Defect detection method and system for metal-composite material mixed structure device

Abstract

The invention discloses a defect detection system for a metal-composite material mixed structure device. The defect detection system comprises an industrial control computer, a mechanical scanning arm, a polarization detection probe, an ultrasonic detector assembly and a power supply, the invention further provides a defect detection method for the metal-composite material mixed structure device. The defect detection method specifically comprises the following steps: acquiring polarization image information of the metal surface; solving polarization image parameter information; carrying out surface defect detection; metal internal ultrasonic detection information is obtained; performing internal defect detection; and performing multi-mode data fusion detection. According to the invention, two information sources with complementary advantages, namely polarization and ultrasound, are selected as detection information sources of the multi-modal detection method, so that all-directional defect detection and visualization on the surface-interior of the metal-based composite material mixed structure device are effectively realized; and a feasible technical method is provided for engineering application of polarized light / phased array ultrasonic multichannel integrated detection.

Description

Technical Field

[0001] The present invention relates to the technical field of defect detection, and in particular to a defect detection method and system for a metal-composite material hybrid structure device. Background Art

[0002] Metal-composite hybrid structural devices are composed of a metal or alloy matrix, artificially bonded to one or more reinforcements (such as fibers, whiskers, and particles). The metal provides structural strength, stability, and mechanical properties on the surface, while the composite material enhances specific internal properties. These hybrid structural devices, combining the advantages of metal and composite materials, are widely used in the manufacture of high-performance structural components in aerospace, automotive, advanced weapon systems, and other fields due to their high strength and lightweight properties.

[0003] Currently, commonly used defect detection methods for structural devices can be categorized into five main categories: acoustic, optical, electromagnetic, radiographic, and infrared. In reality, while traditional defect detection methods vary and focus on specific areas, defects in metal-composite hybrid devices are diverse, some of which can be highly complex. It's difficult to find a method that is both suitable for metals and composite materials like ceramics and fibers, while also meeting the requirements for rapid detection and high sensitivity. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the present invention provides a metal-composite material hybrid structure device defect detection system, which includes an industrial control computer, a mechanical scanning arm, a polarization detection probe, an ultrasonic detector component, and a power supply;

[0005] 1) Industrial control computers

[0006] The industrial control computer uses the Adantech IPC-610 model, which is connected to the mechanical scanning arm, polarization detection probe, and ultrasonic detector via data cables. It is responsible for transmitting detection instructions and detection information. At the same time, the built-in high-performance data processing unit and industrial-grade image processing software are responsible for processing and analyzing defect information.

[0007] 2) Mechanical scanning arm

[0008] A mechanical scanning arm (hereinafter referred to as the "robot arm") has a polarization detection probe fixed at its end. The position of the robot arm is adjusted by instructions from an industrial control computer to change the working distance and scanning path of the polarization detection probe.

[0009] 3) Polarization detection probe

[0010] The polarization detection probe includes a light source, a lens, a polarization camera, and a light source driver, as follows:

[0011] (1) Combined light source

[0012] The combined light source is a combination of a circularly polarized light source and four angle-adjustable strip light sources. The circularly polarized light source emits linearly polarized light directly toward the surface to be measured. The four angle-adjustable strip light sources are approximately located in the upper, lower, left, and right directions of the circularly polarized light source and are rotatably fixed to the approximate center of the four edges of the front end of the housing. They can be rotated around the corresponding edges to adjust the light emission angle. The circularly polarized light source is used as the directly incident light source, and the obliquely incident strip light sources are used to supplement the brightness.

[0013] (2) Lens

[0014] The lens is located behind and adjacent to the circularly polarized light source, the center of the circularly polarized light source is on the optical axis of the lens, and the circularly polarized light source is arranged perpendicular to the optical axis;

[0015] (3) Polarization camera

[0016] The polarization camera is located behind the lens, and the optical axis of the camera coincides with the optical axis of the lens. The polarization camera is used to capture images and generate image information at four different polarization angles: 0°, 45°, 90°, and 135°.

[0017] (4) Light source driver

[0018] The light source driver provides drive for the combined light source;

[0019] (5) Support shell

[0020] The supporting shell is used to accommodate the lens, polarization camera, and light source driver, and the combined light source is arranged at the front end of the supporting shell;

[0021] 4) Ultrasonic detector components

[0022] The ultrasonic detector assembly includes an ultrasonic detector, a phased array scanner, and an encoder; the ultrasonic detector assembly is placed at an appropriate position around the robotic arm;

[0023] 5) Power supply

[0024] A power supply provides power to the system.

[0025] In a specific embodiment of the present invention,

[0026] 1) The industrial control computer adopts Adantech IPC-610 model;

[0027] 2) The mechanical scanning arm uses the lightweight 6-DOF robotic arm AUBO-i10;

[0028] 3) In the polarization detection probe,

[0029] (1) Combined light source

[0030] The circular polarized light source model is CST-POR10090-w, and the bar light source model is CST-ROS140-w; all four bar light sources emit white natural light toward the surface to be measured;

[0031] (2) Lens

[0032] Use JC-0.4X10Y liquid focus lens as the electric focus telecentric lens;

[0033] (3) Polarization camera

[0034] A BFS-U3-51S5PC polarization camera was used for image acquisition;

[0035] (4) Light source driver

[0036] The light source driver adopts CST-DPS20-CM-TD model;

[0037] (5) Support shell

[0038] The supporting shell was designed using Solidwork software, and a lightweight shell model was produced using 3D printing technology;

[0039] 4) Ultrasonic detector

[0040] The ultrasonic detector uses the PHASCAN II portable ultrasonic phased array detector, which is equipped with a 5S64 wheel-type phased array scanner and a Lemo encoder.

[0041] A method for detecting defects in a metal-composite hybrid structure device is also provided. The method is based on the above-mentioned metal-composite hybrid structure device defect detection system. The method is specifically as follows:

[0042] Step 1: Obtain polarization image information of the metal surface;

[0043] The industrial control computer sends an image acquisition command to the robotic arm, and the robotic arm follows a predetermined path to reach the designated inspection point. After reaching the designated point, the combined light source is turned on to illuminate the component to be inspected. The component to be inspected generates reflected light, which is converged by the telecentric lens to form a fixed field of view and projected to the polarization camera, generating four images with different polarization angles of 0°, 45°, 90°, and 135°. The image in each polarization direction provides the intensity information I0, I1 of the polarized light in the four different directions. 45 , I 90 and I 135 ;

[0044] Step 2: Calculate polarization image parameter information;

[0045] After all points are captured, the polarization images are transmitted to an industrial control computer. The polarization parameter information of the target, including the degree of polarization (DoP), angle of polarization (AoP), and three Stokes vectors (S0, S1, and S2), is calculated using the following formulas, effectively improving the characteristic contrast of defects on the metal surface.

