Rapid terahertz time domain scanning imaging system and image reconstruction method

Abstract

The invention discloses a rapid terahertz time domain scanning imaging system and an image reconstruction method, and relates to the technical field of terahertz imaging. The system comprises a femtosecond pulse generation device used for outputting periodic ultrashort femtosecond laser pulses; the single-mode optical fiber transmission device is used for transmitting the periodic ultrashort femtosecond laser pulses generated by the femtosecond pulse generation device and converting the periodic ultrashort femtosecond laser pulses into two paths of femtosecond laser pulses; the terahertz wave generation device is used for generating terahertz waves under the action of one path of femtosecond laser pulse; the two-dimensional scanning device is used for carrying a sample and performing two-dimensional scanning on the sample by using generated terahertz waves; the terahertz wave receiving device is used for detecting time domain electric field information of terahertz waves generated after two-dimensional scanning of the sample; and the data acquisition device is used for performing signal acquisition and processing on the terahertz time-domain electric field information of the sample to obtain image information of the sample. The imaging time can be shortened, and the imaging definition is high.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of terahertz imaging, and in particular to a fast terahertz time-domain scanning imaging system and an image reconstruction method. BACKGROUND

[0002] The fast terahertz scanning imaging technology has great potential in the fields of biomedical diagnosis, industrial nondestructive testing, security inspection and security protection due to its characteristics of nondestructiveness, penetrability and fingerprinting. However, in the process of pursuing fast scanning imaging to improve real-time performance, the system faces multiple technical challenges: first, the image signal-to-noise ratio is reduced due to the thermal noise of the detector, environmental interference and the reduction of the terahertz light spot, and part of the details is submerged in the noise; second, the vibration of the mechanical scanning platform and the fast scanning cause motion artifacts and stripe distortion, affecting the geometric fidelity of the image; third, the integral time between the terahertz light spot and the sample under high-speed scanning aggravates signal attenuation and motion blur, making it difficult to identify weak contrast targets. Therefore, it is necessary to combine adaptive image processing methods to effectively improve the accuracy of fast terahertz scanning images. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a fast terahertz time-domain scanning imaging system and an image reconstruction method capable of shortening the imaging time and having high imaging clarity.

[0004] To solve the above technical problems, the technical solution adopted by the present application is: a fast terahertz time-domain scanning imaging system, comprising:

[0005] A femtosecond pulse generating device for outputting periodic ultrashort femtosecond laser pulses;

[0006] A single-mode optical fiber transmission device for receiving the periodic ultrashort femtosecond laser pulses generated by the femtosecond pulse generating device and converting them into double-path femtosecond laser pulses;

[0007] A terahertz wave generating device for generating terahertz waves under the action of one path of femtosecond laser pulses;

[0008] A two-dimensional scanning device for carrying a sample and using the generated terahertz waves to perform two-dimensional scanning on the sample;

[0009] A terahertz wave receiving device for detecting the time-domain electric field information of the terahertz waves generated after two-dimensional scanning of the sample;

[0010] A data acquisition device for signal acquisition and processing of the terahertz time-domain electric field information of the sample to obtain image information of the sample.

[0011] The present application also discloses an image reconstruction method of a fast terahertz time-domain scanning imaging system, which uses the scanning imaging system and comprises the following steps:

[0012] Based on the results of the bow-shaped two-dimensional scanning process of the scanning sample by the fast terahertz time-domain scanning imaging system, the terahertz wave intensity is fixed at the time-domain maximum value using an optical fiber delay line, and a two-dimensional data matrix of the terahertz time-domain maximum intensity change of the scanning sample is obtained;

[0013] Performing denoising and image enhancement processing on the two-dimensional matrix data to obtain a high signal-to-noise ratio image of the sample;

[0014] Based on the two-dimensional space adaptive local variance optimized Wiener filtering deconvolution method, the two-dimensional matrix of the denoised high signal-to-noise ratio image is adaptively deconvolution filtered to obtain a high-quality restored image of the sample.

