A crack identification method, device, equipment, medium and product

Abstract

This application discloses a crack identification method, apparatus, device, medium, and product. The method includes: generating a linear array scan image based on ultrasonic imaging data acquired from a test object under preset multi-condition constraints; preprocessing the linear array scan image using a preset method to obtain an enhanced image; inputting the enhanced image into a preset crack identification network to extract crack shared features; and inputting the crack shared features into a preset crack segmentation head and a preset crack size quantization head respectively to determine the crack identification result. The technical solution of this application embodiment can improve the accuracy of crack identification.

Description

Technical Field

[0001] This application relates to the field of image processing and detection technology, and in particular to a crack recognition method, apparatus, device, medium and product. Background Technology

[0002] Long-distance steel pipelines are the core mode of oil and gas energy transportation, and the integrity of their circumferential welds directly determines the safe operation of the pipeline. During the welding process, due to factors such as technology and environment, welds are prone to defects such as porosity and cracks. Among them, longitudinal cracks can easily lead to pipeline leakage or even breakage. However, existing detection technologies still have shortcomings in crack detection. Ultrasonic phased array testing has become one of the mainstream non-destructive testing methods due to its high sensitivity and applicability to complex workpieces.

[0003] Currently, crack assessment of ultrasonic phased array linear scan images relies on manual interpretation, which suffers from high subjectivity, low efficiency, limited accuracy, and poor repeatability, making it difficult to meet the needs of large-scale, high-precision inspection of natural gas pipelines. Therefore, the industry is committed to researching technologies such as deep learning to replace manual inspection and improve the automation level and accuracy of pipeline crack detection. Summary of the Invention

[0004] This application provides a crack identification method, apparatus, device, medium, and product to improve the accuracy of crack detection.

[0005] According to one aspect of this application, a crack identification method is provided, comprising: Based on the ultrasound imaging data collected from the object under test under preset multi-condition constraints, a linear scan image is generated. The linear scan images are preprocessed using a preset method to obtain enhanced images; The enhanced image is input into a preset crack recognition network to extract shared crack features; The crack shared features are input into the preset crack segmentation head and the preset crack size quantization head respectively to determine the crack recognition result.

[0006] According to another aspect of this application, a crack detection device is provided, comprising: The scanning image generation module is used to generate linear scan images based on the ultrasound imaging data acquired from the object under test under preset multi-condition constraints. The enhanced image determination module is used to preprocess the linear scan image in a preset manner to obtain an enhanced image; The shared feature extraction module is used to input the enhanced image into a preset crack recognition network and extract the crack shared features; The crack result recognition module is used to input the shared crack features into a preset crack segmentation head and a preset crack size quantization head to determine the crack recognition result.

[0007] According to another aspect of this application, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the crack identification method according to any embodiment of this application.

[0008] According to another aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the crack identification method according to any embodiment of this application.

[0009] According to another aspect of this application, a computer program product is provided, the computer program product including a computer program that, when executed by a processor, implements the crack identification method according to any embodiment of this application.

[0010] In the technical solution of this application embodiment, by acquiring ultrasonic imaging data and generating linear array scan images under preset multi-condition constraints, noise interference can be effectively suppressed, and the relevance and consistency of the original data can be improved, laying a high-quality foundation for subsequent processing. Secondly, the linear array scan images are preprocessed to obtain enhanced images, which can further highlight crack edges and texture information and reduce the influence of background clutter. Furthermore, crack recognition networks are used to extract shared crack features, and these shared features are fed to the crack segmentation head and crack size quantization head respectively, realizing parallel processing of crack contour localization and size evaluation, avoiding computational redundancy caused by repeated feature extraction, and improving system efficiency. Finally, the collaborative constraints of segmentation and quantization tasks enable the network to learn crack morphology and scale information simultaneously, significantly improving the crack recognition accuracy.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a crack identification method according to Embodiment 1 of this application; Figure 2A This is a flowchart of a crack identification method according to Embodiment 2 of this application; Figure 2B This is a schematic diagram of partial channel convolution provided in Embodiment 2 of this application; Figure 2C This is the original image of the crack provided according to Embodiment 2 of this application; Figure 2D This is an enhanced image of a crack after reinforcement, provided according to Embodiment 2 of this application; Figure 2E This is a schematic diagram of an actual test provided according to Embodiment 2 of this application; Figure 2F These are the original ultrasonic scan images from the actual test provided in Embodiment 2 of this application; Figure 2G These are images processed according to actual tests provided in Embodiment 2 of this application; Figure 2H This is a schematic diagram of the defect number in the actual test provided according to Embodiment 2 of this application; Figure 3 This is a schematic diagram of a crack identification device according to Embodiment 3 of this application; Figure 4 This is a schematic diagram of the structure of an electronic device that implements the crack identification method of the embodiments of this application. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] Example 1 Figure 1 This application provides a flowchart of a crack identification method according to Embodiment 1. This embodiment is applicable to situations requiring crack size detection and quantification. The method can be executed by a crack identification device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method includes: S110. Generate a linear scan image based on the ultrasonic imaging data collected from the object under the preset multi-condition constraints.

