A hyperspectral-based high-temporal and spatial resolution detection device and method

Abstract

The application discloses a kind of high spatial resolution detection device and method based on hyperspectral, wherein, high spatial resolution detection device includes X-ray response module, detection light generation module, synchronous control system, imaging recording module and calibration light path.Detection light generation module generates chirp pulse as detection light;X-ray response module makes X-ray and semiconductor response medium interact, and the space distribution change of the time-space evolution process of X-ray is converted into carrier concentration in semiconductor response medium;Imaging recording module includes imaging system and hyperspectral resolution camera, and the action area of semiconductor response medium is imaged to hyperspectral resolution camera.Under single exposure, multiple high time resolution (femtosecond to picosecond order) two-dimensional X-ray images are obtained, time resolution is adjustable, spatial resolution is high, system structure is simple, cost is low, efficiency is high, anti-interference ability is strong, suitable for X-ray time-space diagnosis in laser fusion and high energy density physics research.

Description

Technical Field

[0001] This invention relates to the field of X-ray measurement technology, and specifically to a high spatiotemporal resolution detection device and method based on hyperspectral imaging. Background Technology

[0002] In laser fusion and high-energy-density physics research, X-ray spatiotemporal diagnostics is a crucial technique. Measuring the X-ray energy spectrum and its spatiotemporal evolution reveals important plasma properties, including key physical parameters such as electron temperature, electron density, and temperature gradient. Therefore, the spatiotemporal evolution of X-rays provides crucial data for understanding processes such as laser-matter interactions, plasma heating and compression, target implosion dynamics, and even combustion.

[0003] Two-dimensional images of the X-ray emission region at different times are fundamental to understanding and studying the state and dynamics of black cavity plasma. They are crucial for investigating the energy coupling efficiency between lasers and black cavities, the transmission and energy deposition locations of target lasers, X-ray conversion and transport, and play a vital role in understanding physical processes such as the growth of hydrodynamic interface instabilities and shock wave propagation. Traditional streak cameras can provide time resolution on the picosecond (ps) level or even higher, but lack two-dimensional spatial resolution. Traditional framing cameras have a time resolution in the range of 30-70 ps. DIXI-type framing cameras based on drift tube technology can achieve a time resolution of up to 5 ps, offering high time resolution but lacking high spatial resolution. Furthermore, the electronic diagnostic systems formed by such technologies are often structurally complex and susceptible to interference from strong neutrons, gamma rays, and electromagnetic noise, which can lead to inaccurate measurements and even component damage. Therefore, additional anti-interference devices are typically designed to protect the diagnostic system. However, as laser fusion and high-energy-density physics experiments progress, various interferences become stronger. The complexity of such devices will not only increase significantly, but the protection effect cannot be fully guaranteed, becoming the main factor limiting the further development of current X-ray diagnostic technology.

[0004] Converting the spatiotemporal evolution of X-rays into the detection of visible light signals is a very effective method. However, the time resolution of CCD and CMOS ultrafast two-dimensional imaging in the optical band is only about 100 ns, corresponding to a readout speed of only 10 ns. 7The current technology cannot fundamentally overcome the limitations of chip storage technology and electronic readout speed. Therefore, it is insufficient to achieve ultrafast two-dimensional imaging using only electronic-based detection techniques; other non-electronic detection techniques also need to be developed. Related detection techniques include frequency-resolved optical shutter (FROG), self-referenced spectral phase coherent electric field reconstruction (SIPDER), and improved pump-probe techniques. Although these methods have very high time resolution (depending on the width of the detection pulse), they can only measure repetitive phenomena and cannot measure the spatiotemporal evolution of X-ray images.

[0005] Solving these problems is now a top priority. Summary of the Invention

[0006] To address the technical problem that while some existing methods have very high temporal resolution, they can only measure repetitive phenomena and cannot measure the spatiotemporal evolution of X-ray images, this invention provides a high spatiotemporal resolution detection device and method based on hyperspectral imaging. By utilizing the spectral-time mapping characteristics of chirped pulses combined with hyperspectral imaging technology, multiple high temporal resolution two-dimensional X-ray images can be acquired in a single exposure.

