the fiber-wound gas cylinder defect detection method based on digital speckle according to an embodiment of the present invention.

[0040] Figure 3 This is a functional block diagram of a fiber entanglement gas cylinder defect detection system based on digital speckle according to an embodiment of the present invention;

[0041] Figure 4 This is a detailed flowchart of the fiber entanglement gas cylinder defect detection method based on digital speckle according to an embodiment of the present invention;

[0042] Figure 5 This is a schematic diagram of the speckle pattern on the surface of the normal area of ​​a gas cylinder, based on the fiber entanglement defect detection method for a gas cylinder according to an embodiment of the present invention.

[0043] Figure 6 This is a schematic diagram of the speckle pattern on the surface of the defect area of ​​a gas cylinder, based on the fiber entanglement gas cylinder defect detection method according to an embodiment of the present invention.

[0044] Figure 7 This is a schematic diagram of the structure of a computer device employing an embodiment of the present invention. Detailed Implementation

[0045] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0046] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. 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. Therefore, they should not be construed as limiting this invention.

[0047] In the description of this invention, "several" means one or more, "multiple" means two or more, "greater than," "less than," "exceeding," etc. are understood to exclude the stated number, and "above," "below," "within," etc. are understood to include the stated number. If "first," "second," etc. are used in the description, they are only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the number of indicated technical features or the order of the indicated technical features.

[0048] Digital speckle correlation (DSC) is an optical non-contact full-field deformation measurement technique that calculates full-field displacement and strain by tracking the grayscale changes of speckle patterns on an object's surface before and after deformation. This technique has been applied in laboratory material mechanical property testing. However, effectively applying it to the detection of defects in the winding layer of fiber-wound gas cylinders faces a series of unique challenges. First, the gas cylinder has a complex hyperboloid structure (body + end cap), and the general assumptions of planar digital speckle correlation introduce significant errors. Second, the strain anomaly signals induced on the surface by internal defects (especially deep defects) are extremely weak, resulting in a low signal-to-noise ratio, making them difficult to effectively identify and extract using conventional digital speckle processing procedures. Third, there is a lack of a dedicated system that integrates loading excitation specifically for gas cylinders, multi-view image acquisition, targeted image enhancement algorithms, and automatic defect identification and classification. Therefore, this technical field urgently needs a dedicated detection method and system that can overcome the above technical difficulties, achieve high sensitivity and efficiency, and be applicable to curved gas cylinders.

[0049] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] This embodiment provides a method for detecting defects in fiber-wound gas cylinders based on digital speckle. Please refer to [link to relevant documentation]. Figure 1 The method for detecting defects in fiber-wound gas cylinders based on digital speckle mainly includes steps S101 to S107:

[0051] S101. Create a high-contrast speckle pattern on the outer surface of the gas cylinder.

[0052] S102. Apply a controllable load to the gas cylinder and acquire image sequences of the gas cylinder surface before and after the load is applied.

[0053] S103. Based on the image sequence of the gas cylinder surface, the full-field displacement field data of the gas cylinder surface is calculated by the digital speckle correlation method.

[0054] S104. Perform surface coordinate transformation correction on the full-field displacement data of the gas cylinder surface to eliminate curvature error.

[0055] S105. Based on the corrected full-field displacement field data of the gas cylinder surface, the image set of full-field strain distribution on the gas cylinder surface is calculated using the difference method.

[0056] S106. Perform image enhancement processing on the full-field strain distribution image set on the gas cylinder surface, including dynamic range gain, image decomposition and detail enhancement, normalization and histogram equalization, to amplify the micro-strain signal caused by gas cylinder defects.

[0057] S107. Based on the enhanced full-field strain distribution image set of the gas cylinder surface, extract quantitative feature parameters and use machine learning models to automatically identify and classify gas cylinder defects.