[0017] The object under test can be any object requiring crack identification and detection, such as online / offline detection of welds or surface cracks in oil and gas pipelines. Ultrasonic imaging data can be obtained by an ultrasonic phased array detection system detecting the object under test, acquiring phased array ultrasonic imaging data corresponding to internal or surface crack defects. Multiple constraints can be constraints on different ultrasonic scanning methods during the ultrasonic phased array detection process, such as, but not limited to, beam deflection angle, focal depth, and scanning direction; these are not limited in this embodiment. The linear array scan image can be a visual image obtained by converting the ultrasonic imaging data into an image. It is understood that ultrasonic imaging data is merely ultrasonic data and does not directly show the crack condition; the linear array scan image obtained by image mapping of the ultrasonic imaging data can be used for subsequent image processing. Of course, this embodiment does not limit the specific model and configuration of the ultrasonic phased array detection system.

[0018] S120. The linear scan image is preprocessed using a preset method to obtain an enhanced image.

[0019] Preprocessing can include, but is not limited to, standardization, normalization, or contour enhancement, with the aim of highlighting potential cracks in the linear scan image, thus aiding subsequent image processing. Correspondingly, the enhanced image can be an image that has undergone a series of preprocessing steps. This application does not exhaustively describe or limit the specific preprocessing procedures.

[0020] S130. Input the enhanced image into the preset crack recognition network and extract the crack shared features.

[0021] The crack recognition network can be any pre-set or improved computer vision and image processing model, such as a deep learning model for semantic segmentation, etc., which is not limited in this embodiment. Through this crack recognition network, shared crack features are extracted from the enhanced image. These shared crack features can be characteristic information representing the shape, size, etc., of the cracks in the tested object, extracted from the enhanced image.

[0022] S140. Input the crack shared features into the preset crack segmentation head and the preset crack size quantization head respectively, and determine the crack recognition result.

[0023] Both the crack segmentation head and the crack size quantization head belong to the deep learning crack detection framework. The crack segmentation head can be a component used to determine the crack outline in an image; the crack size quantization head can be a component used to calculate the actual size of the crack based on the crack outline. After calculation by the crack segmentation head and the crack size quantization head, the crack identification result of the tested object is obtained. Of course, the crack identification result can include descriptions of various dimensions such as crack morphology, shape, size, length, and width, which are not limited in this embodiment.

[0024] In the technical solution of this application embodiment, by acquiring ultrasonic imaging data and generating linear array scan images under preset multi-condition constraints, noise interference can be effectively suppressed, and the relevance and consistency of the original data can be improved, laying a high-quality foundation for subsequent processing. Secondly, the linear array scan images are preprocessed to obtain enhanced images, which can further highlight crack edges and texture information and reduce the influence of background clutter. Furthermore, crack recognition networks are used to extract shared crack features, and these shared features are fed to the crack segmentation head and crack size quantization head respectively, realizing parallel processing of crack contour localization and size evaluation, avoiding computational redundancy caused by repeated feature extraction, and improving system efficiency. Finally, the collaborative constraints of segmentation and quantization tasks enable the network to learn crack morphology and scale information simultaneously, significantly improving the crack recognition accuracy.

[0025] Example 2 Figure 2AThis is a flowchart illustrating a crack identification method according to Embodiment 2 of this application. This embodiment further refines the process of determining shared crack features based on the aforementioned embodiments. The crack identification network includes a preset encoder, a crack structure enhancement module, a pooling module, and a decoder. Figure 2A As shown, the method includes: S210. Generate a linear scan image based on the ultrasound imaging data collected from the object under the preset multi-condition constraints.

[0026] S220. The linear scan image is preprocessed using a preset method to obtain an enhanced image.

