Intelligent detection method based on nondestructive inspection equipment

Abstract

The invention discloses an intelligent detection method based on nondestructive inspection equipment, and relates to the technical field of nondestructive inspection, and the method comprises the following steps: establishing a twin digital model; the method comprises the following steps: acquiring an internal image of a to-be-detected part through an endoscope, obtaining an original surface image, performing multi-dimensional feature extraction, and screening the original surface image to obtain a standard surface image; and acquiring a motion parameter sequence of the endoscope, fitting the standard surface image to the twin digital model based on the motion parameter sequence, and generating a three-dimensional restored image. And based on the shape features of the to-be-tested component model, adaptively dividing the grids of the three-dimensional restored image to obtain a grid three-dimensional image. And in combination with a data dimension reduction method, a multi-channel classification detector is constructed, intelligent flaw detection is carried out on the gridding three-dimensional image, and the multi-channel classification detector comprises a plurality of flaw detection channels and is used for executing different types of flaw detection tasks. Therefore, the technical effects of reducing the detection cost and reducing the operation difficulty are achieved.

Description

Technical Field

[0001] The invention relates to the technical field of nondestructive testing, and in particular to an intelligent testing method based on nondestructive flaw detection equipment. Background Art

[0002] Nondestructive Testing (NDT) technology refers to a method of detecting and evaluating internal and surface defects or performance of the object being tested without destroying or affecting its performance and appearance. Existing nondestructive testing technologies include ultrasonic testing, X-ray testing, magnetic particle testing, and eddy current testing. Although these technical methods can effectively detect defects inside the material, they usually require professional operators and the testing process is complicated. There are technical problems such as high testing costs and difficult operations. Summary of the invention

[0003] The present invention provides an intelligent detection method based on non-destructive flaw detection equipment to solve the technical problems of high detection cost and great operating difficulty in the prior art, thereby achieving the technical effect of reducing detection cost and reducing operating difficulty.

[0004] In a first aspect, the present invention provides an intelligent detection method based on non-destructive testing equipment, wherein the method comprises: Establish a twin digital model, including the model of the component to be tested and the endoscope model.

[0005] The internal image of the component to be tested is collected through an endoscope to obtain the original surface image.

[0006] Perform multidimensional feature extraction on the original surface image to obtain a multidimensional feature set, and perform image screening on the original surface image based on the multidimensional feature set to obtain a standard surface image.

[0007] The interactive control component obtains a motion parameter sequence of the endoscope optical component, and fits the standard surface image to the twin digital model based on the motion parameter sequence to obtain a three-dimensional restored image.

[0008] Based on the shape features of the model of the component to be tested, the three-dimensional restored image is adaptively meshed to obtain a meshed three-dimensional image.

[0009] In combination with the data dimension reduction method, a multi-channel classification detector is constructed to perform intelligent flaw detection of the gridded three-dimensional image, wherein the multi-channel classification detector includes multiple flaw detection channels for performing different types of flaw detection tasks.

[0010] In a feasible implementation, the internal image of the component to be tested is collected by an endoscope to obtain the original surface image, including: Based on the model of the component to be tested, multi-level acquisition parameters are defined; image acquisition path planning of the model of the component to be tested is performed in combination with the multi-level acquisition parameters to obtain an internal image acquisition scheme, wherein the internal image acquisition scheme includes multiple serialized acquisition points; based on the internal image acquisition scheme, an endoscope is activated to perform image acquisition to generate an original surface image.

[0011] In a feasible implementation, multi-dimensional feature extraction is performed on the original surface image to obtain a multi-dimensional feature set, and image screening is performed on the original surface image based on the multi-dimensional feature set to obtain a standard surface image, including: Grayscale the original surface image, and solve the grayscale co-occurrence matrix of the grayscale image; based on the grayscale co-occurrence matrix, calculate and obtain the grayscale information entropy and the first contrast; establish a scatter point coordinate system, use the grayscale information entropy and the first contrast as the first coordinate and the second coordinate of the scatter point coordinate system respectively, and perform scatter point distribution of the original surface image; perform cluster analysis on the scatter point distribution results, and perform image screening on the original surface image according to the cluster analysis results, eliminate multiple original images contained in abnormal clusters, and generate the standard surface image.

[0012] In a feasible implementation, a multi-channel classification detector is constructed by combining data dimensionality reduction methods, including: Acquire historical flaw detection records of a target environment, wherein the historical flaw detection records include historical surface images and associated historical damage information; acquire flaw detection task information in an interactive target flaw detection scenario, and extract a flaw detection target set; extract broad-spectrum feature data of the historical surface images, and based on the flaw detection target set, perform grouping and dimensionality reduction on the broad-spectrum feature data to obtain a multi-target key feature set; construct multiple classification detection channels based on the multi-target key feature set, perform supervised training based on the historical flaw detection records, integrate the multiple classification detection channels after training, and obtain a multi-channel classification detector.

