Multi-frame average shear speckle interference method

Abstract

The invention discloses a multi-frame averaging shear speckle interference method, which is applied to a multi-frame averaging shear speckle interference system, and the system comprises a multi-frame image acquisition device based on hardware synchronous control, a phase recovery device and a single 3 * 3 sine-cosine mean filtering device. According to the invention, the high-frequency synchronous driver generates a multi-step signal and a high-frequency square wave signal to synchronously control the camera and the PZT to achieve millisecond-level synchronization, a multi-frame phase shift diagram can be accurately acquired in a fixed phase shift state, and the dynamic measurement requirement is met; single 3 * 3 sine-cosine mean filtering processing is combined on the basis of a multi-frame average pre-phase recovery method and a multi-frame average post-phase recovery method, so that phase noise is remarkably reduced, and meanwhile, high-frequency details are reserved; the image precision and the calculation speed are balanced through a multi-frame average pre-phase recovery method and a multi-frame average post-phase recovery method, and the calculation time is only 1.2 times of that of a traditional method while high precision is kept.

Description

Technical Field

[0001] The present invention relates to the technical field of optical detection, and particularly to a multi-frame averaging shear speckle interferometry method. Background Art

[0002] Composite materials are widely used in high-performance fields such as aerospace. Internal defects can seriously damage the structural integrity and service life, posing significant safety hazards. Fast, accurate, and non-destructive detection methods are crucial for reducing safety risks and failure rates.

[0003] Shear speckle interferometry is a full-field non-contact and non-destructive optical detection method based on laser interferometry, aiming to detect minute deformations on the surface of an object. It has been widely applied to the defect detection of aerospace-grade composite materials. It can identify subsurface defects in components such as control surfaces and airframe structures, and effectively detect defects such as delamination, debonding, poor adhesion, wrinkles, cracks, and impact damage.

[0004] In traditional speckle interferometry measurements, the shear speckle phase map usually contains a large amount of random noise. The noise mainly comes from the inherent randomness of the speckle pattern and various interference factors during the image acquisition process. When coherent laser irradiates a rough surface, it generates randomly distributed speckles, whose intensity and phase are also random, resulting in multiplicative noise. Other noise sources include optical element defects, ambient light fluctuations, dark current, and detector-related quantum noise. In addition, limitations of image processing algorithms and insufficient spatial sampling introduce post-acquisition artifacts, jointly reducing the fringe quality and measurement accuracy.

[0005] Traditional denoising methods for speckle interferometry images include adaptive filtering methods using anisotropic sine-cosine averaging, complex amplitude domain filtering methods operating in the complex amplitude reconstruction stage, methods combining phase filtering based on Fourier transform with clustering algorithms, methods based on wavelet transform, etc. Modern denoising methods for speckle interferometry images are deep learning methods of Speckle Denoising Convolutional Neural Network (SDCNN).

[0006] In the process of implementing the technical solution of the present invention, the inventors of this patent have at least found the following technical problems in the prior art: 1. Contradiction between noise suppression and detail preservation: Traditional filtering algorithms may introduce new noise or artifacts in the image, especially when the filter parameters are not properly selected. The filtering process may reduce the image resolution, especially when applying some global filters. This is because these filters usually change the frequency content of the image, resulting in the loss of high-frequency detail information and affecting the micro-defect detection accuracy.

[0007] 2. Dynamic measurement limitations: Existing methods rely on single-frame or a small number of frame image acquisitions, making it difficult to balance noise suppression and temporal resolution. Moreover, hardware asynchronous operations are prone to introducing phase shift errors.

[0008] 3. High computational complexity: Multistage filtering or deep learning algorithms require complex computational resources and are difficult to meet the requirements of real-time detection. For example, Convolutional Neural Network (CNN) requires a large amount of training data and has a high computational cost.

[0009] In summary, the existing denoising methods for speckle interferometry images cannot simultaneously take into account image accuracy, computational efficiency, and dynamic measurement requirements. Summary of the Invention

[0010] The present invention provides a multi-frame averaging shear speckle interferometry method for high-precision detection of micro-deformations and internal defects on the surface of composite materials, solving the problem that the existing denoising methods for speckle interferometry images cannot simultaneously take into account image accuracy, computational efficiency, and dynamic measurement requirements. It can meet the requirements of noise suppression and detail retention for high-resolution phase images in the aerospace field, while taking into account high precision, high efficiency, and dynamic adaptability.