[0058] The speckle pattern features speckle spots of 3-8 pixels in size, a distribution density greater than 200 spots / cm², and a local density variation coefficient less than 0.1. The speckle pattern is created using a high-contrast matte paint, and its average grayscale gradient is greater than 50 to ensure the accuracy of digital speckle calculations. First, the outer surface of the gas cylinder is cleaned and dried. Then, a random speckle pattern is prepared on the cylinder surface using a specialized spray gun and matte black and white paint. The speckle spot size is optimized to ensure that each spot occupies approximately 3-8 pixels in the acquired image. A high-resolution initial image I0(x, y) of the unloaded gas cylinder is acquired using an industrial camera under uniform illumination.

[0059] The controllable load can be either low-pressure gas filling or mechanical loading. Low-pressure gas filling involves applying a pressure of 5% to 15% of the cylinder's rated operating pressure, with the loading rate controlled within 0.1 MPa / s. Alternatively, mechanical loading can be used to induce a circumferential strain of 0.1% to 0.5% in the cylinder. For example, with low-pressure gas filling, an inert gas (such as nitrogen) is introduced into the cylinder through a precision pressure control system, and the pressure is slowly increased to 10% of the cylinder's rated operating pressure (e.g., for a 35 MPa hydrogen storage cylinder, loading to 3.5 MPa). During this process, a camera synchronously acquires a series of time-series images I(t) at a fixed frame rate (e.g., 5-10 fps). The loading module and the industrial camera are synchronously triggered by a central controller, ensuring that each load state has a corresponding image.

[0060] The digital speckle correlation algorithm employs the inverse combined Gauss-Newton algorithm, achieving sub-pixel accuracy. The sub-region size is 21×21 pixels, and the number of iterations is 20. The inverse combined Gauss-Newton algorithm is a high-precision sub-pixel displacement search algorithm. Unlike traditional forward algorithms, it iteratively optimizes the displacement field, offering advantages such as noise resistance and fast convergence. This algorithm was chosen to adapt to surface computation. The sub-region is the smallest unit for correlation calculation in the image. The sub-region size, measured in pixels, is used for grayscale matching during digital speckle calculation. The sub-region size affects the accuracy and noise resistance of the displacement calculation. Larger sub-regions improve stability but may reduce detail resolution; smaller sub-regions have the opposite effect. Setting the sub-region size to 21×21 pixels takes into account the texture features of the gas cylinder surface, ensuring the highest correlation coefficient under the gas cylinder speckle pattern. The number of iterations ensures that the displacement calculation converges to sub-pixel accuracy (e.g., 0.01 pixels). Too few iterations may lead to non-convergence, while too many increase computation time. Setting it to 20 iterations controls efficiency while maintaining accuracy, determined through convergence testing.

[0061] Surface coordinate transformation correction is based on the CAD model of the gas cylinder or stereo vision calibration data. A reverse-combined Gaussian-Newton algorithm is used to perform sub-pixel precision digital speckle analysis on the sequence of images to calculate the displacement fields u(x, y, t) and v(x, y, t). The displacement fields are then smoothed using Gaussian filtering, and surface coordinate transformation is performed based on the known CAD model of the gas cylinder or surface information (such as curvature) obtained through stereo vision calibration to calculate the true Green-Lagrange strain tensor component ε. xx、 ε yy and ε xy The radius of curvature R is the radius of curvature of the gas cylinder surface at a certain point, stored in each grid node using a triangular mesh model. The surface is divided into fine triangular meshes, with each node storing its local curvature value for coordinate transformation. In the displacement correction formula, the radius of curvature R is used to map the two-dimensional displacement in the image coordinate system to the three-dimensional displacement in the object coordinate system, eliminating projection errors caused by the surface.