[0046] S0=I0+I 90 ,S1=I0-I 90 ,S2=I 45 -I 135

[0047]

[0048] Step 3: Perform surface defect detection;

[0049] The solved polarization parameter information is used to detect defects on the surface of metal-composite hybrid structure devices through a metal industrial parts inspection algorithm based on deep learning polarization image fusion to obtain surface defect detection results.

[0050] Step 4: Obtaining ultrasonic testing information inside the metal;

[0051] Connect the ultrasonic signal cable of the wheeled phased array scanner to the probe interface of the ultrasonic phased array detector, connect the encoding cable of the Lemo encoder to the encoder interface of the wheeled phased array scanner, and connect the other end of the Lemo encoder encoding cable to the encoder input interface of the ultrasonic phased array detector. Turn on the ultrasonic phased array detector, select the scanning imaging mode, and adjust the detection parameters. Start the ultrasonic phased array detector, apply ultrasonic coupling agent or water as the liquid conductive medium to the structure to be inspected, and use the wheeled phased array scanner to scan the inspection object along a fixed trajectory. During the scanning process, a scanning image is generated on the detector.

[0052] Step 5: Conduct internal defect detection;

[0053] The generated scanned image is transmitted to the industrial control computer for grayscale processing to retain key information and reduce unnecessary color interference. The industrial control computer removes noise and resizes the scanned image to optimize image quality. The industrial control computer uses the U-Net model to perform pixel-level segmentation on the processed image to obtain a binary defect mask, extract the defect geometric features and fit an ellipse. Combined with SIFT feature detection, FLANN matching and RANSAC algorithm, image registration is performed to determine the actual location of the defect, completing internal defect detection, quantification and positioning.

[0054] Step 6: Multi-modal data fusion detection;

[0055] The specific steps include:

[0056] (01) Data conversion and preprocessing;

[0057] Data conversion and preprocessing are the starting point of the entire fusion process. They are responsible for unified processing of data from different detection sensors to ensure data consistency and comparability. The specific operations are as follows:

[0058] a) Data format unification: Different sensors generate data in different formats, and these data need to be converted into a unified format;

[0059] b) Data denoising: Remove noise from data through filtering algorithms to improve data quality;

[0060] c) Data calibration: Calibrate the data of different sensors to ensure that their measurement results on the same physical quantity are comparable;

[0061] d) Data standardization: standardize the numerical ranges of different data sources so that they can be compared and integrated on the same scale;

[0062] e) Data enhancement: Image enhancement technology is used to improve image quality, enhance defect visibility, and assist in subsequent defect detection;

[0063] (02) Coordinate transformation

[0064] Coordinate transformation is responsible for transforming the defect information from different data sources to ensure that the defect locations from different data sources are matched in the same coordinate system; the details are as follows:

[0065] Coordinate system 1: Convert the defect information obtained by the polarization camera and ultrasonic scanner from their respective coordinate systems to a unified robotic arm base coordinate system. This is achieved by the following steps:

[0066] Pixel coordinate system → camera coordinate system: Use the polarization camera intrinsic parameter matrix to convert the pixel coordinates (u, v) into a three-dimensional point (X c ,Y c ,Z c ), the formula is:

[0067]

[0068] Among them, f u ,f v is the equivalent focal length of the camera in the horizontal (u direction) and vertical (v direction), (u0, v0) is the coordinate of the principal point on the camera imaging plane, which refers to the coordinate position of the intersection of the camera optical axis and the imaging plane in the pixel coordinate system.

[0069] Camera coordinate system → Robotic arm base coordinate system: Obtain the extrinsic parameter matrix of the camera and the robotic arm base through hand-eye calibration, including the rotation matrix R and translation vector T, to achieve coordinate transformation:

[0070]

[0071] Among them, (X b ,Y b ,Z b ) is the spatial coordinate of the defect information in the robot arm base coordinate system.

[0072] Ultrasonic scanner coordinate system → robotic arm base coordinate system: The position of the ultrasonic scanner is determined by the tool center point TCP coordinate system at the end of the robotic arm. The defect position detected by ultrasonic inspection is converted from the TCP coordinate system to the base coordinate system through the forward solution of the robotic arm kinematics.

[0073] When detecting a small area, affine transformation is used, and when dealing with large-scale viewing angle changes, perspective transformation is used;

[0074] (03) Defect Registration

[0075] The specific operations are as follows:

[0076] A) Image Registration: Use image processing technology to register images from different data sources to ensure that the location of defects in different images is consistent;

[0077] B) Defect feature extraction: Use image processing and machine learning algorithms to extract defect features, including shape, size, and location information; ensuring that defects from different data sources can be matched based on features;

[0078] C) Defect matching: Based on the extracted features, defects from different data sources are matched to determine whether they are different manifestations of the same defect. This ensures that defects from different data sources can be accurately registered.

[0079] D) Defect classification: Classify defects according to their characteristics and distinguish different types of defects;

[0080] (04) Defect fusion and post-processing

[0081] The specific operations are as follows:

[0082] I) DS Evidence Theory Fusion: DS evidence theory is used for information-level fusion, using the defect area as a credibility indicator. By combining defect information from different data sources, the comprehensive defect credibility is calculated to help determine the authenticity and severity of the defect.

[0083] II) Result output: The fused result output includes the location, type, and size information of the defect.

[0084] In the coordinate conversion (02) of one embodiment of the present invention, error compensation is also required, and the following method is used to improve the conversion accuracy:

[0085] Calibration of calibration parts: Using a high-precision calibration plate, data is collected at multiple locations to fit the transformation parameters between the sensor coordinate system and the robot base coordinate system; the singular value decomposition (SVD) method is used to solve the least squares optimal transformation matrix;

[0086] Dynamic Error Correction: Utilizes feedback from the robotic arm encoder to compensate for positional deviations during motion in real time. For ultrasonic scanners, couplant uniformity testing and scanning speed control are used to reduce detection errors caused by contact position variations. The error compensation algorithm follows industry-standard procedures.

[0087] In another embodiment of the present invention (03) defect registration C) defect matching, when defects from different data sources overlap in position in a unified coordinate system, have high similarity in feature parameters and conform to physical mechanisms, they are determined to be different manifestations of the same defect; otherwise, due to position deviation, feature contradiction or physical logic inconsistency, they are determined to be independent defects or noise artifacts; the matching process includes matching of feature points and calculation of shape similarity.

