Internal structure information detection method and detection system

Abstract

The invention discloses an internal structure information detection method and system. The internal structure information detection method comprises the following steps: constructing a sparse dictionary according to a pulse transmission signal; introducing a layer number X of a tested object as prior information, and constructing a compressed sensing reconstruction model according to the sparse dictionary; wherein the layer number X of the tested object is at least used for setting the sparseness K of the compressed sensing reconstruction model; and solving the compressed sensing reconstruction model, and calculating at least one item of internal structure information of the tested object according to the obtained pulse echo signal. The internal structure information detection method provided by the invention can be suitable for detected objects with various substrate materials, and has the advantages of general applicability, high detection speed and high precision.

Description

Technical Field

[0001] The present invention relates to the technical field of nondestructive testing, and particularly to a method and a system for detecting internal structure information. Background Art

[0002] At present, in the technical field of nondestructive testing, the use of pulse signals to detect the structural information of a measured object has gradually become the mainstream technical solution. Among them, compared with traditional technologies such as ultrasound and ray, terahertz technology has received great attention due to its high penetrability, excellent time resolution, and spatial resolution. In the prior art, the model method is usually adopted to detect the internal structure information of a measured object including thickness information. The technical means is to first model the structure of a sample related to the measured object, and then use the structural rules of the sample summarized by the model to deduce the structural information of other measured objects. However, this solution has high requirements for the cleanliness of the measured object and the echo signal, requires the measured object to have relatively uniform structural characteristics, and especially requires the substrate of the measured object to be metal. Summary of the Invention

[0003] One object of the present invention is to provide a method for detecting internal structure information to solve the technical problems of narrow application range, messy generated data, low detection accuracy and low accuracy rate of the existing nondestructive testing technology.

[0004] One object of the present invention is to provide a system for detecting internal structure information.

[0005] To achieve one of the above-mentioned objects of the present invention, an embodiment of the present invention provides a method for detecting internal structure information, including: constructing a sparse dictionary according to a pulse transmission signal; introducing the number of layers X of the measured object as prior information, and constructing a compressive sensing reconstruction model according to the sparse dictionary; wherein, the number of layers X of the measured object is at least used to set the sparsity K of the compressive sensing reconstruction model; solving the compressive sensing reconstruction model, and calculating at least one item of internal structure information of the measured object according to the obtained pulse echo signal.

[0006] As a further improvement of an embodiment of the present invention, the step of "constructing a compressive sensing reconstruction model according to the sparse dictionary" specifically includes: constructing the compressive sensing reconstruction model based on a greedy algorithm according to the sparse dictionary.

[0007] As a further improvement of an embodiment of the present invention, the "solving the compressive sensing reconstruction model" specifically includes: initializing model parameters, and at least setting an initial residual; calculating the correlation parameter between each atom in the sparse dictionary and the initial residual, determining and storing the most relevant atom in the support set; iteratively updating the support set, and updating the estimated value of the response sequence and the residual according to the support set until a preset iteration condition is satisfied, and reconstructing the pulse echo signal.

[0008] As a further improvement of an embodiment of the present invention, the "constructing the compressive sensing reconstruction model based on the greedy algorithm according to the sparse dictionary" specifically includes: constructing the compressive sensing reconstruction model based on the forward prediction matching pursuit algorithm according to the sparse dictionary.

[0009] As a further improvement of an embodiment of the present invention, the "solving the compressive sensing reconstruction model" specifically includes: initializing model parameters, at least setting an initial residual, setting the sparsity K according to the number of layers of the object under test, and setting the forward prediction parameter L according to the number of layers of the object under test; where K≥L; calculating the inner product between each atom in the sparse dictionary and the initial residual as the correlation parameter, determining and sequentially storing L relatively relevant atoms in the support set, and updating to obtain L candidate support sets; performing forward prediction on the relatively relevant atoms, updating to obtain the corresponding L estimated values of the response sequence according to the candidate support sets, updating to obtain the corresponding L residuals according to the estimated values of the response sequence, and determining and storing the most relevant atom in the support set according to the corresponding residuals; iteratively updating the support set, the estimated value of the response sequence, and the residual until the number of iterations is equal to K, determining the impulse response sequence according to the obtained several estimated values of the response sequence; reconstructing the pulse echo signal according to the impulse response sequence and the support set.

[0010] As a further improvement of an embodiment of the present invention, the "calculating at least one internal structure information of the object under test according to the obtained pulse echo signal" specifically includes: determining several pulse sparse vectors according to the obtained pulse echo signal; calculating the thickness of at least one layer of structure in the object under test according to at least two groups of obtained pulse sparse vectors and their corresponding time differences of flight.

[0011] As a further improvement of an embodiment of the present invention, the pulse transmission signal is configured as a terahertz signal, and the object under test includes a first coating; the step of "calculating the thickness of at least one layer structure in the object under test according to the obtained at least two groups of pulse sparse vectors and their corresponding time differences of flight" specifically includes: determining a first sparse vector and a second sparse vector corresponding to the first coating according to the order of the pulse sparse vectors and the position of the first coating in the object under test; determining the thickness of the first coating according to the time difference of flight, the speed of light, and the refractive index of the first coating; wherein the time difference of flight is the time difference between the first sparse vector and the second sparse vector.

[0012] As a further improvement of an embodiment of the present invention, the pulse transmission signal is configured as a terahertz signal, and the step of "constructing a sparse dictionary according to the pulse transmission signal" specifically includes: constructing the sparse dictionary according to a pulse reference signal formed by the terahertz signal transmitted in an air medium.

[0013] As a further improvement of an embodiment of the present invention, the step of "constructing a sparse dictionary according to the pulse transmission signal" specifically includes: constructing a pulse convolution model of the pulse transmission signal; discretizing the pulse convolution model in accordance with the Nyquist-Shannon sampling theorem to obtain a discrete convolution model; constructing the sparse dictionary according to the discrete convolution model.

[0014] As a further improvement of an embodiment of the present invention, the step of "constructing a pulse convolution model of the pulse transmission signal" specifically includes: constructing a pulse convolution model of the pulse transmission signal according to the relationship between the convolution of the pulse transmission signal and the unit response pulse of the object under test and the pulse echo signal; the step of "discretizing the pulse convolution model to obtain a discrete convolution model" specifically includes: discretizing the pulse convolution model according to the relationship between the convolution of the discrete transmission signal and the discrete response pulse of the object under test and the discrete echo signal to obtain a discrete convolution model; wherein the discrete echo signal is the discrete form of the pulse echo signal, the discrete response pulse is the discrete form of the unit response pulse, and the discrete transmission signal is the discrete form of the pulse transmission signal; the step of "using the sparsity relationship between the pulse transmission signal and the pulse echo signal as prior information, and constructing the sparse dictionary according to the discrete convolution model" specifically includes: arranging the discrete convolution model into a form that the discrete echo signal matrix is equal to the sum of the dot product of a first matrix and the discrete response pulse matrix and noise, and using the first matrix as the sparse dictionary.

[0015] As a further improvement of an embodiment of the present invention, the number of layers X of the object under test is the number of coating layers coated on the substrate of the object under test; the sparsity K is slightly greater than the number of layers X + 1 of the object under test.