[0011] A multi-frame averaging shear speckle interferometry method provided by the present invention is applied to a multi-frame averaging shear speckle interferometry system. The system includes an image acquisition device, a phase recovery device, and a single-shot 3×3 sine-cosine mean filtering device based on hardware synchronous control. The method includes: In a multi-frame image acquisition device based on hardware synchronous control, a multi-step staircase signal and a high-frequency square wave signal are generated by a high-frequency synchronous driver to synchronously control the camera and the PZT. The multi-step staircase signal is used as the PZT drive signal to control the PZT displacement, and the square wave signal is used as the camera trigger signal to trigger the camera to capture images through the rising edge. The rising edge or falling edge of each step of the staircase signal is strictly synchronized with the rising edge of the square wave. Multiple camera frame periods correspond to the duration of a stepping signal. The multi-frame multi-step time phase-shifting method is used to make the PZT introduce multiple discrete phase shifts. Before and after the deformation of the measurement object, the camera captures multiple frame phase shift diagrams of the PZT under the same position in multiple phase shift intensity modes to obtain multiple frame phase shift diagrams in the reference state and multiple frame phase shift diagrams in the deformed state. In the phase recovery device, the multi-frame phase-shifted images in the reference state and the multi-frame phase-shifted images in the deformed state are processed by using the multi-frame average pre-phase recovery method or the multi-frame average post-phase recovery method to obtain a phase reconstruction image; wherein, the multi-frame average pre-phase recovery method calculates the average value of the multi-frame phase-shifted images in the two states before phase recovery to obtain an average phase-shifted image, and then calculates the average phase difference image as the phase reconstruction image; the multi-frame average post-phase recovery method calculates the average value of the multiple phase differences of the multi-frame phase-shifted images in the two states to obtain the final phase difference image as the phase reconstruction image. The single-shot 3×3 sine-cosine mean filtering device is used to perform single-shot 3×3 sine-cosine mean filtering on the phase reconstruction image to obtain a filtered image.

[0012] One or more technical solutions provided by the present invention have at least the following technical effects or advantages: 1. Noise suppression and detail retention: Based on the multi-frame average pre-phase recovery method and the multi-frame average post-phase recovery method, combined with single-shot 3×3 sine-cosine mean filtering, by multi-frame averaging combined with single-shot filtering, the phase noise is significantly reduced while the high-frequency details are retained.

[0013] 2. Dynamic adaptability: The multi-step ladder signal and the high-frequency square wave signal are generated by the high-frequency synchronous driver to synchronously control the camera and the PZT to achieve millisecond-level synchronization, which can accurately collect multi-frame phase-shifted images in the fixed phase-shift state, support high-speed multi-frame acquisition, and meet the dynamic measurement requirements.

[0014] 3. Computational efficiency optimization: The two multi-frame average phase recovery methods (the multi-frame average pre-phase recovery method and the multi-frame average post-phase recovery method) balance the image accuracy and the computational speed. The multi-frame average post-phase recovery method maintains high accuracy while the computational time is only 1.2 times that of the traditional method.

[0015] 4. Multiple applications: In addition to being applicable to shear speckle interferometry, the present invention can also be applied to electronic speckle interferometry, laser speckle interferometry, and laser holographic imaging technology. Description of the Drawings

[0016] Figure 1 It is a schematic diagram of the process of realizing shear speckle interferometry by using the four-frame four-step time phase-shift method; Figure 2 It is a process diagram of obtaining a phase-shifted image by using the four-frame four-step time phase-shift method under high-frequency synchronous triggering; Figure 3 It is the phase reconstruction image obtained by using the multi-frame average pre-phase recovery method and the multi-frame average post-phase recovery method; Figure 4 For Figure 3The filtered image obtained after applying single - time 3×3 sine - cosine mean filtering to the phase reconstruction diagram in Figure 5 is Figure 3 the phase reconstruction diagram in Figure 4 the denoising performance evaluation index diagram corresponding to the filtered image in Figure 6 is a schematic diagram of the calculation time (normalized time) of the multi - frame average pre - phase recovery method and the multi - frame average post - phase recovery method; Figure 7 is the phase reconstruction diagram obtained by using the multi - frame average pre - phase recovery method and the multi - frame average post - phase recovery method in the non - destructive testing experiment; Figure 8 is Figure 7 the filtered image obtained after applying single - time sine - cosine mean filtering to the phase reconstruction diagram in Figure 9 is Figure 7 the phase reconstruction diagram in Figure 8 the denoising performance evaluation index diagram corresponding to the filtered image in Figure 10 is a schematic diagram of the shear - speckle interference system based on hardware - synchronous multi - frame averaging of the present invention; In the figure: Method 1 - multi - frame average pre - phase recovery method; Method 2 - multi - frame average post - phase recovery method. Detailed implementation manner

[0017] The present invention provides a multi - frame averaging shear - speckle interference method for high - precision detection of surface micro - deformation and internal defects of composite materials, which solves the problem that the existing denoising methods for speckle interference images cannot simultaneously take into account image accuracy, computational efficiency, and dynamic measurement requirements, and can meet the requirements of noise suppression and detail retention for high - resolution phase images in the aviation and aerospace fields, while taking into account high precision, high efficiency, and dynamic adaptability.

[0018] First, the terms appearing in the specification will be explained respectively.

[0019] Multi - frame averaging: By collecting multiple speckle interference images in a fixed phase - shift state, taking them as multi - frame phase - shift diagrams, and performing averaging processing on the multi - frame phase - shift diagrams to improve the imaging quality.

[0020] Phase recovery: The process of extracting phase information from speckle interference images for calculating the displacement change of the object surface.

[0021] PZT: Includes piezoelectric materials and mirrors that generate micron - level displacements through voltage control for precisely adjusting the optical path phase.