[0088] In another embodiment of the present invention, (04) defect fusion and post-processing, III) post-processing is also required: post-processing of the fusion results, including defect annotation, unified output format, and deflection angle calculation, to provide support for 3D visualization.

[0089] In (05) three-dimensional visualization of another embodiment of the present invention, the OpenGL three-dimensional engine technology is used to render the three-dimensional model of the metal-composite material structure device, and the 3D display engine is used to draw the defect shape into a three-dimensional shape and superimpose it with the three-dimensional structure model of the metal-composite material structure device in the 3D visualization area, and the visualization display of all detected defects is completed by highlighting.

[0090] The advantages of the present invention are as follows:

[0091] 1. The present invention selects two complementary signal sources - polarization and ultrasound - as the detection signal sources of the multimodal detection method, effectively realizing all-round defect detection and visualization of the surface and interior of metal-based composite hybrid structure devices, providing a more efficient and reliable solution for industrial detection, with certain practical value and promotion and application prospects.

[0092] 2. This paper proposes a multimodal inspection information fusion and visualization technology based on DS evidence theory, integrating the actual defect locations obtained using polarization and phased array ultrasound at the information level. This provides a feasible technical approach for the engineering application of polarization / phased array ultrasound multi-channel integrated inspection, filling a gap in the research on the application of multi-channel integrated automatic inspection using multiple inspection technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 Shows a schematic diagram of the working of the mechanical scanning arm;

[0094] Figure 2 Shows the composition and operation scenario of the polarization detection probe;

[0095] Figure 3 A schematic diagram of a polarization detection probe housing model is shown;

[0096] Figure 4 A schematic diagram of light source illumination is shown;

[0097] Figure 5 Shows a schematic diagram of the working of the ultrasonic detector; DETAILED DESCRIPTION

[0098] The present invention will be described in detail below with reference to the accompanying drawings.

[0099] 1. A defect detection system for metal-composite hybrid structure devices

[0100] The present invention provides a metal-composite material hybrid structure device defect detection system, which mainly consists of five parts: an industrial control computer, a mechanical scanning arm, a polarization detection probe, an ultrasonic detector, and a power supply.

[0101] 1. Industrial control computers

[0102] The industrial control computer, an Adantech IPC-610, connects to the mechanical scanning arm, polarization detection probe, and ultrasonic detector via data cables, responsible for transmitting inspection instructions and information. A built-in high-performance data processing unit and industrial-grade image processing software are also responsible for processing and analyzing defect information.

[0103] 2. Mechanical scanning arm

[0104] The robotic scanning arm (hereinafter referred to as the "robotic arm") uses a lightweight 6-DOF robotic arm AUBO-i10 (developed by AUBO Robotics). The end of the robotic arm is fixed with a polarization detection probe. The position of the robotic arm is adjusted by industrial control computer instructions, and the working distance and scanning path of the polarization detection probe are changed to complete the surface defect detection of complex metal parts, such as Figure 1 shown.

[0105] In terms of scanning path planning, the present invention adopts a quadratic approximation path planning algorithm and uses RobotArt software (well known in the industry) for offline programming to achieve full automation of polarization detection.

[0106] 3. Polarization detection probe

[0107] Polarization detection probes are designed to address the difficulty in detecting complex metal parts due to their structural characteristics. They utilize two-dimensional polarization imaging technology to detect defects. Polarization detection probes are primarily composed of a light source, lens, polarization camera, and light source driver. A schematic diagram of their composition and operation is shown below. Figure 2 shown.

[0108] (1) Combined light source

[0109] The combined light source uses a circular polarized light source of model CST-POR10090-w (manufactured by Kangshida Automation Technology Co., Ltd.) and a combination of four angle-adjustable strip light sources of model CST-ROS140-w (manufactured by Kangshida Automation Technology Co., Ltd.) (color temperature 6000-6500K) for illumination. Figure 4 As shown, a circularly polarized light source emits linearly polarized light directly toward the surface being measured. Four angle-adjustable bar light sources are positioned approximately in the center of the four front edges of the circularly polarized light source, rotating about their respective edges to adjust the light emission angle. Each of the four bar light sources emits white natural light toward the surface being measured. Using the circularly polarized light source as the direct-incidence light source and the oblique-incidence bar light sources as supplemental brightness, this multi-angle combined light source optimizes illumination, achieving uniform brightness and rich polarization information for complex image structures.

[0110] (2) Lens

[0111] The lens is located behind and adjacent to the circularly polarized light source. The center of the circularly polarized light source is on the optical axis of the lens, and the circularly polarized light source is arranged perpendicular to the optical axis as a whole.

[0112] Due to its special optical design, the telecentric lens can ensure that the size of objects in each area of the image remains unchanged, especially objects at the edge of the image, and can provide measurement accuracy consistent with the center area, which is crucial for the quantitative detection of complex metal surface defects. At the same time, when performing defect detection on complex metal surfaces, the surface usually has different reflection and refraction characteristics, and ordinary lenses are easily interfered by these factors. The telecentric lens can better avoid optical distortion at different angles and distances, maintain consistent image quality, and ensure the accuracy of the detection results. Therefore, the present invention adopts the JC-0.4X10Y model liquid focusing lens as an electric focus telecentric lens. The maximum focusing range of this lens can reach 20 to 170 mm, and the lens magnification is 0.4, which prevents fluctuations in the field of view caused by changes in object distance. The telecentricity is less than 0.03° and the distortion is less than 0.08%, which ensures image quality while ensuring a wide range of focus shooting.

[0113] (3) Polarization camera

[0114] The polarization camera is located behind the lens, with the camera's optical axis coinciding with the lens' optical axis. The present invention uses a BFS-U3-51S5PC polarization camera (manufactured by FLIR) for image acquisition, generating image information at four different polarization angles: 0°, 45°, 90°, and 135° (this is well known to those skilled in the art).

[0115] (4) Light source driver

[0116] The light source driver uses the CST-DPS20-CM-TD model (manufactured by Kangshida Automation Technology Co., Ltd.), with dimensions of 116×114×164mm.

[0117] (4) Support shell

[0118] The supporting shell is designed using Solidwork software (well known to industry personnel), and 3D printing technology is used to produce a lightweight shell model. Figure 3 As shown, the shell is 246.5mm long, 142.8mm wide, and 138mm high. This structure makes the system flexible to adjust, easy to integrate, and highly adaptable. Figure 3 The two recessed rectangular bodies on the upper surface of the middle back are for easy hand-held installation on the robotic arm, and the yellow base is for installation and fixation on the robotic arm. These two are not necessary conditions for the present invention, but can improve the performance of the present invention.

