Compression ultrafast imaging device based on time domain discrete illumination

Abstract

The invention discloses a compression ultrafast imaging device based on time domain discrete illumination. The device is composed of a time domain discrete illumination compression ultrafast imaging system, a time sequence control system and an image reconstruction system. Time domain discrete illumination is provided for a to-be-measured scene through a pulse sequence generator, dozens of time slices of the to-be-measured scene are uniformly intercepted, spatial random coding is carried out, then the time slices are transmitted into a streak camera for space-time deflection and superposition imaging, and finally the to-be-measured scene is restored by adopting an image reconstruction algorithm (TV-CD) combining a total variation noise reduction model and a cascade noise reducer. The invention relates to a time domain discrete illumination compression ultrafast imaging technology (DI-CUP) for short. According to the invention, not only can single multi-image ultrafast imaging be realized, but also the data compression ratio can be reduced, the spatial resolution can be improved, and the observation time window is flexible and adjustable. The above advantages provide an accurate and effective measurement tool for non-repeatable ultrafast processes such as light pulse propagation and plasma evolution, and are of great significance to the development of related research and technology of natural science.

Description

Technical Field

[0001] The present invention relates to the fields of ultrafast optics, compressed sensing technology and computational imaging, and in particular to a compressed ultrafast imaging device based on time-domain discrete illumination. Background Art

[0002] Accurately capturing ultrafast phenomena is crucial to fundamental physics, biology, chemistry, and technological development. However, commonly used commercial cameras are limited by the electronic storage and readout speed of complementary metal oxide semiconductors (CMOS) or charge coupled devices (CCDs), which can only reach a speed of 10 9 fps, it is difficult to meet the detection needs of ultrafast phenomena. Especially for non-repeatable or irreversible ultrafast dynamic processes such as optical anomalous waves, solitons, shock waves, radiation molecular transitions, complex light pulses, chemical reactions, etc. Another widely used measurement method is pump-probe, which requires that ultrafast phenomena must be reproducible, and the scenes at different times are obtained by adjusting the time delay multiple times. However, when ultrafast phenomena cannot be accurately repeated or cannot be repeated at all, pump-probe technology is no longer applicable. Therefore, single measurement becomes the key to solving such problems. Compressed ultrafast imaging technology can not only achieve single measurement by combining compressed sensing (CS) and streak camera (SC), but also meet the imaging frame rate, sequence depth and pixel requirements at the same time. The ultrafast space-time conversion of the streak camera makes the imaging speed reach the fs level. By combining with CS, encoding is added during the imaging process to reconstruct the captured two-dimensional image, thereby achieving ultrafast imaging, so that the entire two-dimensional dynamic scene can be restored with a single exposure. In recent years, CUP has played an important role in the measurement of ultrafast scenes such as spot motion, laser light field, fluorescence lifetime, chirped pulse, linear and nonlinear optical effects, laser-induced ultrafast dynamics, etc. With the development of CUP technology, its application field has been continuously broadened, and the measurement dimension has been continuously broken through, from the initial three-dimensional (two-dimensional space, one-dimensional time) to four-dimensional (two-dimensional space, one-dimensional time, one-dimensional spectrum), and even five-dimensional (three-dimensional space, one-dimensional time, one-dimensional spectrum or two-dimensional space, one-dimensional time, one-dimensional spectrum, one-dimensional polarization) and more dimensions. This means that a single two-dimensional image taken by a streak camera is coupled with more dimensional information, the data compression ratio is higher, the reconstruction difficulty increases accordingly, and the image reconstruction quality is difficult to meet higher application requirements. In most applications, it is necessary to restore dozens or even hundreds of images from a two-dimensional image, and the compression ratio is extremely high, which exceeds the limit of the CS reconstruction algorithm. Therefore, reducing the image compression ratio is an effective way to improve image quality and is also the key to the further development of CUP technology. Summary of the invention

[0003] The purpose of the present invention is to provide a compressed ultrafast imaging device based on discrete illumination in time domain in view of the deficiencies of the prior art. The present invention is composed of a discrete illumination compressed ultrafast imaging system, a timing control system and an image reconstruction system. The present invention provides discrete illumination in time domain for the scene to be measured through a pulse sequence generator, uniformly intercepts dozens of time slices of the scene to be measured, and transmits them to the streak camera for spatiotemporal deflection and superposition imaging after spatial random encoding. Finally, the image reconstruction algorithm (TV-CD) combining the total variation denoising model and the cascade denoiser is used to restore the scene to be measured, which is referred to as discrete illumination compressed ultrafast imaging technology in time domain (DI-CUP). The present invention can not only realize single-shot multi-frame ultrafast imaging, but also reduce the data compression ratio, improve the spatial resolution, and the observation time window is flexible and adjustable. The above advantages provide an accurate and effective measurement tool for non-repeatable ultrafast processes such as light pulse propagation and plasma evolution, which is of great significance to the development of relevant research and technology in natural sciences.

