Laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing

Abstract

The present invention discloses a laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing, which relates to the fields of laser ultrasonic testing and material defect detection. The method involves sparse scanning of a sample to be tested using a random sparse sampling strategy; reconstructing the global ultrasonic intensity distribution of a full-matrix scan at a high sampling rate based on the principle of compressed sensing, and achieving global sparse positioning through extreme value detection; then, based on the global sparse positioning, performing local fine scanning with high spatial resolution; utilizing the global sparse positioning and local fine scanning data, generating a low-spatial-resolution sparse imaging result through synthetic aperture focusing technology, and then re-using the principle of compressed sensing to perform high-resolution compressed sensing reconstruction imaging. Therefore, the laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing can achieve efficient imaging of tiny defects, improving imaging accuracy while reducing the number of scanning points.

Description

Technical Field

[0001] The present invention relates to the technical field of laser ultrasonic testing and material defect detection, and in particular to a laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing. Background Art

[0002] Laser ultrasonic testing technology, through a non-contact mechanism whereby pulsed lasers excite ultrasonic waves and laser interferometers receive the acoustic field response, has successfully broken free from the limitations of the coupling medium. It boasts significant advantages such as non-contact, high sensitivity, a wide range of material applicability, and applicability to complex working conditions. However, its engineering application still faces a key bottleneck: under the Nyquist sampling framework, traditional data acquisition modes often require spatial high-resolution scanning to achieve high-precision characterization of micron-level defects. The resulting high-dimensional full-matrix data set not only occupies a large amount of storage resources, but also significantly reduces overall detection efficiency due to the exponential growth in data preprocessing complexity. The low detection efficiency, large data storage volume, and complex processing algorithms caused by this full-matrix high-resolution scanning have become the main technical barriers restricting the application of laser ultrasonic technology in industrial scenarios.

[0003] Against this backdrop, the rise of compressed sensing (CS) theory offers a new technical approach to addressing these issues. First proposed in 2006 by Donoho, Candès, and Tao, CS theory relies on an innovative basis transformation technique that reconfigures the signal into a new sparse representation, where most coefficients are close to zero and only a few critical non-zero coefficients are retained. This sparsity significantly reduces the sampling dimension during data acquisition. Combined with compressed sensing algorithms, it can reconstruct complete signal information from these limited samples, significantly reducing the time required for signal acquisition. CS theory has shown promising application prospects in a variety of fields, including medical image reconstruction, radar signal processing, and wireless sensor networks.

[0004] With the development of laser ultrasonic technology, researchers have begun to combine compressed sensing with laser ultrasonic detection technology to improve detection efficiency and real-time imaging. Although the collaborative application of compressed sensing technology and laser ultrasonic detection has shown significant potential for efficiency improvement, its engineering application still faces two major problems. First, most existing technical systems use the compressed sensing theoretical framework to achieve ultrasonic field reconstruction and calculation. The wave field reconstruction process involves solving complex inverse problems and requires iterative calculation to optimize the objective function. This process usually consumes a lot of computing resources, and the full-field reconstruction time can reach minutes, which is difficult to meet the needs of rapid detection. Second, to ensure the quality of the reconstructed image, the sparse transformation matrix used is usually extremely large in dimension. The resulting computational load seriously restricts the imaging speed. This contradiction is particularly evident in industrial field applications that emphasize detection efficiency. Summary of the Invention

[0005] The purpose of the present invention is to provide a laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing, which utilizes the layered sampling strategy of "global sparse positioning-local fine scanning" and combines it with synthetic aperture focusing imaging technology to provide a feasible solution for efficient imaging of tiny defects.