[0013] In a feasible implementation, the broad-spectrum feature data of the historical surface image is extracted, and based on the flaw detection target set, the broad-spectrum feature data is grouped and dimensionally reduced to obtain a multi-target key feature set, further comprising: Based on the flaw detection target set, a first flaw detection target is obtained, and the first flaw detection target corresponds to any damage type; according to the first flaw detection target, the historical flaw detection records are traversed to extract the first historical flaw detection record, and the first historical flaw detection record includes a first historical surface image and a first historical damage information; the grayscale co-occurrence matrix, grayscale difference, and color moment of the first historical surface image are solved, and feature extraction is performed to obtain first broad-spectrum feature data, wherein the first broad-spectrum feature data at least includes grayscale information entropy, first contrast, energy, differential entropy, second contrast, average value, first-order moment, and second-order moment; dimensionality reduction analysis is performed on the first broad-spectrum feature data, main features are extracted, and stored as a first target key feature set; the flaw detection target set is traversed to obtain a multi-target key feature set.

[0014] In a feasible implementation, multiple classification detection channels are constructed according to the multi-target key feature set, and supervised training is performed based on the historical flaw detection records, and the multiple classification detection channels after training are integrated to obtain a multi-channel classification detector, which also includes: According to the first target key feature set, a first classification detection model is constructed based on a classification algorithm; using the first historical flaw detection record as a training data set, supervised training of the first classification detection model is performed to obtain a first classification detection channel; based on the multi-target key feature set and the historical flaw detection records, a plurality of the classification detection channels are constructed and trained; partition constraints of adaptive grid partitioning are obtained, and based on the partition constraints, an input distributor is configured to connect a plurality of the classification detection channels to the input distributor to construct the multi-channel classification detector.

[0015] In a feasible implementation, performing intelligent flaw detection of the gridded three-dimensional image further includes: Acquire the flaw detection result, parse and obtain multiple edge damages, and output them as an edge damage set; perform continuity analysis on the edge damage set based on damage type, damage location, damage direction, and damage size to obtain continuity analysis results, wherein the continuity analysis results include multiple edge damage groups; based on the continuity analysis results, perform damage merging processing on the edge damage set to obtain a merged edge damage set; and update the merged edge damage set to the flaw detection result.

[0016] In a feasible implementation, based on the damage type, damage location, damage direction, and damage size, the edge damage set is subjected to continuity analysis, and then the method further includes: Analyze the continuity analysis results, perform type difference discrimination on the multiple edge damage groups, and extract multiple difference edge damage groups; merge the multiple difference edge damage groups to obtain multiple difference damage partitions; activate the multi-channel classification detector to perform update detection on the multiple difference damage partitions to obtain an updated damage type set; and perform damage merging processing on the edge damage set in combination with the updated damage type set.

[0017] The present invention discloses an intelligent detection method based on nondestructive flaw detection equipment, including: establishing a twin digital model, including a model of a component to be tested and an endoscope model; collecting an internal image of the component to be tested through an endoscope to obtain an original surface image; performing multidimensional feature extraction on the original surface image to obtain a multidimensional feature set, and performing image screening on the original surface image based on the multidimensional feature set to obtain a standard surface image; an interactive control component obtains a motion parameter sequence of an endoscope optical component, and fits the standard surface image to the twin digital model based on the motion parameter sequence to obtain a three-dimensional restored image; based on the shape features of the component model to be tested, adaptively meshing the three-dimensional restored image to obtain a gridded three-dimensional image; combining a data dimensionality reduction method, constructing a multi-channel classification detector, and performing intelligent flaw detection on the gridded three-dimensional image, wherein the multi-channel classification detector includes multiple flaw detection channels for performing different types of flaw detection tasks. The intelligent detection method based on nondestructive flaw detection equipment disclosed by the present invention solves the technical problems of high detection cost and high operation difficulty, and achieves the technical effect of reducing detection cost and reducing operation difficulty. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the process of the intelligent detection method based on non-destructive flaw detection equipment of the present invention.

[0019] Figure 2 It is a schematic diagram of the process of obtaining a standard surface image in the intelligent detection method based on non-destructive flaw detection equipment of the present invention. DETAILED DESCRIPTION

[0020] The technical solution provided in the embodiments of the present invention is to solve the technical problems of high detection cost and great operation difficulty in the prior art. The overall idea adopted is as follows: First, the historical welding records of the carbon dioxide cylinder are obtained, and the historical welding records include welding parameter records and welding quality evaluation records. Then, according to the welding quality evaluation records, the welding quality coefficient is extracted for bidirectional division, and the historical welding records are divided into a forward solution set and a reverse solution set. Then, based on the forward solution set, the influencing component analysis is performed to obtain the forward key factor set, and the welding quality evaluation function is constructed based on the forward key factor set and the welding quality coefficient. Then, according to the welding task information, the forward solution set is traversed for dual-objective matching, the dual-objective matching results are obtained, and the first population and the second population are initialized. Then, combined with the welding quality evaluation function, a shared path is constructed, the first population and the second population are collaboratively optimized, and the optimal welding parameter combination is iteratively obtained, wherein the shared path is configured with a shared constraint. Finally, welding control is performed based on the optimal welding parameter combination.

[0021] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.

[0022] Example Figure 1 The figure is a flow chart of the intelligent detection method based on non-destructive flaw detection equipment of the present invention, wherein the method comprises: Establish a twin digital model, including the model of the component to be tested and the endoscope model.

[0023] Optionally, the twin digital model includes a model of the component to be tested and an endoscope model registered in the same digital space, in other words, the coordinate systems of the model of the component to be tested and the endoscope model are aligned so that the two can be accurately matched in the same digital space. The component to be tested is an industrial production component with an inner surface and requiring inner surface control.