[0004] The specific technical solution for achieving the purpose of the present invention is:

[0005] A compressed ultrafast imaging device based on time-domain discrete illumination, comprising a time-domain discrete illumination compressed ultrafast imaging system, a timing control system and an image reconstruction system;

[0006] The time-domain discrete illumination compressed ultrafast imaging system is composed of a femtosecond laser, a pulse sequence generator, a scene to be measured, an objective lens, a convex lens, a transmissive spatial encoder, a lens pair, a bandpass filter and a streak camera; the femtosecond laser, the pulse sequence generator, the scene to be measured, the objective lens, the convex lens, the transmissive spatial encoder, the lens pair, the bandpass filter and the streak camera are optically connected in sequence;

[0007] Wherein, the pulse sequence generator includes a first pulse sequence generator and a second pulse sequence generator, the two pulse sequence generators are independent of each other, and one of the pulse sequence generators is selected according to the actual observation time window; wherein, the first pulse sequence generator is composed of an ultrafast reflector, a first BBO frequency doubling crystal, and a dichroic mirror, and the ultrafast reflector, the first BBO frequency doubling crystal, and the dichroic mirror are optically connected in sequence, wherein the ultrafast reflector is connected to a femtosecond laser, and the dichroic mirror is connected to a scene to be measured; the second pulse sequence generator is composed of a second BBO frequency doubling crystal, a linear polarizer, a first beam splitter, a first reflector, a second reflector, a second beam splitter, a third reflector, a fourth reflector, a third beam splitter, a fifth reflector, a sixth reflector, a fourth beam splitter, a seventh reflector, an eighth reflector, a half-wave plate, and a polarization beam splitter, wherein the second BBO frequency doubling crystal is connected to the linear polarizer, and the linear polarizer The oscillator and the first beam splitter are connected in an optical path. The first beam splitter divides the optical path into two paths, one of which is connected to the first reflector and the second reflector, and the other is connected to the second beam splitter, and the second reflector is connected to the second beam splitter. The second beam splitter divides the optical path into two paths again, one of the two paths is connected to the third reflector and the fourth reflector, and the other is connected to the third beam splitter, and the fourth reflector is connected to the third beam splitter. The third beam splitter divides the optical path into two paths for the third time, one of the two paths is connected to the fifth reflector and the sixth reflector, and the other is connected to the fourth beam splitter, and the sixth reflector is connected to the fourth beam splitter. The fourth beam splitter divides the optical path into two paths for the fourth time, one of the two paths is connected to the seventh reflector, the eighth reflector and the half-wave plate, and the other is connected to the polarization beam splitter, and the half-wave plate is connected to the polarization beam splitter. The second BBO frequency doubling crystal is connected to the femtosecond laser, and the polarization beam splitter is connected to the scene to be measured;

[0008] The timing control system is composed of a digital delay generator;

[0009] The image reconstruction system is composed of a computer;

[0010] The digital delay generator of the timing control system is respectively connected to the femtosecond laser and the streak camera of the time-domain discrete illumination compression ultrafast imaging system;

[0011] The computer of the image reconstruction system is respectively connected with the transmission type space encoder and the streak camera of the time domain discrete illumination compression ultrafast imaging system.

[0012] The computer of the image reconstruction system is equipped with a reconstruction algorithm TV-CD that combines a total variation denoising model with a cascade denoiser to reconstruct the compressed image collected by the streak camera of the time-domain discrete illumination compression ultrafast imaging system. The image reconstruction process is as follows:

[0013] The scenario to be tested is represented as I ( x, y, t )( x, yare the horizontal and vertical coordinates of space, t is time), in the time-domain discrete illumination compression ultrafast imaging system, it is first spatially encoded by the transmission spatial encoder, and then enters the streak camera after being intensity modulated. It is vertically deflected under the action of the deflection voltage, and finally the deflected image is superimposed in time and space and collected by the streak camera. The two-dimensional image collected by the streak camera is expressed as E ( x', y' )( x', y' are the spatial horizontal and vertical coordinates of the image collected by the streak camera, respectively). Spatial encoding, time deflection, and spatiotemporal superposition are represented by C, T, and M operators, respectively. Then the imaging process of the time-domain discrete illumination compression ultrafast imaging system is expressed as:

[0014] , (1)

[0015] In order to restore 3D information from compressed images I ( x, y, t ), the inverse problem of equation (1) needs to be solved. The TV-CD algorithm uses the plug-and-play (PnP) framework and the alternating direction multiplier method (ADMM) to decompose the inverse problem into three sub-problems. When solving the denoising sub-problem, four denoisers, TV, FFDNet, DRUNet and FastDVDNet, are used for cascade denoising in the order of [TV, FFDNet, DRUNet, FastDVDNet, TV]. Each algorithm is iterated independently, and after completing the iteration of one algorithm, it enters the next algorithm. After all iterations are completed, we can get I ( x, y, t ).