[0015] The beneficial effects of adopting the above technical solution are as follows: The system described in this application uses bow-shaped two-dimensional rapid scanning in conjunction with a data acquisition card to collect terahertz time-domain maximum data, greatly shortening the imaging time previously required in terahertz point-by-point scanning imaging to minutes. The method described in this application proposes a two-dimensional spatial adaptive local variance-optimized Wiener filter deconvolution algorithm to reconstruct image details, improve the clarity of rapidly scanned images, and adaptively balance the local noise-to-signal ratio, proposing new approaches to image processing for terahertz rapid scanning imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] Figure 1 is a principle block diagram of the system according to an embodiment of the present invention;

[0018] Figure 2 is a main flow chart of the method according to an embodiment of the present invention;

[0019] Among them: 1. Femtosecond pulse generator; 2. Single-mode optical fiber transmission device; 3. Optical fiber delay line; 4. Terahertz wave generator; 5. Two-dimensional translation stage; 6. Current amplifier; 7. Bias voltage source; 8. Terahertz wave detection device; 9. Optical fiber coupler. DETAILED DESCRIPTION

[0020] The following is a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts are within the scope of protection of the present invention.

[0021] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this disclosure. Therefore, the present application is not limited to the details described herein and can be practiced with variations that are within the scope and spirit of the present application.

[0022] Generally, as shown in the accompanying drawings, Figure 1 Embodiments of the present application disclose a fast terahertz time-domain scanning imaging system, comprising:

[0023] A femtosecond pulse generating device 1 is configured to output periodic ultrashort femtosecond laser pulses.

[0024] A single-mode fiber transmission device 2 is configured to receive the periodic ultrashort femtosecond laser pulses generated by the femtosecond pulse generating device and convert the periodic ultrashort femtosecond laser pulses into double-path femtosecond laser pulses.

[0025] A terahertz wave generating device 4 is configured to generate terahertz waves under the action of one of the double-path femtosecond laser pulses.

[0026] A two-dimensional scanning device is configured to carry a sample and perform two-dimensional scanning on the sample using the generated terahertz waves.

[0027] A terahertz wave receiving device 8 is configured to detect time-domain electric field information of the terahertz waves generated after the two-dimensional scanning on the sample.

[0028] A data acquisition device is configured to perform signal acquisition and processing on the time-domain electric field information of the sample to obtain image information of the sample.

[0029] The modules in the system will be described in detail in combination with specific contents as follows:

[0030] The femtosecond pulse generating device 1 adopts an erbium-doped optical fiber as a gain medium, excites erbium ions to an excited state by a 980 nm laser diode, forms population inversion, provides a basis for stimulated radiation, relies on a passive mode-locking technology, and makes the phases of different longitudinal modes in a laser cavity synchronous through a mode-locking mechanism, superimposes to form an ultrashort pulse sequence, and compresses the pulse width to femtosecond level through self-phase modulation.

[0031] The single-mode fiber transmission device 2 mainly comprises a fiber coupler and a single-mode transmission optical fiber. The fiber coupler couples the ultrashort laser pulses emitted by the femtosecond laser device into the single-mode transmission optical fiber, and the coupling efficiency is ≥ 90%; the single-mode transmission optical fiber is composed of multiple optical fibers, and the two beams of femtosecond pulse lasers output by the femtosecond pulse generating device are respectively transmitted to the terahertz photoconductive antenna end of the terahertz wave generating device 4 and the optical fiber delay line end of the terahertz wave receiving device 8 through the transmission optical fibers.

[0032] The terahertz wave generating device 4 mainly comprises photoconductive material (low-temperature gallium arsenide) and a bias voltage source 7. The photoconductive material is excited by super-short pulse laser irradiation to generate photo-carriers, and the bias voltage applied by the bias voltage source 7 makes the carriers move rapidly and form an electric field, thereby generating terahertz waves.

[0033] The two-dimensional scanning device mainly comprises a two-dimensional translation stage 5 and a fixed support. The sample is placed in the terahertz wave converging area, and the fixed support is finely adjusted to fix the sample for two-dimensional rapid non-stop arc scanning and data acquisition.

