Precise target positioning method and system for flaw detection

Abstract

The invention relates to the technical field of material detection, and discloses an accurate target positioning method and system for flaw detection, and the method comprises the steps: synchronously collecting multi-source flaw detection signals, carrying out the preprocessing of the multi-source flaw detection signals, obtaining a standardized signal sequence, carrying out the fusion analysis of the standardized signal sequence, and generating a fusion feature map; recognizing a suspected defect area according to the fused feature map to obtain a preliminary defect positioning coordinate set, sorting the preliminary defect positioning coordinate set based on the coordinate feature saliency of the fused feature map, generating a priority scanning path, and calculating the signal confidence of the priority scanning path in real time, the priority scanning path is dynamically adjusted based on a preset self-adaptive threshold value to obtain a dynamic scanning path, and multi-angle cross validation is performed on a scanning result of the dynamic scanning path to obtain a final positioning result; according to the method, through feature-level dynamic fusion of the multi-source signals, the recognition sensitivity and positioning precision of complex defects are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of materials testing technology, and in particular to a precise target positioning method and system for flaw detection. Background Technology

[0002] In the field of materials testing, conventional flaw detection methods mostly rely on a single type of flaw detection signal for defect identification and location. While these methods are simple to implement, they generally have significant limitations. The single signal source can only reflect one type of physical characteristic of the defect. When faced with complex defects such as debonding and microcracks within composite materials, the characterization information is incomplete, easily leading to missed detections or ambiguous location. The detection process is static and open-loop, with a fixed scanning path and rigid defect judgment criteria. It cannot adaptively optimize based on real-time signal quality and intermediate detection results. When inspecting complex components or in the presence of environmental noise interference, repeated scanning is common, resulting in low efficiency, and the ability to track nonlinearly expanding defects is poor.

[0003] Existing technologies often employ multiple sensors, presenting data in a simple parallel manner or undergoing post-processing fusion. They lack mechanisms for dynamic, weighted fusion at the feature level, failing to deeply explore the complementary and enhancing relationships between multiple signal sources. The scanning process lacks dynamic decision-making capabilities based on real-time confidence feedback; path adjustments often rely on human experience or simple distance optimization rather than the information content of the signals themselves. Confirmation of the final positioning result often depends on judgment from a single angle or simple comparison, lacking an automatic cross-validation process based on different physical principles to ensure spatial geometric consistency. This results in insufficient reliability and automation of the positioning results. These shortcomings limit the application of existing technologies in high-end manufacturing, precision components, and other scenarios requiring high precision, high reliability, and fully automated inspection. Summary of the Invention

[0004] This invention provides a precise target positioning method and system for flaw detection, the main purpose of which is to address the problems raised in the background section above.

[0005] To achieve the above objectives, the present invention provides a precise target localization method for flaw detection, comprising: S1: Simultaneously acquire multi-source flaw detection signals and preprocess the multi-source flaw detection signals to obtain a standardized signal sequence of the multi-source flaw detection signals; S2: Perform fusion analysis on the standardized signal sequence to generate a fusion feature map of the multi-source flaw detection signal; S3: Identify the suspected defect areas of the multi-source flaw detection signal based on the fused feature map to obtain a preliminary set of defect location coordinates of the multi-source flaw detection signal; S4: Based on the saliency of the coordinate features of the fused feature map, sort the preliminary defect location coordinate set to generate the priority scanning path of the multi-source flaw detection signal; S5: Calculate the signal confidence of the priority scanning path in real time, and dynamically adjust the priority scanning path based on a preset adaptive threshold to obtain the dynamic scanning path of the multi-source flaw detection signal. S6: Perform multi-angle cross-verification on the scanning results of the dynamic scanning path to obtain the final positioning result of the multi-source flaw detection signal.

[0006] Preferably, the synchronous acquisition of multi-source flaw detection signals and the preprocessing of the multi-source flaw detection signals to obtain a standardized signal sequence of the multi-source flaw detection signals include: The multi-source flaw detection signal includes ultrasonic flaw detection signal and eddy current flaw detection signal; The ultrasonic flaw detection signal is subjected to signal conditioning processing to obtain a standardized ultrasonic signal sequence. The eddy current flaw detection signal is subjected to signal analysis processing to obtain the standardized eddy current signal sequence of the eddy current flaw detection signal; The standardized ultrasonic signal sequence and the standardized eddy current signal sequence together constitute the standardized signal sequence.

[0007] Preferably, the step of performing fusion analysis on the standardized signal sequence to generate a fusion feature map of the multi-source flaw detection signal includes: The time-frequency domain features of the standardized signal sequence are obtained by performing wavelet packet transform decomposition on the standardized signal sequence. Spatial autocorrelation analysis is performed on the standardized signal sequence to obtain its spatial correlation characteristics. Based on preset fusion rules, the time-frequency domain features and the spatial correlation features are fused at the feature level to generate a fused feature map of the multi-source flaw detection signal.

[0008] Preferably, based on preset fusion rules, the time-frequency domain features and the spatial correlation features are fused at the feature level to generate a fused feature map of the multi-source flaw detection signal: Based on the signal-to-noise ratio of the standardized signal sequence, the time-frequency domain features and the spatial correlation features are dynamically allocated fusion weights to obtain the fusion weights of the standardized signal sequence. Based on the fusion weights, the time-frequency domain features and the spatial correlation features are weighted and synthesized to generate the fusion feature map of the standardized signal sequence.

[0009] Preferably, the step of identifying suspected defect regions of the multi-source flaw detection signals based on the fused feature map to obtain a preliminary set of defect location coordinates for the multi-source flaw detection signals includes: The pixels in the fused feature map are determined based on the comparison results between the preset threshold and the feature values ​​in the fused feature map. By performing region connectivity on the pixels, the suspected defect regions of the multi-source flaw detection signal are obtained; The location information of the suspected defect area is obtained, and the preliminary defect location coordinate set of the multi-source flaw detection signal is determined based on the location information.

[0010] Preferably, the step of sorting the preliminary defect location coordinate set based on the coordinate feature saliency of the fused feature map to generate the priority scanning path of the multi-source flaw detection signal includes: Based on the fused feature map, feature values ​​of the preliminary defect location coordinate set are extracted to obtain the coordinate feature saliency of the fused feature map; Based on the saliency of the coordinate features, the coordinate sequence of the preliminary defect location coordinate set is sorted to obtain the sorted coordinate sequence of the preliminary defect location coordinate set; Based on the spatial distance between adjacent coordinates of the sorted coordinate sequence and the preliminary defect location coordinate set, a connection path of the coordinate sequence is constructed to obtain the priority scanning path of the multi-source flaw detection signal.

[0011] Preferably, the step of calculating the signal confidence of the priority scanning path in real time and dynamically adjusting the priority scanning path based on a preset adaptive threshold to obtain the dynamic scanning path of the multi-source flaw detection signal includes: Based on the degree of matching between the real-time signal of the priority scanning path and the fused feature map, the signal confidence of the priority scanning path is calculated in real time. Based on the signal confidence level and a preset adaptive threshold, the scanning order and scanning focus depth of the priority scanning path are dynamically replanned to obtain the dynamic scanning path of the multi-source flaw detection signal.