[0062] It should be noted that when traditional digital speckle technology is directly applied to curved gas cylinders, the strain calculation error exceeds 15%, and the signal-to-noise ratio is insufficient for defects with strains below 50 microstrain. However, hydrogen storage cylinder defect detection corresponds to single-filament fiber breakage, requiring a sensitivity of 50 microstrain. The curved-surface adaptive digital speckle algorithm proposed in this invention aims to solve the measurement error problem caused by fiber entanglement on the curved surface structure of gas cylinders. Through coordinate transformation and curvature correction, traditional digital speckle technology is adapted to curved surface measurement, significantly improving the accuracy of displacement and strain calculations. The curved-surface adaptive digital speckle algorithm mainly includes two stages: first, obtaining the geometric information (such as curvature) of the gas cylinder surface through stereo vision calibration and CAD models; second, introducing a correction formula in the displacement calculation to eliminate errors caused by the curved surface. The accuracy of the algorithm highly depends on the optimized settings of multiple parameters, which together ensure the effective detection of defects at the microstrain (e.g., 50 microstrain) level. Parameter optimization is based on the curved surface characteristics of the gas cylinder and the digital speckle calculation principle, aiming to balance computational efficiency and accuracy. The surface of the gas cylinder is divided into a triangular mesh, and each mesh node stores the local radius of curvature R. The displacement field correction formula is as follows:

[0063]

[0064] In the formula, u is the displacement component of a point on the surface of the gas cylinder in the x-direction; Δu measured It is the x-direction displacement directly calculated by digital speckle; Δu corrected It is the corrected displacement in the x-direction; u / x is the rate of change of displacement along the x-direction; 2 z / x 2 The curvature term characterizes the local curvature of the gas cylinder surface in the x-direction. 2 z / x 2 The larger the value, the more drastic the curvature change, and the greater the correction amount for a standard cylindrical surface. 2 z / x 2 It is proportional to 1 / R; z(x,y) is a function of the surface height of the gas cylinder, obtained through stereo vision calibration.

[0065] Please see Figure 2 The image enhancement processing of the full-field strain distribution image set on the gas cylinder surface in step S106 above includes:

[0066] S201. The gain factor is adaptively adjusted based on the standard deviation of the strain values ​​across the entire field. The gain formula is α = k / σ, where σ is the standard deviation and k is a constant. A linear gain is applied to the original strain field ε(x, y): ε'(x, y) = α * ε(x, y). The gain factor α is dynamically adjusted based on the standard deviation of the strain values ​​across the entire field. For example, the smaller the standard deviation, the larger the value of α (e.g., 10-50 times) to amplify minute strain changes.

[0067] S202. The guided filtering algorithm is used to decompose the full-field strain distribution map into a base layer and a detail layer, with a filtering radius parameter of 15-25 pixels. The guided filtering algorithm is then used to decompose the amplified image I into a base layer B (reflecting macroscopic deformation) and a detail layer D (containing local abrupt changes caused by defects).

[0068] S203. Apply an enhancement factor γ to the image detail layer, where γ ranges from 2.0 to 5.0. Multiply the detail layer by the enhancement factor γ: D' = γ * D, where γ ranges from 2.0 to 5.0, for example, 2.0, 3.0, 4.0, or 5.0. Finally, merge the magnified detail layer with the base layer: I' = B + D'.

[0069] S204. Enhance image contrast using histogram equalization algorithm. Normalize the grayscale values ​​of the enhanced image I' to the range of [0, 255] and perform histogram equalization to obtain a high-contrast strain field visualization image for defect identification.

[0070] On the enhanced strain field visualization image, abnormal regions with strain exceeding a threshold are automatically identified. Quantitative feature parameters include local strain concentration factor, gradient magnitude entropy, abnormal region area, and shape factor. The machine learning model is a support vector machine or random forest classifier, trained on samples of fiber fracture, delamination, and resin cracking defects. For each abnormal region, its local strain concentration factor (maximum strain in the region / average strain of the surrounding background), maximum strain gradient, and other features are calculated. These feature vectors are then input into a pre-trained support vector machine (SVM) classification model. This model is trained on a large dataset containing known defect types (fiber fracture, delamination, resin cracking) and severity levels, and can output defect type and grade assessment results. Finally, the system generates an inspection report containing the defect location, type, size, and grade, with the defect location marked on the 3D model of the gas cylinder.