[0022] Secondly, the core innovations of the present invention mainly include the following three points.

[0023] 1. In the multi-frame image acquisition device based on hardware synchronous control, a multi-step ladder signal and a high-frequency square wave signal are generated by a high-frequency synchronous driver to synchronously trigger the PZT and the camera, and a multi-frame multi-step time phase-shift method is improved based on the multi-step time phase-shift method, which can accurately acquire multi-frame phase-shift images in a fixed phase-shift state.

[0024] 2. The multi-frame average pre-phase recovery method and the multi-frame average post-phase recovery method are proposed for the first time. These two multi-frame average phase recovery methods can significantly improve the imaging quality, enhance the noise suppression ability, and at the same time retain the key phase detail information. Specifically, by increasing the number of frames, the phase fringe noise is significantly reduced without reducing the image resolution. Through experimental verification, after adopting the multi-frame average phase recovery method, the equivalent number of fixations (ENL) of the phase reconstruction image is enhanced, and the values of the speckle suppression index (SSI) and the speckle mean preservation index (SMPI) are reduced, indicating that the imaging quality improves with the increase in the number of frames.

[0025] Among them, the multi-frame average pre-phase recovery method reduces the calculation amount by directly averaging the multi-frame phase-shift images and then calculating the phase difference, and is applicable to real-time detection scenarios; The multi-frame average post-phase recovery method improves the accuracy by independently calculating each frame of the phase-shift image and then taking the average, and is applicable to static high-precision detection scenarios.

[0026] 3. Single 3×3 sine-cosine mean filtering: After obtaining the phase reconstruction image by the multi-frame average phase recovery method, perform single 3×3 sine-cosine mean filtering on it to avoid the loss of high-frequency details caused by multi-level filtering.

[0027] To better understand the multi-frame average shear speckle interferometry method of the present invention, the following will be described in detail in conjunction with the accompanying drawings of the specification and specific embodiments. Obviously, the embodiments described in the present invention are part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0028] Such as Figure 10As shown in the figure, a multi-frame averaging shear speckle interferometry method provided by the present invention is applied to a multi-frame averaging shear speckle interferometry system. The system mainly includes three component devices, namely a multi-frame image acquisition device based on hardware synchronous control, a phase recovery device, and a single 3×3 sine-cosine mean filtering device. The improvement points of the present invention focus on the first two component devices.

[0029] The multi-frame image acquisition device based on hardware synchronous control of the present invention is a Michelson-type shear interferometer device, which is composed of a laser, a Michelson-type misaligned interferometer, a camera, a shear mirror, and a piezoelectric-driven mirror (PZT-based Mirror, abbreviated as PZT). The laser generates a coherent light beam. The Michelson-type misaligned interferometer divides the incident light beam into two beams. After passing through the object under test, the two beams will be slightly misaligned. The camera captures two misaligned images generated by the Michelson-type misaligned interferometer and records the interference fringe pattern. The shear mirror in the Michelson-type misaligned interferometer introduces the misalignment of the light beam. Therefore, it generates the interference of two coherent light beams. The PZT is used to precisely control the misalignment amount. By adjusting the position of the mirror in the PZT, the degree of misalignment can be adjusted.

[0030] In speckle interferometry measurement, the high-precision and robust TPS (Time Phase Shifting) method is used to measure the interference phenomenon of light. The TPS method can be subdivided into single-step TPS method, three-step TPS method, four-step TPS method, and five-step TPS method. The TPS methods with three steps or more in the non-single-step TPS method are called multi-step TPS methods. Here, the four-step TPS method is taken as an example for illustration.

[0031] As Figure 1 shown, the laser beam emitted by the laser passes through the expander and illuminates the surface of the measurement object. The laser undergoes diffuse reflection on the surface of the measurement object, and the generated light beam is divided into two misaligned light beams by the Michelson-type misaligned interferometer. These two misaligned light beams are respectively reflected by mirror 1 and mirror 2. Subsequently, they interfere with each other to form the speckle interference fringes recorded by the camera. The camera captures four phase-shifted images in the reference state and four phase-shifted images in the deformed state. According to the phase in the reference state and the phase in the deformed state, the phase difference is calculated.

[0032] During the measurement process, the PZT is driven by a high-frequency synchronous driver, and four discrete phase shifts are introduced using the four-step TPS method. At each discrete phase shift, the camera captures one phase-shifted image of the speckle interference fringes in a phase shift intensity mode. The camera captures four phase-shifted images in four phase shift intensity modes for each state before and after deformation.

[0033] First, obtain four phase-shifted images of the object to be measured in the reference state , where i represents the phase-shift intensity pattern. The phase shift between adjacent patterns is , which belongs to the fixed phase shift. Each phase-shifted image in the reference state respectively represents: , where represents the four phase-shifted images obtained in the reference state, represents the image pixel coordinates, is the background image at the pixel coordinates in the speckle interferogram, is the modulated image of the interference term at the pixel coordinates , is the phase image at the pixel coordinates in the reference state.