[0119] 4. Ultrasonic detector

[0120] The ultrasonic detector (the position of the ultrasonic detector is not fixed and can be placed at any appropriate position around the robotic arm) uses a PHASCAN II portable ultrasonic phased array detector (produced by Guangzhou Dopule Electronic Technology Co., Ltd.), which is equipped with a 5S64 wheel-type phased array scanner and a Lemo encoder (this combination is well known to those skilled in the art) to realize defect detection inside metal-composite hybrid structure devices.

[0121] 2. A defect detection method for metal-composite hybrid structure devices

[0122] The method uses the above-mentioned metal-composite material hybrid structure device defect detection system for detection, and the method is specifically as follows:

[0123] Step 1: Obtain polarization image information of the metal surface;

[0124] The industrial control computer sends an image acquisition command to the robotic arm, and the robotic arm follows a predetermined path to reach the designated inspection point. After reaching the designated point, the combined light source is turned on to illuminate the part to be inspected, such as Figure 4 As shown in the figure, the part to be inspected generates reflected light, which is converged by the telecentric lens to form a fixed field of view and projected to the polarization camera, generating four images with different polarization angles of 0°, 45°, 90° and 135°. The image in each polarization direction provides the intensity information I0, I 45 , I 90 and I 135 .

[0125] Step 2: Calculate polarization image parameter information;

[0126] After image acquisition at all points is completed, the collected polarization images are transmitted to the industrial control computer. The polarization parameter information of the measured target, such as the degree of polarization (DoP), angle of polarization (AoP), and three Stokes vectors (S0, S1, and S2), is solved using the following formula, effectively improving the characteristic contrast of defects on the metal surface.

[0127] S0=I0+I 90 ,S1=I0-I 90 ,S2=I 45 -I 135

[0128]

[0129] Step 3: Perform surface defect detection;

[0130] The polarization parameter information obtained is used to detect defects on the surface of metal-composite hybrid structure devices through a metal industrial parts detection algorithm based on deep learning polarization image fusion (Reference: [1] Meng Jianwen. Metal industrial parts detection based on deep learning polarization image fusion [D]. Northwest Agriculture and Forestry University, 2023. DOI: 10.27409 / d.cnki.gxbnu.2023.001828.), and the surface defect detection results are obtained.

[0131] Step 4: Obtaining ultrasonic testing information inside the metal;

[0132] Connect the ultrasonic signal cable from a 5S64 wheeled phased array scanner (a component of the ultrasonic detector) to the probe port of a PHASCAN II portable ultrasonic phased array detector (a component of the ultrasonic detector). Connect the encoder cable from a Lemo encoder (a component of the ultrasonic detector) to the encoder port on the 5S64 wheeled phased array scanner, and the other end to the encoder input port on the PHASCAN II portable ultrasonic phased array detector. Open the PHASCAN II portable ultrasonic phased array detector, access the settings page, select C-scan imaging mode, and adjust the C-scan gate and other inspection parameters. Start the PHASCAN II portable ultrasonic phased array detector, apply ultrasonic couplant and water as liquid conducting media, respectively, to the structure to be inspected. Use the handheld 5S64 wheeled phased array scanner to scan the inspection object in a fixed trajectory. During the scanning process, the ultrasonic analysis software phascan_view_v1.5.6 on the detector generates scanned images.

[0133] Step 5: Conduct internal defect detection;

[0134] The C-scan image generated by phascan_view_v1.5.6 (which comes with the PHASCAN II portable ultrasonic phased array detector) is transferred to the industrial control computer for grayscale processing to retain key information and reduce unnecessary color interference. The industrial control computer then removes noise and resizes the C-scan image to optimize image quality. The industrial control computer then uses the U-Net model to perform pixel-level segmentation on the processed image to obtain a binary defect mask, extract the defect geometric features, and fit an ellipse. Image registration is then performed using SIFT feature detection, FLANN matching, and the RANSAC algorithm to determine the actual defect location, effectively completing internal defect detection, quantification, and location.

[0135] The above-mentioned U-Net model: proposed by Ronneberger et al. in 2015, it is widely used in semantic segmentation tasks. SIFT feature detection: SIFT (Scale-Invariant Feature Transform) is a classic algorithm for extracting image feature points, proposed by Lowe in 2004. FLANN matching: FLANN (Fast Library for Approximate Nearest Neighbors) is an algorithm for fast feature matching, commonly used in computer vision tasks. RANSAC algorithm: RANSAC (Random Sample Consensus) is a robust algorithm for calculating homomorphic transformations, widely used in image registration and feature matching. These methods are widely used in the fields of image processing and computer vision, especially in tasks such as defect detection, image registration, and feature extraction.

[0136] Step 6: Multi-modal data fusion detection;

[0137] The surface defect results detected in step 3 and the internal defect results detected in step 5 are fused through the multimodal detection information fusion and visualization algorithm based on DS evidence theory designed below, ultimately achieving comprehensive identification and positioning of defects in metal-composite hybrid structure devices.

[0138] Multi-source detection data fusion mainly combines polarization detection and phased array ultrasonic detection data to perform defect information detection, coordinate conversion, defect registration and fusion. Specifically, it includes the following steps:

[0139] (01) Data conversion and preprocessing;

[0140] Data conversion and preprocessing are the starting point of the entire fusion process, responsible for unified processing of data from different detection sensors to ensure data consistency and comparability. The specific operations are as follows:

[0141] a) Data Format Standardization: Different sensors (e.g., polarization cameras and ultrasound scanners) generate data in different formats. This data needs to be converted to a unified format (e.g., JSON or CSV) to facilitate subsequent processing and analysis. Format conversion methods are well known to those skilled in the art and will not be detailed here.

[0142] b) Data denoising: Use filtering algorithms (such as median filtering, Gaussian filtering, etc.) to remove noise from the data and improve data quality. The presence of noise will affect subsequent defect detection and fusion effects.

[0143] c) Data Calibration: Data from different sensors are calibrated to ensure comparability of their measurements of the same physical quantity (e.g., temperature, polarization, etc.). The calibration process includes compensating and correcting the sensor's response characteristics. Data calibration methods are known to those skilled in the art.

[0144] d) Data normalization: Normalize the numerical ranges of different data sources so that they can be compared and integrated on the same scale. For example, Z-score normalization is used for ultrasound amplitude (0-65535) and polarization degree (0-1).