[0016] As a further improvement of an embodiment of the present invention, the pulse transmission signal is configured as a terahertz signal, the substrate of the object under test is an anisotropic material, and the internal structure information includes the thickness of at least one layer of structure in the object under test.

[0017] As a further improvement of an embodiment of the present invention, the substrate of the object under test is a carbon fiber reinforced composite material.

[0018] To achieve one of the above-mentioned invention purposes, an embodiment of the present invention provides an internal structure information detection system, including a pulse generating device, a pulse receiving device, and an arithmetic device; the arithmetic device is configured to execute the high-precision pulse method described in any of the above technical solutions.

[0019] Compared with the prior art, the internal structure information detection method provided by the present invention detects internal structure information such as the thickness of a single layer or multiple layers at the object under test by sequentially constructing a sparse dictionary, constructing a compressive sensing reconstruction model, and solving the pulse echo signal. In this way, it is possible to avoid the material requirements for the object under test of the conventional propagation model, reduce the requirements for samples, improve universality, reduce the difficulty of variable control and convergence, and achieve the technical effect of obtaining detailed internal structure information in a short detection time. The internal structure information detection method provided by the present invention improves the detection and reconstruction accuracy of the compressive sensing reconstruction model for such objects under test by introducing the number of layers of the object under test and setting the sparsity accordingly. Even if the internal structure, material, and hierarchical composition of the object under test are complex, it is possible to more completely and accurately restore the actual internal situation of the object under test based on prior information. Description of the Drawings

[0020] Figure 1 It is a structural schematic diagram of the internal structure information detection system in an embodiment of the present invention.

[0021] Figure 2 It is a step schematic diagram of the internal structure information detection method in an embodiment of the present invention.

[0022] Figure 3 It is a step schematic diagram of an embodiment of the internal structure information detection method in the first embodiment of the present invention.

[0023] Figure 4 It is a step schematic diagram of an embodiment of the internal structure information detection method in the second embodiment of the present invention.

[0024] Figure 5 It is a step schematic diagram of an embodiment of the internal structure information detection method in the third embodiment of the present invention.

[0025] Figure 6 It is a step schematic diagram of the internal structure information detection method in the fourth embodiment of the present invention.

[0026] Figure 7 It is a schematic diagram of the steps of an embodiment of the internal structure information detection method in the fourth embodiment of the present invention.

[0027] Figure 8 It is a waveform schematic diagram of the pulse reference signal obtained by implementing the internal structure information detection method in an embodiment of the present invention.

[0028] Figure 9 It is a waveform schematic diagram of the sampled response pulse and the reconstructed pulse echo signal obtained by implementing the internal structure information detection method in an embodiment of the present invention.

[0029] Figure 10 It is a waveform schematic diagram of the pulse sparse vector obtained by implementing the internal structure information detection method in an embodiment of the present invention. Detailed Embodiments

[0030] The present invention will be described in detail below in conjunction with the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, method, or functional transformation made by those of ordinary skill in the art based on these embodiments is included within the protection scope of the present invention.

[0031] It should be noted that the term "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In addition, the terms "first", "second", "third", "fourth", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0032] An embodiment of the present invention provides an internal structure information detection system 100, as Figure 1 shown. The internal structure information detection system 100 includes a pulse generation device 11, a pulse reception device 12, and an arithmetic device 13.

[0033] Among them, "pulse" represents a state of the output signal, and it does not necessarily limit that the signal for detecting the internal structure information can only be output in the form of a pulse. In other cases, the signal can also be configured as a continuous output according to the needs of those skilled in the art.

[0034] The pulse generating device 11 is used to output a signal for internal structure detection to the object under test 200. Among them, the signal preferably has strong penetrability. In one embodiment, it can be a terahertz signal. By using its advantage of being transparent to most non-polar materials, it passes through the material surface to obtain the internal structure information of the material. The signal for internal structure detection can specifically be a pulse transmission signal Ti.

[0035] The internal structure information can specifically be the thickness of a certain structure at a certain position in the object under test 200. In other words, the internal structure information detection system 100 provided by the present invention can be used for thickness measurement. It can be understood that "thickness measurement" only represents one of the functions of the technical solution provided by the present invention, and does not necessarily limit that the system can only be used for thickness measurement. In one embodiment, the thickness can also be detected at multiple positions of the object under test 200, and the thickness is used as intermediate information to comprehensively analyze and obtain other internal structure information (such as porosity, uniformity) of the object under test 200. In other words, the present invention can at least realize the detection of the structure information of the object under test, and will not be elaborated below.

[0036] In one scenario, the object under test 200 can be a multi-layer structure including a substrate 20 and several coatings. Among them, the several coatings can be coated on a certain surface of the substrate 20 and stacked thereon, or can be uniformly arranged on two opposite surfaces of the substrate 20. Based on this, the internal structure information detection system 100 can specifically be used to detect the thickness of at least one of the several coatings.

[0037] In one embodiment, the substrate 20 can be a conventional metal material or a non-metal composite material. Specifically, the substrate 20 can be a carbon fiber reinforced polymer (CFRP). It has anisotropic properties, and the pulse transmission signal Ti will generate a response signal or echo signal with a complex path on its surface. The internal structure information detection system 100 provided by the present invention processes such a signal with a complex path to analyze and obtain the internal structure information of the object under test 200. When the object under test 200 is configured as a multi-layer structure, the internal structure information includes but is not limited to the thickness of the coating.

[0038] In other words, on the one hand, the internal structure information detection system 100 provided by the present invention can detect the internal structure information of the object under test 200 configured with CFRP as the substrate; on the other hand, the internal structure information detection system 100 provided by the present invention can detect the internal structure information of the object under test 200 configured as a multi-layer structure.

[0039] After the pulse transmission signal Ti is output from the pulse generating device 11, it passes through the surface of the object under test 200 layer by layer and is reflected on the surface of each layer of the object under test 200 to form a pulse response signal. Ideally, the pulse response signal can be directly used to calculate the internal structure information of the object under test 200, that is, as an ideal pulse echo signal; in the actual detection process, since the pulse response signal is interfered by factors such as waveform deflection and superposition, it can be processed first (for example, after reconstructing the pulse echo signal according to the pulse response signal) and then analyzed to obtain the internal structure information of the object under test 200.

[0040] For example, the pulse transmission signal Ti is reflected on the surface of the first coating 21 facing away from the substrate 20 to form a first pulse response signal Te1; the pulse transmission signal Ti is reflected on the surface of the second coating 22 facing away from the substrate 20 to form a second pulse response signal Te2; the pulse transmission signal Ti is reflected on the surface of the third coating 23 facing away from the substrate 20 to form a third pulse response signal Te3. In this way, the internal structure information detection system 100 can at least analyze the internal structure of the object under test 200 according to the above three groups of pulse response signals, and at least obtain information such as the arrangement and thickness of its internal coatings.

[0041] Although in the above description, the pulse transmission signal Ti corresponding to the three pulse response signals is unified, this does not limit the protection scope of the present invention. The pulse transmission signals corresponding to the three pulse response signals can also have three, and the corresponding relationship between them is based on the properties of the coating or the internal structure of the object under test 200 itself.