[0006] To achieve the above object, the present invention provides a laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing, comprising the following steps:

[0007] S1. Within the global detection range of the sample to be tested, a random sparse sampling strategy is used to select The measurement points are sparsely scanned by the excitation-detection laser source fixed-distance scanning detection method, and the time domain A-scan waveform data is recorded synchronously;

[0008] S2. Mapping the sparsely scanned A-scan waveform data to three-dimensional spatial coordinates, reconstructing the global ultrasonic intensity distribution of the full matrix scan at a high sampling rate based on the principle of compressed sensing, and identifying the feature points of the global ultrasonic intensity distribution through extreme value analysis to achieve global sparse positioning;

[0009] S3. Use the global sparse positioning result as the center position of the local fine scan, and perform local fine scans with high spatial resolution stepping within the length range L on both sides of the center position;

[0010] S4, based on the A-scan signal data recorded by global sparse positioning and local fine scanning, uses synthetic aperture focusing technology, combined with signal delay correction and coherent superposition methods, to achieve dynamic focusing on each pixel point in the detection area, generating a sparse imaging result with low spatial resolution;

[0011] S5. Utilizing the sparse imaging results with low spatial resolution, the compressed sensing principle is further adopted to perform high-resolution compressed sensing reconstruction imaging to obtain high-precision laser ultrasonic defect imaging.

[0012] Preferably, in S1, the number of measurement points of the random sparse sampling strategy is , according to the imaging signal-to-noise ratio, detection efficiency and sparse sampling theorem constraints, and the fixed-distance scanning detection method is used for sparse scanning.

[0013] Preferably, S2 includes:

[0014] S21, extracting the defect echo intensity of each sampling point as a characteristic parameter, constructing a mapping relationship matrix between the three-dimensional spatial coordinate system and the ultrasonic intensity, and forming a discretized representation of the ultrasonic intensity distribution of sparse scanning at a low sampling rate;

[0015] S22. Selecting a sparse basis matrix and a measurement matrix, recovering a sparse coefficient vector from the low-dimensional measurement value using a compressed sensing reconstruction algorithm, and obtaining a reconstructed high-precision defect image through an inverse transform, thereby obtaining a reconstructed global ultrasonic intensity distribution at a high sampling frequency;

[0016] S23. Perform extreme value analysis on the global ultrasonic intensity distribution to identify the characteristic points with the strongest energy in the global ultrasonic intensity distribution at a high sampling frequency as global sparse positioning points of material defects.

[0017] Preferably, in S3, a single variable control method is used to optimize the value of L according to the signal-to-noise ratio variation trend of the standard defect imaging.

[0018] Preferably, S5 includes:

[0019] S51, using a sparse basis matrix, converting a sparse imaging result of low spatial resolution into a sparse coefficient vector of a sparse representation;

[0020] S52, determining a measurement matrix, sampling the sparse coefficient vector, and obtaining a low-dimensional measurement value;

[0021] S53. Use the compressed sensing reconstruction algorithm to recover the sparse coefficients from the low-dimensional measurement values, and finally obtain the reconstructed high-precision defect imaging through inverse transformation.

[0022] Preferably, in S2 or S5, the compressed sensing principle uses a measurement matrix and a sparse basis matrix to reconstruct the signal;

[0023] The measurement matrix options include but are not limited to Gaussian measurement matrix and Toeplitz measurement matrix;

[0024] Sparse basis matrix options include, but are not limited to, fast Fourier transform, discrete wavelet transform, and discrete cosine transform.

[0025] Preferably, in S2 or S5, the compressed sensing reconstruction algorithm options include but are not limited to convex optimization methods and greedy algorithms;

[0026] Among them, the convex optimization method includes iterative reweighted least squares method;

[0027] Greedy algorithms include orthogonal matching pursuit, compressed sampling matching pursuit, and subspace tracking imaging methods.

[0028] Therefore, the present invention adopts the above-mentioned laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing, which has the following technical effects:

[0029] (1) The present invention adopts a layered sampling strategy of "global sparse positioning-local fine scanning" to generate low-resolution sparse imaging through synthetic aperture focusing, and then uses the compressed sensing algorithm to complete super-resolution reconstruction, which can obtain high-precision defect images. It effectively combines the imaging advantages of synthetic aperture focusing with the sparse reconstruction capabilities of compressed sensing, and provides a feasible solution for efficient imaging of tiny defects.

[0030] (2) The present invention optimizes the number of measurement points of the random sparse sampling strategy , at the sampling frequency ( ) is 25%, the preliminary spatial positioning of the defect can be achieved, and high-resolution focused fine scanning is implemented in the double-peak feature area, organically integrating sparse scanning data with fine scanning data. The positioning accuracy of tiny defects of 0.35mm exceeds 90%, while the number of scanning points is reduced by 67%.