[0024] Optionally, based on the design data of the component to be tested (such as CAD drawings, design stability, etc.), the geometric shape and size of the component to be tested are obtained, and a geometric model of the component to be tested is established using 3D modeling software (such as SolidWorks, CATIA, etc.) to obtain the model of the component to be tested.

[0025] Optionally, a geometric model of the endoscope is established based on the physical size and structure of the endoscope. The endoscope model includes the probe, fiber bundle, lens and other parts. Then, based on the motion constraints of the motion components of the endoscope, the motion range and motion mode of the endoscope are defined, such as rotation, forward, backward, etc., to ensure that the endoscope can cover all areas of the component to be tested.

[0026] Optionally, the endoscope model also defines the optical center position, phase plane coordinates and internal parameter matrix of the endoscope lens or image acquisition component. The optical center position determines the center of the endoscope's field of view during detection. The phase plane coordinates refer to the coordinates of the endoscope lens or image acquisition component on the imaging plane, which are used to describe the position and size of the image. Through the optical center position phase plane coordinates and the internal parameter matrix, the image mapping path of the endoscope lens can be determined in combination with the external parameter matrix, which helps to subsequently map the image to the model of the component to be tested. Among them, the external parameter matrix describes the position and posture of the endoscope in space, including the rotation matrix and translation vector.

[0027] For example, in the optical system of an endoscope, the optical center position and the internal parameter matrix are determined by optical calibration technology. The optical center refers to the center point of the endoscope lens, and the internal parameter matrix is ​​used to describe the internal parameters of the endoscope imaging system, including focal length, principal point coordinates (the position of the optical center on the imaging plane) and distortion coefficients. Specifically, optical calibration technology includes laser calibration, calibration plate, etc.

[0028] Exemplarily, the phase plane refers to the plane where the image sensor is located. In an endoscope, the phase plane is usually the plane of the image sensor chip. The position and direction of the phase plane are determined by the optical design parameters of the endoscope.

[0029] Furthermore, the position and direction of the phase plane are defined in the same three-dimensional coordinate system as the optical center to ensure that the spatial relationship between the two is clear.

[0030] The internal image of the component to be tested is collected through an endoscope to obtain the original surface image.

[0031] Optionally, the endoscope traverses the internal space of the component to be tested to collect images based on preset collection routes and collection parameters, converts optical signals into electrical signals through image sensors, processes the electrical signals through an embedded ISP unit, and outputs the processing results as the original surface image.

[0032] Optionally, the original surface image is a series of local surface images with acquisition marks, and the acquisition marks include acquisition sequence marks and acquisition orientation marks. Among them, the acquisition sequence mark is used to mark the acquisition sequence of the image; the acquisition orientation mark is used to mark the orientation information when the image is acquired, such as the angle and position of the endoscope. In other words, the acquisition orientation mark represents the external parameter matrix when the endoscope acquires the image. The acquisition marks including the acquisition sequence mark and the acquisition orientation mark facilitate the management of the original surface image and the three-dimensional restoration of the subsequent image.

[0033] In some embodiments, collecting an internal image of the component to be tested through an endoscope to obtain an original surface image includes: Based on the model of the component to be tested, multi-level acquisition parameters are defined.

[0034] The image acquisition path planning of the component model to be tested is performed in combination with the multi-level acquisition parameters to obtain an internal image acquisition scheme, wherein the internal image acquisition scheme includes a plurality of serialized acquisition points.

[0035] The endoscope is activated to collect images based on the internal image collection scheme to generate an original surface image.

[0036] Optionally, for areas of different shape complexity of the component to be tested, multi-level acquisition parameters are adaptively configured. Specifically, the higher the shape complexity and the more stringent the damage control requirements are, the higher the accuracy of the corresponding acquisition parameters. Exemplarily, the acquisition parameters include: resolution parameters, that is, setting appropriate resolutions for different areas, and high-precision areas require higher resolutions; acquisition angle parameters, that is, determining the best angle of the endoscope at each acquisition point to ensure full coverage of the inner surface; acquisition distance parameters, that is, the distance between the endoscope and the acquisition point is set according to the size and shape of the component to ensure image quality. Lighting parameters, including the intensity and direction of the endoscope light source, ensure uniform lighting in different areas and improve image clarity.

[0037] Furthermore, combined with multi-level acquisition parameters, the image acquisition path planning is performed on the model of the component to be tested to obtain the internal image acquisition plan. First, a suitable path planning algorithm (such as A* algorithm, Dijkstra algorithm or genetic algorithm) is used to generate the image acquisition path to ensure that the endoscope can cover all internal surface areas. Then, based on the multi-level acquisition parameters, the acquisition points are adaptively distributed on the image acquisition path, and the generated acquisition points are sorted to form serialized acquisition points to ensure the optimal path. Finally, a complete image acquisition plan is formed based on the path planning results, serialized acquisition points and multi-level acquisition parameters. The image acquisition plan includes parameter settings for each acquisition point, including position, angle, distance, lighting and other information.

[0038] Optionally, based on the internal image acquisition scheme, the endoscope is moved to the first acquisition point, the acquisition angle and distance are adjusted, and then the endoscope camera and light source are activated to acquire the image of the first acquisition point to obtain the original surface image. Next, the endoscope is moved to each acquisition point in turn, and the positioning and image acquisition process is repeated until all points are acquired, and the acquired original surface image is stored to obtain the original surface image.