[0042] In one embodiment, the pulse transmission signal Ti is reflected on the surface of the substrate 20 close to the pulse generating device 11 to form a fourth pulse response signal Te4. The present invention does not exclude configuring the function of analyzing the substrate 20 and its internal structure on this basis, nor does it exclude using the difference between the fourth pulse response signal Te4 and the first three pulse response signals to determine whether the analysis of the coating part has been completed.

[0043] The pulse receiving device 12 is used to receive a plurality of pulse response signals in response to the pulse transmission signal Ti, analyze the pulse response signals and hand them over to the computing device 13; finally, the internal structure information of the object under test 200 required is calculated.

[0044] It can be understood that the positional relationship between the pulse receiving device 12 and the pulse generating device 11, or among the pulse receiving device 12, the pulse generating device 11 and the computing device 13, or the positional relationship among the pulse receiving device 12, the pulse generating device 11, the computing device 13 and the object under test 200 is not limited to Figure 1Under the shown scheme; any way of setting the positional relationship that can generate a pulse response signal corresponding to the pulse transmission signal Ti and can be obtained by the pulse receiving device 12 can be alternatively implemented in the present invention. In other words, since there must be a certain positional relationship between several coatings and the substrate 20, as long as the pulse transmission signal Ti is incident from one side of the coating, the internal structure information detection system 100 provided by the present invention will necessarily be able to analyze and obtain the internal structure information.

[0045] For example, in one embodiment, the pulse transmission signal Ti generated by the pulse generating device 11 is perpendicularly incident on the object under test 200, and the pulse receiving device 12 perpendicularly receives the corresponding pulse response signal. At this time, the pulse generating device 11 and the pulse receiving device 12 can be arranged on any straight line perpendicular to the surface of the object under test 200.

[0046] Based on this, the object under test 200 can be regarded as a physical system, and the subsequent analysis of internal structure information such as thickness can be carried out by using the property that the pulse transmission signal Ti only changes at the interfaces between multiple layers. Specifically, the process of analyzing the internal structure information can be constructed in the time domain, and in particular, the time-of-flight method (ToF, Time of Flight) can be used to complete or assist the process of analyzing the internal structure.

[0047] Continue as Figure 1 As shown, if the time when the first pulse response signal Te1 is received by the pulse receiving device 12 is set as t, then the second pulse response signal Te1, the third pulse response signal Te3, and the fourth pulse response signal Te4 successively have received times later than t. The reason for this time difference is that the transmission paths of the pulse transmission signal Ti inside the object under test 200 are not the same; in particular, it lies in the different setting positions and coating thicknesses of each coating; thus, the coating thickness can be inversely deduced based on the arrangement order of the response signals.

[0048] The computing device 13 is used to obtain the pulse response signal obtained at the pulse receiving device 12 and analyze and obtain the structure information of the object under test 200. In one embodiment, the computing device is used to execute the internal structure information detection method described in any of the following technical solutions, and in particular, execute the steps: constructing a sparse dictionary, introducing the number of layers of the object under test 200 as prior information, constructing and solving a compressive sensing reconstruction model, and calculating at least one internal structure information such as thickness according to the reconstructed pulse echo signal.

[0049] On the one hand, looking at the internal structure information detection system provided by the present invention as a whole, it can correspondingly have a detection method, including the steps of: operating the pulse generating device 11 to output a pulse transmission signal Ti; operating the pulse receiving device 12 to receive a pulse response signal corresponding to the pulse transmission signal Ti and sending it to the arithmetic device 13 for analysis; the arithmetic device 13 executes the internal structure information detection method provided by any one of the following technical solutions to analyze and obtain at least one item of internal structure information of the object under test 200.

[0050] On the other hand, in the embodiment where the arithmetic device 13 is configured to introduce the number of layers of the object under test 200 as prior information, the compressive sensing reconstruction model can be clearly determined, and based on the number of coating layers, the structural information of the coating can be analyzed quickly and accurately, especially the thickness of the coating can be analyzed; and this analysis process is independent of the material of the substrate 20. Even in the scenario where the substrate 20 is configured as an anisotropic material such as CFPR, the internal structure information detection system provided by the present invention can still distinguish and calculate quickly and accurately.

[0051] It can be understood that the arithmetic device 13 can also be integrated into the pulse receiving device 12. The pulse receiving device 12 can also be integrated with the pulse generating device 11. Based on this, in a preferred embodiment of the present invention, the internal structure information detection system 100 is integrated into an internal structure information detection device, which can simultaneously implement all the technical features in this article.

[0052] One embodiment of the present invention further provides an internal structure information detection method, as Figure 2 shown. This internal structure information detection method can be executed by any one of the above-mentioned arithmetic devices 13, internal structure information detection systems 100 or the internal structure information detection devices, or can also be executed by the pulse receiving device 12 or the pulse generating device 11. The internal structure information detection method specifically includes the following steps.

[0053] Step 31, constructing a sparse dictionary according to the pulse transmission signal.

[0054] Among them, "pulse" represents a state of the output signal, and it does not necessarily limit that the signal for detecting internal structure information can only be output in the form of a pulse; in other cases, this signal can also be configured as a continuous output according to the needs of those skilled in the art.

[0055] The pulse transmission signal can be configured as a terahertz signal, which has stronger penetration performance and is suitable for detection processes with high-precision requirements. When applying this internal structure information detection method to the scenario of detecting the internal structure thickness of a test object, by carrying the terahertz signal, it is possible to detect a test object with a thinner single-layer thickness, a thinner total thickness, and a smaller volume. In particular, by applying the internal structure information detection method provided by the present invention, it is possible to achieve thickness detection of a test object with a single-layer thickness of less than 40 μm and / or a total thickness of less than 105 μm.

[0056] In the initial stage of implementing the technical solution of the present invention, it is mainly based on the sparse representation theory to denoise the pulse response signal and reconstruct the pulse echo signal, so as to effectively improve the accuracy of pulse echo signal positioning. In this way, the problem of low detection accuracy caused by the overlap of terahertz and other pulse response signals, multiple reflections, and large noise can be solved, and the credibility of the pulse echo signal used to calculate the internal structure information can be improved; among them, the reasons for the overlap of pulse response signals, multiple reflections, and large noise may be one of the situations such as the sample being thin, the sample having defects, or the substrate material being anisotropic.

[0057] Preferably, prior information can be introduced in the construction process of the sparse dictionary, which can further improve the accuracy of subsequent operations. In a preferred embodiment, step 31 further includes the step of constructing the sparse dictionary according to the pulse reference signal formed by the terahertz signal transmitted in the air medium.

[0058] In this way, the transmission process in the air can be used as prior information, and an algorithm can be constructed according to the actual transmission process of the pulse transmission signal (or specifically, the terahertz signal) without considering the characteristics of the test object or its sample.

[0059] Step 32, introduce the number of layers of the test object as prior information, and construct a compressed sensing reconstruction model according to the sparse dictionary.