[0031] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The sparse distribution characteristics of the defect reflection echo intensity curve in an embodiment of the laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing;

[0033] Figure 2 : This is the relationship between the imaging signal-to-noise ratio and L in an embodiment of a laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing. DETAILED DESCRIPTION

[0034] The present invention can be explained in more detail by the following examples. The purpose of disclosing the present invention is to protect all changes and improvements within the scope of the present invention. The present invention is not limited to the following examples.

[0035] The present invention prepared 7075 aluminum specimens with artificial through holes, with dimensions of 250×20×20 mm³. Internal through-hole defects with diameters of 0.35 mm, 0.7 mm, 1.5 mm, and 3 mm were machined 10 mm below the sample surface by Shandong Ruixiang Mould Co., Ltd., which is certified by ISO 9001:2015. The defects were evenly distributed to verify the effectiveness of the method.

[0036] Considering that the fixed-distance scanning mode of the excitation-probe laser source maintains a constant distance between the excitation point and the probe point and moves them synchronously, the arrival time of the surface direct wave remains stable and the arrival time of the defect reflection echo shows regular changes, which not only effectively suppresses the aliasing of multimodal waveforms, but also simplifies the signal extraction algorithm through the spatiotemporal decoupling strategy. Although the movement of the probe point will introduce surface roughness interference, the reflectivity difference can be effectively compensated by signal normalization processing. Therefore, the present invention adopts a method of fixing the distance between the excitation-probe laser source and fixing both on a six-axis robotic arm, and realizing fixed-distance scanning by controlling the robotic arm. To avoid aliasing of the longitudinal wave and the surface wave, the spacing is set to 6mm; at the same time, considering the size of the smallest defect, the full-matrix high-resolution scanning step is set to 0.1mm.

[0037] Based on the above settings, the present invention carried out a laser ultrasonic B-scan experiment, and selected artificial through-hole defects with diameters of 3mm and 0.35mm as research objects, and systematically analyzed the evolution of the defect reflection echo signal intensity with scanning distance. Figure 1 As shown in the figure, the defect reflection echo intensity curve presents a typical bimodal distribution feature with the defect center as the symmetry axis, which verifies that the defect reflection echo intensity has a spatial sparse distribution characteristic. This characteristic is highly consistent with the prior conditions of the sparse reconstruction algorithm and conforms to the sparse law of compressed sensing, providing a reliable experimental basis for the quantitative characterization of defects based on compressed sensing theory.

[0038] The compressed sensing theory used in this paper is a theory that uses the sparsity or compression properties of signals to achieve efficient reconstruction. When a signal has a sparse representation in a certain transform domain, that is, most of the transform coefficients are close to zero and only a few are significantly non-zero, only a small number of measurements are needed to accurately restore the original signal. Let the signal to be restored (original signal) be dimensional vector , the measured value is dimensional vector ( Much smaller than ), the linear mapping relationship between the two can be expressed by the measurement matrix Expressed as:

[0039] ;

[0040] in, is an M×N matrix.

[0041] Signal to be restored Available sparse basis matrices express:

[0042] ;

[0043] in, is a sparse coefficient vector with the number of non-zero elements being , and satisfies Bringing the sparse representation into the measurement relationship, we can get , is the perception matrix. In practical applications, we can first use Solve to obtain sparse coefficient vector , and then calculate the original signal .

[0044] Based on the above, the present invention proposes a laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing. This method effectively combines the focused imaging advantages of synthetic aperture and the sparse reconstruction capability of compressed sensing, providing an efficient and reliable solution for intelligent defect identification under complex working conditions. The details are as follows:

[0045] S1. Within the global detection range of the sample to be tested, a random sparse sampling strategy is used to select measurement points, of which , is the number of scanning points required for full matrix high spatial resolution sampling, corresponding to the sparse sampling rate The selected scanning points are subjected to laser ultrasonic testing by the manipulator motion mechanism, and the time domain A-scan waveform data is recorded synchronously. The number of sparse sampling points must satisfy the sparse sampling theorem:

[0046] ;

[0047] Where c is a constant related to the signal-to-noise ratio and algorithm efficiency, and is usually between 1 and 4. The sparsity K is the number of peak points in the ultrasound signal intensity distribution curve, that is, K=2.