[0027] S230. Input the enhanced image into the encoder to obtain convolutional output features and low-level features.

[0028] The encoder can be any encoder in the relevant field. In this embodiment, the crack is implemented by the network using DeepLabv3+, which includes a lightweight channel convolutional encoder / decoder, a crack structure enhancement module, and a lightweight hole spatial pyramid pooling module. Correspondingly, the convolutional output features can be features output after convolutional computation, while the low-level features can be features directly extracted by the encoder.

[0029] S240. Input the convolution output features into the crack structure enhancement module, and determine the enhancement features through the preset crack structure enhancement mapping function.

[0030] After completing the feature extraction of the lightweight encoder, to further enhance the network's ability to express effective crack features in terms of slender crack boundaries, directional continuity, and complex backgrounds, a crack structure enhancement module is set up based on the encoder output features to obtain enhanced features. : in, This represents a crack structure enhancement mapping function, used to strengthen the crack backbone structure, boundary continuity, and slender morphological features; The output features are from the convolution. The crack structure enhancement module is mainly used to highlight the effective structural response in the crack region and reduce the interference of complex background, artifacts, and irrelevant scattering features on subsequent recognition.

[0031] S250. Input the enhanced features into the pooling module, and determine the high-level features through the preset lightweight void space pyramid pooling mapping function.

[0032] After obtaining the enhanced features, they are input into a lightweight, hollow spatial pyramid pooling module for multi-scale context aggregation to extract crack semantic information at different scales. Let the high-level features output by the lightweight, hollow spatial pyramid pooling module be... : in, This represents a lightweight void space pyramid pooling mapping function.

[0033] S260. Input the high-level features and low-level features into the decoder for fusion to obtain crack shared features.

[0034] The high-level features processed by the lightweight void space pyramid pooling module are fused with the low-level detail features output by the encoder into the decoder to recover crack boundary details and spatial distribution information, thus obtaining crack shared features. : in, This represents the high-level features after processing by the lightweight void space pyramid pooling module. This represents the low-level features output by the encoder. This represents a lightweight decoder mapping function.

[0035] S270. Input the crack shared features into the preset crack segmentation head and the preset crack size quantization head respectively, and determine the crack recognition result.

[0036] In the technical solution of this application embodiment, by making lightweight improvements to the encoder convolution kernel structure and combining it with a crack structure enhancement module and a lightweight void space pyramid pooling module, the crack recognition network can effectively extract crack boundaries, slender structures and multi-scale contextual features while reducing the number of parameters and computational complexity, thereby improving the overall model's recognition efficiency and engineering application capabilities.

[0037] In one optional implementation, the step of inputting the enhanced image into the encoder in S230 to obtain convolutional output features may include: inputting the enhanced image into the encoder to encode the basic input features; and processing the basic input features based on a preset lightweight convolution operator to obtain convolutional output features.

[0038] Let the enhanced image output in step S2 be The features extracted by the lightweight encoder are represented as follows: in, This represents the lightweight encoder mapping function.

[0039] Since the feature extraction network plays a decisive role in the overall model, its complexity has a significant impact on the overall performance and deployment efficiency of the model. However, the output feature map usually has a certain degree of redundancy in the channel dimension, with high similarity between different channels, and some channels contribute little to the current crack recognition task. Therefore, the standard convolution operator in the encoder is improved by lightweighting.

[0040] Specifically, let the input features be... It is divided into features along the channel dimension to participate in the convolution calculation. and directly retained features Then we have: in, , And satisfy: For the channels involved in convolution calculation Apply convolution operations to obtain convolution output features. : in, This is a standard convolution operator. Then, the convolution output features are... With preserved features Concatenating the features along the channel dimension yields lightweight convolutional output features. : in, This represents the feature concatenation operation. Through this method, the encoder can retain effective information from the input features while reducing the number of parameters and computational cost, thus improving feature extraction efficiency. The partial channel convolution calculation process is as follows: Figure 2B As shown.

[0041] In an optional implementation, the step S270 of inputting the crack shared features into a preset crack segmentation head and a preset crack size quantization head respectively to determine the crack identification result may include: S271. Input the crack shared features into the crack segmentation head to obtain the crack region probability map.

[0042] Among them, the crack region probability map can be used to characterize the probability that cracks exist in each region of the image.

[0043] The shared features extracted by S3 Input the crack segmentation head to obtain the crack region probability map. : in, This represents the crack segmentation head mapping function.