[0060] Constructing a compressed sensing reconstruction model based on the sparse dictionary can make full use of the correspondence between the non-zero values in the sparse dictionary and the coating surface (or the interface between coatings), quickly and accurately construct a compressed sensing reconstruction model corresponding to the current test object, and improve the operation speed and result accuracy in the subsequent solution process.

[0061] Since the number of layers of the test object is introduced as prior information for constructing the compressed sensing reconstruction model, it is possible to greatly shorten the operation time and improve the result accuracy in the subsequent process of solving the compressed sensing reconstruction model; the reconstruction and determination process of the pulse echo signal or the pulse sparse vector has a faster convergence speed and accuracy under the guidance of the number of layers of the test object.

[0062] In one implementation, solving the compressive sensing reconstruction model can be interpreted as a zero-norm solving (NP-hard) problem, specifically:

[0063]

[0064] Among them, y represents the discrete echo signal matrix, A represents the sparse dictionary, and h represents the discrete response pulse matrix. Solving ||h||0 can be understood as an NP-hard problem, that is, a problem to which all non-deterministic polynomial (NP. Non-deterministic Polynomial) problems can be reduced within polynomial time complexity.

[0065] In one implementation, the number of layers X of the object under test is at least used to set the sparsity K of the compressive sensing reconstruction model. In this way, the construction and solution process of the compressive sensing reconstruction model can be determined based on the number of layers of the object under test. For objects under test with different numbers of layers, the adaptability of the algorithm to the object under test and the accuracy of the output results can be improved. When the sparsity is used to guide the jump-out iteration condition in the process of solving the compressive sensing reconstruction model, the defect of long processing cycle caused by repeated iteration can also be avoided.

[0066] In a preferred implementation, the number of layers X of the object under test can be interpreted as the number of coating layers coated on the substrate 20 of the object under test. Taking Figure 1 the shown scenario as an example, at this time the number of coating layers of the object under test 200 is 3, that is, it is considered that the number of layers X of the object under test = 3. In this way, the structure of the substrate 20 can be ignored, and the substrate 20 can be distinguished from the surface coating. Regardless of its isotropy or anisotropy, regardless of how many layers it has, the internal structure information detection method provided by the present invention can accurately measure the coating thickness.

[0067] Of course, the above configuration scheme is not limited to the scenario of measuring the thickness of multi-layer structures, and can also be applied to other scenarios where the internal structure of the object under test needs to be analyzed.

[0068] The relationship between the sparsity K and the number of layers X of the object under test can be set to be greater than or less than. When K < X is set, the effect of only detecting the thickness of some structural layers can be achieved. In the implementation of using the time-of-flight method for thickness calculation, K ≥ 2.

[0069] Preferably, the sparsity K is slightly greater than the number of layers X + 1 of the object under test. Taking Figure 1Taking the shown scenario as an example, the number of layers of the object under test is 3. Therefore, the sparsity K set for the compressive sensing reconstruction model should be slightly greater than 4 to prevent the operator from being unable to promptly understand the credibility of the current detection result. The slightly greater amplitude can be set according to the number of layers or properties of the substrate 20; when the structure of the substrate 20 is relatively complex, the amplitude of sparsity K > 4 can be set larger. For example, sparsity K = 7 can be set to improve the robustness of the algorithm; when the structure of the substrate 20 is relatively simple, the amplitude of sparsity K > 4 can be set smaller. For example, sparsity K = 5 can be set to accelerate the iterative convergence speed.

[0070] In one implementation, the pulse transmission signal Ti is configured as a terahertz signal, which can realize the detection of the internal structure information of most objects, especially the on-line thickness detection.

[0071] In one implementation, the substrate 20 of the object under test is an anisotropic material. In other words, compared with the prior art, the internal structure information detection method provided by the present invention can ignore the complex response signal path caused by the anisotropic material and accurately detect the internal structure information of the object under test 200 with the anisotropic material as the substrate, especially the thickness detection of the thin coating applied thereon.

[0072] In one implementation, the substrate 20 of the object under test is a carbon fiber reinforced composite material, that is, the technical solution provided by the present invention can at least perform internal structure analysis on the object under test 200 with such an anisotropic material as the substrate.

[0073] Step 33: Solve the compressive sensing reconstruction model, and calculate at least one internal structure information of the object under test according to the obtained pulse echo signal.

[0074] The compressive sensing reconstruction model based on the sparse decomposition theory can be configured into various types, such as being constructed based on convex optimization algorithms or greedy algorithms. The convex optimization algorithms include but are not limited to the gradient descent method, the coordinate descent method, and the Newton iteration method. The greedy algorithms include but are not limited to the matching pursuit algorithm (MP, Matching Pusuit) and the orthogonal matching pursuit algorithm (OMP, Orthogonal Matching Pursuit).

[0075] The solution of the compressive sensing reconstruction model can be further explained as: solving the compressive sensing reconstruction model according to the pulse response signal corresponding to the pulse transmission signal, and reconstructing the pulse echo signal. It can be seen that the compressive sensing reconstruction model can be explained as a model that at least characterizes the theoretical relationship between the pulse transmission signal and the pulse echo signal.

[0076] Calculating at least one piece of internal structure information of the object under test can be directly based on the pulse echo signal or based on the pulse sparse vector. Therefore, what is obtained by solving the compressive sensing reconstruction model can be a pulse sparse vector containing at least part of the content of the pulse echo signal. In other words, the "pulse echo signal" summarized in step 33 of the present invention can be interpreted as a signal containing at least part of the content of the pulse echo signal.

[0077] When there are multiple layers inside the object under test, at least one piece of internal structure information of the object under test can be interpreted as the structural information of at least one layer in the object under test, or as the internal structural situation of at least one dimension of the whole object under test.

[0078] At least one layer in the object under test can be interpreted as at least one layer in several coating layers of the object under test. Of course, for an object under test with other internal structures, it can also be interpreted as at least one layer in several other encapsulation layer structures; or replace the number of layers of the object under test with other prior information and calculate another structural data corresponding to this prior information. The structural information of this layer can be the uniformity, porosity or thickness as described above. In other words, at least one piece of internal structure information of the object under test; in one embodiment, can include the thickness of at least one layer structure in the object under test.

[0079] Thus, the internal structure information detection method provided by the present invention can, on the one hand, solve for more accurate structural information based on the sparse representation theory, reducing the difficulty of variable control and convergence; on the other hand, it can set the sparsity by introducing the number of layers as prior information, improving the accuracy of detection and reconstruction. Even if the structure of the object under test is relatively complex, it can completely and accurately restore the actual internal situation and accelerate the convergence speed.

[0080] In addition, by solving the compressive sensing reconstruction model, the present invention obtains a reconstructed pulse echo signal. Even if the pulse response signal is superimposed and confused due to various factors, by implementing the above technical solution, it is still possible to analyze a relatively clean pulse echo signal from the chaotic signal, thereby improving the accuracy of internal structure information calculation.

[0081] In the first embodiment provided by the present invention, "constructing a compressive sensing reconstruction model according to the sparse dictionary" in step 32 can specifically include the steps of: constructing the compressive sensing reconstruction model based on the greedy algorithm according to the sparse dictionary.