[0145] e) Data Enhancement: Image enhancement techniques (such as contrast adjustment and sharpening) are used to improve image quality, enhance defect visibility, and assist in subsequent defect detection. Data enhancement methods are well known to those skilled in the art and will not be described in detail here.

[0146] (02) Coordinate transformation

[0147] Coordinate transformation is responsible for transforming the defect information from different data sources to ensure that the defect locations from different data sources are matched in the same coordinate system. The specific functions are as follows:

[0148] Coordinate system 1: Convert the defect information obtained by the polarization camera and ultrasonic scanner from their respective coordinate systems (such as pixel coordinate system, TCP coordinate system, robot arm base coordinate system, etc.) to a unified robot arm base coordinate system. This coordinate conversion ensures that the defect locations of different data sources can be accurately matched, avoiding errors caused by different coordinate systems. This is achieved through the following steps:

[0149] Pixel coordinate system → camera coordinate system: Use the polarization camera internal parameter matrix (focal length, principal point coordinates, etc.) to convert the pixel coordinates (u, v) into a three-dimensional point (X c ,Y c ,Z c ), the formula is:

[0150]

[0151] Among them, f u ,f v is the equivalent focal length of the camera in the horizontal (u direction) and vertical (v direction), (u0, v0) is the coordinate of the principal point on the camera imaging plane, which refers to the coordinate position of the intersection of the camera optical axis and the imaging plane in the pixel coordinate system.

[0152] Camera coordinate system → Robot arm base coordinate system: Obtain the extrinsic parameter matrix (rotation matrix R and translation vector T) of the camera and the robot arm base through hand-eye calibration to achieve coordinate transformation:

[0153]

[0154] Among them, (X b ,Y b ,Z b ) is the spatial coordinate of the defect information in the robot arm base coordinate system.

[0155] Ultrasonic scanner coordinate system → robotic arm base coordinate system: The position of the ultrasonic scanner is determined by the TCP (tool center point) coordinate system at the end of the robotic arm. The defect position detected by ultrasonic detection is converted from the TCP coordinate system to the base coordinate system through the forward solution of the robotic arm kinematics (this coordinate conversion is well known to those skilled in the art).

[0156] At the same time, affine transformation (maintaining parallelism, suitable for small-scale detection) or perspective transformation (handling large-scale changes in viewing angle, such as the formula x' = Hx, where H is a 3×3 homogeneous transformation matrix) is used to ensure conversion accuracy. For relevant geometric transformation theory, please refer to the book "Multi-Coordinate Calibration Technology in Machine Vision" (ISBN: 978-7-111-59876-2).

[0157] Error compensation: Considering factors such as sensor installation error (such as the position deviation between the camera and the end of the robot arm) and robot arm motion error (joint positioning accuracy), the following methods are used to improve conversion accuracy:

[0158] Calibration of calibration parts: Using a high-precision calibration plate (such as a checkerboard or spherical target), data is collected from multiple locations to fit the transformation parameters between the sensor coordinate system and the robot base coordinate system. Specifically, the singular value decomposition (SVD) method can be used to solve the least squares optimal transformation matrix. This method is widely used in robot hand-eye calibration (reference: "A Review of Robot Hand-Eye Calibration Technology", Robot, 2020, 42(3): 365-376).

[0159] Dynamic Error Correction: Utilizing feedback from the robot arm's encoder, positional deviations during motion are compensated in real time. For ultrasonic scanners, couplant uniformity testing and scanning speed control are used to reduce inspection errors caused by contact position variations. The error compensation algorithm adheres to industry-standard procedures, such as the ISO 9283 robot arm positioning accuracy test method, ensuring coordinate conversion errors of less than 0.1 mm (set according to inspection accuracy requirements).

[0160] (03) Defect Registration

[0161] Defect registration is responsible for registering defect information from different data sources to determine the specific location and shape of the defect. The specific operations are as follows:

[0162] A) Image Registration: Images from different data sources are registered using image processing techniques (such as feature point matching and the SIFT feature matching algorithm) to ensure consistent defect locations across images. The registration process involves geometric transformations such as rotation, scaling, and translation, enabling comparison of images from different data sources within the same coordinate system. The techniques described in this section are well known to those skilled in the art and will not be reiterated here.

[0163] B) Defect Feature Extraction: Image processing and machine learning algorithms (such as edge detection, morphological operations, and YOLO) are used to extract defect features, including shape, size, and location. Feature extraction is the foundation of registration, ensuring that defects from different data sources can be matched based on their features. The techniques described in this section are well known to those skilled in the art and will not be repeated here.

[0164] C) Defect matching: Based on the extracted features, defects from different data sources are matched to determine whether they are different manifestations of the same defect (when the defects from different data sources have overlapping positions in a unified coordinate system, the feature parameters (shape / size / feature points) are highly similar, and conform to the physical mechanism, they are judged to be different manifestations of the same defect; otherwise, due to position deviation, feature contradictions, or physical logic inconsistencies, they are judged to be independent defects or noise artifacts). The matching process includes matching feature points, calculating shape similarity, etc., to ensure that defects from different data sources can be accurately aligned. The technology described in this paragraph is known to those skilled in the art and will not be repeated here.

[0165] D) Defect Classification: Defects are classified based on their characteristics, distinguishing different types of defects (such as cracks, voids, and delaminations), providing a foundation for subsequent fusion. This classification process helps the system better understand the nature of the defects and improves fusion accuracy. The techniques described in this section are well known to those skilled in the art and will not be repeated here.

[0166] (04) Defect fusion and post-processing

[0167] Defect fusion and post-processing are the core of multimodal detection information fusion and visualization technology, responsible for fusing defect information from different data sources to improve the accuracy and reliability of detection. The specific operations are as follows:

[0168] I) DS Evidence Theory Fusion: DS evidence theory is used for information-level fusion, with the defect area used as a credibility indicator. Defect information from different data sources is combined to calculate the comprehensive defect credibility. The techniques described in this paragraph are known to those skilled in the art and will not be repeated here. It helps determine the authenticity and severity of defects. DS evidence theory was proposed by Dempster in 1967 and further developed by Shafer in 1976. It is an imprecise reasoning theory that is widely used in information fusion, target recognition, fault diagnosis and other fields. The techniques described in this paragraph are known to those skilled in the art and will not be repeated here.

[0169] II) Result output: The fused results are output in JSON format, including information such as the location, type, and size of the defect for subsequent analysis and decision-making.