[0044] S272. Based on the crack region probability map, generate a crack mask and extract the initial size of the crack from the crack mask.

[0045] The initial dimensions of the crack may include initial geometric measurements of its length and height.

[0046] Specifically, a crack mask is generated based on the crack region probability map. : in, The preset segmentation threshold is used. The crack mask is used to characterize the spatial distribution, boundary location, and extension area of ​​the crack in the image, providing a geometric basis for subsequent size quantization.

[0047] After obtaining the crack mask, a geometric analysis is performed on the crack region to extract initial measurements of the crack length and height.

[0048] Let the initial crack length be _____. The initial value of the crack height is Among them, the initial value of the crack length It is obtained by the distance between the two endpoints of the crack principal axis, and its calculation formula is: in, and These represent the coordinates of the two endpoints of the crack principal axis.

[0049] Initial value of crack height The distance between the boundaries of the crack region in the direction perpendicular to the principal axis is obtained by measurement, and its calculation formula is: in, Indicates the crack masking area. Represents pixels The boundary distance perpendicular to the crack principal axis. Using the above method, the initial geometric measurements of the crack length and height, i.e., the initial crack dimensions, can be obtained.

[0050] S273. Input the crack shared features into the crack size quantization head to perform pooling at different scales, and obtain pooling features at different scales.

[0051] To preserve the crack statistical information at different scales in the shared crack features, the crack size quantization head does not use a single global pooling method, but instead performs multi-scale pooling processing on the shared features.

[0052] Preferably, respectively using , ,and Pooling methods for shared features Pooling was performed to obtain pooling features at different scales. , ,and : in, , and These represent the pooling functions for the corresponding scales. Pooling features are used to characterize the overall global information of cracks, the and Pooling features are used to characterize the spatial distribution information of cracks at mesoscale and local scales, thereby enhancing the ability of the size quantization head to perceive the overall extension of cracks and changes in local boundaries.

[0053] S274. The pooling features are spliced ​​and fused to obtain multi-scale fused features.

[0054] Pooling features at different scales are concatenated and fused to obtain multi-scale fused features. : in, This indicates a feature splicing operation.

[0055] S275. Determine the crack size compensation amount based on the multi-scale fusion characteristics.

[0056] Similarly, crack size compensation can also be provided in two ways: length and height.

[0057] Furthermore, the multi-scale fusion feature is input into the size quantization mapping layer, which outputs crack length compensation and crack height compensation. in, This represents the crack size quantization head mapping function. This indicates the amount of crack length compensation. This represents the crack height compensation amount. The compensation amount is used to correct geometric measurement errors caused by blurred crack boundaries, localized fractures, noise interference, or weak echo regions.

[0058] S276. Determine the crack identification result based on the initial size and crack size compensation amount.

[0059] The crack identification results can include the physical length and physical height of the crack.

[0060] Obtaining the initial geometric dimensions of the crack and and size compensation amount and Then, the final crack length is calculated. and crack height : Therefore, the final crack size is determined by the geometric measurement value provided by the crack segmentation result and the compensation information output by the crack size quantization head.

[0061] After obtaining the pixel-scale results of the crack length and height, the results are converted into actual physical dimensions by combining the mapping relationship between pixel coordinates and physical coordinates.

[0062] Let the mapping coefficients for the length and height directions be respectively and Then the physical length of the crack and crack physical height They are respectively: in, and It can be determined by the calibration parameters of the imaging system, scanning step distance, sound velocity parameters, wedge parameters, or pixel scale calibration information, etc., and the embodiments of this application do not limit it again.

[0063] In the above embodiments, the crack segmentation head provides the spatial geometric basis, and the crack size quantization head retains global and local statistical information through multi-scale pooling and outputs size compensation, thereby realizing the joint output of crack region identification and crack size quantization, improving the stability and accuracy of automated crack size measurement under complex backgrounds, blurred boundaries and various crack morphologies.

[0064] In one optional embodiment, the preprocessing of the linear scan image in step S220 to obtain an enhanced image may include: S221. Perform preset image standardization processing and preset wavelet denoising processing based on the input image entropy value on the linear array scan image to obtain a denoised image.

[0065] Image standardization processing may include, but is not limited to, grayscale conversion, size unification, and normalization to ensure consistency between subsequent image processing and model input. Preferably, the image is converted into a single-channel grayscale image and scaled to a preset size H×W. Simultaneously, the image pixel values ​​are normalized to map them to a preset range, thereby reducing the impact of differences in brightness range and imaging intensity between different images.