[0082] Since the computational complexity of the greedy algorithm is relatively low and it is easy to implement, the basic idea of solving the sparse solution can be explained as: transforming the original NP-hard problem (solving the zero norm) into an approximate solution of solving this problem to achieve the purpose of approximating the target signal. Thus, when it is applied to the scenario of analyzing the internal structure of the object under test, it has a more significant effect.

[0083] Preferably, as Figure 3 combined with Figure 2 shown, in an embodiment of the first embodiment of the present invention, the internal structure information detection method may specifically include the following steps.

[0084] Step 31, construct a sparse dictionary according to the pulse transmission signal.

[0085] Step 32, introduce the number of layers of the object under test as prior information, and construct a compressive sensing reconstruction model according to the sparse dictionary.

[0086] Step 331, initialize the model parameters, and at least set the initial residual.

[0087] Step 332, calculate the correlation parameter between each atom in the sparse dictionary and the initial residual, determine and store the most relevant atom in the support set.

[0088] Step 333, iteratively update the support set, and update the estimated value of the response sequence and the residual according to the support set until the preset iteration condition is met, and reconstruct the pulse echo signal.

[0089] Step 334, calculate at least one item of internal structure information of the object under test according to the obtained pulse echo signal.

[0090] Among them, the step 32 may be specifically configured to include step 320: introduce the number of layers of the object under test as prior information, and construct a compressive sensing reconstruction model based on the greedy algorithm according to the sparse dictionary.

[0091] Among them, the initialization of the model parameters in the step 331 includes, but is not limited to, the initialization of the residual, and may also include the initialization of the sparse dictionary and the initialization of the number of iterations. Based on this, the setting of the sparse dictionary and the number of iterations may be combined with other embodiments provided by the present invention; for example, according to the pulse transmission signal, especially using the pulse reference signal formed during its transmission in the air as prior information to construct the sparse dictionary to realize the initialization process of the sparse dictionary; or for another example, introduce the number of layers X of the object under test as prior information, set the sparsity K according to the number of layers X of the object under test, and set the number of iterations of the model according to the sparsity K to realize the initialization of the number of iterations.

[0092] In step 332, the correlation parameter is used to evaluate the degree of correlation between the corresponding atom in the sparse dictionary and the initial residual. On the one hand, the magnitude of the degree of correlation between the two is reflected by the numerical value of the correlation parameter; on the other hand, during the iteration of the operation, the initial residual is continuously updated to form a new residual, and in each iteration process, the calculation of the correlation parameter is always based on the updated residual.

[0093] In a preferred embodiment, the correlation parameter may be the inner product between each atom in the sparse dictionary and the residual. In other words, step 332 may specifically include the steps of: performing an inner product operation on each atom in the sparse dictionary and the residual to obtain the correlation parameter.

[0094] If the magnitude of the correlation parameter is positively correlated with the magnitude of the correlation between the atom and the residual in the sparse dictionary, then the most relevant atom represents the one with the largest corresponding correlation parameter; conversely, it represents the one with the smallest corresponding correlation parameter.

[0095] Storing the most relevant atom in the support set does not necessarily mean storing the atom in the corresponding sparse dictionary in the support set. In one implementation, the support set is interpreted as a set containing indices pointing to several specific atoms in the sparse dictionary. In this case, step 332 may include the steps of: determining and storing the index of the most relevant atom in the support set. Of course, the present invention does not exclude the existence of a support set containing specific atoms in the sparse dictionary and a support set containing indices of the specific atoms in some scenarios.

[0096] In step 333, other data in the compressive sensing reconstruction model can be updated based on the update of the support set, so as to realize the iteration of the support set.

[0097] During the iteration of the support set, it includes but is not limited to updating the estimated value of the response sequence; the estimated value of the response sequence represents the reconstruction result of the pulse echo signal or the pulse sparse vector at the current stage. During the iteration of the support set, it includes but is not limited to updating the residual; the residual represents the difference between the pulse response signal and the estimated value of the response sequence; preferably, it can be the difference between the two; thus, it represents the difference between the actual signal and the reconstructed signal.

[0098] After performing the above update of the residual, the updated residual can be calculated again with each atom in the sparse dictionary for the correlation parameter, and the most relevant atom can be continuously stored in the support set until a preset iteration condition is met. In a preferred implementation, the preset iteration condition is set to the number of iterations being equal to the sparsity. Since the sparsity is set according to the parameters of the object under test, at least the data corresponding to each layer of the object under test can be obtained, so as to judge the thickness of at least one layer.

[0099] In the second implementation provided by the present invention, in step 32, "constructing the compressive sensing reconstruction model based on the greedy algorithm according to the sparse dictionary" may specifically include the steps of: constructing the compressive sensing reconstruction model based on the forward prediction matching pursuit algorithm according to the sparse dictionary.

[0100] Compared with the matching pursuit algorithm, the forward prediction matching pursuit algorithm will not project the residual back to the previous direction again during the subsequent projection process. Compared with the orthogonal matching pursuit algorithm, the forward prediction matching pursuit algorithm can take into account the global optimum, that is, it can consider the influence of the selected atoms on the final iterative residual, and can improve the credibility of the output result by adjusting the residual projection and the update basis of the support set on the basis of avoiding the adjustment of a large number of parameters.

[0101] Preferably, as Figure 4 Combined with Figure 2 shown, in an embodiment of the second embodiment of the present invention, the internal structure information detection method may specifically include the following steps.

[0102] Step 31, construct a sparse dictionary according to the pulse transmission signal.

[0103] Step 32, introduce the number of layers of the object under test as prior information, and construct a compressive sensing reconstruction model according to the sparse dictionary.

[0104] Step 331A, initialize the model parameters, at least set the initial residual, set the sparsity K according to the number of layers of the object under test, and set the forward prediction parameter L according to the number of layers of the object under test.

[0105] Step 332A, calculate the inner product between each atom in the sparse dictionary and the initial residual as the correlation parameter, determine and sequentially store L relatively relevant atoms in the support set, and update to obtain L candidate support sets.

[0106] Step 332B, perform forward prediction on the relatively relevant atoms, update to obtain the corresponding L response sequence estimated values according to the candidate support set, update to obtain the corresponding L residuals according to the response sequence estimated values, and determine and store the most relevant atom in the support set according to the corresponding residuals.

[0107] Step 333A, iteratively update the support set, the response sequence estimated value, and the residual until the number of iterations is equal to K, and determine the impulse response sequence according to the obtained several response sequence estimated values.

[0108] Step 333B, reconstruct the pulse echo signal according to the impulse response sequence and the support set.

[0109] Step 334, calculate at least one item of internal structure information of the object under test according to the obtained pulse echo signal.

[0110] Wherein, K≥L.

[0111] Overall, the above steps 331A to 333B are equivalent to implementing the following iterative process: residual projection; constructing L candidate support sets each containing atoms with relatively large correlation parameters and / or their indices; predicting the atom with the minimum residual and / or its index; updating the support set; response sequence estimation; residual update. Its overall architecture lies in adding a forward prediction strategy during the selection process of the sparse vector to observe the influence of each preselected atom and the atoms in the existing support set on the global residual, avoiding the problem of low accuracy caused by only considering the influence of atoms on the current residual.