[0034] The terahertz wave detecting device 8 mainly comprises a focusing lens, an optical fiber delay line 3 and a terahertz receiving antenna. Another laser beam at the output end of the femtosecond laser is transmitted to the optical fiber delay line 3 through an optical fiber, the time delay of the reference light and the terahertz pulse is changed to fix the terahertz intensity at the maximum value in the time domain, the sampling process of the sample after the terahertz pulse is completed, and the light beam carrying the sample information is converged to the terahertz receiving antenna through the focusing lens.

[0035] The data acquisition device mainly comprises a current amplifier and a data acquisition card. The terahertz time-domain maximum value data collected by the terahertz receiving antenna 897 is input into the current amplifier to amplify the signal strength, so that the subsequent data analysis can effectively analyze the intensity data. The amplified data is temporarily stored in the data acquisition card buffer area and then stored in the computer, the sample data acquisition process of rapid terahertz time-domain scanning imaging is completed, and rapid terahertz time-domain scanning imaging is realized.

[0036] Correspondingly, as shown in Figure 2 The application also discloses an image reconstruction method of the rapid terahertz time-domain scanning imaging system, the method uses the scanning imaging system, and the method comprises the following steps:

[0037] Based on the result of the arc two-dimensional scanning processing of the scanning sample by the rapid terahertz time-domain scanning imaging system, the terahertz wave intensity is fixed at the maximum value in the time domain by using the optical fiber delay line 3, and a two-dimensional data matrix of the terahertz time-domain maximum value intensity change of the scanning sample is obtained.

[0038] The two-dimensional matrix data is subjected to denoising and image enhancement processing, and a high signal-to-noise ratio image of the sample is obtained.

[0039] Based on the two-dimensional spatial adaptive local variance optimization Wiener filter deconvolution method, the two-dimensional matrix of the high signal-to-noise ratio image after denoising is subjected to adaptive deconvolution filtering, and a high-quality recovery image of the sample is obtained.

[0040] In the method, the arch-shaped two-dimensional scanning of the sample is processed based on a fast terahertz time-domain scanning imaging system, the terahertz wave intensity is fixed at a time-domain maximum value by using a fiber delay line, and a two-dimensional data matrix of the terahertz time-domain maximum intensity variation of the scanned sample is obtained. The two-dimensional matrix data is denoised and image-enhanced by using a two-dimensional discrete wavelet transform and other processing methods, a high signal-to-noise ratio image is obtained, the denoised two-dimensional matrix is adaptively deconvoluted and filtered based on a two-dimensional spatial adaptive local variance optimization Wiener filter deconvolution method, artifacts and blur are removed, local details are improved, and a high-quality recovered image is obtained.

[0041] In view of problems such as inaccurate local information splicing, noise and excessively low contrast of the original reconstructed image, the two-dimensional matrix data is denoised, local information is aligned and image-enhanced by using a two-dimensional discrete wavelet transform, histogram enhancement, fog removal processing, morphological processing and other processing methods, and a high signal-to-noise ratio image is obtained. The image processing part parameter calculation formula is as follows:

[0042] First, the two-dimensional matrix is normalized, and the two-dimensional data matrix is decomposed by using a two-dimensional discrete wavelet transform:

[0043]

[0044] Wherein ψ is a wavelet base function, j is a decomposition layer number, W ψ [f](j,m,n) is a wavelet coefficient matrix, and (m,n) is a spatial position coordinate.