[0039] Through the above method steps, reasonable path planning and multi-level parameter setting are carried out to ensure that the endoscope can fully cover the inner surface of the component to be tested, and adjust the acquisition parameters according to the detection requirements of different areas to ensure the imaging quality of high-precision areas and meet the detection requirements.

[0040] Perform multidimensional feature extraction on the original surface image to obtain a multidimensional feature set, and perform image screening on the original surface image based on the multidimensional feature set to obtain a standard surface image.

[0041] Optionally, during the image acquisition process, due to the complex operating environment of the endoscope, various invalid images with poor acquisition effects may be generated. In order to ensure the accuracy and reliability of subsequent analysis, these invalid images need to be filtered. Exemplary invalid images with poor acquisition effects include abnormal exposure (oversaturation or underexposure), glare spots, distortion (geometric distortion or geometric distortion), blur (motion blur or out-of-focus blur).

[0042] Optionally, the multidimensional feature set refers to a set of feature indicators related to the grayscale information of the original image, exemplarily including average value, contrast, information entropy, energy, second-order moment, third-order moment, etc. The quality and details of the original surface image can be fully described by the multidimensional feature set.

[0043] In some embodiments, Figure 2 As shown, performing multidimensional feature extraction on the original surface image to obtain a multidimensional feature set, and performing image screening on the original surface image based on the multidimensional feature set to obtain a standard surface image, including: The original surface image is grayed, and a gray-level co-occurrence matrix of the gray-level image is obtained by solving the gray-level co-occurrence matrix.

[0044] Based on the gray level co-occurrence matrix, gray level information entropy and a first contrast are calculated and obtained.

[0045] A scatter point coordinate system is established, and the grayscale information entropy and the first contrast are respectively used as the first coordinate and the second coordinate of the scatter point coordinate system to perform scatter point distribution of the original surface image.

[0046] A cluster analysis is performed on the scatter point distribution results, and the original surface image is screened according to the cluster analysis results to eliminate multiple original images included in abnormal clusters to generate the standard surface image.

[0047] Specifically, a multidimensional feature set of the original surface image is obtained. First, the original surface image is grayed to convert the original surface image, which is a chromaticity image, into a gray image, wherein the graying method includes an average method, a weighted average method, a component method, a summation method, etc., and graying is performed based on an RGB model or an HIS model, wherein the gray value is obtained by weighted summation of the three RGB channels based on the RGB model, which is suitable for most cases. Preferably, graying is performed based on an IHS model, and the gray value is obtained by extracting the intensity component, thereby retaining the image brightness information.

[0048] Optionally, based on the grayscale image, the grayscale co-occurrence matrix (GLCM) is solved. The grayscale co-occurrence matrix is ​​a matrix that describes the spatial relationship between pixel value pairs in the grayscale image. It can reflect the comprehensive information of the image and grayscale about the direction and the amplitude of the change of adjacent intervals. Specifically, the elements in the grayscale co-occurrence matrix represent the probability of a certain grayscale value appearing around the pixel with grayscale i.

[0049] Optionally, based on a statistical method, the grayscale information entropy and the first contrast ratio are calculated based on the grayscale co-occurrence matrix, wherein the grayscale information entropy is used to reflect the randomness and complexity of the grayscale value of the image, and the first contrast ratio is used to measure the degree of change of the grayscale value of the image. Specifically, the grayscale information entropy is calculated based on the following formula: ; in, Represents the normalized value of the i-th row and j-th column in the gray-level co-occurrence matrix.

[0050] Specifically, the first contrast is calculated based on the following formula: ; Furthermore, a scatter point coordinate system is established, with grayscale information entropy as the first coordinate and contrast as the second coordinate, and each original surface image is represented as a point in the scatter point coordinate system. In other words, the grayscale information entropy and contrast of each image are used as coordinates and represented in the scatter plot. Then, the scatter plot is analyzed using a clustering algorithm to identify different categories of images, and each image category corresponds to a cluster cluster. Exemplarily, the scattered points in the scatter plot are divided into clusters of arbitrary shapes and noise points are detected through density-based DBSCAN clustering.

[0051] Optionally, the original image corresponding to the abnormal cluster in the clustering result is removed, and the image retained after removal is the standard surface image. Through the above method, high-quality standard surface images can be effectively screened out from the original surface images, and invalid images that may be generated during the acquisition process can be removed. This not only improves the quality of image data, but also provides a more reliable data basis for subsequent detection and analysis.

[0052] The interactive control component obtains a motion parameter sequence of the endoscope optical component, and fits the standard surface image to the twin digital model based on the motion parameter sequence to obtain a three-dimensional restored image.

[0053] Optionally, the imaging image can be restored and mapped to the inner surface of the model of the component to be tested by determining the position of the phase plane and the optical center. Specifically, first, according to the optical system parameters of the endoscope (i.e., the internal parameter matrix), the imaging image is geometrically corrected through a polynomial distortion model or other optical distortion correction algorithm to eliminate image distortion caused by optical distortion. Then, through the interactive control component, the motion parameters of the endoscope during the image acquisition process are called to obtain a motion parameter sequence, and the initial position of the coordinate system of the endoscope optical component is initialized based on the motion parameter sequence, wherein the motion parameter sequence includes multiple sets of serialized position parameters, angle parameters and other information, and the position parameters define the specific coordinates of the optical center of the endoscope optical component in three-dimensional space. The angle parameter refers to the rotation angle (such as pitch angle, yaw angle and roll angle) of the coordinate system of the endoscope optical component relative to the model coordinate system.