[0170] III) Post-processing: Post-processing the fusion results, including defect annotation, unified output format, deflection angle calculation, and other functions, to support 3D visualization. The specific implementation of this step is well known to those skilled in the art and will not be repeated here.

[0171] (05) Three-dimensional (3D) visualization

[0172] 3D visualization utilizes OpenGL 3D engine technology (well-known to industry professionals) to render the 3D model of the metal-composite structural device. A 3D display engine is used to render the defect shape into a three-dimensional shape and overlay it with the 3D structural model of the metal-composite structural device in the 3D visualization area. All detected defects are visualized through highlighting and other means. "Rendering" refers to the process of using computer graphics technology to transform the 3D model data of the metal-composite structural device through a series of complex calculations and processing, ultimately resulting in a realistic visual image displayed on the screen. This process is well-known to those skilled in the art. Specific embodiments

[0174] Example 1. See Figure 2 The polarization detection probe provided by the present invention is used to detect surface defects of metal-composite material structural devices whose surface material is metal.

[0175] Step 1: Environment and software installation. First, you need to configure the Matlab 2019 compilation environment and Visual Studio 2019 development environment on the industrial control computer to ensure the normal operation of the defect detection system. In addition, you need to install the FLIR polarization camera driver and JC-0.4 liquid zoom lens driver.

[0176] Step 2: Combined light source settings: The polarization detection probe light source is equipped with a bar-shaped combined light source, a circular polarization light source, and a power driver. Power the driver and adjust the illumination angle of the bar-shaped combined light source according to actual needs to change the optical path and provide sufficient illumination to the defect. At the same time, set the basic parameters of the optical components of the polarization detection probe, as shown in the table below.

[0177] Table 1 Basic parameters of optical components

[0178]

[0179]

[0180] Step 3: Wiring Installation: Connect the polarization detection probe to one camera driver cable, one liquid zoom lens driver cable, one light source driver power cable, and five light source connection cables. Before turning on the power, connect the camera driver cable and the liquid lens driver cable to the camera and lens connectors, respectively, and then to the industrial control computer. Connect the light source driver power cable from the light source driver to a 220V outlet (keep the light source driver switch off when not in use). Connect the five light source connection cables from the five light sources to the light source driver connector to provide power.

[0181] Step 4: Distance and angle setting: For situations where the workpiece to be measured has a complex structure and the optical path is difficult to enter the defect, due to the limitations of the lens focal length and polarization camera pixels, in order to meet the accuracy indicators, the system working distance should be specified within 80-120mm.

[0182] Step 5: Obtain polarization image information of metal surface: The industrial control computer first sends an image acquisition instruction to the robotic arm, and the robotic arm follows the predetermined path to reach the designated detection point. After reaching the designated point, the combined light source is turned on to illuminate the part to be detected, such as Figure 4 As shown in the figure, the part to be inspected generates reflected light, which is converged by the telecentric lens to form a fixed field of view and projected to the polarization camera, generating four images with different polarization angles of 0°, 45°, 90° and 135°. The image in each polarization direction provides the intensity information I0, I 45 , I 90 and I 135 .

[0183] Step 6: Solve the polarization image parameter information. After the image acquisition of all points is completed, the collected polarization image is transmitted to the industrial control computer. The polarization parameter information of the measured target, such as the degree of polarization (DoP), angle of polarization (AoP), Stokes vector (S0, S1 and S2), is solved by the following formula (well known in the industry), effectively improving the characteristic contrast of defects on the metal surface. S0=I0+I 90 ,S1=I0-I 90 ,

[0184] S2=I 45 -I 135

[0185]

[0186] Step 7: Surface defect detection: The obtained polarization parameter information is used to detect defects on the surface of metal-composite hybrid structure devices using the metal surface damage defect intelligent detection algorithm based on combined polarized light source detection designed by the present invention.

[0187] Example 2. See Figure 5The ultrasonic detector provided by the present invention performs internal inspection on metal-composite material structural components whose internal material is polyvinyl chloride.

[0188] Step 1: First click on the menu wizard.

[0189] Step 2: Workpiece material: polyvinyl chloride, others are default.

[0190] Step 3: Set the probe steering angle to 90°, the transceiver mode to PE pulse echo, the starting transmit channel to 1, series 5s64, model 5s64, probe type linear, frequency 5 MHz, number of primary axis array elements 64, number of secondary axis array elements 1, primary axis array element spacing 0.6 mm, secondary axis array element spacing 0.6 mm, wedge: click Select after loading, click Contact.opw in the OTHER series, and click OK.

[0191] Step 4: Select the longitudinal wave type for the focusing rule. The type is line scan, not fan scan and full focus. The starting angle is 0°, the Zhuzhou aperture is 8, the spindle start and end 1-64 array element step is 1, the focusing type is true depth, the depth is 5mm, the reflection and weld reinforcement are all closed, the simulated reflection, the scanning offset, and the stepping offset are all 0. The fourth step is scanning: single-line scanning, encoder 1, the maximum scanning speed is 150mm / s, the encoder 1 mode is orthogonal, the resolution is 34, and the scanning resolution is 1mm.

[0192] Step 5: Click the menu, select Ultrasonic Detection Settings, click General, set the wedge delay to 29.21us (fine-tunable), and the ultrasonic axis unit to true depth.

[0193] Step 6: (Set parameters according to the specific detection object) Click the gate curve gate A in the menu, turn on the switch, the gate width is about 0.8 to 1, the gate is 14%, the mode is depth, the synchronization is 1 input, the measurement method is peak gate 1, the switch on and off seems to have no effect, the gate starting point is 0mm, the gate width is about 10mm, the gate is 25%, the mode is depth, the synchronization is pulse, and the measurement method is peak. Adjust the locking screw of the probe angle (fine-tune clockwise and counterclockwise after loosening it). When the A scan amplitude is the largest, make sure the adjustment position is fixed and tighten the screw.

[0194] Step 7: Prepare tap water as a coupling agent and apply it evenly to the surface of the sample to be inspected (flat or curved panel). Use a handheld 5S64 wheeled phased array scanner to scan the inspection object in a fixed trajectory. The imaging results can be imported into an industrial control computer via a storage device such as a USB flash drive. A U-Net model is used for pixel-level segmentation to obtain a binary defect mask. The defect's geometric features are extracted and an ellipse is fitted. Image registration is performed using SIFT feature detection, FLANN matching, and the RANSAC algorithm to determine the actual defect location, effectively completing internal defect detection, quantification, and localization.