[0012] Preferably, the formula for calculating the signal confidence level is: in, The confidence level of the signal. This represents the distribution density of singular points in the intrinsic mode function obtained after performing a Hilbert-Huang transform on the real-time signal. The effective echo width extracted from the real-time signal. The reference feature width extracted at the corresponding position in the fused feature map. The weighted fusion feature value of the real-time signal, This represents the variance of the energy values ​​of the historical scan signals.

[0013] Preferably, the step of performing multi-angle cross-verification on the scanning results of the dynamic scanning path to obtain the final positioning result of the multi-source flaw detection signal includes: The scanning results of the dynamic scanning path include defect contour features and defect depth features; The defect boundary position and extension depth of the multi-source flaw detection signal are determined based on the defect contour features and the defect depth features to obtain the final positioning result of the multi-source flaw detection signal.

[0014] A precise target positioning system for flaw detection, the system comprising: The signal acquisition and preprocessing module is used to synchronously acquire multi-source flaw detection signals and preprocess the multi-source flaw detection signals to obtain a standardized signal sequence of the multi-source flaw detection signals. The fusion analysis module is used to perform fusion analysis on the standardized signal sequence and generate a fusion feature map of the multi-source flaw detection signal; The defect identification module is used to identify suspected defect areas of the multi-source flaw detection signal based on the fused feature map, so as to obtain a preliminary set of defect location coordinates of the multi-source flaw detection signal; The path planning module is used to sort the preliminary defect location coordinate set based on the coordinate feature saliency of the fused feature map, and generate the priority scanning path of the multi-source flaw detection signal; The dynamic adjustment module is used to calculate the signal confidence of the priority scanning path in real time and dynamically adjust the priority scanning path based on a preset adaptive threshold to obtain the dynamic scanning path of the multi-source flaw detection signal. The cross-validation module is used to perform multi-angle cross-validation on the scanning results of the dynamic scanning path to obtain the final positioning result of the multi-source flaw detection signal.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention effectively overcomes the problem of incomplete defect characterization caused by relying on a single signal source in existing technologies by introducing a feature-level dynamic fusion analysis mechanism for multi-source flaw detection signals. By simultaneously acquiring flaw detection signals from multiple physical principles such as ultrasonic waves and eddy currents and performing standardized processing, wavelet packet transform and spatial autocorrelation analysis are used to extract their time-frequency domain features and spatial correlation features, respectively. Dynamic weighted fusion is then performed based on the signal-to-noise ratio and the prominence of the defect. This achieves complementarity and enhancement of sensitive information for different types of defects, generating a more comprehensive and salient fusion feature map. This significantly improves the sensitivity and positioning accuracy for identifying complex defects such as debonding and microcracks within composite materials, while also enhancing the system's anti-interference capability.

[0016] This invention transforms the inefficient traditional static scanning and manual judgment model by constructing a closed-loop decision-making process of real-time evaluation, dynamic adjustment, and cross-validation. Based on intelligent planning of the initial priority scanning path using fused feature maps, signal confidence is calculated in real-time during scanning. This confidence level is then used to dynamically adjust the scanning order and focusing depth, achieving adaptive optimization of detection resources towards high-probability defect areas. This significantly reduces invalid scans and improves detection efficiency. Through geometric consistency cross-validation based on different physical signals, the validity and precise spatial location of defects are automatically confirmed, significantly reducing the uncertainty of results and reliance on human experience, achieving highly reliable, fully automated, and accurate target localization. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a precise target localization method for flaw detection according to an embodiment of the present invention. Figure 2 This is a functional block diagram of a precise target positioning system for flaw detection provided in an embodiment of the present invention; The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] This application provides a precise target localization method for flaw detection. The executing entity of this precise target localization method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the cloud computing-based resource scheduling method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks, and big data and artificial intelligence platforms.

[0020] Reference Figure 1 The diagram shown is a flowchart illustrating a precise target localization method for flaw detection according to an embodiment of the present invention. In this embodiment, the precise target localization method for flaw detection includes: S1: Simultaneously acquire multi-source flaw detection signals and preprocess the multi-source flaw detection signals to obtain a standardized signal sequence of the multi-source flaw detection signals; In this embodiment of the invention, the step of synchronously acquiring multi-source flaw detection signals and preprocessing the multi-source flaw detection signals to obtain a standardized signal sequence of the multi-source flaw detection signals includes: The multi-source flaw detection signal includes ultrasonic flaw detection signal and eddy current flaw detection signal; The ultrasonic flaw detection signal is subjected to signal conditioning processing to obtain a standardized ultrasonic signal sequence. The eddy current flaw detection signal is subjected to signal analysis processing to obtain the standardized eddy current signal sequence of the eddy current flaw detection signal; The standardized ultrasonic signal sequence and the standardized eddy current signal sequence together constitute the standardized signal sequence.

[0021] Specifically, multi-source flaw detection signals are multiple sets of raw detection data that reflect the internal or surface state of a workpiece, acquired by various flaw detection devices based on different physical principles at the same time or within a strictly synchronized time window for the same workpiece. In this embodiment, it specifically refers to the collection of ultrasonic flaw detection signals and eddy current flaw detection signals, as defined later.

[0022] Specifically, ultrasonic flaw detection signals are formed by using an ultrasonic probe to emit high-frequency mechanical ultrasonic waves into the workpiece being inspected, and receiving the reflected waves, refracted waves, or attenuated waveforms generated when the ultrasonic waves encounter defects or interfaces during propagation inside the workpiece. The resulting signals are a one-dimensional sequence of voltage or current changes over time.

[0023] Specifically, eddy current testing signal refers to a complex sequence of electrical parameters formed in an industrial non-destructive testing environment. This is achieved by using an excitation coil inside an eddy current probe to induce eddy currents on the surface of the workpiece being tested, and then measuring the changes in coil impedance or induced voltage caused by changes in the eddy current field due to defects on or near the workpiece surface through a detection coil.

[0024] Specifically, signal conditioning is a general term for a series of standardized operations performed on ultrasonic flaw detection signals. The purpose of this processing is to suppress interference such as electrical noise and mechanical vibration noise mixed in the original signal, and to adjust the signal amplitude to a uniform quantization range suitable for subsequent analysis.

[0025] Specifically, signal analysis processing is a general term for a series of standardized operations performed on eddy current testing signals to extract effective information from the original induction parameters. The purpose of this processing is to decouple complex signals containing amplitude and phase information, separate parameters that are more sensitive to defects, and normalize them to standard dimensions.

[0026] Specifically, the standardized ultrasonic signal sequence is the output result obtained after the ultrasonic flaw detection signal has undergone complete signal conditioning processing. It is a sequence of amplitude values ​​at discrete time points that has filtered out major environmental and system noise and has a consistent signal gain level.

[0027] Specifically, the standardized eddy current signal sequence is the output result obtained after the eddy current flaw detection signal has undergone complete signal analysis processing. It is a discrete, dimensionless or dimensionless sequence of eigenvalues ​​that transforms the original impedance or voltage information into a normalized real part, imaginary part, or specific derived parameter of the impedance.