[0048] This embodiment also optimizes the number of sparse sampling points, and uses the imaging signal-to-noise ratio (SNR) as an evaluation index to evaluate the imaging quality of different sparse sampling point numbers. According to the compressed sensing sparse sampling theorem, the number of sparse sampling points is set. The value range is [20, 200], and the sampling is performed at an interval of 20, and different Perform synthetic aperture focusing imaging on the laser ultrasonic detection data under the sparse sampling rate ( ) curve, where the signal-to-noise ratio shows an approximately exponential change law with the increase of sampling rate. Taking into account the signal-to-noise ratio requirements, detection efficiency and sparse sampling theorem constraints, the optimal number of sparse sampling points is finally determined in this embodiment. =50 (random sampling rate 25%), this value satisfies both The theoretical lower limit of ≥16 can capture the maximum value of the intensity change curve, ensuring good imaging quality while avoiding efficiency loss caused by oversampling.

[0049] S2. Based on the principle of compressed sensing, the global ultrasonic intensity distribution of the full matrix scan at a high sampling rate is reconstructed, and the feature points of the global ultrasonic intensity distribution are identified through extreme value analysis to achieve global sparse positioning. The specific steps are as follows:

[0050] S21. Establish mapping relationship: Based on the sparse scan A-scan waveform data, extract the defect echo intensity of each spatial sampling point as the characteristic parameter and construct a three-dimensional spatial coordinate system The mapping relationship matrix with the ultrasonic intensity I forms a discrete representation of the global ultrasonic intensity distribution.

[0051] S22. Signal reconstruction: Use the compressed sensing reconstruction algorithm to recover the sparse coefficients from the low-dimensional measurement values, and finally obtain the reconstructed high-precision defect imaging through inverse transformation, thereby reconstructing the ultrasonic intensity distribution in the global area.

[0052] During this process, it is necessary to select appropriate sparse basis matrices and measurement matrices to establish a sparse coefficient vector that closely matches the intensity distribution of the defect echo signal. Sparsity is key to signal reconstruction. Selecting an appropriate sparse basis matrix, such as the fast Fourier transform (FFT), discrete wavelet transform (DWT), or discrete cosine transform (DCT), can ensure that the signal exhibits sparse characteristics under the transform. Given that the DWT can effectively capture the multi-layered structural features of an image, a property crucial for improving synthetic aperture focusing imaging quality, this embodiment selects the DWT as the sparse basis matrix. Furthermore, given that the elements of the Gaussian measurement matrix follow independent and identical distributions, they can satisfy the restricted isometry property (RIP) condition with extremely high probability. Furthermore, the Gaussian matrix exhibits asymptotic independence from any sparse basis, making it significantly superior to structured measurement matrices in terms of universality. Therefore, this embodiment selects the Gaussian measurement matrix. In other embodiments, different measurement matrices or sparse basis matrices may be selected based on practical circumstances.

[0053] Signal reconstruction algorithms fall into two main categories: convex optimization methods, such as iteratively reweighted least squares (IRLS), transform the signal reconstruction problem into a linear program with minimum norm. The other type is greedy algorithms, which gradually approximate the signal by selecting atoms most relevant to the observation. Typical greedy algorithms include orthogonal matching pursuit (OMP), compressed sampling matching pursuit (CoSaMP), and subspace pursuit (SP). These algorithms each offer advantages in computational efficiency and accuracy and are widely used for the rapid recovery of sparse signals. In practical applications, the signal reconstruction algorithm can be optimized based on the image signal-to-noise ratio.

[0054] S3. Use the global sparse positioning result as the center position of the local fine scan, and perform local fine scans with high spatial resolution stepping within the length range L on both sides of the center position.

[0055] In order to determine the optimal L value, this embodiment adopts a single variable control method to optimize the selection based on the signal-to-noise ratio variation trend of standard defect imaging. Specifically, by comparing the signal-to-noise ratio of synthetic aperture imaging images under different scanning lengths L, it is found that the imaging signal-to-noise ratio shows a non-monotonic trend of first increasing and then decreasing with the increase of L value. Figure 2 . In the initial scanning stage, the newly added scanning points are mainly concentrated in the effective area where the signal intensity is concentrated, which significantly improves the quality of synthetic aperture focusing imaging. However, when the scanning range exceeds the optimal threshold, the proportion of scanning points in the edge area with low contribution rate increases, which in turn leads to a decrease in the signal-to-noise ratio. Taking into account the quantitative analysis of scanning efficiency and imaging quality, L=1.5mm was finally selected as the optimal scanning parameter. This value maximizes the scanning efficiency while ensuring the imaging signal-to-noise ratio. The entire scanning process is completed through a high-resolution focusing mechanism, ensuring fine coverage of local areas.