[0054] Further, based on the position of the optical center and the phase plane, the pixel coordinates in the imaging image are mapped to the inner surface of the model of the component to be tested. This includes converting the coordinates of each pixel in the imaging image into the phase plane coordinate system to obtain the spatial coordinates of the pixel on the phase plane. Next, starting from the optical center, ray tracing is performed through the spatial coordinates of each pixel to determine the intersection of the ray on the inner surface of the component to be tested. Furthermore, the spatial coordinates of the intersection of the ray are converted into the inner surface coordinate system of the model of the component to be tested to obtain the mapping position of the imaging image on the inner surface.

[0055] Optionally, the mapping results of multiple imaging images are fused to generate a complete inner surface image of the component to be tested. Exemplarily, the images are superimposed and spliced ​​according to the motion trajectory of the endoscope to form a complete three-dimensional inner surface image of the component to be tested. During the fusion process, image splicing, image registration and other technologies can be used to ensure seamless connection between images.

[0056] Through the above method, the motion parameters of the endoscope are used to accurately map the collected images to the digital model, so as to achieve accurate three-dimensional restoration of the inner surface. The obtained three-dimensional restored image can more intuitively reflect the internal defect status of the component to be tested, which is helpful to discover and deal with potential problems.

[0057] Based on the shape features of the model of the component to be tested, the three-dimensional restored image is adaptively meshed to obtain a meshed three-dimensional image.

[0058] Optionally, adaptive meshing of the 3D restored image is performed. First, based on the 3D model of the component to be tested, its shape feature information is extracted, including surface features (the curvature change of the model, used to identify the main surfaces and inflection points); boundary features (used to identify the boundary lines and boundary points of the model, especially the edges with sharp changes); angle features: detect areas in the model with significant angle changes, such as sharp corners or corners.

[0059] Optionally, preliminary meshing is performed based on the overall size of the model of the component to be tested. The initial mesh can be uniform or based on the rough features of the model. Then, based on the above-mentioned shape features, the initial mesh is adaptively adjusted. The adjustment process includes: refining the mesh, refining the mesh division in areas with large curvature changes and significant boundary and angle features, and increasing the mesh density. Optimizing the mesh shape, optimizing the shape and size of the mesh based on the feature points and feature lines, so that the mesh better adapts to the shape characteristics of the model.

[0060] Furthermore, the grid nodes are mapped to the coordinate system of the three-dimensional restored image. According to the grid division, the three-dimensional restored image is cut into multiple grid units.

[0061] In combination with the data dimension reduction method, a multi-channel classification detector is constructed to perform intelligent flaw detection of the gridded three-dimensional image, wherein the multi-channel classification detector includes multiple flaw detection channels for performing different types of flaw detection tasks.

[0062] In some embodiments, a multi-channel classification detector is constructed in combination with a data dimension reduction method, including: Acquire historical flaw detection records of the target environment, wherein the historical flaw detection records include historical surface images and associated historical damage information.

[0063] The interactive target flaw detection scenario obtains the flaw detection task information and extracts the flaw detection target set.

[0064] The broad-spectrum feature data of the historical surface image is extracted, and based on the flaw detection target set, the broad-spectrum feature data is grouped and dimensionally reduced to obtain a multi-target key feature set.

[0065] According to the multi-target key feature set, a plurality of classification detection channels are constructed, and supervised training is performed based on the historical flaw detection records, and the plurality of classification detection channels after training are integrated to obtain a multi-channel classification detector.

[0066] Specifically, first, collect and organize the historical flaw detection records of the target environment, including historical surface images and related historical damage information, and clean and organize the collected historical flaw detection records to ensure the consistency and integrity of the data. Then, interact with the target flaw detection scene to obtain the current flaw detection task information. For example, the flaw detection task information includes the type of the component to be tested, the flaw detection requirements, the type of damage to be detected, etc. Next, set the multiple detected damage types in the flaw detection task information as flaw detection targets to form a flaw detection target set. For example, the damage types include cracks, corrosion, dents, deformation, etc.

[0067] Furthermore, image processing technology is used to extract broad-spectrum feature data of multiple historical surface images, and the broad-spectrum feature data includes all feature index items in the historical images of the component to be tested. Exemplarily, the broad-spectrum feature data includes texture features, color features, morphological features, grayscale features, etc.

[0068] Optionally, the broad-spectrum feature data is grouped according to the detection target set, and each group of feature indicators corresponds to a detection target. In other words, each group of feature indicators is a discriminant feature indicator of a possible type of damage. Then, using dimensionality reduction methods such as principal component analysis (PCA), the group feature indicators are reduced in dimension, key features are extracted, and output as a multi-target key feature set. Through the above steps, feature indicators that actively contribute to damage identification can be obtained, which helps to improve the efficiency of subsequent detection and analysis processes and reduce irrelevant interference features.