[0016] The beneficial effects of the present invention are as follows: on the basis of compressed ultrafast imaging technology, the present invention introduces a pulse sequence generator to provide time-domain discrete illumination for the scene to be measured, which is referred to as time-domain discrete illumination compressed ultrafast imaging technology. Through time-domain discrete illumination, dozens of time slices are uniformly intercepted during the occurrence of dynamic events, and after spatial random encoding, they are transmitted to the streak camera for time-space deflection and superposition imaging. Finally, an image reconstruction algorithm combining a total variation noise reduction model and a cascade noise reducer is used to restore the event evolution images at different times. The present invention can not only realize single-shot multi-frame ultrafast imaging, effectively avoid the errors caused by pump-probe methods, but also reduce the data compression ratio, improve the spatial resolution, and flexibly select different pulse sequence generation methods to adjust the observation time window. The above advantages provide an accurate and effective measurement tool for non-repeatable ultrafast dynamic processes such as time focusing, propagation of ultra-strong light pulses in media, and laser-induced plasma, which is of great significance to the research of basic physics, biology, and chemistry and the development of related technologies. The flexible adjustment mode of sub-pulse interval enables the technology to be applied to observe ultrafast dynamic processes of various time scales above picoseconds, meet different application requirements, and expand the application scope of compressed ultrafast imaging technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a structural schematic diagram of the present invention;

[0018] Figure 2 is a schematic diagram of a first pulse sequence generator;

[0019] Figure 3 is a schematic diagram of a second pulse train generator;

[0020] Figure 4 Schematic diagram of the internal structure of the streak camera;

[0021] Figure 5 It is a transmissive spatial coding diagram;

[0022] Figure 6 The spatiotemporal evolution diagram of the femtosecond laser-induced air plasma measured by the present invention and the corresponding pump-probe results;

[0023] Figure 7 The spatiotemporal evolution diagram of the femtosecond laser-induced shock wave measured by the present invention and the corresponding pump-probe results.

[0024] In the figure: 100-time domain discrete illumination compression ultrafast imaging system; 200-timing control system; 300-image reconstruction system; 400-first pulse sequence generator; 500-second pulse sequence generator; 101-femtosecond laser; 102-pulse sequence generator; 103-scene to be measured; 104-objective lens; 105-convex lens; 106-transmission space encoder; 107-lens pair; 108-bandpass filter; 109-streak camera; 201-digital delay generator; 301-computer; 401-ultrafast reflection Mirror; 402-first BBO frequency doubling crystal; 403-dichroic mirror; 501-second BBO frequency doubling crystal; 502-linear polarizer; 503-first beam splitter; 504-first reflector; 505-second reflector; 506-second beam splitter; 507-third reflector; 508-fourth reflector; 509-third beam splitter; 510-fifth reflector; 511-sixth reflector; 512-fourth beam splitter; 513-seventh reflector; 514-eighth reflector; 515-half-wave plate; 516-polarization beam splitter. DETAILED DESCRIPTION

[0025] See also Figure 1 , Figure 2 , Figure 3 , the present invention includes a time-domain discrete illumination compression ultrafast imaging system 100, a timing control system 200 and an image reconstruction system 300;

[0026] The time-domain discrete illumination compressed ultrafast imaging system 100 is composed of a femtosecond laser 101, a pulse sequence generator 102, a scene to be measured 103, an objective lens 104, a convex lens 105, a transmissive spatial encoder 106, a lens pair 107, a bandpass filter 108 and a streak camera 109; the femtosecond laser 101, the pulse sequence generator 102, the scene to be measured 103, the objective lens 104, the convex lens 105, the transmissive spatial encoder 106, the lens pair 107, the bandpass filter 108 and the streak camera 109 are optically connected in sequence;

[0027] The pulse sequence generator 102 includes a first pulse sequence generator 400 and a second pulse sequence generator 500, the two pulse sequence generators are independent of each other, and one of the pulse sequence generators is selected according to the actual observation time window; the first pulse sequence generator 400 is composed of an ultrafast reflector 401, a first BBO frequency doubling crystal 402, and a dichroic mirror 403, and the ultrafast reflector 401, the first BBO frequency doubling crystal 402, and the dichroic mirror 403 are optically connected in sequence, wherein the ultrafast reflector 401 is connected to the femtosecond laser 101, and the dichroic mirror 403 is connected to the femtosecond laser 101. The second pulse sequence generator 500 is composed of a second BBO frequency doubling crystal 501, a linear polarizer 502, a first beam splitter 503, a first reflector 504, a second reflector 505, a second beam splitter 506, a third reflector 507, a fourth reflector 508, a third beam splitter 509, a fifth reflector 510, a sixth reflector 511, a fourth beam splitter 512, a seventh reflector 513, an eighth reflector 514, a half-wave plate 515, and a polarization beam splitter 516. The second BBO frequency doubling crystal 501 is connected to the linear polarizer 502. , the linear polarizer 502 and the first beam splitter 503 are connected in an optical path, the first beam splitter 503 divides the optical path into two paths, one of which is connected to the first reflector 504 and the second reflector 505, and the other is connected to the second beam splitter 506, the second reflector 505 is connected to the second beam splitter 506, the second beam splitter 506 again divides the optical path into two paths, one of the two paths is connected to the third reflector 507 and the fourth reflector 508, and the other is connected to the third beam splitter 509, the fourth reflector 508 is connected to the third beam splitter 509, and the third beam splitter 509 divides the optical path into two paths for the third time Path, one of the two paths is connected to the fifth reflector 510 and the sixth reflector 511, and the other path is connected to the fourth beam splitter 512, the sixth reflector 511 is connected to the fourth beam splitter 512, the fourth beam splitter 512 divides the optical path into two paths for the fourth time, one of the two paths is connected to the seventh reflector 513, the eighth reflector 514 and the half-wave plate 515, and the other path is connected to the polarization beam splitter 516, the half-wave plate 515 is connected to the polarization beam splitter 516, the second BBO frequency doubling crystal 501 is connected to the femtosecond laser 101, and the polarization beam splitter 516 is connected to the scene to be measured 103;

[0028] The timing control system 200 is composed of a digital delay generator 201;

[0029] The image reconstruction system 300 is composed of a computer 301;

[0030] The digital delay generator 201 of the timing control system 200 is respectively connected to the femtosecond laser 101 and the streak camera 109 of the time-domain discrete illumination compression ultrafast imaging system 100;

[0031] The computer 301 of the image reconstruction system 300 is connected to the transmission-type spatial encoder 106 and the streak camera 109 of the time-domain discrete illumination compression ultrafast imaging system 100 respectively.