[0066] Let the original linear scan image be... The normalized image It can be represented as: in, and These represent the minimum and maximum pixel values ​​in the image, respectively.

[0067] After standardization, wavelet denoising is performed on the image to reduce background noise and artifact interference, while preserving crack boundaries and detailed structures as much as possible. First, the grayscale entropy of the input image is calculated. This is used to characterize the complexity of image grayscale distribution and the degree of noise interference. The calculation method is as follows: in, Represents grayscale level The probability of appearing in an image. This represents the total number of gray levels. Then, based on the entropy value... Dynamically determine the number of wavelet decomposition levels : in, and Preset entropy threshold Furthermore, after determining the number of wavelet decomposition layers, a threshold prediction neural network is used to adaptively determine the wavelet denoising threshold. Specifically, the normalized input image... The threshold prediction neural network is fed into the wavelet denoising neural network to estimate the threshold, and outputs the wavelet threshold. : in, This represents a wavelet threshold prediction neural network. Then, based on the threshold... Thresholding shrinkage is applied to the wavelet decomposition coefficients, and the denoised enhanced image is obtained through wavelet reconstruction. For example, when using soft thresholding shrinkage, the wavelet coefficients... Processing results It can be represented as: =sign( max(| |- ,0) S222. Process the denoised image according to the preset crack-directed enhancement algorithm to obtain the crack-directed enhancement map.

[0068] After denoising, crack-oriented enhancement is performed on the image to further enhance crack boundaries and slender structures.

[0069] Let the enhanced responses in different directions be as follows: Then the multi-directional enhanced response diagram It can be represented as: The above methods can enhance the boundary continuity and endpoint identification of cracks in different extension directions.

[0070] S223. Process the denoised image according to the preset local contrast enhancement algorithm to obtain a local contrast enhancement image.

[0071] To improve the grayscale difference between the cracked area and the background area, local contrast enhancement is performed on the denoised image, resulting in a local contrast enhancement image. This enhances the grayscale difference between the crack target and the background, thereby improving the visibility of weak crack areas and areas with blurred boundaries.

[0072] The embodiments of this application do not limit the crack-directed enhancement algorithm and the local contrast enhancement algorithm used.

[0073] S224. The denoised image, crack-directed enhancement image, and local contrast enhancement image are weighted and fused according to preset weights to determine the enhanced image.

[0074] To enhance crack boundary and local discriminability while preserving the denoising results, the denoised image, crack-directed enhancement image, and local contrast enhancement image are weighted and fused according to preset weights to obtain the final enhanced image. Let the final enhanced image be... Then we have: in, , , Preset weight parameters to satisfy .like Figure 2C and Figure 2D As shown, Figure 2C For the original image, Figure 2D The enhanced image shows that... Figure 2C The cracks are more intuitive and clear.

[0075] The enhanced image obtained after the above processing This image serves as the input for subsequent crack identification and size quantization models. After the above processing steps, the input image can more fully preserve the boundary, contour, and spatial extension features of the crack region while suppressing background noise and artifact interference. This provides a more reliable data foundation for subsequent pixel-level segmentation of the crack region and accurate prediction of dimensional parameters such as crack length and height.

[0076] In one optional implementation, the step of generating a linear array scan image based on the ultrasonic imaging data acquired from the tested object under preset multi-condition constraints in S210 may include: performing repeated ultrasonic phased array scans on the area to be detected of the tested object under preset multi-condition constraints to obtain ultrasonic imaging data; wherein, the multi-condition constraints include at least one of different beam deflection angles, different focal depths, and different scanning directions; and converting the ultrasonic imaging data into a linear array scan image.

[0077] The object to be tested is identified to determine the area to be inspected. The object to be inspected includes at least one of metal components, weld areas, and pipe structures; the inspection area includes internal areas and / or surface areas where cracks or defects may exist.

[0078] The detection area is scanned using an ultrasonic phased array detection system. During the scanning process, the area under test is scanned in a sector or by electronic line scanning through the coordinated transmission and reception of multiple array elements, delayed focusing, and beam deflection, thereby obtaining the ultrasonic imaging data corresponding to the area under test.