[0028] Specifically, the standardized signal sequence is a multidimensional data set composed of the standardized ultrasonic signal sequence and the standardized eddy current signal sequence after they are aligned with the same time or spatial sampling points.

[0029] Furthermore, the ultrasonic probe and the eddy current probe are installed at specific positions on the multi-axis scanning device or robotic arm according to the geometry of the workpiece being inspected and the inspection requirements, ensuring that the inspection areas of the two probes overlap in space or have a definite mapping relationship.

[0030] Secondly, the ultrasonic transmitting or receiving unit and the eddy current excitation unit are activated simultaneously by a unified system clock or hardware trigger signal.

[0031] Finally, as the scanning device moves the probe along the preset initial path, the two systems continuously and in pairs record the original waveforms or parameters of the ultrasonic flaw detection signal and the eddy current flaw detection signal at the same sampling frequency or spatial trigger interval, and send these two sets of raw data streams that are strictly synchronized in time and space to the central processing unit.

[0032] Furthermore, the ultrasonic flaw detection signal is subjected to signal conditioning processing. After receiving the original ultrasonic flaw detection signal, the central processing unit first applies a digital bandpass filter to filter out low-frequency vibration noise and high-frequency electronic switching noise located outside the operating frequency band of the ultrasonic transducer.

[0033] Subsequently, the filtered signal is normalized by gain. That is, based on the system gain curve obtained by calibration with standard test blocks before the test, the amplification factor of the signal is dynamically adjusted so that the amplitude of reflected waves from different depths or material regions of the workpiece is compensated to a uniform reference level, so as to eliminate the amplitude deviation caused by differences in sound wave attenuation or coupling.

[0034] After the above noise reduction and gain normalization operations, we get a clear and standardized ultrasonic signal sequence with standard amplitude.

[0035] Furthermore, the eddy current testing signal is subjected to signal analysis processing. After receiving the original eddy current testing signal, the central processing unit first performs phase demodulation. Using orthogonal demodulation technology or digital phase detection method, it separates the impedance phase angle change that is sensitive to the surface crack orientation and the impedance amplitude change that reflects the size of the defect from the complex signal.

[0036] Next, impedance normalization is performed by dividing the demodulated impedance amplitude by the average value measured in the defect-free reference region to obtain the dimensionless normalized impedance change rate; at the same time, the phase angle difference is normalized or standardized in a similar way.

[0037] Finally, the normalized amplitude variation parameters and phase variation parameters are combined to form a standardized eddy current signal sequence.

[0038] Furthermore, the specific implementation of the standardized signal sequence, in which the standardized ultrasonic signal sequence and the standardized eddy current signal sequence together constitute the standardized signal sequence, is as follows: The processed, standardized ultrasonic signal sequence is used as the first dimension of this matrix. Simultaneously, the processed, standardized eddy current signal sequence is used as the second dimension and potentially more dimensions of the matrix. It is ensured that the data points of the two sequences are precisely arranged in a one-to-one correspondence based on the synchronization timestamp or spatial coordinates at the time of acquisition.

[0039] In summary, the synchronous acquisition of multi-source flaw detection signals ensures the consistency of information reflecting the internal structure and surface condition of the workpiece in time and space from the data source, providing a comparable and common-source basis for subsequent fusion analysis.

[0040] In summary, the signal conditioning process performed on ultrasonic flaw detection signals aims to extract effective waveforms that purely reflect the acoustic discontinuities within the material, eliminating the instability of the detection system itself and environmental interference. This allows subsequent analysis to focus on the characteristics of the defects themselves, resulting in a standardized ultrasonic signal sequence with high signal-to-noise ratio and comparable amplitude.

[0041] In summary, the signal analysis processing performed on eddy current testing signals aims to transform complex electromagnetic induction signals into standardized characteristic parameters that are more sensitive to defect geometry and decoupled from testing conditions. The result is a standardized eddy current signal sequence that can stably characterize changes in the electrical properties of surface and near-surface defects.

[0042] In summary, combining the two types of standardized sequences into a standardized signal sequence creates a unified, multimodal data representation framework. This eliminates the differences in dimensions, magnitudes, and physical meanings between signals from different sensors, enabling heterogeneous data from the acoustic and electromagnetic domains to be correlated and fused on the same platform.

[0043] S2: Perform fusion analysis on the standardized signal sequence to generate a fusion feature map of the multi-source flaw detection signal; The step of performing fusion analysis on the standardized signal sequence to generate a fusion feature map of the multi-source flaw detection signal includes: The time-frequency domain features of the standardized signal sequence are obtained by performing wavelet packet transform decomposition on the standardized signal sequence. Spatial autocorrelation analysis is performed on the standardized signal sequence to obtain its spatial correlation characteristics. Based on preset fusion rules, the time-frequency domain features and the spatial correlation features are fused at the feature level to generate a fused feature map of the multi-source flaw detection signal.

[0044] Based on preset fusion rules, the time-frequency domain features and the spatial correlation features are fused at the feature level to generate a fused feature map of the multi-source flaw detection signal: Based on the signal-to-noise ratio of the standardized signal sequence, the time-frequency domain features and the spatial correlation features are dynamically allocated fusion weights to obtain the fusion weights of the standardized signal sequence. Based on the fusion weights, the time-frequency domain features and the spatial correlation features are weighted and synthesized to generate the fusion feature map of the standardized signal sequence.

[0045] Specifically, fusion analysis refers to an advanced information processing procedure performed on a preprocessed standardized signal sequence within the detection environment of this invention. Through specific technical actions, it deeply mines and integrates complementary defect characterization information contained in different types of signals, thereby synthesizing a more comprehensive and feature-rich integrated characterization result.

[0046] Specifically, time-frequency domain features are a set of data that, after specific analysis of the standardized signal sequence, can simultaneously reflect the distribution characteristics of signal energy or intensity in both time and frequency dimensions.

[0047] Specifically, spatial correlation features are another set of data obtained after specific analysis of the standardized signal sequence, which can reflect the degree of correlation between signal amplitude or shape at different spatial locations on or inside the surface of the workpiece.

[0048] Specifically, feature-level fusion refers to the operational level in the analysis process of this invention, which combines and integrates two heterogeneous but complementary feature information—time-frequency domain features and spatial correlation features—according to certain rules.

[0049] Specifically, the fused feature map is the final output generated after feature-level fusion operations. It is a two-dimensional or three-dimensional data matrix or image, where the value of each pixel or data unit integrates the time-frequency characteristics and spatial correlation information from multiple source signals.

[0050] Further, the specific steps for performing wavelet packet transform decomposition on the standardized signal sequence to obtain the time-frequency domain features of the standardized signal sequence are as follows: First, a suitable mother wavelet function is selected for each type of standardized signal sequence. Then, multi-level wavelet packet decomposition is performed on the signal sequence, with each level further subdividing the current frequency band into two sub-bands, thereby generating a complete time-frequency analysis tree.

[0051] Next, the analysis tree is traversed to extract the wavelet packet coefficient sequence corresponding to each terminal node, that is, a specific time-frequency sub-band.