[0071] It should be noted that machine learning classification models include:

[0072] Feature library: Contains 12-dimensional features including strain concentration factor, gradient magnitude entropy, and morphological factor, which are reduced to 5 dimensions through PCA principal component analysis. Among them, strain concentration factor is the ratio of strain in the defect region to the background; gradient magnitude entropy characterizes the complexity of the defect edges; morphological factor represents the compactness of the defect, such as... MF = A / P 2 In the formula, A is the area and P is the perimeter.

[0073] Training data: Contains 500 defect samples (200 fiber fractures, 150 delaminations, and 150 cracks), and cross-validation (k=10) is used to evaluate model performance.

[0074] Please see Figure 4 The detailed steps of the fiber entanglement gas cylinder defect detection method based on digital speckle provided in this embodiment of the invention are as follows:

[0075] (1) Create speckle patterns and collect original speckle images.

[0076] Random black and white speckle patterns are uniformly sprayed onto the outer surface of the fiber-wound gas cylinder, and the original image I0(x, y) of the gas cylinder in an unloaded state is captured using an industrial camera.

[0077] (2) Apply a controllable load and acquire time series images.

[0078] Low-pressure inflation or mechanical loading is used to apply load to the gas cylinder, and time series image set I(t) is acquired during the loading process. (3) Calculation of displacement difference

[0079] By matching the grayscale changes of each pixel using a digital speckle correlation algorithm, a displacement difference sequence is obtained: Δu(x,y,t) = u t (x,y) - u0(x,y) .

[0080] Among them, u t (x,y) represents the displacement of the pixel at time t, and u0(x,y) is the initial reference state displacement.

[0081] The displacement field correction formula described above is used to perform surface coordinate transformation correction on the displacement difference sequence data to eliminate curvature error, thereby obtaining the corrected full-field displacement field data of the gas cylinder surface.

[0082] (4) Strain field calculation

[0083] Based on the corrected full-field displacement data of the gas cylinder surface, the strain tensor components are calculated using the finite difference method;

[0084] ε xx = u / x, εyy = v / y ,

[0085] ε xy = 1 / 2( u / y + v / x).

[0086] (5) Enhanced dynamic range of strain field

[0087] Apply a linear gain to the strain field image set: ε'(x,y,t) = α·ε(x,y,t);

[0088] In the formula, α is the gain factor, used to amplify local strain changes.

[0089] (6) Image decomposition and enhancement

[0090] The enhanced strain image is decomposed into a base layer B(x,y) and a detail layer D(x,y) using a filtering algorithm: I(x,y) = B(x,y) + D(x,y).

[0091] An enhancement factor γ is introduced to amplify the detail layer: D'(x,y) = γ·D(x,y).

[0092] The final enhanced image is: I'(x,y) = B(x,y) + D'(x,y).

[0093] (7) Normalization and histogram equilibrium

[0094] The grayscale range of the enhanced strain image is normalized to [0, 255], and the contrast is improved by using a histogram equalization algorithm to obtain the final strain field image.

[0095] (8) Defect identification and classification

[0096] By extracting strain anomaly features, such as principal strain concentration factors and gradient outliers, machine learning models are used to automatically classify and assess the types of defects in gas cylinders, and automatically identify defects such as fiber breakage, delamination, or resin cracking.

[0097] Please see Figure 3This embodiment also provides a digital speckle-based fiber-wound gas cylinder defect detection system to implement the aforementioned digital speckle-based fiber-wound gas cylinder defect detection method. The system includes a speckle fabrication module 100, a loading module 200, an image acquisition module 300, and a data processing module 400. The speckle fabrication module 100 is used to prepare a speckle pattern that meets the requirements on the gas cylinder surface; the loading module 200 is used to apply precisely controlled low-pressure inflation or mechanical load to the gas cylinder; the image acquisition module 300 is used to acquire image sequences of the gas cylinder surface before and after the load is applied; and the data processing module 400 is used to process displacement field calculations, strain field calculations, image enhancement, and defect identification.