[0069] Optionally, a classification detection channel is constructed for each flaw detection target based on the multi-target key feature set, and the input of the channel is the key feature after dimensionality reduction. Then, supervised training is performed on each classification detection channel using the labeled data in the historical flaw detection records. During the training process, the parameters of the classifier are adjusted to improve the detection accuracy. Then, the trained multiple classification detection channels are integrated to form a multi-channel classification detector, and then the integrated multi-channel classification detector is optimized and verified to ensure its accuracy and robustness in various flaw detection tasks.

[0070] In some implementations, extracting the broad-spectrum feature data of the historical surface image, and grouping and reducing the broad-spectrum feature data based on the flaw detection target set to obtain a multi-target key feature set, further includes: A first flaw detection target is acquired based on the flaw detection target set, where the first flaw detection target corresponds to any damage type.

[0071] According to the first flaw detection target, the historical flaw detection records are traversed to extract a first historical flaw detection record, where the first historical flaw detection record includes a first historical surface image and first historical damage information.

[0072] Solve the grayscale co-occurrence matrix, grayscale difference, and color moment of the first historical surface image, and perform feature extraction to obtain first broad-spectrum feature data, wherein the first broad-spectrum feature data at least includes grayscale information entropy, first contrast, energy, differential entropy, second contrast, average value, first-order moment, and second-order moment.

[0073] Perform dimensionality reduction analysis on the first broad-spectrum feature data, extract main features, and store them as a first target key feature set.

[0074] The detection target set is traversed to obtain a multi-target key feature set.

[0075] Specifically, first, based on the flaw detection target, the corresponding historical flaw detection record is obtained, and the historical flaw detection record includes the historical surface image and historical damage information consistent with the type of the flaw detection target. Then, the gray level co-occurrence matrix, gray level difference, and color moment are solved for the first historical surface image, and feature extraction is performed to obtain the first broad spectrum feature data.

[0076] Optionally, the first broad spectrum feature data includes at least grayscale information entropy, first contrast, energy, differential entropy, second contrast, average value, first-order moment, and second-order moment. Specifically, grayscale information entropy is an indicator for measuring the complexity of image grayscale distribution. The larger the grayscale information entropy, the richer the texture information of the image; the first contrast is an indicator for measuring image grayscale contrast calculated based on the grayscale co-occurrence matrix. The larger the contrast, the more drastic the grayscale change of the image; energy (ASM) is an indicator for measuring image texture uniformity. The larger the energy, the more uniform the texture of the image; differential entropy is an indicator for measuring image grayscale differential complexity based on grayscale difference. The larger the differential entropy, the richer the grayscale differential information of the image; the second contrast is an indicator for measuring image grayscale contrast based on grayscale difference calculation; the average value is an indicator for measuring the average grayscale level of the image; the first-order moment is an indicator for representing the mean of the grayscale distribution; the second-order moment is an indicator for measuring the second-order statistical characteristics (such as variance) of the grayscale distribution of the image.

[0077] Furthermore, based on the corresponding flaw detection target, the broad-spectrum feature data is subjected to dimensionality reduction analysis, and the main features are extracted and stored as the first target key feature set.

[0078] In some implementations, multiple classification detection channels are constructed based on the multi-target key feature set, and supervised training is performed based on the historical flaw detection records, and the multiple classification detection channels after training are integrated to obtain a multi-channel classification detector, which also includes: According to the first target key feature set, a first classification detection model is constructed based on a classification algorithm.

[0079] The first historical flaw detection record is used as a training data set to perform supervised training on the first classification detection model to obtain a first classification detection channel.

[0080] Based on the multi-target key feature set and the historical flaw detection records, a plurality of the classification detection channels are constructed and trained.

[0081] A partition constraint of the adaptive grid partition is obtained, and based on the partition constraint, an input distributor is configured, a plurality of the classification detection channels are connected to the input distributor, and the multi-channel classification detector is constructed.

[0082] Optionally, first, multiple feature indicators in the first target key feature set are used as input data classes, and a first classification detection model is constructed based on a classification algorithm, wherein the classification algorithm includes a support vector machine, a decision tree, a neural network (such as KNN), etc. Then, historical flaw detection records are used as a training data set to identify damage labels and train each classification detection model so that it can predict the corresponding flaw detection category based on the input features.

[0083] Furthermore, the above process is repeated for each target key feature set to construct multiple classification detection channels. Each channel is an independent classification detection model trained for a specific target key feature set. Finally, an input distributor is constructed to connect all classification detection channels to form a multi-channel classification detector. The input distributor can distribute input data to appropriate classification detection channels according to the partitioning constraints of the adaptive grid partitioning. Exemplarily, if the partitioning constraints include the position of the component to be tested and the corresponding multiple flaw detection targets, through the above partitioning constraints, the input distributor transmits the input data to the classification detection channels corresponding to the multiple flaw detection targets to perform flaw detection.

[0084] Through the above method steps, special training is performed for different feature sets to construct multiple classification detection channels specifically suitable for a single flaw detection target. In other words, each classification detection channel is specially trained for a specific flaw detection target, and can detect its dedicated target more accurately and effectively without being interfered by other targets, which helps to improve the interpretability and detection efficiency of the multi-channel classification detector.

[0085] In addition, the design of the multi-channel classification detector allows it to be deployed more flexibly. For example, if a new flaw detection target needs to be added, only a new classification detection channel needs to be added without retraining the entire system. At the same time, if a channel has a problem, it can be repaired or replaced separately without affecting the operation of other channels.