[0052] Finally, the energy of the coefficient sequence at each terminal node is calculated, for example, the sum of squares of the coefficients, and all these energy values ​​are organized into a multidimensional array or feature vector according to their corresponding frequency bands and time periods. This reflects the structure of the signal energy distribution across different fine time-frequency resolution subbands, which is the time-frequency domain feature.

[0053] Further, the specific steps for performing spatial autocorrelation analysis on the standardized signal sequence to obtain its spatial correlation characteristics are as follows: First, each type of standardized signal sequence is rearranged according to its spatial scanning coordinates at the time of acquisition to form a two-dimensional spatially distributed data field.

[0054] Then, within this data field, a sliding window is used to calculate the correlation coefficient between the signal amplitude at the center point of the window and the signal amplitudes at other points within the window. This correlation coefficient measures the similarity in signal variation patterns between the center point and its neighboring points.

[0055] Next, the correlation coefficients of all the calculated locations are aggregated to form a spatial autocorrelation function matrix or a correlation coefficient map. In this matrix or map, a continuous region of high correlation indicates the possible existence of a spatially continuous defect.

[0056] Finally, the set of quantized parameters extracted from the matrix or graph that can characterize this spatial continuity pattern is the spatial correlation feature.

[0057] Furthermore, based on preset fusion rules, the specific steps for feature-level fusion of the time-frequency domain features and the spatial correlation features are as follows: First, the fusion weights are dynamically assigned. The system evaluates the signal-to-noise ratio (SNR) of each standardized signal sequence in the current scanning area in real time. For signal sources with high SNR, the information they provide is considered more reliable in the current area, so higher fusion weights are assigned to their extracted time-frequency domain features and spatial correlation features; conversely, lower weights are assigned.

[0058] Then, weighted synthesis is performed. The assigned weights are multiplied by the corresponding time-frequency domain feature data array and spatial correlation feature data array, respectively, to achieve weighting.

[0059] Finally, the weighted time-frequency domain feature array and the spatial correlation feature array are added or combined according to the spatial position of the elements to generate a new fusion feature map of a standardized signal sequence that integrates time-frequency and spatial information.

[0060] In summary, fusion analysis, as the core step of this invention, elevates heterogeneous information from different physical sensors to a unified feature level for intelligent integration, overcoming the limitations of a single signal perspective.

[0061] In summary, extracting time-frequency domain features through wavelet packet transform decomposition can precisely extract frequency components in the signal that are related to the defect type and size, as well as the time or location of their occurrence. This enhances the method's ability to distinguish complex defect patterns.

[0062] In summary, the spatial autocorrelation analysis extracts spatial correlation features, which helps to capture the continuity and structural morphology of defects in physical space. The effect is to improve the ability to delineate the boundaries and overall contours of defects and reduce the misjudgment of noise points.

[0063] In summary, feature-level fusion adaptively trusts the feature contributions of high-quality signal sources through dynamic weight allocation and weighted synthesis, and optimally combines time-frequency and spatial information. This generates a fused feature map with significantly improved information density and reliability. This map serves as the foundation for all subsequent steps and directly determines the accuracy and robustness of the entire method in achieving precise target localization in complex industrial inspection environments.

[0064] S3: Identify the suspected defect areas of the multi-source flaw detection signal based on the fused feature map to obtain a preliminary set of defect location coordinates of the multi-source flaw detection signal; The step of identifying suspected defect regions of the multi-source flaw detection signal based on the fused feature map to obtain a preliminary set of defect location coordinates for the multi-source flaw detection signal includes: The pixels in the fused feature map are determined based on the comparison results between the preset threshold and the feature values ​​in the fused feature map. By performing region connectivity on the pixels, the suspected defect regions of the multi-source flaw detection signal are obtained; The location information of the suspected defect area is obtained, and the preliminary defect location coordinate set of the multi-source flaw detection signal is determined based on the location information.

[0065] Specifically, the preset threshold is one or more threshold values ​​that are pre-set before the detection process of the present invention, based on the material type of the workpiece being detected, the detection accuracy requirements, historical experience data or relevant industry standards.

[0066] Specifically, the feature value refers to the specific numerical value corresponding to each pixel or data unit in the fused feature map.

[0067] Specifically, the comparison result refers to the logical judgment result generated after comparing the feature value of each pixel in the fused feature map with a preset threshold.

[0068] Specifically, in the context of this invention, a pixel refers to each basic data unit that constitutes the fused feature map. This map can be viewed as a digital image, where the coordinates of each pixel precisely correspond to a specific physical location on the surface of the workpiece being inspected, and the grayscale value or color depth of that pixel is the feature value.

[0069] Specifically, region connectivity refers to the process of clustering and merging all pixels marked as suspicious on the fused feature map according to their spatial adjacency. This process aims to aggregate spatially adjacent, discrete anomalous pixels into a continuous and complete spatial entity.

[0070] Specifically, a suspected defect region is a set of one or more consecutive pixels formed after region connectivity operations. Each set represents an independent, spatially connected anomalous signal region identified on the fused feature map.

[0071] Specifically, location information refers to the quantitative data extracted from each suspected defect area that can characterize the spatial position and geometric shape of the area in the workpiece coordinate system.

[0072] Specifically, the preliminary defect location coordinate set is an ordered or unordered data list generated based on the location information of all suspected defect areas. Each entry in this set corresponds to an independent suspected defect and provides its preliminary spatial location estimate in coordinate form. It is a key data interface connecting defect identification with subsequent steps such as path planning and dynamic scanning.

[0073] Furthermore, based on the comparison results between the preset threshold and the feature values ​​in the fused feature map, the specific process for determining the pixels in the fused feature map is as follows: Each pixel in the fused feature map is traversed sequentially. For the currently traversed pixel, its stored feature value is read. Simultaneously, a preset threshold, either pre-stored by the system or dynamically calculated based on the current detection parameters, is invoked.

[0074] Next, a comparison operation is performed to determine if the feature value is greater than a preset threshold. If the feature value is greater than the preset threshold, the pixel is determined to be a significant point or a candidate point, and a binary labeling map of the same size as the fused feature map is created in memory. The corresponding position of this pixel is marked as 1, representing a suspected point. If the feature value is less than the preset threshold, the corresponding position on the labeling map is marked as 0, representing background or a non-suspected point. After traversal, a binary labeling map is obtained, in which all pixels marked as 1 are the set of candidate pixels determined by threshold comparison that need further analysis.

[0075] Furthermore, the specific steps for performing region connectivity on the pixels to obtain the suspected defect regions of the multi-source flaw detection signal are as follows: Using the obtained binary labeled image as input, a region growing or connected component labeling algorithm is employed. First, the binary labeled image is scanned to find a pixel that has not yet been assigned to any region and is labeled as 1, serving as a seed point.

[0076] Then, starting from the seed point, examine all its spatially adjacent pixels, such as those above, below, left, right, and diagonally (i.e., its eight-neighborhood). If a neighboring pixel is also marked as 1, it is grouped into the same region as the current seed point. This process is recursively or iteratively applied to all newly added pixels in the region until no new neighboring pixels with a value of 1 can be included. At this point, all the included pixels collectively form a connected suspected defect region.