[0079] To improve the coverage of acquired data across different operating conditions, repeated scanning acquisitions are performed on the same detection area under multiple constraints. These constraints include at least one of different beam deflection angles, different focal depths, and different scanning directions. By acquiring multiple sets of phased array imaging data under different scanning conditions, the acquired samples can cover the imaging characteristics of cracks under different echo intensities, boundary sharpness, artifact interference levels, and spatial distribution states, thereby improving the adaptability of subsequent models to complex detection conditions. The phased array ultrasonic imaging data is converted into corresponding linear array scan images, which serve as input data for subsequent image optimization processing, crack identification, and size quantization.

[0080] This application also provides a specific example of a dimensional measurement test on a type B ultrasonic phased array test block with artificial defects. The specific process and results are as follows: The test conditions include pre-selected boards, probes, and wedges. The test procedure is as follows: After coating the probe surface with a coupling agent, the ultrasonic phased array ultrasonic scanning imaging mode is activated, and a row of standard reflective apertures is reciprocated for scanning, acquiring corresponding linear array scan image data. During the test, multiple sets of phased array imaging results are acquired under different scanning directions and different probe movement positions. The obtained images are then input into the method provided in the above embodiments and implementation methods of this application for processing, in order to compare the crack / defect imaging effect and size quantization effect before and after using the method of this invention. The test process is as follows: Figure 2E As shown.

[0081] Specifically, ultrasonic scanning images are first acquired; then, the original images are standardized, subjected to adaptive wavelet denoising based on the input image entropy, crack-directed enhancement, local contrast enhancement, and fusion processing to obtain enhanced images; then, the enhanced images are input into a lightweight crack recognition network to extract crack shared features, and crack region segmentation results and crack size compensation results are obtained through crack segmentation head and crack size quantization head, respectively; finally, the final quantized size of the target defect is obtained by combining the initial geometric size of the crack and the size compensation amount.

[0082] By suppressing background noise and artifact interference through adaptive wavelet denoising based on the input image entropy, and combining crack-directed enhancement, local contrast enhancement and fusion processing, the distinguishability of defect boundaries and contours is improved. On this basis, a lightweight crack recognition network is used to identify and segment the defect region, and then a crack size quantization head is used to compensate and predict the defect length and / or height, thereby realizing automatic quantization of defect size.

[0083] Figure 2F and Figure 2G Comparison images of the imaging effects before and after using the method of this invention are provided. Figure 2F The original ultrasonic scan image shows that the defect edges are relatively rough, the background noise and artifacts are obvious, and there is a certain deviation between the defect contour and the actual boundary. Figure 2G The result is obtained after processing using the above method. It can be seen that the defect boundary is clearer, the outline is more complete, and the slender structure and local boundary features are more prominent, which is more conducive to the subsequent automatic measurement and quantitative analysis of defect size.

[0084] To verify the dimensional quantization accuracy of the method of the present invention, a phased array B-shaped test block was selected as a calibration test block for the experiment. Since the position parameters of each standard reflector in the B-shaped test block are known, they can be used as the reference true value for the quantization result.

[0085] First, phased array imaging results of the B-shaped test block were acquired, and the corresponding numbers for each defect were as follows: Figure 2HAs shown. Then, the acquired phased array linear scan images are input into the method of this invention, and image optimization processing, crack region identification and segmentation, and size quantization are completed sequentially to obtain the pixel size results corresponding to each standard reflector; then, combined with the mapping relationship between pixel coordinates and physical coordinates, the physical position parameters and / or equivalent size parameters of each reflector are calculated. Finally, the quantization results obtained by the method of this application are compared with the nominal values ​​of the B-shaped test block, and the absolute error and relative error are calculated to evaluate the accuracy of the method of this invention in automatic defect size quantization. A comparison table of defect depth quantization results and a comparison table of defect size quantization results for B-shaped test blocks are shown, wherein Table 1 is the comparison table of defect depth quantization results for B-shaped test blocks; Table 2 is the comparison table of defect size quantization results for B-shaped test blocks.

[0086] Table 1 Table 2 Table 1 presents the quantitative measurement results for different artificial defect depth directions, while Table 2 provides a quantitative comparison of the geometric dimensions (such as aperture and depth) of hole-shaped defects. In ultrasonic phased array imaging and quantitative analysis, the accuracy of the measurement results directly reflects the performance of the system and algorithm. That is, the closer the defect depth measured by ultrasonic imaging is to the nominal depth of the test block, and the closer the measured hole size is to the nominal size of the test block, the higher the imaging accuracy, the smaller the quantitative error, and the more reliable the system calibration status and defect identification algorithm.