[0032] The computer 301 of the image reconstruction system 300 is equipped with a reconstruction algorithm TV-CD combining a total variation denoising model and a cascade denoiser, which reconstructs the compressed image collected by the streak camera 109 of the time-domain discrete illumination compression ultrafast imaging system 100 to restore the scene to be measured.

[0033] The present invention works like this:

[0034] See also Figure 1 , Figure 2 , Figure 3 The first pulse sequence generator 400 in the time-domain discrete illumination compressed ultrafast imaging system 100 of the present invention is suitable for observation scenarios with shorter pulse intervals, that is, faster ultrafast dynamic processes; the second pulse sequence generator 500 is suitable for observation scenarios with longer pulse intervals, corresponding to relatively slower ultrafast dynamic processes.

[0035] See also Figure 1 , Figure 2 When the time-domain discrete illumination compression ultrafast imaging system 100 of the present invention selects the first pulse sequence generator 400 to work, the femtosecond laser pulse generated by the femtosecond laser 101 enters the Fabry-Perot resonant cavity composed of the ultrafast reflector 401, the first BBO frequency doubling crystal 402, and the dichroic mirror 403. After passing through the ultrafast reflector 401, the pulse is frequency-doubled by the first BBO frequency doubling crystal 402 to generate frequency-doubled light, and then passes through the dichroic mirror 403, wherein the fundamental frequency light transmits out of the resonant cavity, a part of the frequency-doubled light is reflected, and another part transmits out of the resonator to form the first sub-pulse, and then the reflected frequency-doubled light passes through the first BBO frequency doubling crystal 402 and is reflected by the ultrafast reflector 401, passes through the first BBO frequency doubling crystal 402 again, and hits the dichroic mirror 403 for the second time, wherein a part of it is reflected, and another part is transmitted to form the second sub-pulse, and sequentially, a series of sub-pulses with proportional energy attenuation are formed. The pulse sequence is irradiated onto the scene to be measured 103, carrying the scene information at different times into the optical imaging system composed of the objective lens 104 and the convex lens 105, and then encoded by the transmission space encoder 106. The encoded scene passes through the lens pair 107 and the bandpass filter 108 and enters the streak camera 109 for deflection and superposition, and is finally collected by the built-in CCD of the streak camera 109. The collected two-dimensional image is transmitted to the computer 301 of the image reconstruction system 300 for calculation and reconstruction.

[0036] See also Figure 1 , Figure 2The ultrafast reflector 401 in the first pulse sequence generator 400 of the present invention can provide low group delay dispersion for femtosecond laser pulses, with an incident angle of 0°, high transmittance for fundamental frequency light, and high reflectance for frequency doubled light, with a reflectance greater than 99.6%, thereby maximizing the utilization rate of frequency doubled light. The energy of the frequency doubled light is determined by the frequency doubled efficiency of the first BBO frequency doubled crystal 402. The incident angle of the dichroic mirror 403 is 0°, and the cut-off wavelength is between the wavelengths of the fundamental frequency light and the frequency doubled light. The fundamental frequency light is transmitted, and the transmission and reflection ratio of the frequency doubled light determines the attenuation ratio of the sub-pulse. In order to minimize the intensity difference between sub-pulses, the dichroic mirror generally selects a transmission and reflection ratio of 1:9 for the frequency doubled light. At this time, the intensity of the eighth sub-pulse is about half of the intensity of the first sub-pulse. Therefore, in order to minimize the impact of intensity attenuation on the image reconstruction result, the first 8 sub-pulses are used. The distance between the ultrafast reflector 401 and the dichroic mirror 403 determines the time interval between adjacent sub-pulses. , where c is the speed of light, 3.0*10 8 m / s.