[0077] Repeat the above process until all pixels marked as 1 in the binary labeled image are assigned to a unique connected region. Finally, the output is a list of multiple independent suspected defect regions, each defined by the set of coordinates of all pixels it contains.

[0078] Furthermore, the specific steps for obtaining the location information of the suspected defect area and determining the preliminary defect location coordinate set of the multi-source flaw detection signals based on the location information are as follows: First, obtain the coordinates of all pixels in the region within the workpiece coordinate system. Then, calculate the bounding rectangle or minimum bounding rectangle of the region. Typically, the coordinates of the center point of this rectangle are taken as the initial location coordinates of the suspected defect. Next, these coordinates, along with the unique identifier of the region, are compiled into a single data entry.

[0079] Finally, all such data entries corresponding to the suspected defect areas are summarized into a list or array, which is the preliminary set of defect location coordinates.

[0080] In summary, the preset threshold provides an objective and unified judgment standard. Its role is to transform subjective and ambiguous feature saliency judgments into repeatable and automatically executable binarization operations, which is a key control parameter for automated recognition processes.

[0081] In summary, by comparing feature values ​​with thresholds to determine candidate pixels, the function is to quickly and initially filter out all noteworthy abnormal locations from the entire feature map, filtering out a large amount of background and noise information, and greatly narrowing the data range for subsequent fine processing.

[0082] In summary, the role of region connectivity operations is to aggregate spatially discrete outliers into meaningful wholes based on their physical adjacency. The effect is to group pixels representing potentially identical physical defects into one category, forming a preliminary estimate of the spatial extent of potential defects, i.e., suspected defect regions, thus avoiding the misclassification of a large defect as multiple smaller defects.

[0083] In summary, acquiring regional location information and determining the initial set of defect location coordinates transforms the image-level, pixel-based regional description into a geometric, coordinate-based location description. This generates a list of spatial coordinate-based task instructions that can be directly understood and executed by the subsequent automated scanning system, laying a clear spatial target foundation for the subsequent adaptive scanning and verification process of the entire precise target localization method.

[0084] S4: Based on the saliency of the coordinate features of the fused feature map, sort the preliminary defect location coordinate set to generate the priority scanning path of the multi-source flaw detection signal; The step of sorting the preliminary defect location coordinate set based on the coordinate feature saliency of the fused feature map to generate the priority scanning path of the multi-source flaw detection signal includes: Based on the fused feature map, feature values ​​of the preliminary defect location coordinate set are extracted to obtain the coordinate feature saliency of the fused feature map; Based on the saliency of the coordinate features, the coordinate sequence of the preliminary defect location coordinate set is sorted to obtain the sorted coordinate sequence of the preliminary defect location coordinate set; Based on the spatial distance between adjacent coordinates of the sorted coordinate sequence and the preliminary defect location coordinate set, a connection path of the coordinate sequence is constructed to obtain the priority scanning path of the multi-source flaw detection signal.

[0085] Specifically, coordinate feature saliency refers to a quantified value extracted from the fused feature map for each specific coordinate position within the preliminary defect location coordinate set.

[0086] Specifically, the coordinate sequence refers to an initial list consisting of all coordinate points in the preliminary defect location coordinate set. The initial order of this list may be arbitrary and does not contain any logic regarding scan priority.

[0087] Specifically, the sorted coordinate sequence refers to the new list obtained by rearranging the coordinate points in the coordinate sequence according to the magnitude of their corresponding coordinate feature values. Typically, they are arranged in descending order to ensure that the positions with the most significant features and the highest probability of defects are placed in the highest priority scanning order.

[0088] Specifically, the spatial distance between adjacent coordinates refers to the actual Euclidean distance in physical space between any two coordinate points that are sequentially adjacent in the list in the sorted coordinate sequence.

[0089] Specifically, the connection path refers to a continuous movement trajectory that can be executed by a scanning device, established between all these coordinate nodes by taking the aforementioned sorted coordinate sequence as the node access order and considering the spatial distance between adjacent coordinates, through specific optimization or construction rules.

[0090] Specifically, the priority scan path is the final generated sequence of motion instructions that can be directly used by the scanning actuator.

[0091] Furthermore, based on the fused feature map, the specific process for extracting feature values ​​from the preliminary defect location coordinate set to obtain the coordinate feature saliency of the fused feature map is as follows: For each coordinate entry in the set, the system uses it as an index to locate the corresponding pixel position on the fused feature map. Then, it directly reads the stored feature value from that pixel position and assigns this value to the current coordinate, recording it as the coordinate feature saliency. This operation is repeated for all coordinates in the set, ultimately resulting in a list of coordinate feature saliency values ​​corresponding to the initial defect location coordinate set. Essentially, this operation maps information from the feature map to a list of coordinate points.

[0092] Further, based on the saliency of the coordinate features, the specific steps for sorting the coordinate sequence of the preliminary defect location coordinate set to obtain the sorted coordinate sequence of the preliminary defect location coordinate set are as follows: Each coordinate is paired with its corresponding coordinate significance value as input. A sorting algorithm is then used to sort all coordinates in descending order based on their coordinate significance values.

[0093] After sorting, a new list of coordinates is generated. The order of the coordinates in this list is no longer arbitrary, but strictly arranged from high to low significance of the fusion feature at their corresponding positions. This new list is the sorted coordinate sequence.

[0094] Further, based on the spatial distance between adjacent coordinates of the sorted coordinate sequence and the preliminary defect location coordinate set, the specific steps for constructing the connection path of the coordinate sequence to obtain the preferred scanning path of the multi-source flaw detection signal are as follows: First, the system uses the sorted coordinate sequence as the basis for the required node order. Then, it calculates the spatial distance between adjacent coordinates in each pair of consecutive coordinates within the sequence. Next, with the goal of minimizing the total travel distance or avoiding unreasonable long-distance crossings, the path is fine-tuned while strictly maintaining the priority of the node visit order.

[0095] For example, if two consecutive high-priority points are too far apart in actual space, a point that is closer and has a slightly lower priority may be inserted between them as a transition to save scanning time, but a low-priority point will never be accessed before a high-priority point.

[0096] Ultimately, a continuous trajectory is constructed that starts from the scanning start point, passes through all coordinate points in the sorted coordinate sequence in sequence, and finally returns or ends.

[0097] In summary, the purpose of extracting coordinate feature saliency is to assign an objective priority quantification score from the fused feature map to each initially located coordinate point. This effectively links the abstract feature map with specific coordinate points numerically, providing direct and comparable data for subsequent intelligent ranking.

[0098] In summary, sorting the coordinate sequence based on saliency helps to classify the importance of all targets to be inspected according to the probability of defects. The effect is to generate a clear list of action priorities, ensuring that the detection system always allocates its most limited time and resources to areas most likely to find actual defects, thereby improving overall detection efficiency and detection rate.

[0099] In summary, the role of constructing connection paths based on spatial distance is to transform the ideal priority order into a physically executable and relatively efficient mechanical motion scheme. It respects both the urgency of defect detection and execution efficiency. The ultimate result is the generation of a prioritized scanning path, which serves as the direct input to the scanning system, directing the probe to perform refined re-inspection in an optimized manner. This is a key scheduling step in achieving the two objectives of accuracy and efficiency in the precise target localization method described in this invention.