[0056] S4, using global sparse positioning and local fine scanning strategies to obtain A-scan signal data, based on synthetic aperture focusing technology, combined with signal delay correction and coherent superposition methods, to achieve dynamic focusing on each pixel point in the detection area, generating low spatial resolution sparse imaging results.

[0057] The core idea of ​​synthetic aperture focusing imaging is the principle of time-delay superposition. By discretizing the scanning area into grid points in the x, y, and z directions, the coordinates of the laser excitation point and the detection point are defined. Assume that the internal defect position is C and the depth from the surface is z. The laser excitation point Located at coordinates , the detection point is , a total of Scanning points. Defect reflection signal S At the excitation point At, after time Arrives at the detection point. If there is a defect at the target location, the sound wave is reflected at the defect and the laser detects a clear signal. Otherwise, there is no obvious reflection in the signal. By superimposing the signals along all scanning points, the signal at the target defect location is enhanced, while the signal in the non-defect area is randomly distributed, achieving the positioning purpose. The mathematical expression is:

[0058] ;

[0059] Where, C Represents the target position point of the circular through hole inside the metal sample in the imaging image, Represents a discretized set of detection areas consisting of grid points. Reflection time By distance Harmony Decide, is the total path from the excitation point to the defect and then to the detection point, and the calculation formula is ,in 、 They are the distance between the excitation point and the defect, and the distance between the defect and the detection point, respectively.

[0060] S5. After completing the sparse imaging with low spatial resolution, in order to improve the image resolution, the compressed sensing technology is further used to achieve super-resolution reconstruction of the image, and finally obtain high-precision defect imaging. The specific steps are as follows:

[0061] S51. Use a sparse basis matrix to convert the sparse imaging results of low spatial resolution into a sparse coefficient vector form of sparse representation.

[0062] S52: Multiply the sparse coefficient vector by the preferred measurement matrix, sample the sparse coefficient vector, and obtain measurement values. Considering that the Toeplitz measurement matrix is ​​relatively simple in construction and highly efficient in computation, it can achieve high-quality image reconstruction at a low sampling rate. Therefore, in this step, the Toeplitz measurement matrix is ​​selected as the measurement matrix.

[0063] S53. Use the compressed sensing reconstruction algorithm to recover the sparse coefficients from the low-dimensional measurement values, and finally obtain the reconstructed high-precision defect imaging through inverse transformation.

[0064] This example also compared three typical reconstruction algorithms: OMP, IRLS, and CoSaMP. Their reconstruction performance for low-spatial-resolution sparse images was systematically evaluated at different downsampling ratios: 60%, 50%, 40%, 30%, and 20%. The results showed that the OMP algorithm maintained an RMSE of less than 0.033 at all downsampling ratios, achieving the best performance. In contrast, the RMSEs of the IRLS and CoSaMP algorithms increased to 0.049 (a 0.5-fold increase) and 0.062 (a 0.9-fold increase), respectively. Taking into account both the downsampling ratio and image reconstruction accuracy, this example ultimately selected the OMP imaging method at a 30% downsampling ratio.

[0065] In addition, this embodiment also performs high-precision imaging detection on internal through-hole defects of different diameters in the test piece sample, thereby obtaining high-precision defect imaging results. In order to quantitatively evaluate the super-resolution reconstruction imaging effect of the compressed sensing reconstruction technology, the reconstructed image is subjected to binary mask defect area representation processing: first, the image is preprocessed using the median filtering technology; then, based on the mean value of the filtered image, the image is processed. and standard deviation Setting thresholds ; Then extract the defect area through threshold segmentation. Specifically, when the pixel value Exceeding the threshold When the binary mask matrix The image is marked as 1 (defective area); otherwise, it is marked as 0 (background area). Analysis of the detection accuracy of defect diameters shows that the super-resolution imaging accuracy of defects with four different diameters remains above 90%, fully ensuring the imaging quality of defect detection.