[0034] Secondly, similar to the previous step, obtain four phase-shifted images of the object to be measured in the deformed state . Where i represents the phase-shift intensity pattern, and the phase shift between adjacent patterns is , which belongs to the fixed phase shift. Each phase-shifted image in the deformed state is respectively expressed as: , where represents the four phase-shifted images obtained in the deformed state, is the phase image at the pixel coordinates in the deformed state.

[0035] To avoid repeated execution of time-consuming arctangent operations, use the following phase difference calculation formula to calculate the phase difference at different pixel coordinates , and this phase difference calculation formula is expressed as: , where represents the displacement in the vertical direction at the pixel coordinates , represents the displacement in the horizontal direction at the pixel coordinates .

[0036] To achieve the purpose of multi-frame averaging, based on the traditional multi-step TPS method, add the processing characteristics of multiple frames, and design a multi-frame multi-step TPS method. Secondly, based on the traditional speckle interference image acquisition device, add a high-frequency synchronous driver component for realizing hardware synchronization control, and construct a multi-frame image acquisition device based on hardware synchronization control.

[0037] The high-frequency synchronous driver controls the camera and the PZT simultaneously. Taking the four-frame four-step TPS method as an example, the following is an illustration. Figure 2 It shows that the high-frequency synchronous driver generates a high-frequency square wave signal and a four-step staircase signal simultaneously. The rising edge or falling edge of each step of the four-step staircase signal is strictly synchronized with the rising edge of the square wave. In addition, the four-step staircase signal is used to drive the PZT to generate correct interference fringes, while the rising edge of the square wave triggers the camera to capture images.

[0038] One step signal in the multi-step staircase signal corresponds to multiple square wave signals, that is, the camera frame period is a small part of the duration of one step signal. . According to this hardware synchronous control method, multiple phase-shifted images corresponding to multiple phase-shift intensity patterns of the PZT at the same position can be accurately captured multiple times. The multi-frame phase-shifted diagrams in the reference state are denoted as , where i represents the phase-shift intensity pattern, and k represents the number of frames. Correspondingly, the multi-frame phase-shifted diagrams in the deformed state are denoted as . In Figure 2 , one step signal corresponds to 3 square wave signals ( n = 3), represents the camera frame period, represents the duration of one step signal. Therefore, the camera frame period is 1 / 3 of the duration of one step signal . The solid dots represent the moments when the phase-shifted images are completely captured, and λ represents the wavelength of the laser.

[0039] The high-frequency synchronization of the PZT and the camera can ensure that multi-frame phase-shifted diagrams can be accurately obtained for each phase-shift intensity in the reference state and the deformed state, so as to be used to perform subsequent phase recovery steps.

[0040] In the phase recovery device, the present invention improves the conventional phase recovery algorithm and proposes two improved phase recovery methods, namely the multi-frame average pre-phase recovery method and the multi-frame average post-phase recovery method.

[0041] The first one: the multi-frame average pre-phase recovery method; The multi-frame average pre-phase recovery method is characterized by obtaining the result of the average value of multi-frame phase-shifted diagrams before phase recovery. It calculates the average value of the multi-frame phase-shifted diagrams in two states before phase recovery to obtain the average phase-shifted diagram, and then calculates the average phase difference image as the phase reconstruction diagram. The average phase difference image is caused by the displacement change of the measurement object.

[0042] The multi-frame average pre-phase recovery method specifically includes the following steps: First, calculate the average value of the multi-frame phase-shift maps generated under each phase-shift intensity pattern in the reference state to obtain multiple average phase-shift maps in the reference state; Second, calculate the average value of the multi-frame phase-shift maps generated under each phase-shift intensity pattern in the deformed state to obtain multiple average phase-shift maps in the deformed state; Finally, based on the multiple average phase-shift maps in the reference state and the multiple average phase-shift maps in the deformed state, calculate the average phase difference image caused by the displacement change of the measurement object.

[0043] Taking the multi-frame four-step TPS method as an example, first, calculate the average value of the multi-frame phase-shift maps generated under the four phase-shift intensity patterns in the reference state respectively to obtain four average phase-shift maps in the reference state , and the calculation formula is: , where k represents the number of frames, n represents the maximum number of frames, represents the multi-frame phase-shift maps generated under the four phase-shift intensity patterns in the reference state.

[0044] Then, similarly calculate the average value of the multi-frame phase-shift maps generated under the four phase-shift intensity patterns in the deformed state respectively to obtain four average phase-shift maps in the deformed state , and the calculation formula is: , where represents the multi-frame phase-shift maps generated under the four phase-shift intensity patterns in the deformed state.

[0045] Finally, based on the four average phase-shift maps in the reference state , the four average phase-shift maps in the deformed state , calculate the average phase difference image caused by the displacement change of the measurement object , and the calculation formula is: , where represents the average displacement amount in the vertical direction at the pixel coordinate , represents the average displacement amount in the horizontal direction at the pixel coordinate .

[0046] The second one: the multi-frame average post-phase recovery method; The multi-frame average post-phase recovery method is an alternative method to the multi-frame average pre-phase recovery method. The multi-frame average post-phase recovery method calculates the average value of multiple phase differences of the multi-frame phase-shift maps in two states to obtain the final phase difference image as the phase reconstruction map.