[0112] Among them, step 32 can be specifically configured to include step 321: introducing the number of layers of the object under test as prior information, and based on the sparse dictionary, constructing a compressive sensing reconstruction model using the forward prediction matching pursuit algorithm.

[0113] In step 331A, the setting of the initial residual and the sparsity K can refer to the first embodiment. The forward prediction parameter L represents the number of iterations of the forward prediction strategy. The larger the value of the forward prediction parameter L, the better the reconstruction effect of the pulse echo signal or the pulse sparse vector.

[0114] In step 332A, the step of calculating the correlation parameter according to the initial residual can refer to the first embodiment. In this embodiment, during each iteration (a total of K times), instead of directly storing the most relevant atom or its index into the support set, L relatively relevant atoms or their indices are taken and stored into the support set. After obtaining L candidate support sets each containing L relatively relevant atoms or their indices, the residual is updated and evaluated accordingly. In this way, before storing the relatively relevant atoms, since the support set already contains several most relevant atoms or their indices obtained from previous iterations, global judgment can be achieved during the overall evaluation and update process.

[0115] It can be understood that the process of storing one of the L relatively relevant atoms into the support set and updating to obtain a candidate support set can be interpreted as one iteration. Based on this, in step 332A, the set forward prediction parameter is used as the number of iterations to set the iteration condition. For example, when L = 5, it is equivalent to iterating five times.

[0116] The step of performing forward prediction on the more relevant atoms is described in step 332B. Among them, for the step of updating the response sequence estimate according to the candidate support set, reference can be made to the step of updating the response sequence estimate according to the support set in the first embodiment; since there are L candidate support sets, the finally obtained response sequence estimates can correspondingly include L. For the step of updating the residual according to the response sequence estimate, reference can be made to the first embodiment; since there are L response sequence estimates, there are also L corresponding residuals. In this way, the corresponding relationship among "candidate support set - response sequence estimate - residual" is constructed. For the step of determining and storing the most relevant atom in the support set, reference can also be made to the first embodiment, and a support set including the atom itself or a support set including the index of the atom can be constructed.

[0117] For step 332A and step 332B, in one embodiment, the L residuals can be determined one by one, that is, the above steps can be specifically: determining and storing the first atom or its index among the L more relevant atoms in the support set, and updating to obtain the first candidate support set; performing forward prediction on the first atom, updating to obtain the corresponding first response sequence estimate according to the first candidate support set, and updating to obtain the corresponding first residual accordingly;...; determining and storing the Lth atom or its index among the more relevant atoms in the support set, and updating to obtain the Lth candidate support set; performing forward prediction on the Lth atom, updating to obtain the corresponding Lth response sequence estimate according to the Lth candidate support set, and updating to obtain the corresponding Lth residual accordingly.

[0118] Furthermore, step 332A, step 332B or their variations actually correspond to the process of the first iteration in step 333A (a total of K times).

[0119] For step 333A, in each iteration, it is necessary to calculate the correlation parameter between each atom in the sparse dictionary and the updated residual, and select L more relevant atoms among them. After corresponding L internal iterations, L candidate support sets, L corresponding sequence estimates are formed, and the obtained L residuals are screened to obtain the most relevant atom or its index as the result of this iteration and stored in the support set; the above iteration process is executed a total of K times.

[0120] In this way, an impulse response sequence composed of several response sequence estimates can be reconstructed, and based on this sequence and the support sets of the K atoms obtained by iteration, the reconstruction of the impulse echo signal or the impulse sparse vector can be completed. Due to the global residual evaluation, the obtained reconstruction result can achieve global optimality.

[0121] As a supplementary explanation, in the technical solution provided by the present invention, the atoms in the sparse dictionary can be interpreted as feature sequences in the sparse dictionary. When the sparse dictionary is configured in matrix form, its atoms can be interpreted as matrix columns; each column is an atom.

[0122] Next, taking two layers of coating as an example, i.e., X = 2, the implementation process of the present invention will be described in conjunction with Figure 1 , and the implementation process of the present invention will be described. Assume that the refractive index of the first coating 21 is 2.56 and the thickness is 66.04 μm; assume that the refractive index of the second coating 22 is 2.10 and the thickness is 35.56 μm.

[0123] First, perform step 31 and its derivative steps, and use the pulse transmission signal or the pulse reference signal Te(ref) formed by it in the air to form a sparse dictionary; among them, the reference signal is as Figure 8 shown, the vertical axis is the amplitude AMP, and the horizontal axis is the time point Time Point.

[0124] Secondly, the sparsity required to calculate the thickness of the two layers of coating is at least 3; when the pulse transmission signal is configured as a terahertz signal, considering that the terahertz signal is sensitive to environmental humidity and temperature, the signal collected in the laboratory still contains a small amount of noise. Therefore, the sparsity K can be set to 5 (>X + 1 = 3), and the forward prediction parameter L is also set to 5 (K≥L).

[0125] When iterating to the second time (i.e., K = 2), at this time, there should be one atom selected in the first iteration (K = 1) in the support set.

[0126] Based on this, through the inner product operation, calculate the correlation parameter between each atom in the sparse dictionary and the residual updated after the first iteration (K = 1). Select the L atoms with the highest correlation parameters. For each atom, use the forward prediction strategy to calculate the influence of the atom on the overall residual.

[0127] Specifically, sequentially add the atoms with the highest correlation to the support set to form a candidate support set containing 2 atoms (when K = n, there should be n - 1 atoms in the corresponding candidate support set). Subsequently, update the residual according to the candidate support set. After repeating the above steps to obtain L residuals, select the residual, the corresponding support set, and the response sequence estimate value with the largest corresponding correlation parameter as the result of this residual iteration.

[0128] Since the set sparsity is 5 and two residual iterations have been performed so far, it is necessary to repeat steps 332A to 332B a total of 3 times; each time, update the support set, the response sequence estimate value, and the residual, and finally obtain the global prediction residual, and determine the response sequence estimate value and the support set accordingly.

[0129] The pulse echo signal Te(rec) reconstructed by the above technical solution of the present invention is as Figure 9 (b) shown, and the actual sampled response pulse Te(spl) at the object under test is as Figure 9(a). In the figure, the vertical axis is the amplitude AMP and the horizontal axis is the time point Time Point. Compared with the prior art, the pulse waveform reconstructed in this application has obvious peaks, and it is relatively easy to distinguish the time point corresponding to the coating boundary position in the object under test.

[0130] In one embodiment, a pulse sparse vector Space-Vector as shown in Figure 10 can be further obtained, so as to calculate the internal structure information of the object under test according to the distribution of the pulse sparse vector Space-Vector at different time points; in particular, calculate the thickness of at least one layer of coating on the substrate 20. Preferably, the time-of-flight method is used to calculate the coating thickness. According to the Figure 10 data in, the thickness of the first coating can be finally calculated to be 64.93 microns, and the thickness of the second coating is 38.13 microns. Compared with the actual thickness, the deviation is extremely small.