[0007] The technical solution is as follows:

[0008] The first aspect of this application relates to a high spatiotemporal resolution detection device based on hyperspectral imaging, comprising:

[0009] The X-ray response module includes a target laser, a vacuum target chamber, and a semiconductor response medium. An X-ray imaging structure is provided on the vacuum target chamber. The target laser emitted by the target laser is focused on the target in the vacuum target chamber to generate X-rays. The X-rays are imaged on the front surface of the semiconductor response medium by the X-ray imaging structure.

[0010] The probe light generation module includes a probe laser, an adjustable attenuator, a dispersive medium, a beam expander, and a first semi-transparent and semi-reflective mirror. The probe laser emitted by the probe laser is attenuated by the adjustable attenuator, and the spectrum is stretched in time by the dispersive medium and converted into a chirped pulse as the probe light. The probe light is expanded by the beam expander and then reflected by the first semi-transparent and semi-reflective mirror to the back surface of the semiconductor response medium.

[0011] A synchronous control system is used to control the relative delay between the target laser emitted by the target laser and the probe laser emitted by the probe laser, so that X-rays and chirped pulses interact within a semiconductor response medium.

[0012] The imaging recording module includes an imaging system and a hyperspectral resolution camera. The probe light reflecting back from the semiconductor response medium, carrying information on the change in carrier concentration, passes through the first semi-transparent mirror and is then imaged onto the hyperspectral resolution camera by the imaging system, recording a two-dimensional image containing multispectral information in a single exposure.

[0013] The calibration optical path includes a spectrometer, a streak camera, and a second semi-transparent mirror positioned between a dispersive medium and a beam expander. The probe light emitted from the dispersive medium is split in two after hitting the second semi-transparent mirror. One beam of probe light passes through the second semi-transparent mirror and hits the beam expander, while the other beam of probe light reflected by the second semi-transparent mirror passes through the spectrometer and hits the streak camera.

[0014] The second aspect of this application relates to a high spatiotemporal resolution detection method based on the above-mentioned high spatiotemporal resolution detection device, which is carried out according to the following steps:

[0015] S1. The probe laser emitted from the probe laser has its intensity adjusted by an adjustable attenuator. The spectrum is then stretched in time by the dispersive medium and converted into a chirped pulse as the probe light. The probe light is guided to the spectrometer by the second semi-transparent and semi-reflective mirror and finally recorded by the streak camera. The mapping relationship between the spectrum and time is then calibrated.

[0016] S2. Using a spectrally calibrated tungsten halogen lamp as the calibration light source, the light is irradiated onto the diffuse scattering plate to produce uniformly distributed light. After filtering out unwanted spectra and adjusting the light intensity, a hyperspectral resolution camera is used to record the light spectrum and establish the mapping relationship between gray values ​​and spectral response.

[0017] S3. After the target is placed in the vacuum target chamber, the synchronous control system controls the target laser and the probe laser to emit the target laser and the probe laser respectively according to the set relative delay. The X-ray generated by the target laser is focused on the target and the probe laser is converted into chirped pulse probe light through the dispersive medium. They arrive at the front surface and the back surface of the semiconductor response medium synchronously and interact within the semiconductor response medium.

[0018] S4. The probe light carrying information on the change in carrier concentration reflected from the back surface of the semiconductor response medium passes through the first semi-transparent mirror and is imaged onto the hyperspectral resolution camera by the imaging system. The hyperspectral resolution camera records a snapshot spectral image containing multispectral information in a single exposure.

[0019] S5. First, the snapshot spectral images recorded by the hyperspectral resolution camera are processed to obtain an image sequence of X-ray evolution over time.

[0020] The above-mentioned high spatiotemporal resolution detection device and method based on hyperspectral imaging have achieved the following technical results:

[0021] 1. High spatiotemporal resolution capability: This invention utilizes the spectral-time mapping characteristics of chirped pulses and combines them with hyperspectral imaging technology to achieve high temporal and spatial resolution two-dimensional imaging at the femtosecond to picosecond level. It can simultaneously obtain a large number of high temporal resolution two-dimensional multispectral images, forming an image sequence of X-ray evolution over time. This breaks through the temporal resolution limit of traditional electronic detection technology and meets the detection requirements of ultrafast physical processes such as laser fusion.