[0098] The speckle pattern creation module 100 employs an automated spraying robot or a handheld precision spray gun to ensure uniform speckle distribution. The loading module 200 includes a precision gas source, pressure regulating valve, safety relief valve, and pressure sensor for precise low-pressure inflation loading; it can also be equipped with a mechanical loading mechanism, such as an actuator, for special inspection scenarios. The image acquisition module 300 consists of a camera array composed of multiple industrial cameras, coupled with LED diffused light sources, ensuring uniform illumination and no shadows across the entire gas cylinder surface; multiple industrial cameras undergo high-precision stereo calibration to cover the entire gas cylinder and achieve 3D deformation measurement; the industrial camera resolution is no less than 12 megapixels, and the frame rate is adjustable from 5-20fps. The data processing module 400 is the core computing unit, a high-performance workstation or industrial computer with built-in dedicated software integrating a digital speckle calculation engine, image enhancement algorithm library, and machine learning model. It can automatically execute all the above processing steps and provides a user-friendly interface for parameter setting and result visualization. The data processing module 400 also includes a result output unit for generating 3D defect distribution maps and inspection reports.

[0099] Please see Figure 5 and Figure 6 , Figure 5 This is a speckle pattern on the surface of the normal area of ​​the gas cylinder. Figure 6 It is a speckle pattern on the surface of the defect area of ​​the gas cylinder. Figure 6 This visually demonstrates that under uniform loading, strain concentration occurs in the defect region due to stiffness differences. This is achieved by introducing a curvature term, which characterizes the local geometry of the surface. 2 z / x 2 and the displacement gradient term characterizing the degree of deformation ( u / x) 2Displacement field correction effectively eliminates systematic errors introduced by surface effects, ensuring that strain concentration phenomena can be accurately measured rather than being masked by surface errors. Surface coordinate transformation correction of the full-field displacement data on the gas cylinder surface is key to achieving high-precision surface measurement, enabling digital speckle technology to accurately detect minute defects on fiber-wound gas cylinders.

[0100] The embodiments of the present invention have the following beneficial effects:

[0101] 1. A leap in detection sensitivity and reliability: Through the synergistic effect of load excitation and image enhancement, it can effectively detect micron-level strain changes and identify defects at the micro-strain level of 50 microstrain. The sensitivity is 4 times higher than that of ultrasonic technology, thus discovering tiny and deep defects that are difficult to detect by traditional ultrasound and X-rays, and greatly reducing the false negative rate.

[0102] 2. Significantly improved detection efficiency: The system uses a camera array for single imaging or rapid scanning, replacing point-scan ultrasonic detection, to achieve rapid full-field measurement of gas cylinders. The fully automated process reduces the detection time from 30 minutes to 8 minutes, making it suitable for rapid detection on production lines.

[0103] 3. Objectivity and visualization of test results: Provides full-field, color strain distribution cloud map, making the location, size and relative severity of defects clear at a glance; combined with quantitative features and machine learning, it reduces human interference and makes the judgment results more objective and traceable.

[0104] 4. Applicability throughout the entire life cycle: This testing method is a non-destructive testing method that uses coordinate transformation algorithms to solve surface measurement errors and is compatible with various specifications of gas cylinders. It is safe to use for gas cylinders throughout their entire life cycle, from factory inspection and periodic calibration to end-of-life assessment, providing continuous data support for the safe use of gas cylinders.