[0045] The sym6 wavelet base is selected, the denoising improves the symmetry and reduces the boundary effect, the local information is aligned, and the high-frequency noise is filtered out. Then, the histogram enhancement is performed to adjust the pixel intensity distribution and expand the dynamic range to enhance the contrast. The cumulative distribution function mapping is as follows:

[0046]

[0047] Wherein, n i is the number of pixels of the gray level i, N is the total number of pixels, and L is the number of gray levels. The function stretches the intensity distribution and enhances the visibility of details. Then, the fog removal processing is performed:

[0048]

[0049] Wherein, w k is a scale weight, k is a scale number, and I(x,y) is an original image intensity, Gaussian kernel function; the reflection component is extracted in the logarithmic domain difference after multi-scale Gaussian filtering of the RGB three channels, and the channels are combined and linearly stretched to the display range to improve the image clarity and contrast. Then, the target image is subjected to morphological processing, erosion and expansion, and open operation to effectively remove the noise points in the image, and finally, the closed operation fills the holes in the image to obtain a target image with high signal-to-noise ratio and contrast for two-dimensional spatial adaptive local variance optimization Wiener filter deconvolution processing.

[0050] Before two-dimensional spatial adaptive local variance optimization Wiener filter deconvolution processing, a terahertz scanning image degradation model needs to be established first, and the two-dimensional PSF (point spread function) is modeled around the terahertz spot size and the corresponding scanning parameters in the fast terahertz time-domain scanning imaging system. Then, the two-dimensional PSF is Fourier transformed and brought into the Wiener filter transfer function, and the local noise-signal ratio mapping is adaptively improved. The local noise variance and energy estimate are calculated in the sliding window, and finally, the results are fused. Among them:

[0051] The terahertz degradation model can be modeled as:

[0052] g(x, y) = f(x, y) * h(x, y) + n(x, y) (4)

[0053] Where f(x, y) is the original image, h(x, y) is the two-dimensional point spread function, which describes the blurring characteristics of the system, n(x, y) is the additive noise, and g(x, y) is the blurred image.

[0054] The two-dimensional PSF (point spread function) of the fast terahertz time-domain scanning imaging system can be modeled as the blurring effect in the horizontal scanning direction and the vertical stepping direction. The horizontal PSF is determined by the spot size and can be modeled as:

[0055]

[0056] The vertical PSF can be modeled as:

[0057]

[0058] Where D is the diameter of the terahertz spot, y blur is the blurring diffusion width. Therefore, the synthesized two-dimensional PSF is:

[0059] h(x, y) = h x (x) · h y (y) (9)

[0060] The transfer function of Wiener filter in frequency domain is:

[0061]

[0062] where H(u, v) is the Fourier transform of the two-dimensional PSF, H * (u, v) is the complex conjugate matrix of H(u, v), is the global noise-signal ratio.

[0063] The local NSR mapping is adaptively improved, and the local noise variance is calculated within the sliding window W:

[0064]

[0065] where, N w is the total number of pixels in the window. Then the local signal energy estimation is performed:

[0066]

[0067] where,

[0068] Therefore, the local NSR mapping is:

[0069]

[0070] Therefore, the final adaptive Wiener filter frequency domain transfer function is:

[0071]

[0072] The image is divided into overlapping sub-blocks B K , and the overlapping area is the width of the PSF. The NSR(x, y) of each pixel in the sub-block is calculated and filtered in the frequency domain:

[0073] F k (u, v) = G k (u, v) · H adaptive (u, v, x, y) (15)

[0074] The inverse Fourier transform is obtained f k (x, y), and finally the fusion result is obtained. The Hamming window weighted average is used for the overlapping area:

[0075]

[0076] where w k (x, y) is the Hamming window function, with high center weight and low edge weight.

[0077] The present application aims at the image quality problem generated under the condition of fast terahertz time-domain scanning imaging, proposes a complete image processing flow, and proposes a two-dimensional space adaptive local variance optimization Wiener filter deconvolution algorithm, which can define a sliding window to calculate the local noise variance according to the characteristics of the input image matrix data, realize adaptive deconvolution image quality improvement, and propose a new idea for fast terahertz scanning image reconstruction.