[0131] In the third embodiment provided by the present invention, step 33 "calculating at least one internal structure information of the object under test according to the obtained pulse echo signal" may specifically include the steps of: determining a number of pulse sparse vectors according to the obtained pulse echo signal; calculating the thickness of at least one layer of structure in the object under test according to at least two sets of obtained pulse sparse vectors and their corresponding time-of-flight differences. In this way, the layer thickness inside the multi-layer structure can be calculated more accurately. It should be noted that this embodiment will be described by taking the calculation of thickness according to the pulse sparse vector as an example. The scheme of directly calculating the thickness according to the reconstructed pulse echo signal included in the present invention, as well as the scheme of calculating other internal structure information, is similar thereto and will not be elaborated here.

[0132] Preferably, as shown in Figure 5 , in an embodiment of the third embodiment of the present invention, the pulse transmission signal is configured as a terahertz signal, and the object under test includes a first coating. Based on this, the internal structure information detection method may specifically include the following steps.

[0133] Step 31, constructing a sparse dictionary according to the pulse transmission signal.

[0134] Step 32, introducing the number of layers of the object under test as prior information, and constructing a compressive sensing reconstruction model according to the sparse dictionary.

[0135] Step 330, solving the compressive sensing reconstruction model.

[0136] Step 3341, determining a number of pulse sparse vectors according to the obtained pulse echo signal.

[0137] Step 3342A: Determine a first sparse vector and a second sparse vector corresponding to the first coating according to the order of the obtained pulse sparse vector and the position of the first coating in the object under test.

[0138] Step 3342B: Determine the thickness of the first coating according to the time difference of flight, the speed of light, and the refractive index of the first coating.

[0139] Specifically, in Step 3342A, the distribution order of the pulse sparse vectors is basically consistent with the arrangement order of the coatings in the direction from the side close to the pulse generating device to the side far from it. For example, if the first coating 21, as Figure 1 shown, is located at the position closest to the pulse generating device 11 in the object under test 200, then Figure 9 (b) or Figure 10 the first pulse echo signal or the first pulse sparse vector reconstructed in [b] corresponds to the surface of the first coating 21 close to the pulse generating device 11; and so on.

[0140] Since the surface of the first coating 21 facing away from the pulse generating device 11 is equivalent to the surface of the second coating 22 close to the pulse generating device 11, therefore, corresponding to the first coating 21, there will be two pulse echo signals or two pulse sparse vectors. Thus, the thickness of the first coating 11 is determined by using the difference between the two. For example, the first pulse response signal Te1 corresponds to the first pulse sparse vector, and the second pulse response signal Te2 corresponds to the second pulse sparse vector. On the side of the pulse receiving device 12, there is a time difference Δt between them.

[0141] Based on this, when the pulse transmission signal is configured to be incident perpendicular to the object under test (or at least perpendicular to the first coating), the thickness d of the first coating can at least satisfy:

[0142]

[0143] where c is the speed of light, Δt is the time difference of flight of the coating, and n is the refractive index of the coating.

[0144] In the fourth embodiment provided by the present invention, as Figure 6 shown, the internal structure information detection method may specifically include the following steps.

[0145] Step 311: Construct a pulse convolution model of the pulse transmission signal.

[0146] Step 312: Discretize the pulse convolution model in accordance with the Nyquist-Shannon sampling theorem to obtain a discrete convolution model.

[0147] Step 313: Construct a sparse dictionary according to the discrete convolution model.

[0148] Step 32: Introduce the number of layers of the object under test as prior information, and construct a compressive sensing reconstruction model based on the sparse dictionary.

[0149] Step 33: Solve the compressive sensing reconstruction model, and calculate at least one piece of internal structure information of the object under test according to the obtained pulse echo signal.

[0150] For step 312, the Nyquist-Shannon sampling theorem reveals that: if a system uniformly samples an analog signal at a rate at least twice the highest frequency of the signal, then the original analog signal can be completely recovered from the discrete values generated by the sampling. Based on this, it is possible to set the output of the pulse transmission signal and the reception of the pulse response signal to satisfy this theorem, and thus discretize the continuous convolution model accordingly.

[0151] In this way, a sparse dictionary can be constructed quickly and accurately.

[0152] Preferably, as Figure 7 shown, in an embodiment of the fourth embodiment of the present invention, the internal structure information detection method may specifically include the following steps.

[0153] Step 311A: Construct a pulse convolution model of the pulse transmission signal according to the relationship between the convolution of the pulse transmission signal and the unit response pulse of the object under test and the pulse echo signal.

[0154] The response of the object under test to the pulse transmission signal can be regarded as conforming to a linear relationship; in other words, the reflection and propagation of the pulse transmission signal on the multi-layer object under test can be regarded as a system. Based on this, the reflected signal (pulse echo signal) in the time domain can be obtained by convolving the incident signal (pulse transmission signal) with the transfer function of the system (i.e., the impulse response function of the system). Therefore, the pulse echo signal y(t) can at least satisfy:

[0155]

[0156] where y(t) is the pulse echo signal; x(t) is the pulse transmission signal; h(t) is the unit response pulse; is convolution; t is the delay time; e(t) represents noise.

[0157] Step 312A: Discretize the pulse convolution model according to the relationship between the convolution of the discrete transmission signal and the discrete response pulse of the object under test and the discrete echo signal, and obtain a discrete convolution model.

[0158] where the discrete echo signal is the discrete form of the pulse echo signal; the discrete response pulse is the discrete form of the unit response pulse; the discrete transmission signal is the discrete form of the pulse transmission signal.

[0159] Referring to Step 312, following the Nyquist-Shannon sampling theorem, the discrete echo signal y(n) can at least satisfy:

[0160]

[0161] wherein, the discrete echo signal y(n) is the discrete form of the pulse echo signal y(t); the discrete transmission signal x(n) is the discrete form of the pulse transmission signal x(t); the discrete response pulse h(n) is the discrete form of the unit response pulse h(t).

[0162] Step 313A, arrange the discrete convolution model into: the form that the discrete echo signal matrix is equal to the dot product of the first matrix and the discrete response pulse matrix plus noise, and use the first matrix as the sparse dictionary.

[0163] Regarding the multi-layer structure coating as a physical system, the propagation path of the pulse transmission signal only changes at the coating interfaces. The pulse reference signal formed by the pulse transmission signal in the air can be used as prior information to construct the sparse dictionary. Since the number of layers of the coating in the object under test is often much smaller than the sampling points of the internal structure information detection system or its method, most of the elements in the sequence formed by the unit response pulse h are 0, and only a few non-zero values correspond to each coating interface. Considering the actual use environment, the discrete convolution model can be arranged into:

[0164] y = Ah(n) + e.

[0165] wherein, A is the first matrix, that is, the sparse dictionary; y is the pulse echo signal, and e is the noise.

[0166] Furthermore, according to the several discrete signals obtained in Step 312A, the discrete convolution model can be specifically arranged into:

[0167]

[0168] wherein, y(n) is an m×1 dimensional matrix; A is an m×m dimensional matrix; h(n) is an m×1 dimensional matrix. Preferably, m is 2400.

[0169] wherein, Step 311A is included in Step 311, Step 312A is included in Step 312, and Step 313A is included in Step 313.