[0047] The multi-frame average phase retrieval method specifically includes the following steps: First, calculate multiple phase images of multiple phase-shifted images generated under multiple phase-shift intensity patterns in the reference state to obtain multiple phase images in the reference state; then, calculate multiple phase images of multiple phase-shifted images generated under multiple phase-shift intensity patterns in the deformed state to obtain multiple phase images in the deformed state; next, calculate multiple phase difference images generated under multiple phase-shift intensity patterns based on the multiple phase images in the reference state and the multiple phase images in the deformed state; finally, calculate the average value of the multiple phase difference images to calculate the final phase difference image.

[0048] Taking the multi-frame four-step TPS method as an example, calculate multiple phase images of multiple phase-shifted images generated under four phase-shift intensity patterns in the reference state to obtain multiple phase images in the reference state , and the calculation formula is: , where k represents the number of frames, n represents the maximum number of frames, represents multiple phase-shifted images generated under four phase-shift intensity patterns in the reference state; Calculate multiple phase images of multiple phase-shifted images generated under four phase-shift intensity patterns in the deformed state to obtain multiple phase images in the deformed state , and the calculation formula is: , where represents multiple phase-shifted images generated under four phase-shift intensity patterns in the deformed state; Based on the multiple phase images in the reference state and the multiple phase images in the deformed state , calculate multiple phase difference images generated under four phase-shift intensity patterns ; Calculate the average value of the multiple phase difference images to calculate the final phase difference image , and the calculation formula is: .

[0049] The multiple phase difference images correspond to two calculation formulas and can be selected according to the actual situation.

[0050] The multiple phase difference images The first calculation formula is: .

[0051] The multiple phase difference images The second calculation formula is: , Among them, represents the multi-frame average displacement in the vertical direction, represents the multi-frame average displacement in the horizontal direction, represents the multi-frame phase-shifted images generated in four phase-shift intensity modes in the reference state, represents the multi-frame phase-shifted images generated in four phase-shift intensity modes in the deformed state, represents the pixel coordinates.

[0052] The second calculation formula improves the accuracy by averaging at the phase level. However, it requires n times of arctangent operations, which increases the computational complexity and reduces the processing speed, making it less suitable for real-time or dynamic measurements. To improve the computational efficiency, simplify the formula for calculating the final phase difference image . After the simplification operation, the calculation formula is: , Among them, represents the multi-frame simplified phase difference image.

[0053] Finally, in the single 3×3 sine-cosine mean filtering device, perform single 3×3 sine-cosine mean filtering on the phase reconstruction map after phase recovery processing to obtain the filtered image.

[0054] To evaluate the computational efficiency of the two multi-frame average phase recovery methods of the present invention, after obtaining the phase reconstruction image, it further includes: using the normalized time to evaluate the computational efficiency of the multi-frame average pre-phase recovery method and the multi-frame average post-phase recovery method.

[0055] To evaluate the denoising performance of the multi-frame average shear speckle interferometry method provided by the present invention, after obtaining the filtered image, it further includes: using ENL, SSI , SMPI at least one of the indicators in to evaluate the denoising performance of the filtered image.

[0056] The following introduces these three denoising performance indicators respectively: ENL (equivalent number of looks), SSI (speckle suppression index) and SMPI (speckle mean retention index).

[0057] ENL Measure the noise smoothing effect in the uniform image area. A higher ENL value indicates a stronger speckle noise reduction ability in the uniform area. Its calculation formula is as follows: , Among them, represents the denoised phase image, represents the average value of the denoised phase image, represents the variance of the denoised phase image.

[0058] SSI The noise suppression is quantified by comparing the denoised image with the original speckle image. SSI The lower the value, the stronger the noise suppression ability. The calculation formula is as follows: , Among them, represents the original speckle image, represents the average speckle intensity in the uniform region of the image, represents the average value of the original speckle image, represents the variance of the original speckle image.

[0059] SMPI It can effectively evaluate the noise suppression performance and the average retention performance. SMPI The lower the value, the better the performance of the denoising method in terms of mean retention and noise reduction. The calculation formula is as follows: , Among them, the coefficient .

[0060] Deformation measurement embodiment; To verify the feasibility of the proposed multi-frame averaging shear speckle interferometry method, a shear mirror, PZT, synchronous drive circuit, camera, beam splitter, micro-deformation device, and laser are used as the optical setup of the shear speckle deformation measurement device.

[0061] During the implementation, the composite panel is fixed at its four edges to ensure uniform laser irradiation and reduce the influence of environmental noise. A micrometer screw is used to manually introduce a small displacement. The eight-frame four-step TPS method is used to capture multiple phase-shifted speckle patterns of the composite panel before and after deformation. A custom-designed synchronous drive circuit is used as a high-frequency synchronous driver, and eight phase-shifted maps are captured as eight-frame phase-shifted maps at each phase before and after deformation. The eight-frame phase-shifted maps in the reference state captured before deformation are denoted as and the eight-frame phase-shifted maps in the deformed state captured after deformation are denoted as .