[0100] S5: Calculate the signal confidence of the priority scanning path in real time, and dynamically adjust the priority scanning path based on a preset adaptive threshold to obtain the dynamic scanning path of the multi-source flaw detection signal. The real-time calculation of the signal confidence of the priority scanning path and the dynamic adjustment of the priority scanning path based on a preset adaptive threshold to obtain the dynamic scanning path of the multi-source flaw detection signal include: Based on the degree of matching between the real-time signal of the priority scanning path and the fused feature map, the signal confidence of the priority scanning path is calculated in real time. Based on the signal confidence level and a preset adaptive threshold, the scanning order and scanning focus depth of the priority scanning path are dynamically replanned to obtain the dynamic scanning path of the multi-source flaw detection signal.

[0101] The formula for calculating the signal confidence level is as follows: in, The confidence level of the signal. This represents the distribution density of singular points in the intrinsic mode function obtained after performing a Hilbert-Huang transform on the real-time signal. The effective echo width extracted from the real-time signal. The reference feature width extracted at the corresponding position in the fused feature map. The weighted fusion feature value of the real-time signal, This represents the variance of the energy values ​​of the historical scan signals.

[0102] Specifically, real-time signals refer to the latest standardized ultrasonic signal sequence and standardized eddy current signal sequence that are immediately generated when the flaw detection scanning device moves along the preferred scanning path and arrives at a certain coordinate point to be measured on the path, and is collected in real time by the ultrasonic probe and eddy current probe at that point and preprocessed.

[0103] Specifically, the degree of matching is a qualitative description used to measure the consistency, similarity, or conformity between the real-time signals acquired at the current point of the priority scan path and the expected defect features represented by the fused feature map.

[0104] Specifically, signal confidence is a quantitative indicator that comprehensively evaluates the degree of matching mentioned above, as well as the quality of the real-time signal itself. It is a scalar value between 0 and 1; the higher the value, the more reliable the real-time signal acquired at the current point. Specifically, the preset adaptive threshold is a threshold value that is pre-set based on the overall detection environment, historical data, or statistical patterns, and can be dynamically fine-tuned based on the overall confidence distribution of the scanned area during a single scan.

[0105] Specifically, the scanning order refers to the order in which the probe visits the coordinates of each suspected defect, as specified in the priority scanning path.

[0106] Specifically, scanning focusing depth refers to the depth setting at which the probe focuses sound wave energy in ultrasonic testing, or more generally, the sensitivity configuration for detecting defects at different depths.

[0107] Specifically, dynamic replanning refers to the system's real-time, online recalculation and modification of the priority scan path for the remaining unexecuted parts during the closed-loop detection process of this invention, based on the signal confidence results from real-time signal feedback.

[0108] Specifically, a dynamic scan path is the scan trajectory and parameter sequence that is ultimately executed after dynamic replanning. It starts from the initial priority scan path, but during the scan, its subsequent parts evolve continuously based on real-time feedback, thus forming a non-fixed final path that adapts to the actual detection situation.

[0109] Furthermore, the specific process for calculating the signal confidence of the priority scanning path in real time based on the matching degree between the real-time signal of the priority scanning path and the fused feature map is as follows: Once the probe moves along the path to the target point and completes signal acquisition and preprocessing, the resulting real-time signal, including ultrasonic and eddy current characteristics, is sent to an evaluation unit.

[0110] The feature vector corresponding to the coordinate point is extracted from the fused feature map and used as a baseline template. Through a series of predefined metric rules, the differences between the features of the real-time signal and the features of the baseline template are analyzed, while the clarity and stability of the real-time signal itself are evaluated. The calculation process is completed within milliseconds, thus achieving real-time feedback.

[0111] Furthermore, based on the signal confidence level and a preset adaptive threshold, the scanning order and scanning focus depth of the priority scanning path are dynamically reprogrammed to obtain the dynamic scanning path of the multi-source flaw detection signal. The specific steps are as follows: The calculated current signal confidence level is compared with a preset adaptive threshold. If the signal confidence level is greater than the preset adaptive threshold, the detection result of the current point is considered to meet expectations, and the process continues to execute the next point according to the original priority scanning path.

[0112] If the signal confidence level is greater than the preset adaptive threshold, dynamic replanning logic is triggered.

[0113] Scanning order adjustment: The system may move another suspected point with high coordinate feature significance and geographical location adjacent to the current point in the subsequent path to the next scan position in order to verify whether it is an extension of the defect area or to check the related area.

[0114] Scanning focus depth adjustment: For the current point, the system may automatically switch the focusing law of the ultrasound probe or adjust the eddy current detection frequency to obtain information at different depth levels in an attempt to confirm or eliminate defects.

[0115] After adjustment, motion commands and detection parameters from the current point to the next target point are immediately generated and updated, forming a new path segment. This path, which is continuously iterated and adjusted based on real-time feedback until all planned points are scanned or reasonably skipped, is the final dynamic scanning path actually executed.

[0116] In summary, the purpose of real-time signal confidence calculation is to provide a quantitative quality report for the scanning process. Its effect is to transform subjective, vague judgments into precise numerical data that can be used for programmed decision-making, which is a prerequisite for achieving closed-loop control.

[0117] In summary, the role of dynamic replanning based on confidence levels and thresholds is to endow the detection system with online self-correction and adaptive capabilities. When the confidence level is low during execution according to the predetermined plan, the system can immediately and intelligently adjust subsequent action plans, rather than mechanically executing to the end. This significantly improves the robustness of detection and resource utilization efficiency under complex operating conditions.

[0118] In summary, the above steps ensure that the precise target positioning method described in this invention can still flexibly and reliably guide detection resources and complete high-confidence defect positioning tasks when faced with uncertainties and interference from actual workpieces.

[0119] Specifically, The signal confidence level is used to quantitatively evaluate the quality of the real-time signal acquired at the current scanning point and its degree of matching with the expected model.

[0120] Specifically, The distribution density of singular points is obtained by performing a Hilbert-Huang transform on the real-time signal. First, the Hilbert-Huang transform decomposes the signal into a series of intrinsic mode functions (IMFs). Then, singular points, i.e., zero-crossing points and local extrema of the signal, are identified in each IMF.

[0121] Specifically, The effective echo width is a geometric feature extracted from the real-time signal. It is calculated by determining the time interval during which the echo envelope exceeds a preset noise threshold, and then measuring the width of that interval.

[0122] Specifically The baseline feature width is a feature value extracted from the fused feature map corresponding to the coordinate position of the current scan point.

[0123] Specifically, The weighted fusion feature value is a comprehensive feature scalar calculated at the current point after processing by the real-time signal feature extraction and fusion rules. It integrates the time-frequency domain features and spatial correlation features of the current real-time ultrasonic signal and eddy current signal, and synthesizes them according to dynamically allocated weights.

[0124] Specifically, The variance of the historical scan signal energy values ​​reflects the fluctuations in the historical scan signal energy values ​​within the neighborhood of the current scan point. Signal energy is typically obtained by summing the squares of the signal amplitudes. Variance It measures the extent to which these historical energy values ​​deviate from their average.