[0087] Example 3 Figure 3 This is a schematic diagram of a crack identification device provided in Embodiment 3 of this application. Figure 3 As shown, the device 300 includes: The scanning image generation module 310 is used to generate linear scan images based on the ultrasound imaging data acquired from the object under test under preset multi-condition constraints. The enhanced image determination module 320 is used to preprocess the linear scan image in a preset manner to obtain an enhanced image; The shared feature extraction module 330 is used to input the enhanced image into a preset crack recognition network and extract the crack shared features; The crack result recognition module 340 is used to input the crack shared features into the preset crack segmentation head and the preset crack size quantization head respectively to determine the crack recognition result.

[0088] In the technical solution of this application embodiment, by acquiring ultrasonic imaging data and generating linear array scan images under preset multi-condition constraints, noise interference can be effectively suppressed, and the relevance and consistency of the original data can be improved, laying a high-quality foundation for subsequent processing. Secondly, the linear array scan images are preprocessed to obtain enhanced images, which can further highlight crack edges and texture information and reduce the influence of background clutter. Furthermore, crack recognition networks are used to extract shared crack features, and these shared features are fed to the crack segmentation head and crack size quantization head respectively, realizing parallel processing of crack contour localization and size evaluation, avoiding computational redundancy caused by repeated feature extraction, and improving system efficiency. Finally, the collaborative constraints of segmentation and quantization tasks enable the network to learn crack morphology and scale information simultaneously, significantly improving the crack recognition accuracy.

[0089] In one optional implementation, the crack recognition network includes a preset encoder, a crack structure enhancement module, a pooling module, and a decoder; The shared feature extraction module 330 may include: The convolutional feature determination unit is used to input the enhanced image into the encoder to obtain convolutional output features and low-level features; The enhancement feature determination unit is used to input the convolution output features into the crack structure enhancement module and determine the enhancement features through a preset crack structure enhancement mapping function. The high-level feature determination unit is used to input the enhanced features into the pooling module and determine the high-level features through a preset lightweight holed space pyramid pooling mapping function. The shared feature determination unit is used to input high-level features and low-level features into the decoder for fusion to obtain crack shared features.

[0090] In one optional implementation, the convolutional feature determination unit may include: The basic feature transformation subunit is used to input the enhanced image into the encoder and encode it into basic input features; The convolutional feature processing subunit is used to process the basic input features based on a preset lightweight convolution operator to obtain the convolutional output features.

[0091] In one optional embodiment, the crack result identification module 340 may include: The crack probability map determination unit is used to input the crack shared features into the crack segmentation head to obtain the crack region probability map. The initial size determination unit is used to generate a crack mask based on the crack region probability map and extract the initial size of the crack from the crack mask; The pooling feature determination unit is used to input the crack shared features into the crack size quantization head for pooling at different scales, thereby obtaining pooling features at different scales. The fusion feature determination unit is used to concatenate and fuse various pooling features to obtain multi-scale fusion features; The size compensation determination unit is used to determine the crack size compensation amount based on the multi-scale fusion characteristics. The crack identification and compensation unit is used to determine the crack identification result based on the initial size and the crack size compensation amount.

[0092] In one alternative embodiment, the enhanced image determination module 320 may include: The image denoising unit is used to perform preset image normalization processing and preset wavelet denoising processing based on the input image entropy value on the linear scan image to obtain a denoised image. The crack enhancement unit is used to process the denoised image according to the preset crack-directed enhancement algorithm to obtain the crack-directed enhancement map; The contrast enhancement unit is used to process the denoised image according to a preset local contrast enhancement algorithm to obtain a local contrast enhancement map; The image fusion unit is used to weight and fuse the denoised image, the crack-oriented enhancement image, and the local contrast enhancement image according to preset weights to determine the enhanced image.

[0093] In one optional embodiment, the scanned image generation module 310 may include: The ultrasonic imaging data acquisition unit is used to repeatedly scan the area to be detected of the object under test under preset multi-condition constraints to obtain ultrasonic imaging data; wherein, the multi-condition constraints include at least one of different beam deflection angles, different focal depths and different scanning directions. The linear scan image conversion unit is used to convert ultrasound imaging data into linear scan images.

[0094] The crack identification device provided in this application embodiment can execute the crack identification method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects for executing each crack identification method.