[0086] Furthermore, the intelligent flaw detection of the gridded three-dimensional image also includes: Obtain the flaw detection results, parse and obtain multiple edge damages, and output them as an edge damage set.

[0087] Based on the damage type, damage location, damage direction, and damage size, a continuity analysis is performed on the edge damage set to obtain a continuity analysis result, wherein the continuity analysis result includes a plurality of edge damage groups.

[0088] Based on the continuity analysis result, damage merging processing of the edge damage set is performed to obtain a merged edge damage set.

[0089] The combined edge damage set is updated to the flaw detection result.

[0090] Optionally, due to the a posteriori nature of feature recognition and damage detection, when obtaining a gridded three-dimensional image, a complete damage image is often divided into multiple discrete parts, which affects the accuracy of the flaw detection analysis and the continuity and integrity of the damage in the analysis results. Therefore, based on the damage located at the edge of the grid in the gridded three-dimensional image in the flaw detection results, an edge damage set is obtained to merge connected damages, reduce repeated and redundant damages, and provide clearer and more accurate flaw detection results.

[0091] Specifically, first, the flaw detection results are obtained from a multi-channel classification detector. The flaw detection results include a series of damages, each of which includes its position information, size information and type information. Then, multiple edge damages are parsed from the flaw detection results, and these damages are output as an edge damage set. Next, based on the damage type, damage location, damage direction and damage size, the edge damage set is subjected to continuity analysis. Exemplarily, it involves checking whether multiple damages are continuous or close in space, direction or type, and then these continuous damages are grouped into an edge damage group, and the analysis process is repeated to output multiple edge damage groups as continuity analysis results.

[0092] Furthermore, after the continuity analysis is performed, the edge damage set is subjected to damage merging processing. Specifically, each edge damage set in the continuity analysis result is merged into a single damage to obtain a merged edge damage set; finally, the merged edge damage set is updated to the flaw detection result to replace the edge damage set.

[0093] Through the above steps, clearer and more continuous damage images can be provided, thereby improving the accuracy and interpretability of flaw detection.

[0094] In some implementations, the edge damage set is subjected to continuity analysis based on damage type, damage location, damage direction, and damage size, and then the method further includes: The continuity analysis result is analyzed, type difference discrimination is performed on the plurality of edge damage groups, and a plurality of difference edge damage groups are extracted.

[0095] A plurality of the differential edge damage groups are merged to obtain a plurality of differential damage partitions.

[0096] The multi-channel classification detector is activated to perform update detection on the plurality of differential damage partitions to obtain an updated damage type set.

[0097] The edge damage set is subjected to damage merging processing in combination with the updated damage type set.

[0098] Optionally, due to the incompleteness of edge damage in the edge damage group, the multi-channel classification detector may have deviations in the flaw detection of edge damage, thereby misjudging the damage type of edge damage. Therefore, multiple differential edge damage groups are merged separately, that is, multiple differential edge damage groups are regarded as new grids, and multiple differential damage partitions are obtained. Then, the multi-channel classification detector is activated to update the detection of multiple differential damage partitions to obtain more accurate damage type information (i.e., update the damage type set). Finally, combined with the updated damage type set, the edge damage set is subjected to damage merging processing, including marking the damage type of the merged edge damage set with the updated damage type set. This allows for better identification and understanding of the characteristics of damage, and more effective processing and organization of damage information.

[0099] In summary, the intelligent detection method based on non-destructive testing equipment provided by the present invention has the following technical effects: By establishing a twin digital model, the model includes a model of the component to be tested and an endoscope model. The internal image of the component to be tested is collected through the endoscope to obtain the original surface image. Multidimensional feature extraction is performed on the original surface image to obtain a multidimensional feature set. Based on this multidimensional feature set, the original surface image is screened to obtain a standard surface image. Interact with the interactive control component to obtain a motion parameter sequence of the endoscope optical component. Based on these motion parameter sequences, the standard surface image is fitted to the twin digital model to generate a three-dimensional restored image. Based on the shape characteristics of the model of the component to be tested, the three-dimensional restored image is adaptively meshed to obtain a gridded three-dimensional image. Combined with the data dimensionality reduction method, a multi-channel classification detector is constructed for intelligent flaw detection of the gridded three-dimensional image. This multi-channel classification detector includes multiple flaw detection channels, each of which is responsible for performing different types of flaw detection tasks. Thereby achieving the technical effect of reducing detection costs and reducing operation difficulty.

[0100] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. An intelligent detection method based on non-destructive testing equipment, characterized in that: The method comprises: Establish a twin digital model, including the model of the component to be tested and the endoscope model; Collect the internal image of the component to be tested through the endoscope to obtain the original surface image; Performing multidimensional feature extraction on the original surface image to obtain a multidimensional feature set, and performing image screening on the original surface image based on the multidimensional feature set to obtain a standard surface image; The interactive control component obtains a motion parameter sequence of the endoscope optical component, and fits the standard surface image to the twin digital model based on the motion parameter sequence to obtain a three-dimensional restored image; Based on the shape characteristics of the model of the component to be tested, adaptively meshing the three-dimensional restored image to obtain a meshed three-dimensional image; In combination with the data dimension reduction method, a multi-channel classification detector is constructed to perform intelligent flaw detection of the gridded three-dimensional image, wherein the multi-channel classification detector includes multiple flaw detection channels for performing different types of flaw detection tasks.