[0125] Further, calculate the width matching factor. The specific process is as follows: First, calculate the measured effective echo width. With reference feature width The absolute value of the difference. This difference measures the deviation between the measured signal width and the expected width. Then, this difference is used as the negative exponent of an exponential function to calculate... The exponential function maps the width difference to a value between 0 and 1. When and When they are completely equal, the exponent is 1, indicating a perfect match; as the difference between the two increases, the value of the exponent rapidly decays and approaches 0, indicating a decrease in the degree of matching.

[0126] Furthermore, the molecular part is calculated. The specific steps are as follows: The singular point distribution density, which characterizes the richness of transient features of the signal... The width matching factor, which represents the matching degree of the above-mentioned features, is multiplied.

[0127] Further, calculate the denominator. The specific steps are as follows: First, calculate the comprehensive feature strength of the current point. The square of this amplifies the contrast of the feature intensity. Then, the variance of the historical signal energy, which characterizes local background stability, is added. Finally, the square root of the summation is taken. The denominator is designed to normalize the confidence level. Denominator terms This reduction balance is achieved as a scale that measures the combined effect of salient feature intensity and regional background noise variation.

[0128] In summary, this formula can generate a stable and reliable quantitative indicator in real time. The signal confidence level is not based on a simple threshold judgment of a single feature, but rather originates from a comprehensive calculation process that simulates complex judgment logic. It provides a solid and precise numerical basis for subsequent dynamic replanning decisions and is one of the core algorithmic supports for this invention to achieve intelligent adaptive scanning and improve the reliability of accurate target positioning.

[0129] S6: Perform multi-angle cross-verification on the scanning results of the dynamic scanning path to obtain the final positioning result of the multi-source flaw detection signal.

[0130] The step of performing multi-angle cross-verification on the scanning results of the dynamic scanning path to obtain the final positioning result of the multi-source flaw detection signal includes: The scanning results of the dynamic scanning path include defect contour features and defect depth features; The defect boundary position and extension depth of the multi-source flaw detection signal are determined based on the defect contour features and the defect depth features to obtain the final positioning result of the multi-source flaw detection signal.

[0131] Specifically, the scan result refers to the complete set of data collected and stored by the system after the dynamic scan path has been fully executed.

[0132] Specifically, the defect contour features are a set of geometric and morphological parameters extracted from the scan results, primarily based on the analysis of standardized ultrasonic signal sequences.

[0133] Specifically, the defect depth feature is another set of parameters extracted from the scan results, mainly based on time-of-flight information analysis of the standardized eddy current signal sequence and the standardized ultrasonic signal sequence.

[0134] Specifically, the defect boundary position is the precise spatial range occupied by the defect in the workpiece coordinate system, determined by analyzing and calculating the defect contour features.

[0135] Specifically, the extension depth is the dimension of the defect in the direction perpendicular to the workpiece surface, determined by analyzing and calculating the defect depth characteristics.

[0136] Specifically, the final location result is an authoritative and comprehensive spatial description of the defect output by the system after multi-angle cross-verification.

[0137] Furthermore, the specific process for the scanning results of the dynamic scanning path, including defect contour features and defect depth features, is as follows: The ultrasonic scanning results of all points along the path are retrieved, especially the amplitude envelope and arrival time data of the echo signals at each point. Through spatial interpolation and region synthesis techniques, the echo intensity information of discrete points is reconstructed into a two-dimensional image or data field reflecting the cross-sectional characteristics of the defect. From this image, a set of quantized parameters describing the shape, size, and boundary of the defect is automatically extracted, which is the defect contour feature.

[0138] Simultaneously, eddy current scanning results and ultrasonic time-of-flight data for all points along the path are retrieved. Utilizing the known relationship between the sensitivity of eddy current signals to near-surface defects and their depth, as well as the propagation speed of ultrasonic waves in materials, the depth estimates of the defects at various locations are calculated and fused to form a set of parameters describing the three-dimensional depth information of the defects.

[0139] Furthermore, based on the defect contour features, the projected boundary position of the defect is accurately calculated using boundary tracking and geometric fitting algorithms, and the output is a set of planar coordinate sequences. Based on the defect depth features, the extension depth of the defect is determined through statistical analysis, and the output is one or a set of depth values.

[0140] Subsequently, multi-angle cross-validation is performed: the core performs a spatial logic consistency check on the calculated boundary position and extension depth.

[0141] Furthermore, for example, examine whether the eddies and ultrasonic depth estimates at each point within the boundary region determined by the ultrasonic profile are spatially continuous and reasonable; or verify whether the edge regions of the profile correspond to regions with abrupt changes in depth. If the verification passes, the independent evidence from the ultrasonic waves and eddies is considered mutually supportive, and the result is deemed reliable.

[0142] Finally, the system fuses the verified defect boundary locations with the extended depth information to generate a complete 3D defect space description with depth information, and outputs it in a formatted manner.

[0143] In summary, by separating and extracting defect contour and depth features from the scanning results, and by structuring the attribute information about the defect in different dimensions decoded from signals based on different physical principles, a clear and independent chain of evidence is prepared for subsequent comprehensive judgment.

[0144] In summary, the purpose of determining the location and depth of the defect boundary based on its features is to transform the extracted feature parameters into precise geometric spatial quantities that can be directly understood in engineering.

[0145] In summary, multi-angle cross-validation leverages the complementarity of different detection technologies to cross-verify and confirm preliminary judgment results. This significantly enhances the reliability and authority of the final positioning results, effectively preventing misjudgments or inaccurate positioning caused by limitations of a single technology or accidental interference. It ensures the accuracy of the method described in this invention at the output stage, thereby providing solid data support for high-confidence decision-making.

[0146] like Figure 2 The diagram shown is a functional block diagram of a precise target positioning system for flaw detection provided in an embodiment of the present invention.

[0147] The precise target positioning system 100 for flaw detection described in this invention can be installed in an electronic device. Depending on the functions implemented, the precise target positioning system 100 for flaw detection may include a signal acquisition and preprocessing module 101, a fusion analysis module 102, a defect identification module 103, a path planning module 104, a dynamic adjustment module 105, and a cross-validation module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.

[0148] In this embodiment, the functions of each module / unit are as follows: The signal acquisition and preprocessing module 101 is used to synchronously acquire multi-source flaw detection signals and preprocess the multi-source flaw detection signals to obtain a standardized signal sequence of the multi-source flaw detection signals. The fusion analysis module 102 is used to perform fusion analysis on the standardized signal sequence and generate a fusion feature map of the multi-source flaw detection signal; The defect identification module 103 is used to identify the suspected defect area of ​​the multi-source flaw detection signal based on the fused feature map, so as to obtain the preliminary defect location coordinate set of the multi-source flaw detection signal; Path planning module 104 is used to sort the preliminary defect location coordinate set based on the coordinate feature saliency of the fused feature map and generate the priority scanning path of the multi-source flaw detection signal. The dynamic adjustment module 105 is used to calculate the signal confidence of the priority scanning path in real time, and dynamically adjust the priority scanning path based on a preset adaptive threshold to obtain the dynamic scanning path of the multi-source flaw detection signal. The cross-validation module 106 is used to perform multi-angle cross-validation on the scanning results of the dynamic scanning path to obtain the final positioning result of the multi-source flaw detection signal.