[0095] Example 4 Figure 4A schematic diagram of an electronic device 10, which can be used to implement embodiments of this application, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the application described and / or claimed herein.

[0096] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0097] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0098] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as crack identification methods.

[0099] In some embodiments, the crack identification method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or mounted on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the crack identification method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the crack identification method by any other suitable means (e.g., by means of firmware).

[0100] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0101] Computer programs used to implement the methods of this application may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0102] In the context of this application, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0103] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0104] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0105] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0106] This application also discloses a computer program product, which includes a computer program that, when executed by a processor, implements the crack identification method provided in any embodiment of this application. This program product shares the same inventive concept as the crack identification methods disclosed in the embodiments of this application, and therefore will not be described in detail here.

[0107] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this application can be achieved, and this is not limited herein.

[0108] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A crack identification method characterized by, include: Based on the ultrasound imaging data collected from the object under test under preset multi-condition constraints, a linear scan image is generated. The linear scan image is preprocessed using a preset method to obtain an enhanced image; The enhanced image is input into a preset crack recognition network to extract shared crack features; The crack shared features are input into a preset crack segmentation head and a preset crack size quantization head, respectively, to determine the crack identification result.

2. The method of claim 1, wherein, The crack recognition network includes a preset encoder, a crack structure enhancement module, a pooling module, and a decoder; The step of inputting the enhanced image into a preset crack recognition network to extract crack shared features includes: The enhanced image is input into the encoder to obtain convolutional output features and low-level features; The convolutional output features are input into the crack structure enhancement module, and the enhancement features are determined through a preset crack structure enhancement mapping function. The enhanced features are input into the pooling module, and high-level features are determined through a preset lightweight void space pyramid pooling mapping function. The high-level features and the low-level features are input into the decoder for fusion to obtain the crack shared features.

3. The method of claim 2, wherein, The step of inputting the enhanced image into the encoder to obtain convolutional output features includes: The enhanced image is input into the encoder and encoded as basic input features; The basic input features are processed based on a preset lightweight convolution operator to obtain the convolution output features.

4. The method of claim 1, wherein, The step of inputting the shared crack features into a preset crack segmentation head and a preset crack size quantization head respectively to determine the crack identification result includes: The crack shared features are input into the crack segmentation head to obtain a crack region probability map; Based on the crack region probability map, a crack mask is generated, and the initial size of the crack is extracted from the crack mask; The crack shared features are input into the crack size quantization head for pooling at different scales, resulting in pooling features at different scales. The pooling features described above are concatenated and fused to obtain multi-scale fused features; Based on the multi-scale fusion characteristics, the crack size compensation amount is determined; The crack identification result is determined based on the initial size and the crack size compensation amount.

5. The method of claim 1, wherein, The step of preprocessing the linear scan image in a preset manner to obtain an enhanced image includes: The linear array scan image is subjected to preset image normalization processing and preset wavelet denoising processing based on the input image entropy value to obtain a denoised image. The denoised image is processed according to a preset crack-oriented enhancement algorithm to obtain a crack-oriented enhancement map; The denoised image is processed according to a preset local contrast enhancement algorithm to obtain a local contrast enhancement image; The denoised image, the crack-oriented enhancement image, and the local contrast enhancement image are weighted and fused according to preset weights to determine the enhanced image.

6. The method of claim 1, wherein, The step of generating a linear scan image based on the ultrasonic imaging data acquired from the object under preset multi-condition constraints includes: Under preset multi-condition constraints, repeated ultrasonic phased array scanning is performed on the area to be detected of the object under test to obtain ultrasonic imaging data; wherein, the multi-condition constraints include at least one of different beam deflection angles, different focal depths and different scanning directions; The ultrasound imaging data is converted into the linear scan image.

7. A crack identification device characterized by comprising: include: The scanning image generation module is used to generate linear scan images based on the ultrasound imaging data acquired from the object under test under preset multi-condition constraints. An enhanced image determination module is used to preprocess the linear scan image in a preset manner to obtain an enhanced image; A shared feature extraction module is used to input the enhanced image into a preset crack recognition network and extract shared crack features; The crack result recognition module is used to input the shared crack features into a preset crack segmentation head and a preset crack size quantization head respectively, and determine the crack recognition result.

8. An electronic device, comprising: The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the crack identification method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the crack identification method according to any one of claims 1-6.

10. A computer program product, characterised in that, The computer program product includes a computer program that, when executed by a processor, implements the crack identification method according to any one of claims 1-6.

Images