2. The intelligent detection method based on non-destructive testing equipment according to claim 1, characterized in that: The internal image of the component to be tested is collected through the endoscope to obtain the original surface image, including: Based on the model of the component to be tested, defining multi-level acquisition parameters; Perform image acquisition path planning of the component model to be tested in combination with the multi-level acquisition parameters to obtain an internal image acquisition scheme, wherein the internal image acquisition scheme includes a plurality of serialized acquisition points; The endoscope is activated to collect images based on the internal image collection scheme to generate an original surface image.

3. The intelligent detection method based on non-destructive testing equipment according to claim 2 is characterized in that: Extracting multidimensional features of the original surface image to obtain a multidimensional feature set, and performing image screening on the original surface image based on the multidimensional feature set to obtain a standard surface image, including: Graying the original surface image, and solving a gray-level co-occurrence matrix of the gray-level image; Based on the gray level co-occurrence matrix, gray level information entropy and a first contrast ratio are calculated and obtained; Establishing a scatter point coordinate system, using the grayscale information entropy and the first contrast as the first coordinate and the second coordinate of the scatter point coordinate system respectively, and performing scatter point distribution of the original surface image; A cluster analysis is performed on the scatter point distribution results, and the original surface image is screened according to the cluster analysis results to eliminate multiple original images included in abnormal clusters to generate the standard surface image.

4. The intelligent detection method based on non-destructive testing equipment according to claim 3 is characterized in that: Combined with the data dimensionality reduction method, a multi-channel classification detector is constructed, including: Acquire historical flaw detection records of the target environment, wherein the historical flaw detection records include historical surface images and associated historical damage information; The interactive target flaw detection scenario obtains the flaw detection task information and extracts the flaw detection target set; Extracting broad-spectrum feature data of the historical surface image, and grouping and reducing the dimension of the broad-spectrum feature data based on the flaw detection target set to obtain a multi-target key feature set; According to the multi-target key feature set, a plurality of classification detection channels are constructed, and supervised training is performed based on the historical flaw detection records, and the plurality of classification detection channels after training are integrated to obtain a multi-channel classification detector.

5. The intelligent detection method based on non-destructive testing equipment according to claim 4, characterized in that: Extracting the broad-spectrum feature data of the historical surface image, and grouping and reducing the dimension of the broad-spectrum feature data based on the flaw detection target set to obtain a multi-target key feature set, further comprising: Acquire a first flaw detection target based on the flaw detection target set, where the first flaw detection target corresponds to any damage type; According to the first flaw detection target, traverse the historical flaw detection records to extract a first historical flaw detection record, where the first historical flaw detection record includes a first historical surface image and first historical damage information; Solving the grayscale co-occurrence matrix, grayscale difference, and color moment of the first historical surface image, and performing feature extraction to obtain first broad-spectrum feature data, wherein the first broad-spectrum feature data at least includes grayscale information entropy, first contrast, energy, differential entropy, second contrast, average value, first-order moment, and second-order moment; Performing dimensionality reduction analysis on the first broad-spectrum feature data, extracting main features, and storing them as a first target key feature set; The detection target set is traversed to obtain a multi-target key feature set.

6. The intelligent detection method based on non-destructive testing equipment according to claim 4, characterized in that: According to the multi-target key feature set, multiple classification detection channels are constructed, and supervised training is performed based on the historical flaw detection records, and the multiple classification detection channels after training are integrated to obtain a multi-channel classification detector, which also includes: According to the first target key feature set, construct a first classification detection model based on a classification algorithm; Using the first historical flaw detection record as a training data set, performing supervised training on the first classification detection model to obtain a first classification detection channel; Based on the multi-target key feature set and the historical flaw detection records, construct and train a plurality of the classification detection channels; A partition constraint of the adaptive grid partition is obtained, and based on the partition constraint, an input distributor is configured, a plurality of the classification detection channels are connected to the input distributor, and the multi-channel classification detector is constructed.

7. The intelligent detection method based on non-destructive testing equipment according to claim 1, characterized in that: The intelligent flaw detection of the gridded three-dimensional image also includes: Obtain the flaw detection results, parse and obtain multiple edge damages, and output them as an edge damage set; Based on the damage type, damage location, damage direction, and damage size, a continuity analysis is performed on the edge damage set to obtain a continuity analysis result, wherein the continuity analysis result includes a plurality of edge damage groups; Based on the continuity analysis result, performing damage merging processing on the edge damage set to obtain a merged edge damage set; The combined edge damage set is updated to the flaw detection result.

8. The intelligent detection method based on non-destructive testing equipment according to claim 7, characterized in that: Based on the damage type, damage location, damage direction, and damage size, the edge damage set is subjected to continuity analysis, and then, the following is further included: Analyzing the continuity analysis results, distinguishing the types of the plurality of edge injury groups, and extracting a plurality of difference edge injury groups; Merging a plurality of the differential edge damage groups to obtain a plurality of differential damage partitions; activating the multi-channel classification detector to perform update detection on the plurality of differential damage partitions to obtain an updated damage type set; The edge damage set is subjected to damage merging processing in combination with the updated damage type set.

Images