[0170] It can be understood that the above four embodiments and several examples provided by the present invention can be combined to form a better embodiment or implementation manner.

[0171] In summary, the internal structure information detection method provided by the present invention constructs a sparse dictionary, a compressive sensing reconstruction model, and solves the pulse echo signal in sequence to detect the internal structure information such as the thickness of a single layer or multiple layers at the object under test. In this way, the material requirements for the object under test of the conventional propagation model can be avoided, the requirements for samples can be reduced, the universality can be improved, and the difficulties of variable control and convergence can be reduced, achieving the technical effect of obtaining detailed internal structure information in a short detection time. The internal structure information detection method provided by the present invention improves the detection and reconstruction accuracy of the compressive sensing reconstruction model for such objects under test by introducing the number of layers of the object under test and setting the sparsity accordingly. Even if the internal structure, material, and hierarchical composition of the object under test are complex, the actual internal conditions of the object under test can be restored more completely and accurately based on the prior information.

[0172] It should be understood that although this specification is described according to the embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0173] The series of detailed descriptions listed above are only specific descriptions of the feasible embodiments of the present invention, and they are not intended to limit the protection scope of the present invention. Any equivalent embodiments or modifications made without departing from the technical spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting internal structure information, characterized in that, Comprising: Construct a sparse dictionary according to the pulse transmission signal; Introduce the number of layers X of the object under test as prior information, and construct a compressive sensing reconstruction model according to the sparse dictionary; wherein, the number of layers X of the object under test is at least used to set the sparsity K of the compressive sensing reconstruction model; Solve the compressive sensing reconstruction model, and calculate at least one internal structure information of the object under test according to the obtained pulse echo signal.

2. The internal structure information detection method according to claim 1, characterized in that The "constructing a compressive sensing reconstruction model according to the sparse dictionary" specifically includes: Construct the compressive sensing reconstruction model based on the greedy algorithm according to the sparse dictionary.

3. The internal structure information detection method according to claim 2, characterized in that The "solving the compressive sensing reconstruction model" specifically includes: Initialize the model parameters, and at least set the initial residual; Calculate the correlation parameter between each atom in the sparse dictionary and the initial residual, determine and store the most relevant atom in the support set; Iteratively update the support set, and update the response sequence estimate value and the residual according to the support set until the preset iteration condition is met, and reconstruct the pulse echo signal.

4. The internal structure information detection method according to claim 2, wherein The "constructing the compressive sensing reconstruction model based on the greedy algorithm according to the sparse dictionary" specifically includes: Construct the compressive sensing reconstruction model based on the forward prediction matching pursuit algorithm according to the sparse dictionary.

5. The internal structure information detection method according to claim 4, wherein The "solving the compressive sensing reconstruction model" specifically includes: Initialize the model parameters, at least set the initial residual, set the sparsity K according to the number of layers of the object under test, and set the forward prediction parameter L according to the number of layers of the object under test; where K≥L; Calculate the inner product between each atom in the sparse dictionary and the initial residual as the correlation parameter, determine and sequentially store L relatively relevant atoms in the support set, and update to obtain L candidate support sets; Perform forward prediction on the relatively relevant atoms, update to obtain the corresponding L response sequence estimate values according to the candidate support sets, update to obtain the corresponding L residuals according to the response sequence estimate values, and determine and store the most relevant atom in the support set according to the corresponding residuals; Iteratively update the support set, the response sequence estimate value, and the residual until the number of iterations is equal to K, and determine the impulse response sequence according to the obtained several response sequence estimate values; Reconstruct the pulse echo signal according to the impulse response sequence and the support set.

6. The internal structure information detection method according to claim 1, characterized in that The "calculating at least one internal structure information of the object under test according to the obtained pulse echo signal" specifically includes: Determine several pulse sparse vectors according to the obtained pulse echo signal; Calculate the thickness of at least one layer of structure in the object under test according to at least two groups of obtained pulse sparse vectors and their corresponding time-of-flight differences.

7. The internal structure information detection method according to claim 6, characterized in that, The pulse transmission signal is configured as a terahertz signal, and the object under test includes a first coating; the "calculating the thickness of at least one layer of structure in the object under test according to at least two groups of obtained pulse sparse vectors and their corresponding time-of-flight differences" specifically includes: Determine a first sparse vector and a second sparse vector corresponding to the first coating according to the order of the pulse sparse vectors and the position of the first coating in the object under test; Determine the thickness of the first coating based on the time difference of flight, the speed of light, and the refractive index of the first coating; wherein the time difference of flight is the time difference between the first sparse vector and the second sparse vector.

8. The internal structure information detection method according to claim 1, wherein The pulse transmission signal is configured as a terahertz signal, and the "constructing a sparse dictionary according to the pulse transmission signal" specifically includes: Construct the sparse dictionary according to the pulse reference signal formed by the terahertz signal transmitted in the air medium.

9. The internal structure information detection method according to claim 1, wherein The "constructing a sparse dictionary according to the pulse transmission signal" specifically includes: Construct a pulse convolution model of the pulse transmission signal; Follow the Nyquist-Shannon sampling theorem to discretize the pulse convolution model to obtain a discrete convolution model; Construct the sparse dictionary according to the discrete convolution model.

10. The internal structure information detection method according to claim 9, characterized in that, The "constructing a pulse convolution model of the pulse transmission signal" specifically includes: Construct a pulse convolution model of the pulse transmission signal according to the relationship between the convolution of the pulse transmission signal and the unit response pulse of the object under test and the pulse echo signal. The "discretizing the pulse convolution model to obtain a discrete convolution model" specifically includes: Discretize the pulse convolution model according to the relationship between the convolution of the discrete transmission signal and the discrete response pulse of the object under test and the discrete echo signal to obtain a discrete convolution model; wherein the discrete echo signal is the discrete form of the pulse echo signal, the discrete response pulse is the discrete form of the unit response pulse, and the discrete transmission signal is the discrete form of the pulse transmission signal. The "using the sparsity relationship between the pulse transmission signal and the pulse echo signal as prior information and constructing the sparse dictionary according to the discrete convolution model" specifically includes: Arrange the discrete convolution model into the form that the discrete echo signal matrix is equal to the sum of the dot product of the first matrix and the discrete response pulse matrix and the noise, and use the first matrix as the sparse dictionary.

11. The internal structure information detection method according to claim 1, characterized in that The number of layers X of the object under test is the number of coating layers coated on the substrate of the object under test; the sparsity K is slightly greater than the number of layers X + 1 of the object under test.

12. The internal structure information detection method according to claim 1, characterized in that The pulse transmission signal is configured as a terahertz signal, the substrate of the object under test is an anisotropic material, and the internal structure information includes the thickness of at least one layer of structure in the object under test.

13. The internal structure information detection method according to claim 12, characterized in that The substrate of the object under test is a carbon fiber reinforced matrix composite material.

14. An internal structure information detection system, characterized in that It includes a pulse generating device, a pulse receiving device, and an arithmetic device; the arithmetic device is configured to execute the high-precision pulse method according to any one of claims 1-13.

Images