[0195] The detection effect of the wheel probe requires the following conditions to be guaranteed:

[0196] In actual use, it is mainly necessary to ensure sufficient water volume;

[0197] The test object is usually a flat or curved panel;

[0198] Adjust the probe angle to the maximum echo value;

[0199] The output Bscandata and Cscandata can both intuitively reflect the quantitative information of defects. Bscandata focuses on depth, while Cscandata focuses on defect shape and size.

[0200] The present invention involves a large amount of neural networks and mathematical knowledge. Anything not specifically emphasized is a common method used by those skilled in the art and is well known to those skilled in the art.

[0201] The present invention proposes a polarization detection probe, the innovation of which is reflected in (1) the design of a combined polarized light source: general polarization detection only uses a single polarized light source. The present invention adopts a technical solution of combining a circular polarized light source (direct) with a four-angle adjustable strip light source (oblique), and optimizes the brightness uniformity and polarization information richness of complex metal surfaces through multi-angle illumination, solving the problem of defect omission caused by uneven illumination in traditional single light source detection; (2) combining a combined polarized light source with a telecentric lens: using a telecentric lens to shoot polarization images, and realizing a wide range of focus of 20 to 170 mm through liquid focusing technology, with a telecentricity of <0.03° , distortion <0.08%, ensuring the dimensional measurement accuracy of complex surface defects and image distortion-free; (3) Applying multi-polarization angle imaging to metal polarization defect detection: Although the application of multi-polarization angle imaging technology in target detection and complex environments has been studied, no one has applied multi-polarization angle imaging to metal polarization defect detection. The BFS-U3-51S5PC polarization camera is used to synchronously collect polarization images in four directions of 0°, 45°, 90°, and 135°, and the degree of polarization (DoP) and angle of polarization (AoP) are calculated in combination with the Stokes vector, which can improve the characteristic contrast of tiny defects. In addition, the present invention adopts a lightweight integrated structure: a lightweight shell (size 246.5×142.8×138mm) designed by SolidWorks and 3D printed, supports robotic arm mounting and flexible scanning, and is suitable for high-mobility detection scenarios such as aerospace.

Claims

1. A metal-composite material hybrid structure device defect detection system, characterized in that: The system includes an industrial control computer, a mechanical scanning arm, a polarization detection probe, an ultrasonic detector component, and a power supply; 1) Industrial control computers The industrial control computer uses the Adantech IPC-610 model, which is connected to the mechanical scanning arm, polarization detection probe, and ultrasonic detector via data cables. It is responsible for transmitting detection instructions and detection information. At the same time, the built-in high-performance data processing unit and industrial-grade image processing software are responsible for processing and analyzing defect information. 2) Mechanical scanning arm A mechanical scanning arm (hereinafter referred to as the "robot arm") has a polarization detection probe fixed at its end. The position of the arm is adjusted by commands from an industrial control computer, changing the working distance and scanning path of the polarization detection probe. 3) Polarization detection probe The polarization detection probe includes a light source, a lens, a polarization camera, and a light source driver, as follows: (1) Combined light source The combined light source is a combination of a circularly polarized light source and four angle-adjustable strip light sources. The circularly polarized light source emits linearly polarized light directly toward the surface to be measured. The four angle-adjustable strip light sources are approximately located in the upper, lower, left, and right directions of the circularly polarized light source and are rotatably fixed to the approximate center of the four edges of the front end of the housing. They can be rotated around the corresponding edges to adjust the light emission angle. The circularly polarized light source is used as the directly incident light source, and the obliquely incident strip light sources are used to supplement the brightness. (2) Lens The lens is located behind and adjacent to the circularly polarized light source, the center of the circularly polarized light source is on the optical axis of the lens, and the circularly polarized light source is arranged perpendicular to the optical axis; (3) Polarization camera The polarization camera is located behind the lens, and the optical axis of the camera coincides with the optical axis of the lens. The polarization camera is used to capture images and generate image information at four different polarization angles: 0°, 45°, 90°, and 135°. (4) Light source driver The light source driver provides drive for the combined light source; (5) Support shell The supporting shell is used to accommodate the lens, polarization camera, and light source driver, and the combined light source is arranged at the front end of the supporting shell; 4) Ultrasonic detector components The ultrasonic detector assembly includes an ultrasonic detector, a phased array scanner, and an encoder; the ultrasonic detector assembly is placed at an appropriate position around the robotic arm; 5) Power supply A power supply provides power to the system.

2. The metal-composite material hybrid structure device defect detection system according to claim 1, characterized in that: 1) The industrial control computer adopts Adantech IPC-610 model; 2) The mechanical scanning arm uses the lightweight 6-DOF robotic arm AUBO-i10; 3) In the polarization detection probe, (1) Combined light source The circular polarized light source model is CST-POR10090-w, and the bar light source model is CST-ROS140-w; all four bar light sources emit white natural light toward the surface to be measured; (2) Lens Use JC-0.4X10Y liquid focus lens as the electric focus telecentric lens; (3) Polarization camera A BFS-U3-51S5PC polarization camera was used for image acquisition; (4) Light source driver The light source driver adopts CST-DPS20-CM-TD model; (5) Support shell The supporting shell was designed using Solidwork software, and a lightweight shell model was produced using 3D printing technology; 4) Ultrasonic detector The ultrasonic detector uses the PHASCAN II portable ultrasonic phased array detector, which is equipped with a 5S64 wheel-type phased array scanner and a Lemo encoder.