[0037] See also Figure 1 , Figure 3When the time-domain discrete illumination compression ultrafast imaging system 100 of the present invention selects the second pulse sequence generator 500 to work, the horizontally polarized laser pulse generated by the femtosecond laser 101 passes through the second BBO frequency doubling crystal 501 to generate vertically polarized frequency doubling light, which is then filtered by the linear polarizer 502 to be pure vertically polarized frequency doubling light. The first beam splitter 503, the second beam splitter 506, the third beam splitter 509, and the fourth beam splitter 512 are all polarization beam splitters, and the splitting ratio of the vertical polarized light is 50:50, with an error of ±2%. The polarization beam splitter 516 reflects the horizontal polarized light and transmits the vertical polarized light. Here, two adjacent beam splitters and the two reflectors between them respectively form four time delay lines, and the distances between the five beam splitters are equal. In order to make the time intervals between the sub-pulses equal, the length of the next time delay line is twice the length of the previous time delay line. The vertically polarized frequency-doubled light enters the first beam splitter 503 and is split into two paths. The two paths of light are respectively incident directly to and incident through the first time delay line to the second beam splitter 506. The second beam splitter 506 splits the light path into two paths again. The two paths are respectively incident directly to and incident through the second time delay line to the third beam splitter 509. The third beam splitter 509 splits the light path into two paths for the third time. The two paths are respectively incident directly to and incident through the third time delay line to the fourth beam splitter 512. The fourth beam splitter 512 splits the light path into two paths for the fourth time. The fourth time delay line is incident on the polarization beam splitter 516. The eight vertically polarized sub-pulses emitted upward from the fourth beam splitter 512 are deflected 90° into horizontal polarization after passing through the half-wave plate 515. Together with the eight vertically polarized sub-pulses emitted to the right, they are incident on the polarization beam splitter 516. The eight horizontally polarized sub-pulses are reflected, and the eight vertically polarized sub-pulses are transmitted. The 16 sub-pulses are combined here to form a pulse sequence with equal sub-pulse energy. The first eight sub-pulses are vertically polarized, and the last eight sub-pulses are horizontally polarized. The pulse sequence is irradiated onto the scene to be measured 103, carrying the scene information at 16 different times into the optical imaging system composed of the objective lens 104 and the convex lens 105, and then encoded by the transmission space encoder 106. The encoded scene passes through the lens pair 107 and the bandpass filter 108 and enters the streak camera 109 for deflection and superposition, and is finally collected by the built-in CCD of the streak camera 109. The collected two-dimensional image is transmitted to the computer 301 of the image reconstruction system 300 for calculation and reconstruction.

[0038] See also Figure 1 , Figure 3 The lengths of the four time delay lines in the second pulse sequence generator 500 are d 1 ,d 2 ,d 3 ,d 4 , then d 4 =2*d 3 =4*d2 =8*d 1 , and the time interval between adjacent sub-pulses is d 1 / c where c is the speed of light, 3.0*10 8 m / s. The energy of the doubled frequency light is determined by the efficiency of the second BBO doubled frequency crystal 501. Compared with using five beam splitters with a splitting ratio of 50:50, the method of first splitting the light with four beam splitters with a splitting ratio of 50:50 and then combining the light with one polarization beam splitter can make the sub-pulse energy higher.

[0039] See also Figure 1 The central wavelength of the bandpass filter 108 in the time-domain discrete illumination compression ultrafast imaging system 100 of the present invention is the wavelength of the frequency-doubled light, the transmittance is greater than 90%, and the full width at half maximum (FWHM) of the passband is 10 nm, which filters out the plasma self-luminescence and ambient light generated by the laser-induced dynamic process and improves the contrast of the imaging.

[0040] See also Figure 1 In the time-domain discrete illumination compression ultrafast imaging system 100 of the present invention, the objective lens 104 and the convex lens 105 are selected with appropriate specifications according to the size of the dynamic scene and the number of pixels of the CCD built in the streak camera, and a suitable magnification is matched. In order to obtain a better imaging effect, the lens pair 107 is selected to match the achromatic lens pair, and a suitable magnification is selected according to the needs.

[0041] See also Figure 1 , Figure 5 The transmissive spatial encoder 106 of the time-domain discrete illumination compression ultrafast imaging system 100 of the present invention is a customized random coding plate, which is made by coating a metal chromium film on optical glass and patterning it into random coding. The transmittance of the light-transmitting part is greater than 95%, and about 50% of the pixels are completely transparent, which can achieve intensity modulation of dynamic scenes. The pixel size and coding pattern of the transmissive spatial encoder 106 can be designed and customized according to the resolution of the time-domain discrete illumination compression ultrafast imaging system 100. Compared with the digital micromirror device, the transmissive spatial encoder 106 is small in size and low in price. The basic unit size of the transmissive spatial encoder 106 used in the present invention is 80μm.

[0042] See also Figure 1 , Figure 4 The streak camera 109 of the time-domain discrete illumination compression ultrafast imaging system 100 of the present invention is C7700 produced by Hamamatsu Corporation. The encoded scene enters the streak camera 109, and the photoelectric effect occurs on the photocathode to generate electrons. The electrons are accelerated by the acceleration gate into the linear scanning area and linearly deflected under the action of the scanning voltage. Subsequently, the deflected electrons are multiplied in the microchannel plate and finally bombard the fluorescent screen to generate photons. Finally, the three-dimensional information of time and space is converted into two-dimensional spatial information and captured by the built-in CCD.

[0043] See also Figure 1 The digital delay generator 201 of the timing control system 200 of the present invention adopts DG645 of Stanford Company. The digital delay generator 201 is simultaneously connected to the femtosecond laser 101 and the streak camera 109 of the time-domain discrete illumination compression ultrafast imaging system 100, and accurately controls the timing on a picosecond scale to compensate for the inherent delay of the system and the signal transmission delay, so that the work starts at the same time as the laser arrives, thereby recording the complete signal to be measured.