[0062] To evaluate the influence of the number of frames on the quality of the phase map, phase recovery is performed using different numbers of frames. Using the multi-frame averaging pre-phase recovery method, Figure 3The phase reconstruction diagrams shown in A1 to A8, where "Am" (m = 1 - 8) represents the phase reconstruction diagram restored with m-frame phase shift diagrams at each offset step. Similarly, using the multi-frame average post-phase recovery method, Figure 3 the phase reconstruction diagrams shown in B1 to B8 are obtained, where "Bm" (m = 1 - 8) represents the phase reconstruction diagram restored with m-frame phase shift diagrams at each offset step. Since when m = 1, the two multi-frame average pre-phase recovery methods are consistent with the traditional four-step TPS method, Figure 3 the phase reconstruction diagram shown in A1 in Figure 3 is the same as the phase reconstruction diagram shown in B1 in

[0063] From Figure 3 it can be seen that both multi-frame average phase recovery methods significantly improve the quality of the phase reconstruction diagram. Nevertheless, it is still recommended to use a simple post-processing filter, and the post-processing filter can be a single 3×3 sine-cosine mean filtering device. In this embodiment, a sine-cosine mean filter with a 3×3 window is applied to all reconstructed phase diagrams. After the filtering process, Figure 3 A1 to A8 in Figure 4 respectively obtain to . Similarly, Figure 3 B1 to B8 in Figure 4 respectively obtain to .

[0064] To further evaluate the denoising performance of the proposed method, three denoising performance metrics, namely ENL , SSI and SMPI , are used to evaluate the image quality before and after filtering shown in Figure 3 and Figure 4 . It should be noted that when the filter exceeds the average value, ENL and SSI are no longer reliable. However, SMPI can effectively evaluate the noise suppression performance and the average preservation performance.

[0065] Calculate the Figure 3 and Figure 4 values of each sub-graph in ENL , SSI and SMPI , and the results are as shown in Figure 5 . The SSI and SMPI values are normalized by dividing each value of the metric by the first value of the metric.

[0066] In addition, this embodiment also evaluates the computational efficiency of two multi-frame average phase retrieval methods. The test environment mainly includes a computer with a Windows 11 system equipped with an AMD Ryzen 9 5900HS CPU and 16 GB of RAM, and MATLAB R2024a software is running on the computer. Figure 6 Shows the time-consuming results of the two phase retrieval methods proposed in the present invention. Figure 6 Shows the variation of the normalized calculation time, which is defined as the ratio of the execution time of each multi-frame average phase retrieval method to the execution time of the traditional single-frame phase retrieval method.

[0067] Non-destructive testing embodiment; To verify the effectiveness of the proposed multi-frame average shear speckle interferometry method in practical applications, a non-destructive testing experiment was conducted on a metal-rubber bonded specimen. Circular debonding defects were artificially embedded in the sample. The surface of the specimen was thermally excited for 5 seconds using a 1kW halogen lamp. A shear speckle deformation measurement device was used to capture the resulting speckle images, and two multi-frame average phase retrieval methods were applied to the phase reconstruction stage. The initial phase reconstruction diagrams obtained by the multi-frame average pre-phase retrieval method at different numbers of frames are shown as A1 to A8 in Figure 7 ; The initial phase reconstruction diagrams obtained by the multi-frame average post-phase retrieval method at different numbers of frames are shown as B1 to B8 in Figure 7 .

[0068] For Figure 7 in each of the phase reconstruction diagrams, a single 3×3 sine-cosine mean filtering process was applied to obtain Figure 8 each of the filtered images in. Among them, Figure 7 A1 to A8 in Figure 8 obtained to , Figure 7 B1 to B8 in Figure 8 obtained to .

[0069] Similar to the deformation measurement embodiment, the denoising performance evaluation metrics ENL , SSI and SMPI of the two multi-frame average phase retrieval methods are shown as Figure 9 .

[0070] Next, the multi-frame average shear speckle interferometry method of the present invention will be verified in combination with the experimental results in the deformation measurement embodiment and the non-destructive testing embodiment.

[0071] As Figure 3 and Figure 7As shown, the experimental results indicate that there is an obvious inverse relationship between the number of frames and the noise level in the phase reconstruction diagram. Similarly, using the denoising performance metrics shown in Figure 5 and Figure 9 for quantitative analysis, it is confirmed that an increase in the number of frames leads to an enhancement of ENL , SSI and SMPI , and a decrease in the ENL and SSI and SMPI values. These results together verify the denoising effects of the two multi-frame average phase retrieval methods.

[0072] The enhancement of noise suppression is positively correlated with the number of frames used. However, due to hardware limitations, the maximum number of speckle patterns obtained by the present invention in the same phase shift state is limited to 8 frames. Despite this limitation, the trend lines from the left figures in Figure 5 and the left figures in Figure 9 show that a further increase in the number of frames will continuously improve the noise suppression performance.

[0073] For dynamic measurements, increasing the frame acquisition rate is crucial for ensuring the time resolution. Therefore, the optimal number of frames should be determined according to the specific time and spatial resolution requirements of the application.