[0105] 5. System integration and intelligence: It integrates optical measurement, precision control, image processing and artificial intelligence technologies to form an integrated solution, which lowers the operation threshold, the equipment investment is 1 / 3 of that of an X-ray inspection system, and there is no need for consumables, resulting in low inspection costs and improved repeatability and consistency of inspection.

[0106] The digital speckle-based defect detection method and system for fiber-wound gas cylinders provided in this embodiment can be widely used in all stages of the entire life cycle of fiber-wound gas cylinders, including manufacturing quality inspection, damage assessment after type testing, and in-service periodic inspection. It is particularly suitable for the safety inspection of high-pressure containers such as hydrogen storage cylinders for hydrogen fuel cell vehicles, breathing gas cylinders, and CNG cylinders, and has enormous market application prospects and industrial value.

[0107] Please see Figure 7This application also provides a computer device 600, which includes a memory 601 and a processor 602. The processor 602 is used to execute computer program instructions stored in the memory 601 to implement a method for detecting defects in fiber-wound gas cylinders based on digital speckle.

[0108] The memory 601 includes at least one type of storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 601 can be an internal storage unit of a computer device, such as a hard disk. In other embodiments, the memory 601 can be an external storage device of a computer device, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. The memory 601 can also include both internal and external storage units of a computer device. The memory 601 can be used not only to store application software and various types of data installed on the computer device, but also to temporarily store data that has been output or will be output.

[0109] Computer device 600 also includes bus 603. Bus 603 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0110] Computer device 600 may also include a display component 604. The display component 604 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display component 604 may also be appropriately referred to as a display device or display unit, used to display information processed in computer device 600 and to display a visual user interface.

[0111] Computer device 600 may also include communication component 605. Communication component 605 may optionally include wired communication component and / or wireless communication component (such as Wi-Fi communication component, Bluetooth communication component, etc.), and is typically used to establish communication connections between computer device 600 and other computer devices.

[0112] Figure 7 Only a computer device 600 with some components is shown; those skilled in the art will understand that... Figure 7 The structure shown does not constitute a limitation on the computer device 600, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0113] This application also provides a storage medium that, when executed by a computer processor, enables the computer to perform the fiber-wound gas cylinder defect detection method based on digital speckle provided in the above embodiments. For example, the storage medium can be a ROM, RAM, CD-ROM, magnetic tape, floppy disk, USB flash drive, or optical data storage device. It is worth noting that the storage medium mentioned in this application embodiment can be a non-volatile storage medium or a non-transient storage medium.

[0114] It should be understood that all or part of the steps of the above embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented in whole or in part as a computer program product. A computer program product includes one or more computer instructions; the computer instructions can be stored in the storage medium described above. That is, in some embodiments, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute the fiber-wound gas cylinder defect detection method based on digital speckle provided in the above embodiments.

[0115] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A digital speckle based method for defect detection of filament wound gas cylinders, characterized in that, The method comprises the following steps: Making a high-contrast speckle pattern on the outer surface of the gas cylinder; Applying a controllable load to the gas cylinder and collecting a sequence of gas cylinder surface images before and after the load is applied; Based on the sequence of gas cylinder surface images, the full-field displacement field data of the gas cylinder surface is calculated by a digital speckle correlation method; The full-field displacement field data of the gas cylinder surface is corrected by surface coordinate transformation to eliminate curvature error; According to the corrected full-field displacement field data of the gas cylinder surface, the full-field strain distribution image set of the gas cylinder surface is calculated by a difference method; The full-field strain distribution image set of the gas cylinder surface is processed by image enhancement, including dynamic range gain, image decomposition and detail enhancement, normalization and histogram equalization, to amplify the micro-strain signal caused by the defects of the gas cylinder; Based on the enhanced full-field strain distribution image set of the gas cylinder surface, the quantitative feature parameters are extracted, and a machine learning model is used to automatically identify and classify the defects of the gas cylinder; The digital speckle correlation method uses a reverse combination Gauss-Newton algorithm, and the quantitative feature parameters include local strain concentration factor, gradient amplitude entropy, abnormal area and shape factor; The surface of the gas cylinder is divided into triangular meshes based on the CAD model of the gas cylinder and the stereo vision calibration data, and each mesh node stores the local curvature radius R. The displacement field correction formula is: where u is the displacement component of a certain point on the surface of the cylinder in the x direction; Δu measured is the x direction displacement directly calculated by digital speckle; Δu corrected is the corrected x direction displacement; u / x is the rate of change of displacement along the x direction; 2 z / x 2 is the curvature term, representing the local curvature of the surface of the cylinder in the x direction, 2 z / x 2 , the greater the curvature change, the greater the correction, for the standard cylindrical surface 2 z / x 2 is proportional to 1 / R; z(x, y) is the height function of the surface of the cylinder, which is obtained by stereo vision calibration.