[0066] Therefore, the present invention adopts the above-mentioned laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing, which can solve the problem of low efficiency of traditional laser ultrasonic synthetic aperture focusing technology under high-resolution scanning detection, and fully ensure the imaging quality of defect detection.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing is characterized by: The following steps are involved: S1. Within the global detection range of the sample to be tested, a random sparse sampling strategy is used to select The measurement points are sparsely scanned by the excitation-detection laser source fixed-distance scanning detection method, and the time domain A-scan waveform data is recorded synchronously; S2. Mapping the sparsely scanned A-scan waveform data to three-dimensional spatial coordinates, reconstructing the global ultrasonic intensity distribution of the full matrix scan at a high sampling rate based on the principle of compressed sensing, and identifying the feature points of the global ultrasonic intensity distribution through extreme value analysis to achieve global sparse positioning; S3. Use the global sparse positioning result as the center position of the local fine scan, and perform local fine scans with high spatial resolution stepping within the length range L on both sides of the center position; S4, based on the A-scan signal data recorded by global sparse positioning and local fine scanning, uses synthetic aperture focusing technology, combined with signal delay correction and coherent superposition methods, to achieve dynamic focusing on each pixel point in the detection area, generating a sparse imaging result with low spatial resolution; S5. Utilizing the sparse imaging results with low spatial resolution, the compressed sensing principle is further adopted to perform high-resolution compressed sensing reconstruction imaging to obtain high-precision laser ultrasonic defect imaging.

2. The laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing according to claim 1, characterized in that: In S1, the number of measurement points of the random sparse sampling strategy , according to the imaging signal-to-noise ratio, detection efficiency and sparse sampling theorem constraints, and the fixed-distance scanning detection method is used for sparse scanning.

3. The laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing according to claim 1, characterized in that S2 include: S21, extracting the defect echo intensity of each sampling point as a characteristic parameter, constructing a mapping relationship matrix between the three-dimensional spatial coordinate system and the ultrasonic intensity, and forming a discretized representation of the ultrasonic intensity distribution of sparse scanning at a low sampling rate; S22. Selecting a sparse basis matrix and a measurement matrix, recovering a sparse coefficient vector from the low-dimensional measurement value using a compressed sensing reconstruction algorithm, and obtaining a reconstructed high-precision defect image through an inverse transform, thereby obtaining a reconstructed global ultrasonic intensity distribution at a high sampling frequency; S23. Perform extreme value analysis on the global ultrasonic intensity distribution to identify the characteristic points with the strongest energy in the global ultrasonic intensity distribution at a high sampling frequency as global sparse positioning points of material defects.

4. The laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing according to claim 1, characterized in that: In S3, the single variable control method is used to optimize the value of L according to the signal-to-noise ratio change trend of standard defect imaging.

5. The laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing according to claim 1, characterized in that S5 include: S51, using a sparse basis matrix, converting a sparse imaging result of low spatial resolution into a sparse coefficient vector of a sparse representation; S52, determining a measurement matrix, sampling the sparse coefficient vector, and obtaining a low-dimensional measurement value; S53. Use the compressed sensing reconstruction algorithm to recover the sparse coefficients from the low-dimensional measurement values, and finally obtain the reconstructed high-precision defect imaging through inverse transformation.

6. The laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing according to claim 3 or claim 5, characterized in that: In S2 or S5, the compressed sensing principle uses the measurement matrix and sparse basis matrix to reconstruct the signal; The measurement matrix options include but are not limited to Gaussian measurement matrix and Toeplitz measurement matrix; Sparse basis matrix options include, but are not limited to, fast Fourier transform, discrete wavelet transform, and discrete cosine transform.

7. The laser ultrasonic defect imaging method based on sampling-imaging dual-order compressed sensing according to claim 3 or claim 5, characterized in that: In S2 or S5, the compressed sensing reconstruction algorithm options include but are not limited to convex optimization methods and greedy algorithms; Among them, the convex optimization method includes iterative reweighted least squares method; Greedy algorithms include orthogonal matching pursuit, compressed sampling matching pursuit, and subspace tracking imaging methods.

Images