3. A method for detecting defects in a metal-composite hybrid structure device, based on the metal-composite hybrid structure device defect detection system according to claim 1 or 2, characterized in that: The method is as follows: Step 1: Obtain polarization image information of the metal surface; The industrial control computer sends an image acquisition command to the robotic arm, and the robotic arm follows a predetermined path to reach the designated inspection point. After reaching the designated point, the combined light source is turned on to illuminate the component to be inspected. The component to be inspected generates reflected light, which is converged by the telecentric lens to form a fixed field of view and projected to the polarization camera, generating four images with different polarization angles of 0°, 45°, 90°, and 135°. The image in each polarization direction provides the intensity information I0, I1 of the polarized light in the four different directions. 45 , I 90 and I 135 ; Step 2: Calculate polarization image parameter information; After all points are captured, the polarization images are transmitted to an industrial control computer. The polarization parameter information of the target, including the degree of polarization (DoP), angle of polarization (AoP), and three Stokes vectors (S0, S1, and S2), is calculated using the following formulas, effectively improving the characteristic contrast of defects on the metal surface. S0=I0+I 90 ,S1=I0-I 90 ,S2=I 45 -I 135 Step 3: Perform surface defect detection; The solved polarization parameter information is used to detect defects on the surface of metal-composite hybrid structure devices through a metal industrial parts inspection algorithm based on deep learning polarization image fusion to obtain surface defect detection results. Step 4: Obtaining ultrasonic testing information inside the metal; Connect the ultrasonic signal cable of the wheeled phased array scanner to the probe interface of the ultrasonic phased array detector, connect the encoding cable of the Lemo encoder to the encoder interface of the wheeled phased array scanner, and connect the other end of the Lemo encoder encoding cable to the encoder input interface of the ultrasonic phased array detector. Turn on the ultrasonic phased array detector, select the scanning imaging mode, and adjust the detection parameters. Start the ultrasonic phased array detector, apply ultrasonic coupling agent or water as the liquid conductive medium to the structure to be inspected, and use the wheeled phased array scanner to scan the inspection object along a fixed trajectory. During the scanning process, a scanning image is generated on the detector. Step 5: Conduct internal defect detection; The generated scanned image is transmitted to the industrial control computer for grayscale processing to retain key information and reduce unnecessary color interference. The industrial control computer removes noise and resizes the scanned image to optimize image quality. The industrial control computer uses the U-Net model to perform pixel-level segmentation on the processed image to obtain a binary defect mask, extract the defect geometric features and fit an ellipse. Combined with SIFT feature detection, FLANN matching and RANSAC algorithm, image registration is performed to determine the actual location of the defect, completing internal defect detection, quantification and positioning. Step 6: Multi-modal data fusion detection; The specific steps include: (01) Data conversion and preprocessing; Data conversion and preprocessing are the starting point of the entire fusion process. They are responsible for unified processing of data from different detection sensors to ensure data consistency and comparability. The specific operations are as follows: a) Data format unification: Different sensors generate data in different formats, and these data need to be converted into a unified format; b) Data denoising: Remove noise from data through filtering algorithms to improve data quality; c) Data calibration: Calibrate the data of different sensors to ensure that their measurement results on the same physical quantity are comparable; d) Data standardization: standardize the numerical ranges of different data sources so that they can be compared and integrated on the same scale; e) Data enhancement: Image enhancement technology is used to improve image quality, enhance defect visibility, and assist in subsequent defect detection; (02) Coordinate transformation Coordinate transformation is responsible for transforming the defect information from different data sources to ensure that the defect locations from different data sources are matched in the same coordinate system; the details are as follows: Coordinate system 1: Convert the defect information obtained by the polarization camera and ultrasonic scanner from their respective coordinate systems to a unified robotic arm base coordinate system. This is achieved by the following steps: Pixel coordinate system → camera coordinate system: Use the polarization camera intrinsic parameter matrix to convert the pixel coordinates (u, v) into a three-dimensional point (X c ,Y c ,Z c ), the formula is: Among them, f u ,f v is the equivalent focal length of the camera in the horizontal (u direction) and vertical (v direction), (u0, v0) is the coordinate of the principal point on the camera imaging plane, which refers to the coordinate position of the intersection of the camera optical axis and the imaging plane in the pixel coordinate system. Camera coordinate system → Robotic arm base coordinate system: Obtain the extrinsic parameter matrix of the camera and the robotic arm base through hand-eye calibration, including the rotation matrix R and translation vector T, to achieve coordinate transformation: Among them, (X b ,Y b ,Z b ) is the spatial coordinate of the defect information in the robot arm base coordinate system. Ultrasonic scanner coordinate system → robotic arm base coordinate system: The position of the ultrasonic scanner is determined by the tool center point TCP coordinate system at the end of the robotic arm. The defect position detected by ultrasonic inspection is converted from the TCP coordinate system to the base coordinate system through the forward solution of the robotic arm kinematics. When detecting a small area, affine transformation is used, and when dealing with large-scale viewing angle changes, perspective transformation is used; (03) Defect Registration The specific operations are as follows: A) Image Registration: Use image processing technology to register images from different data sources to ensure that the location of defects in different images is consistent; B) Defect feature extraction: Use image processing and machine learning algorithms to extract defect features, including shape, size, and location information; ensuring that defects from different data sources can be matched based on features; C) Defect matching: Based on the extracted features, defects from different data sources are matched to determine whether they are different manifestations of the same defect. This ensures that defects from different data sources can be accurately registered. D) Defect classification: Classify defects according to their characteristics and distinguish different types of defects; (04) Defect fusion and post-processing The specific operations are as follows: I) DS Evidence Theory Fusion: DS evidence theory is used for information-level fusion, using the defect area as a credibility indicator. By combining defect information from different data sources, the comprehensive defect credibility is calculated to help determine the authenticity and severity of the defect. II) Result output: The fused result output includes the location, type, and size information of the defect.

4. The defect detection method for metal-composite material hybrid structure device according to claim 3, characterized in that: In (02) coordinate transformation, error compensation is also required. The following methods are used to improve the transformation accuracy: Calibration of calibration parts: Using a high-precision calibration plate, data is collected at multiple locations to fit the transformation parameters between the sensor coordinate system and the robot base coordinate system; the singular value decomposition (SVD) method is used to solve the least squares optimal transformation matrix; Dynamic Error Correction: Utilizes feedback from the robotic arm encoder to compensate for positional deviations during motion in real time. For ultrasonic scanners, couplant uniformity testing and scanning speed control are used to reduce detection errors caused by contact position variations. The error compensation algorithm follows industry-standard procedures.

5. The defect detection method for metal-composite material hybrid structure device according to claim 3, characterized in that: In (03) defect registration C) defect matching, when defects from different data sources have overlapping positions in a unified coordinate system, high similarity in feature parameters, and conform to the physical mechanism, they are determined to be different manifestations of the same defect; On the contrary, due to position deviation, feature contradiction or physical logic inconsistency, it is judged as an independent defect or noise artifact; the matching process includes matching of feature points and shape similarity calculation.

6. The method for detecting defects in a metal-composite hybrid structure device according to claim 3, wherein: In (04) defect fusion and post-processing, III) post-processing is also required: post-processing of the fusion results, including defect annotation, unified output format, and deflection angle calculation to provide support for 3D visualization.

7. The defect detection method for metal-composite material hybrid structure device according to claim 3, characterized in that: In (05) three-dimensional 3D visualization, the OpenGL 3D engine technology is used to render the 3D model of the metal-composite structural device, and the 3D display engine is used to draw the defect shape into a three-dimensional shape and superimpose it with the 3D structural model of the metal-composite structural device in the 3D visualization area. The visualization display of all detected defects is completed by highlighting.

Images