Claims

1. A fast terahertz time-domain scanning imaging system, characterized in that include: A femtosecond pulse generating device (1) for outputting periodic ultrashort femtosecond laser pulses; A single-mode optical fiber transmission device (2) is used to transmit the periodic ultrashort femtosecond laser pulses generated by the femtosecond pulse generating device and convert them into dual-path femtosecond laser pulses; A terahertz wave generating device (4) is used to generate terahertz waves under the action of a femtosecond laser pulse; A two-dimensional scanning device, used for carrying a sample and performing two-dimensional scanning on the sample using the generated terahertz wave; A terahertz wave receiving device (8) is used to detect the time-domain electric field information of the terahertz wave generated after two-dimensional scanning of the sample; The data acquisition device is used to collect and process the terahertz time-domain electric field information of the sample to obtain image information of the sample.

2. The fast terahertz time-domain scanning imaging system according to claim 1, characterized in that: The femtosecond pulse generating device (1) comprises a femtosecond laser, uses erbium-doped optical fiber as a gain medium, excites erbium ions to an excited state through a 980nm laser diode, forms a population inversion, and utilizes a mode-locking mechanism to generate an ultrashort pulse sequence output at the femtosecond level.

3. The fast terahertz time-domain scanning imaging system according to claim 1, characterized in that: The single-mode optical fiber transmission device (2) comprises an optical fiber coupler (9) and a single-mode transmission optical fiber. The optical fiber coupler is used to couple the ultrashort femtosecond laser pulses emitted by the femtosecond pulse generating device into the single-mode transmission optical fiber. The single-mode transmission optical fiber comprises a plurality of single-mode optical fibers, which are used to transmit two beams of femtosecond pulse lasers output by the femtosecond pulse generating device (1) to the terahertz photoconductive antenna end of the terahertz wave generating device (4) and the optical fiber delay line end of the terahertz wave receiving device (8), respectively.

4. The fast terahertz time-domain scanning imaging system according to claim 1, characterized in that: The terahertz wave generating device (4) comprises a photoconductive material and a bias voltage source (7). The photoconductive material excites photogenerated carriers under ultrashort pulse laser irradiation, and the bias voltage applied by the bias voltage source (7) causes the carriers to move rapidly and form an electric field, thereby generating terahertz waves.

5. The fast terahertz time-domain scanning imaging system according to claim 1, characterized in that: The two-dimensional scanning device comprises a two-dimensional translation stage (5) and a fixed bracket. The two-dimensional translation stage (5) is located in a terahertz wave convergence area and its position is finely adjusted by the fixed bracket to perform two-dimensional rapid non-stop bow scanning on a fixed sample, while completing data acquisition.

6. The fast terahertz time-domain scanning imaging system according to claim 1, characterized in that: The terahertz wave detection device (8) comprises a focusing lens, an optical fiber delay line (3) and a terahertz receiving antenna. Another laser beam output by a femtosecond pulse generating device is transmitted to the optical fiber delay line (3) via a single-mode optical fiber transmission device (2). The terahertz intensity is fixed at a time domain maximum value by changing the time delay between the reference light and the terahertz pulse, completing the sampling process of the sample to be tested after the terahertz pulse, and converging the light beam carrying the sample information to the terahertz receiving antenna via the focusing lens.

7. The fast terahertz time-domain scanning imaging system according to claim 1, characterized in that: The data acquisition device comprises a current amplifier (6) and a data acquisition card, which connects terahertz intensity data collected by a terahertz receiving antenna in a terahertz wave detection device (8) to the current amplifier to amplify the signal intensity, temporarily stores the amplified data in a buffer of the data acquisition card, and then transfers it to a computer, thereby completing the sample data acquisition process of fast terahertz time-domain scanning imaging and realizing fast terahertz time-domain scanning imaging.

8. An image reconstruction method for a fast terahertz time-domain scanning imaging system, the method using the scanning imaging system according to any one of claims 1 to 7, characterized in that: The method comprises the following steps: Based on the result of the fast terahertz time-domain scanning imaging system performing a two-dimensional bow scanning process on the scanning sample, the terahertz wave intensity is fixed at the time-domain maximum value using an optical fiber delay line (3), and a two-dimensional data matrix of the terahertz time-domain maximum intensity change of the scanning sample is obtained; Performing denoising and image enhancement processing on the two-dimensional matrix data to obtain a high signal-to-noise ratio image of the sample; Based on the two-dimensional space adaptive local variance optimized Wiener filtering deconvolution method, the two-dimensional matrix of the denoised high signal-to-noise ratio image is adaptively deconvolution filtered to obtain a high-quality restored image of the sample.