[0044] See also Figure 1 The computer 301 of the image reconstruction system 300 of the present invention processes the two-dimensional image collected by the streak camera 109 of the time-domain discrete illumination compression ultrafast imaging system 100, and uses the TV-CD algorithm to restore the dynamic scene. The specific process is as follows:

[0045] The scenario to be tested is represented as I ( x, y, t )( x, y are the horizontal and vertical coordinates of space, t is time), in the time-domain discrete illumination compression ultrafast imaging system, it is first spatially encoded by the transmission spatial encoder, and then enters the streak camera after intensity modulation. It is vertically deflected under the action of the deflection voltage, and finally the deflected image is superimposed in time and space and collected by the built-in camera. The two-dimensional image collected by the streak camera is expressed as E ( x', y' )( x', y' are the spatial horizontal and vertical coordinates of the image collected by the streak camera, respectively). Spatial encoding, time deflection, and spatiotemporal superposition are represented by C, T, and M operators, respectively. Then the imaging process of the time-domain discrete illumination compression ultrafast imaging system can be expressed as:

[0046] , (1)

[0047] In order to restore the three-dimensional information from the compressed image I ( x, y, t ), we need to solve the inverse problem of equation (1). However, the inverse problem is an ill-posed problem and cannot be directly inverted. Traditional reconstruction algorithms usually use a regularization term to restrict the solution of the inverse problem to an ideal signal space. I The best estimate of By minimizing the objective function f(I) Find, let A=MTC, then

[0048] (2)

[0049] in represents the two-norm, is the fidelity term, indicating that the reconstructed image must satisfy the sampling equation, is a regularization term, indicating that the reconstructed image needs to satisfy the prior information. b is the regularization parameter that balances the two terms. The TV-CD algorithm is based on the plug-and-play framework and the alternating direction multiplier method (PnP-ADMM) as an iterative framework. It combines the model-based total variation denoising model (TV) with three advanced deep learning-based denoising models, including the fast and flexible denoising convolutional neural network (FFDNet), the deep residual U-Net denoising neural network (DRUNet), and the fast deep video denoising network (FastDVDNet). It retains the advantages of TV that does not require pre-training, has strong generalization and high robustness, and adds three reliable deep denoisers, which can not only effectively improve the quality of the reconstructed image, but also increase the calculation speed. This combination is not only flexible and efficient, but also has good stability and generalization. In the PnP framework, by introducing auxiliary variables v , equation (2) can be written as

[0050] (3)

[0051] According to the Lagrange (AL) multiplier method, the optimal solution to problem (3) can be obtained by minimizing an AL function, which is expressed as

[0052] (4)

[0053] in u is another auxiliary variable, is the penalty factor. Further, problem (4) can be solved by the ADMM algorithm and decomposed into the following sub-problems:

[0054] (5)

[0055] (6)

[0056] (7)

[0057] Here, superscript k Indicates k Iterations. Subproblem (5) is a quadratic equation with a closed solution

[0058] , (8)

[0059] And subproblem (6) is a question about the regularization term ( v )right v This problem can be solved by using an off-the-shelf denoising algorithm, which can be expressed as

[0060] , (9)

[0061] in D stands for Denoiser, is the noise standard error. In the TV-CD algorithm, D Specifically, the four algorithms of TV, FFDNet, DRUNet and FastDVDNet are cascaded in the order of [TV, FFDNet, DRUNet, FastDVDNet, TV]. Each algorithm iterates an independent number of times. After one algorithm is iterated, it enters the next algorithm. After the iteration is completed, it can be obtained I ( x, y, t ).

[0062] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments, including the process, conditions, experimental methods, etc. for implementing the present invention. Except for the contents specifically mentioned below, they are all common knowledge and common common sense in the field, and the present invention has no special limitations.

[0063] Example 1

[0064] This embodiment observes the spatiotemporal evolution of air plasma excited by femtosecond laser. This embodiment selects the first pulse sequence generator and works in combination with the pump-probe technology. A femtosecond laser pulse with a central wavelength of 800 nm, a single pulse energy of 2.8 mJ, a repetition frequency of 1000 Hz, and a pulse width of 35 fs is provided by a mode-locked titanium sapphire laser amplifier. The 800 nm fundamental frequency light emitted from the dichroic mirror 403 is used as pump light and focused through an objective lens in a direction perpendicular to the detection light path to excite the air to generate plasma. The pulse sequence of 400 nm detection light is irradiated onto the scene to obtain 8 time slices carrying spatial information of air plasma at different times. In the first pulse sequence generator, the distance between the ultrafast reflector and the dichroic mirror is set to 3 cm, the corresponding sub-pulse interval is 200 ps, ​​the 12-fold frequency efficiency of the first BBO frequency-doubling crystal is 29%, the transmission and reflection ratio of the dichroic mirror to the doubled frequency light is 1:9, and the energies of the eight sub-pulses are 12 mw, 10.8 mw, 9.72 mw, 8.74 mw, 7.87 mw, 7.08 mw, 6.37 mw, and 5.74 mw, respectively. By adjusting the time delay between the pump light and the probe light, the moment when the plasma is about to appear is determined as the time zero point. The pulse sequence records the information of 67 ps, 267 ps, 467 ps, 667 ps, 867 ps, 1067 ps, 1267 ps, and 1467 ps in sequence, and the time slices enter the subsequent time-domain discrete illumination compression ultrafast imaging system and are recorded.