2. The digital speckle based fiber-wound gas cylinder defect detection method according to claim 1, characterized by, The speckle size of the speckle pattern is 3-8 pixels, the distribution density is greater than 200 points / cm², the local density variation coefficient is less than 0.1, the speckle pattern is made of high-contrast matte paint, and the average gray gradient of the speckle pattern is greater than 50 to ensure the calculation accuracy of the digital speckle.

3. The digital speckle based fiber-wound gas cylinder defect detection method according to claim 2, characterized by, The controllable load is low-pressure inflation loading or mechanical loading. The pressure value of the low-pressure inflation loading is 5% to 15% of the rated working pressure of the gas cylinder, and the loading rate is controlled to be less than 0.1 MPa / s; or mechanical loading is used to make the gas cylinder produce 0.1% to 0.5% circumferential strain.

4. The digital speckle based fiber-wound gas cylinder defect detection method according to claim 1, characterized by, The accuracy of the reverse combination Gauss-Newton algorithm is sub-pixel accuracy, the sub-area size is 21×21 pixels, and the iteration number is 20.

5. The digital speckle based fiber-wound gas cylinder defect detection method according to claim 1, characterized by, The image enhancement processing of the full-field strain distribution image set of the gas cylinder surface comprises: Adaptively adjusting the gain factor according to the standard deviation of the full-field strain value, and the gain formula is α = k / σ, where σ is the standard deviation and k is a constant; Using a guided filtering algorithm to decompose the full-field strain distribution image into an image basic layer and an image detail layer, and the filtering radius parameter is 15-25 pixels; Applying an enhancement coefficient γ to the image detail layer, and γ is in the range of 2.0-5.0; Enhancing the image contrast by histogram equalization algorithm.

6. The digital speckle based fiber-wound gas cylinder defect detection method according to claim 1, characterized by, The machine learning model is a support vector machine or a random forest classifier, and the training data contains fiber fracture, delamination and resin cracking defect samples.

7. A digital speckle based fiber-wound gas cylinder defect detection system, characterized by, A digital speckle-based fiber-wound gas cylinder defect detection method according to any one of claims 1 to 6, comprising: A speckle making module for preparing a required speckle pattern on the surface of the gas cylinder; A loading module for applying a precisely controlled low-pressure inflation or mechanical load to the gas cylinder; An image acquisition module for acquiring a sequence of gas cylinder surface images before and after the load is applied; The data processing module is used for processing displacement field calculation, strain field calculation, image enhancement and defect identification.

8. The digital speckle based fiber-wound gas cylinder defect detection system according to claim 7, characterized in that, The loading module comprises a precision gas source, a pressure regulating valve, a safety pressure relief valve and a pressure sensor, and is used for realizing accurate low-pressure inflation loading; the image acquisition module comprises a plurality of industrial cameras and illumination light sources, the plurality of industrial cameras form a camera array, and are calibrated in three dimensions to realize panoramic coverage of the gas cylinder; the data processing module further comprises a result output unit, which is used for generating a three-dimensional defect distribution map and a detection report.

Images