[0074] In addition, Figure 4 and Figure 8 show that a single-pass mean filtering operation with a 3×3 window is sufficient to achieve high-quality phase retrieval. Traditional methods usually rely on multiple filtering stages or more computationally intensive denoising algorithms, which need to reduce the phase map noise by suppressing fine structural details. For example, in defect detection applications, small defects can be filtered out during the filtering process. The multi-frame average shearing speckle interferometry method proposed by the present invention only requires one mean filtering operation to achieve similar effects to multiple filtering and complex filtering algorithms, and can improve the image quality while largely retaining the original phase detail information. The filtered denoising performance metrics, as shown in the right figures in Figure 5 and the right figures in Figure 9 , further verify that the imaging quality improves with the increase in the number of frames.

[0075] The present invention obtains multi-frame phase shift diagrams through a multi-frame image acquisition device based on hardware synchronization control, and proposes two multi-frame average phase retrieval methods. As shown in Figure 5 and Figure 9 , Method 2 is superior to Method 1 in terms of performance metrics, indicating better imaging quality. In addition, the experiment also compared the calculation times of these two multi-frame average phase retrieval methods. As shown in Figure 6 , the calculation time required for phase retrieval increases with the increase in the number of frames. In addition, Figure 6It is also shown that the phase recovery time of Method 1 is slightly lower than that of Method 2. Therefore, the choice between these two methods should be guided by the specific application trade-off between processing time and recovery fidelity.

[0076] Although the multi-frame averaging shearography method proposed in the present invention can significantly improve the imaging quality, it requires the acquisition of multiple phase-shifted speckle patterns as multi-frame phase-shifted maps, which is several times that of traditional methods. This poses strict requirements on the system hardware, especially on the camera frame rate and the PZT response speed. For real-time or dynamic measurement application scenarios, it is imperative to deploy high-speed cameras and fast-response PZTs.

[0077] The present invention solves the noise problem in shearography phase recovery through a hardware-assisted multi-frame averaging phase recovery method. By synchronizing the PZT and the camera to obtain multi-frame phase-shifted maps, two new multi-frame averaging phase recovery methods can be applied. Method 1 calculates the phase difference by directly averaging the multi-frame phase-shifted maps, while Method 2 initially calculates the phase maps of each frame of the phase-shifted map independently and then averages them to calculate the phase difference. The experimental results show that Method 2 has excellent performance in ENL , SSI and SMPI aspects, although the calculation time increases slightly.

[0078] The integration of the multi-frame averaging phase recovery method and the single-pass 3×3 sine-cosine mean filtering processing effectively suppresses the phase noise while retaining the key phase details. This two-layer denoising strategy achieves results comparable to traditional multi-pass or complex filtering techniques, with a significantly reduced risk of detail loss.

[0079] However, the dependence of the method of the present invention on obtaining multiple high-fidelity frames introduces hardware limitations, requiring high-speed imaging and high-frequency synchronous driving to meet dynamic measurement scenarios. Despite these limitations, the proposed method provides a robust and high-quality imaging solution for shearography and is equally applicable to other interferometric methods. The method proposed in the present invention can also be applied to electronic speckle pattern interferometry, laser speckle interferometry, and laser holographic imaging technology.

[0080] In summary, the present invention advances the phase recovery in shearography by developing a multi-frame image acquisition device based on hardware synchronization control and two multi-frame averaging phase recovery methods, balancing the denoising effect, computational efficiency, and phase detail retention, providing a scalable and general solution for both static measurement scenarios and dynamic measurement scenarios.

[0081] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A multi-frame averaging shearography method, characterized in that, Applied to a shearography interferometry system for multi-frame averaging, the system includes an image acquisition device, a phase recovery device, and a single-shot 3×3 sine-cosine mean filtering device based on hardware synchronous control. The method includes: In the multi-frame image acquisition device based on hardware synchronous control, a multi-step staircase signal and a high-frequency square wave signal are generated by a high-frequency synchronous driver to synchronously control the camera and the PZT. The multi-step staircase signal is used as the PZT drive signal to control the PZT displacement, and the square wave signal is used as the camera trigger signal to trigger the camera to capture images through the rising edge. The rising edge or falling edge of each step of the staircase signal is strictly synchronized with the rising edge of the square wave, and multiple camera frame periods correspond to the duration of a single-step signal. Using the multi-frame multi-step time phase-shifting method, the PZT is made to introduce multiple discrete phase shifts. Before and after the deformation of the measurement object, the camera captures multiple frames of phase-shift diagrams in multiple phase-shift intensity modes at the same position of the PZT, obtaining multiple frames of phase-shift diagrams in the reference state and multiple frames of phase-shift diagrams in the deformed state. In the phase recovery device, the multi-frame average pre-phase recovery method or the multi-frame average post-phase recovery method is used to process the multiple frames of phase-shift diagrams in the reference state and the deformed state to obtain the phase reconstruction diagram. Among them, the multi-frame average pre-phase recovery method calculates the average value of the multiple frames of phase-shift diagrams in the two states before phase recovery to obtain the average phase-shift diagram, and then calculates the average phase difference image as the phase reconstruction diagram. The multi-frame average post-phase recovery method calculates the average value of the multiple phase differences of the multiple frames of phase-shift diagrams in the two states to obtain the final phase difference image as the phase reconstruction diagram. The single-shot 3×3 sine-cosine mean filtering device is used to perform single-shot 3×3 sine-cosine mean filtering on the phase reconstruction diagram to obtain the filtered image.