[0065] See also Figure 1 , Figure 2 , Figure 4 , Figure 5 , Figure 6 , the computer 301 of the image reconstruction system 300 uses the TV-CD algorithm to reconstruct the dynamic scene, such as Figure 4 As shown in the figure, it is a reconstructed image of air plasma detected by DI-CUP. It can be clearly seen from the reconstruction results that at 67 ps, the laser-induced plasma channel can be observed to form an elliptical area. Subsequently, the plasma channels expand and widen along their axes. It is worth noting that at 67 ps, there is a clear elliptical hollow structure inside the plasma channel, which then gradually weakens until it disappears completely at 1067 ps. This is formed by the interaction between the diffraction caused by the plasma and the refocusing caused by the Kerr effect. The reconstructed internal structure and development trend are consistent with the results obtained by multiple pump-probe shots. The present invention successfully obtains spatial information at multiple times through a single measurement, proving the reliability of the present invention.

[0066] Example 2

[0067] This embodiment observes the spatiotemporal evolution of shock waves induced by femtosecond laser focusing on silicon surface. This embodiment selects a second pulse sequence generator and works in combination with pump-probe technology. A mode-locked titanium sapphire laser amplifier provides femtosecond laser pulses with a central wavelength of 800 nm, a single pulse energy of 2.8 mJ, a repetition frequency of 1000 Hz, and a pulse width of 35 fs. A dichroic mirror is inserted after the linear polarizer 502, and the 800 nm fundamental frequency light emitted from the dichroic mirror is used as pump light to be focused on the silicon target surface through an objective lens in a direction perpendicular to the detection light path to induce shock waves. The pulse sequence composed of 16 sub-pulses in the detection light path is irradiated onto the scene to obtain 16 time slices carrying spatial information of shock waves at different times. In the second pulse sequence generator, the lengths of the four time delay lines are set to 12 cm, 24 cm, 48 cm, and 96 cm, respectively, and the corresponding sub-pulse intervals are about 0.4 ns. The frequency doubling efficiency of the second BBO frequency doubling crystal 501 is 29%, and the energy of each sub-pulse is about 20 mw. By adjusting the time delay between the pump light and the detection light, the moment when the shock wave is about to appear but has not yet appeared is determined as time zero. The moments corresponding to the 16 sub-pulses are 0 ns, 0.4 ns, 0.9 ns, 1.3 ns, 1.7 ns, 2.1 ns, 2.6 ns, 3.0 ns, 3.4 ns, 3.8 ns, 4.2 ns, 4.6 ns, 5.0 ns, 5.4 ns, 5.9 ns, and 6.3 ns. 16 time slices carrying spatial information at different moments will be generated. The time slices enter the subsequent time-domain discrete illumination compression ultrafast imaging system and are recorded.

[0068] See also Figure 1 , Figure 3 , Figure 4 , Figure 5 , Figure 7 , the computer 301 of the image reconstruction system 300 uses the TV-CD algorithm to reconstruct the dynamic scene, such as Figure 5 The figure shows the reconstructed image of the shock wave detected by DI-CUP. The vertical dividing line between the scene area and the non-scene area near the right side of the image is the silicon target surface. The pump light is focused on this surface. The strong laser exceeds the ionization threshold of silicon, and the silicon atoms on the surface are ionized to form a high-temperature and high-pressure dense plasma. The plasma continues to expand outward and squeeze the air, forming a hemispherical shock wave. Figure 4It can be clearly observed that at 0.4 ns, an opaque area appeared on the target surface, and there was a high-intensity cone structure in front of the opaque area. Due to the rapid deposition of strong laser energy, the atoms on the target surface were ionized to form a dense plasma, which shielded the light and led to the generation of the opaque area. Since the target surface is located at the beam waist of the laser focal spot, the high-intensity cone area is considered to be the leading edge of the plasma channel formed by the laser ionization of air. Over time, the high-temperature plasma expands outward, compressing the surrounding air to form a bright shock wave. From 1.3 to 3.0 ns, there is a bulge on the front surface of the shock wave, which is due to the strong laser-induced air breakdown, and the local pressure is low, providing a fast channel for the propagation of the shock wave. Subsequently, from 3.4 to 6.3 ns, as the dense plasma expands and thins, the detection light gradually penetrates the plasma. At this stage, the opaque area near the surface is mainly caused by debris, particles and droplets ejected from the target. At the same time, another fairly large dark area can be observed behind the shock wave, which may be due to the partial reflection caused by the collision of the vapor in the ablation plume with the surrounding air, resulting in a rise in local density and pressure. It can be seen that the reconstruction results have restored the generation and change process of the shock wave well. The internal structure and development trend are consistent with the results obtained through multiple pump-probe shooting, realizing the single measurement of the shock wave.

[0069] In general, the present invention addresses the problem of insufficient resolution of images reconstructed by CUP technology. It can not only achieve single-shot multiple-frame imaging and effectively avoid errors caused by traditional measurement methods such as pump-probe, but also reduce data compression ratio and improve spatial resolution. This provides an accurate and effective measurement tool for non-repeatable ultrafast dynamic processes such as time focusing, propagation of ultra-intense light pulses in media, and laser-induced plasma, which is of great significance to the research of basic physics, biology, and chemistry and the development of related technologies.