[0149] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0150] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0151] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0152] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0153] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A precise target localization method for flaw detection, characterized in that, The method includes: S1: Simultaneously acquire multi-source flaw detection signals and preprocess the multi-source flaw detection signals to obtain a standardized signal sequence of the multi-source flaw detection signals; S2: Perform fusion analysis on the standardized signal sequence to generate a fusion feature map of the multi-source flaw detection signal; S3: Identify the suspected defect areas of the multi-source flaw detection signal based on the fused feature map to obtain a preliminary set of defect location coordinates of the multi-source flaw detection signal; S4: Based on the saliency of the coordinate features of the fused feature map, sort the preliminary defect location coordinate set to generate the priority scanning path of the multi-source flaw detection signal; S5: Calculate the signal confidence of the priority scanning path in real time, and dynamically adjust the priority scanning path based on a preset adaptive threshold to obtain the dynamic scanning path of the multi-source flaw detection signal. S6: Perform multi-angle cross-verification on the scanning results of the dynamic scanning path to obtain the final positioning result of the multi-source flaw detection signal.

2. The precise target localization method for flaw detection as described in claim 1, characterized in that, The simultaneous acquisition of multi-source flaw detection signals and the preprocessing of these signals to obtain a standardized signal sequence include: The multi-source flaw detection signal includes ultrasonic flaw detection signal and eddy current flaw detection signal; The ultrasonic flaw detection signal is subjected to signal conditioning processing to obtain a standardized ultrasonic signal sequence. The eddy current flaw detection signal is subjected to signal analysis processing to obtain the standardized eddy current signal sequence of the eddy current flaw detection signal; The standardized ultrasonic signal sequence and the standardized eddy current signal sequence together constitute the standardized signal sequence.

3. The precise target localization method for flaw detection as described in claim 1, characterized in that, The step of performing fusion analysis on the standardized signal sequence to generate a fusion feature map of the multi-source flaw detection signal includes: The time-frequency domain features of the standardized signal sequence are obtained by performing wavelet packet transform decomposition on the standardized signal sequence. Spatial autocorrelation analysis is performed on the standardized signal sequence to obtain its spatial correlation characteristics. Based on preset fusion rules, the time-frequency domain features and the spatial correlation features are fused at the feature level to generate a fused feature map of the multi-source flaw detection signal.

4. The precise target localization method for flaw detection as described in claim 3, characterized in that, The method, based on preset fusion rules, performs feature-level fusion of the time-frequency domain features and the spatial correlation features to generate a fused feature map of the multi-source flaw detection signal, including: Based on the signal-to-noise ratio of the standardized signal sequence, the time-frequency domain features and the spatial correlation features are dynamically allocated fusion weights to obtain the fusion weights of the standardized signal sequence. Based on the fusion weights, the time-frequency domain features and the spatial correlation features are weighted and synthesized to generate the fusion feature map of the standardized signal sequence.

5. The precise target localization method for flaw detection as described in claim 1, characterized in that, The step of identifying suspected defect regions of the multi-source flaw detection signal based on the fused feature map to obtain a preliminary set of defect location coordinates for the multi-source flaw detection signal includes: The pixels in the fused feature map are determined based on the comparison results between the preset threshold and the feature values ​​in the fused feature map. By performing region connectivity on the pixels, the suspected defect regions of the multi-source flaw detection signal are obtained; The location information of the suspected defect area is obtained, and the preliminary defect location coordinate set of the multi-source flaw detection signal is determined based on the location information.

6. The precise target localization method for flaw detection as described in claim 1, characterized in that, The step of sorting the preliminary defect location coordinate set based on the coordinate feature saliency of the fused feature map to generate the priority scanning path of the multi-source flaw detection signal includes: Based on the fused feature map, feature values ​​of the preliminary defect location coordinate set are extracted to obtain the coordinate feature saliency of the fused feature map; Based on the saliency of the coordinate features, the coordinate sequence of the preliminary defect location coordinate set is sorted to obtain the sorted coordinate sequence of the preliminary defect location coordinate set; Based on the spatial distance between adjacent coordinates of the sorted coordinate sequence and the preliminary defect location coordinate set, a connection path of the coordinate sequence is constructed to obtain the priority scanning path of the multi-source flaw detection signal.

7. The precise target localization method for flaw detection as described in claim 1, characterized in that, The real-time calculation of the signal confidence of the priority scanning path and the dynamic adjustment of the priority scanning path based on a preset adaptive threshold to obtain the dynamic scanning path of the multi-source flaw detection signal include: Based on the degree of matching between the real-time signal of the priority scanning path and the fused feature map, the signal confidence of the priority scanning path is calculated in real time. Based on the signal confidence level and a preset adaptive threshold, the scanning order and scanning focus depth of the priority scanning path are dynamically replanned to obtain the dynamic scanning path of the multi-source flaw detection signal.

8. The precise target localization method for flaw detection as described in claim 7, characterized in that, The formula for calculating the signal confidence level is: in, The confidence level of the signal. This represents the distribution density of singular points in the intrinsic mode function obtained after performing a Hilbert-Huang transform on the real-time signal. The effective echo width extracted from the real-time signal. The reference feature width extracted at the corresponding position in the fused feature map. The weighted fusion feature value of the real-time signal, This represents the variance of the energy values ​​of the historical scan signals.

9. The precise target localization method for flaw detection as described in claim 1, characterized in that, The step of performing multi-angle cross-verification on the scanning results of the dynamic scanning path to obtain the final positioning result of the multi-source flaw detection signal includes: The scanning results of the dynamic scanning path include defect contour features and defect depth features; The defect boundary position and extension depth of the multi-source flaw detection signal are determined based on the defect contour features and the defect depth features to obtain the final positioning result of the multi-source flaw detection signal.

10. A precise target positioning system for flaw detection, used to implement the precise target positioning method for flaw detection as described in any one of claims 1-9, characterized in that, The system includes: The signal acquisition and preprocessing module is used to synchronously acquire multi-source flaw detection signals and preprocess the multi-source flaw detection signals to obtain a standardized signal sequence of the multi-source flaw detection signals. The fusion analysis module is used to perform fusion analysis on the standardized signal sequence and generate a fusion feature map of the multi-source flaw detection signal; The defect identification module is used to identify suspected defect areas of the multi-source flaw detection signal based on the fused feature map, so as to obtain a preliminary set of defect location coordinates of the multi-source flaw detection signal; The path planning module is used to sort the preliminary defect location coordinate set based on the coordinate feature saliency of the fused feature map, and generate the priority scanning path of the multi-source flaw detection signal; The dynamic adjustment module is used to calculate the signal confidence of the priority scanning path in real time and dynamically adjust the priority scanning path based on a preset adaptive threshold to obtain the dynamic scanning path of the multi-source flaw detection signal. The cross-validation module is used to perform multi-angle cross-validation on the scanning results of the dynamic scanning path to obtain the final positioning result of the multi-source flaw detection signal.

Images