2. The method according to claim 1, characterized in that, The multi-frame average pre-phase recovery method specifically includes the following steps: Calculate the average value of the multiple frames of phase-shift diagrams generated in each phase-shift intensity mode in the reference state to obtain multiple average phase-shift diagrams in the reference state. Calculate the average value of the multiple frames of phase-shift diagrams generated in each phase-shift intensity mode in the deformed state to obtain multiple average phase-shift diagrams in the deformed state. Based on the multiple average phase-shift diagrams in the reference state and the multiple average phase-shift diagrams in the deformed state, calculate the average phase difference image caused by the displacement change of the measurement object.

3. The method according to claim 1, wherein The multi-frame average post-phase recovery method specifically includes the following steps: Calculate multiple phase images of the multiple frames of phase-shift diagrams generated in multiple phase-shift intensity modes in the reference state to obtain multiple phase images in the reference state. Calculate multiple phase images of the multiple frames of phase-shift diagrams generated in multiple phase-shift intensity modes in the deformed state to obtain multiple phase images in the deformed state. Based on the multiple phase images in the reference state and the multiple phase images in the deformed state, calculate multiple phase difference images generated in multiple phase-shift intensity modes. Calculate the average value of the multiple phase difference images to calculate the final phase difference image.

4. The method according to claim 2, characterized in that, When the multi-frame multi-step time phase-shifting method is specifically the multi-frame four-step phase-shifting method, the multi-frame average pre-phase recovery method specifically includes the following steps: Calculate the average values of multiple frames of phase-shifted images generated in four phase-shift intensity modes under the reference state respectively to obtain four average phase-shifted images under the reference state , and the calculation formula is as follows: ， Among them, k represents the number of frames, n represents the maximum number of frames, represents multiple phase-shifted diagrams generated under four phase-shift intensity modes in the reference state; Calculate the average values of multiple frames of phase-shifted images generated in four phase-shift intensity patterns under the deformed state respectively to obtain four average phase-shifted images under the deformed state , and the calculation formula is as follows: ， Among them, represents multiple frame phase shift diagrams generated in four phase shift intensity modes in the deformed state; Four average phase-shift maps under the reference state , and four average phase-shift maps under the deformed state , calculate the average phase difference image caused by the displacement change of the measurement object , and the calculation formula is: ， Among them, represents the pixel coordinates and is the average displacement in the vertical direction. represents the pixel coordinates and is the average displacement in the horizontal direction.

5. The method according to claim 3, wherein When the multi-frame multi-step time phase-shifting method is specifically the multi-frame four-step phase-shifting method, the multi-frame average post-phase recovery method specifically includes: Calculate the multi-frame phase images of the multi-frame phase-shifted images generated in four phase-shift intensity modes in the reference state to obtain the multi-frame phase images in the reference state , and the calculation formula is as follows: ， Among them, k represents the number of frames, n represents the maximum number of frames, represents multiple phase-shifted images generated in four phase-shift intensity modes at the reference state; Calculating multiple phase images of multiple phase-shifted images generated in four phase-shift intensity patterns in the deformed state to obtain multiple phase images in the deformed state , and the calculation formula is: ， Among them, represents multiple frame phase shift diagrams generated under four phase shift intensity modes in the deformed state; Multi-frame phase images in the reference state , multi-frame phase images in the deformed state , and calculate multi-frame phase difference images generated in four phase shift intensity modes ; For multiple-frame phase difference images Obtain the average value and calculate the final phase difference image , and the calculation formula is as follows: 。 6. The method according to claim 5, wherein The multi-frame phase difference image The first calculation formula is as follows: 。 7. The method according to claim 5, characterized in that, The second calculation formula of the multi-frame phase difference image is as follows: ， Among them, represents the multi-frame average displacement in the vertical direction, represents the multi-frame average displacement in the horizontal direction, represents the multi-frame phase-shifted images generated in four phase-shift intensity modes in the reference state, represents the multi-frame phase-shifted images generated in four phase-shift intensity modes in the deformed state, represents the pixel coordinates.

8. The method according to claim 7, wherein After calculating the final phase difference image It further includes: For calculating the final phase difference image Perform a simplification operation on the formula, and the calculation formula after the simplification operation is: ， Among them, represents a multi-frame simplified phase difference image.

9. The method according to claim 1, wherein After obtaining the phase reconstruction image, it further includes: Using normalization time to evaluate the calculation efficiency of the multi-frame average pre-phase recovery method and the multi-frame average post-phase recovery method.

10. The method according to claim 1, wherein After obtaining the filtered image, it further includes: Adopt ENL, SSI , SMPI to evaluate the denoising performance of the filtered image.

Images