[0070] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A compressed ultrafast imaging device based on time-domain discrete illumination, characterized in that: It comprises a time-domain discrete illumination compression ultrafast imaging system (100), a timing control system (200) and an image reconstruction system (300); The time-domain discrete illumination compressed ultrafast imaging system (100) is composed of a femtosecond laser (101), a pulse sequence generator (102), a scene to be measured (103), an objective lens (104), a convex lens (105), a transmission-type spatial encoder (106), a lens pair (107), a bandpass filter (108) and a streak camera (109); the femtosecond laser (101), the pulse sequence generator (102), the scene to be measured (103), the objective lens (104), the convex lens (105), the transmission-type spatial encoder (106), the lens pair (107), the bandpass filter (108) and the streak camera (109) are optically connected in sequence; The pulse sequence generator (102) comprises a first pulse sequence generator (400) and a second pulse sequence generator (500), the two pulse sequence generators are independent of each other, and one of the pulse sequence generators is selected according to an actual observation time window; the first pulse sequence generator (400) is composed of an ultrafast reflector (401), a first BBO frequency doubling crystal (402), and a dichroic mirror (403), the ultrafast reflector (401), the first BBO frequency doubling crystal (402), and the dichroic mirror (403) are optically connected in sequence, the ultrafast reflector (401) is connected to a femtosecond laser (101), and the dichroic mirror (403) is connected to a scene to be measured (1 03); the second pulse sequence generator (500) is composed of a second BBO frequency doubling crystal (501), a linear polarizer (502), a first beam splitter (503), a first reflector (504), a second reflector (505), a second beam splitter (506), a third reflector (507), a fourth reflector (508), a third beam splitter (509), a fifth reflector (510), a sixth reflector (511), a fourth beam splitter (512), a seventh reflector (513), an eighth reflector (514), a half-wave plate (515), and a polarization beam splitter (516); the second BBO frequency doubling crystal (501) and the linear polarizer (502) are connected to each other; The linear polarizing plate (502) and the first beam splitter (503) are connected in an optical path, the first beam splitter (503) splits the optical path into two paths, one of which is connected to the first reflector (504) and the second reflector (505), and the other is connected to the second beam splitter (506). The second reflector (505) is connected to the second beam splitter (506). The second beam splitter (506) again splits the optical path into two paths, one of which is connected to the third reflector (507) and the fourth reflector (508), and the other is connected to the third beam splitter (509). The fourth reflector (508) is connected to the third beam splitter (509). The third beam splitter (509) splits the optical path into two paths for the third time. The optical path is divided into two paths, one of which is connected to a fifth reflector (510) and a sixth reflector (511), and the other is connected to a fourth beam splitter (512). The sixth reflector (511) is connected to the fourth beam splitter (512). The fourth beam splitter (512) divides the optical path into two paths for the fourth time. One of the two paths is connected to a seventh reflector (513), an eighth reflector (514) and a half-wave plate (515), and the other is connected to a polarization beam splitter (516). The half-wave plate (515) is connected to the polarization beam splitter (516). The second BBO frequency doubling crystal (501) is connected to a femtosecond laser (101), and the polarization beam splitter (516) is connected to a scene to be measured (103). The timing control system (200) is composed of a digital delay generator (201); The image reconstruction system (300) is composed of a computer (301); The digital delay generator (201) of the timing control system (200) is respectively connected to the femtosecond laser (101) and the streak camera (109) of the time-domain discrete illumination compression ultrafast imaging system (100); The computer (301) of the image reconstruction system (300) is respectively connected to the transmission-type spatial encoder (106) and the streak camera (109) of the time-domain discrete illumination compression ultrafast imaging system (100); The computer (301) of the image reconstruction system (300) is equipped with a reconstruction algorithm TV-CD that combines a total variation denoising model with a cascade denoiser, and reconstructs a compressed image collected by a streak camera (109) of a time-domain discrete illumination compression ultrafast imaging system (100). The image reconstruction process is as follows: The scenario to be tested is represented as I ( x, y, t )( x, y are the horizontal and vertical coordinates of space, t is time), in the time-domain discrete illumination compression ultrafast imaging system (100), it is first spatially encoded by the transmission spatial encoder (106), and then enters the streak camera (109) after being intensity modulated, and is vertically deflected under the action of the deflection voltage. Finally, the deflected image is spatially superimposed and collected by the streak camera (109). The two-dimensional image collected by the streak camera (109) is represented as E ( x ', y ')( x ', y ' are the spatial horizontal and vertical coordinates of the image collected by the streak camera, respectively), spatial encoding, time deflection, and spatiotemporal superposition are represented by C, T, and M operators respectively, so the imaging process of the time-domain discrete illumination compression ultrafast imaging system is expressed as: E ( x', y' )=MTC I ( x, y, t ), (1) In order to restore 3D information from compressed images I ( x, y, t ), the inverse problem of equation (1) needs to be solved; the TV-CD algorithm adopts the plug-and-play PnP framework and the alternating direction multiplier method ADMM to decompose the inverse problem into three sub-problems; when solving the denoising sub-problem, four denoisers, TV, FFDNet, DRUNet and FastDVDNet, are used for cascade denoising in the order of [TV, FFDNet, DRUNet, FastDVDNet, TV]; each algorithm is iterated independently, and after completing the iteration of one algorithm, it enters the next algorithm; after all iterations are completed, we get I ( x, y, t ).

Images