9. The image reconstruction method of a fast terahertz time-domain scanning imaging system according to claim 8, characterized in that: The method for obtaining a high signal-to-noise ratio image comprises the following steps: After normalizing the two-dimensional data matrix, use the two-dimensional discrete wavelet transform to perform soft threshold decomposition on the two-dimensional data matrix: Among them, ψ is the wavelet basis function, j is the number of decomposition layers, W ψ [f](j,m,n) is the wavelet coefficient matrix, (m,n) is the spatial position coordinate; Histogram enhancement is performed to adjust the pixel intensity distribution, expand the dynamic range to enhance the contrast, and its cumulative distribution function mapping is: Among them, n i is the number of pixels at gray level i, N is the total number of pixels, L is the number of gray levels, and then defogging is performed: Among them, w k is the weight of each scale, k is the number of scales, I(x,y) is the original image intensity, The Gaussian kernel function is used. After multi-scale Gaussian filtering is performed on the R, G, and B channels, the reflection component is extracted by difference in the logarithmic domain. After merging the channels, it is linearly stretched to the display range to improve the image clarity and contrast. Then, the target image is morphologically processed. After corrosion and dilation, an opening operation is applied to remove noise in the image. Finally, a closing operation is performed to fill the holes in the image to obtain a target image with high signal-to-noise ratio and contrast. The deconvolution processing is then performed on the two-dimensional spatial adaptive local variance optimized Wiener filter.

10. The image reconstruction method of a fast terahertz time-domain scanning imaging system according to claim 8, characterized in that: Before deconvolution processing using a two-dimensional adaptive local variance-optimized Wiener filter, a degradation model for terahertz scanning images is first established. A two-dimensional point spread function (PSF) is then modeled based on the terahertz spot size and corresponding scanning parameters in a fast terahertz time-domain scanning imaging system. The two-dimensional PSF is then Fourier transformed and incorporated into the Wiener filter transfer function. The local noise-to-signal ratio mapping is adaptively improved, and the local noise variance and energy estimate are calculated within a sliding window to obtain a fusion result. The terahertz degradation model is modeled as: g(x,y)=f(x,y)*h(x,y)+n(x,y) (4) Where f(x,y) is the original image, h(x,y) is the two-dimensional point spread function that describes the blur characteristics of the system, n(x,y) is the additive noise, and g(x,y) is the blurred image; The two-dimensional point spread function (PSF) is modeled as the blurring effect in the transverse scanning direction and the longitudinal stepping direction. The transverse PSF is determined by the spot size and is modeled as: The longitudinal PSF is modeled as: Where, is the diameter of the terahertz spot, y blur For the fuzzy diffusion width, synthesize the two-dimensional PSF: h(x,y)=h x (x)·h y (y) (9) The transfer function of Wiener filtering in the frequency domain is: Where H(u,v) is the Fourier transform of the two-dimensional PSF, H * (u,v) is the complex conjugate matrix of H(u,v), is the global noise-to-signal ratio; Adaptively improve the local NSR map and calculate the local noise variance within the sliding window W: in, N w is the total number of pixels in the window, and then the local signal energy is estimated: in, Therefore, the local NSR mapping is: The final adaptive Wiener filter frequency domain transfer function is: Divide the image into overlapping sub-blocks B K , the overlapping area is the PSF width, and the NSR(x,y) of each pixel in the sub-block is calculated after frequency domain filtering: F k (u,v)=G k (u,v)·H adaptive (u,v,x,y) (15) The inverse Fourier transform gives f k (x,y), and finally fuse the results, using the Hamming window weighted average overlapping area: Among them, w k (x,y) is the Hamming window function, with high center weight and low edge weight.

Images