[0022] 2. Ensure high measurement accuracy through calibration: The mapping relationship between spectrum and time can be obtained simply and efficiently by calibrating the optical path. The high-spectral resolution camera can be calibrated simply and efficiently by using a tungsten halogen lamp in conjunction with a diffuse scattering plate, and the mapping relationship between gray value and spectral response can be established.

[0023] 3. Applicable to single ultrafast physical processes and has a wide range of applications: Unlike methods such as FROG and SPIDER, which can only measure repetitive phenomena, this invention uses a hyperspectral resolution camera to obtain a complete spatiotemporal evolution image sequence in a single laser target experiment, which is suitable for the diagnosis of irreversible ultrafast processes.

[0024] 4. Simple system structure: This method directly uses the laser output from an industrial laser (usually a femtosecond laser) as the probe light, without the need for other complex physical processes (such as optical nonlinear processes) to generate the probe light; it uses a hyperspectral resolution camera with a larger array and higher spectral resolution, which can simultaneously record high temporal and spatial resolution images, making the structure and optical path simpler and more efficient, reducing system complexity and cost.

[0025] 5. Strong anti-interference capability: This invention adopts a non-electronic optical detection method. The main components are a hyperspectral camera and optical elements, which are not easily affected by strong neutrons, gamma rays, electromagnetic noise and other interferences. There is no need for complicated anti-interference protection devices, which improves the reliability and stability of the system. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the optical path of a high spatiotemporal resolution detection device.

[0027] Figure 2 A schematic diagram of a snapshot spectral image recorded by a hyperspectral resolution camera. Detailed Implementation

[0028] The present invention will be further described below with reference to the embodiments and accompanying drawings.

[0029] Example 1:

[0030] like Figure 1As shown, a high spatiotemporal resolution detection device based on hyperspectral imaging mainly includes an X-ray response module, a probe light generation module, a synchronous control system 1, an imaging recording module, and a calibration optical path. By utilizing the spectral-time mapping characteristics of chirped pulses combined with hyperspectral imaging technology, multiple high temporal resolution two-dimensional X-ray images can be acquired in a single exposure.

[0031] In this embodiment, the X-ray response module includes a target laser 2, a vacuum target chamber 3, and a semiconductor response medium 12. The interior of the vacuum target chamber 3 is a vacuum environment, and an X-ray imaging structure 5 is provided on the shell of the vacuum target chamber 3. The target laser emitted from the target laser 2 is focused on the target in the vacuum target chamber 3 to generate X-rays. The X-rays are imaged onto the front surface of the semiconductor response medium 12 by the X-ray imaging structure 5, thereby causing the X-rays to interact with the semiconductor response medium 12 and converting the spatiotemporal evolution process of the X-rays into changes in the carrier concentration in the semiconductor response medium 12.

[0032] Specifically, the high-energy laser emitted from the laser 2 is focused by the focusing lens system and passes through the entrance window of the vacuum target chamber 3, precisely focusing onto the target 4 located at the center of the vacuum target chamber 3. Under the action of the powerful laser, the target 4 instantly generates high-temperature, high-density plasma and radiates X-rays carrying spatiotemporal evolution information. These X-rays then pass through the X-ray imaging structure 5 set on the shell of the vacuum target chamber 3, which images the X-rays onto the front surface of the semiconductor response medium 12 according to a specific spatial distribution pattern. The semiconductor response medium 12 is the core sensing element of this invention. Its body is made of a semiconductor material with ultra-high carrier mobility and fast photoelectric response characteristics. For example, it can be CdSe (gallium selenide) material, or in other embodiments, it can be GaAs (gallium arsenide), Si (silicon), etc., as long as it can meet the functional requirement of rapidly exciting and changing the carrier concentration under X-ray irradiation. When X-rays are incident on the front surface of the semiconductor response medium 12, the photon energy of the X-rays will excite the generation of electron-hole pairs inside the semiconductor, thereby converting the intensity distribution of X-rays into a spatial distribution change of carrier concentration inside the semiconductor response medium 12, and this change in carrier concentration corresponds strictly to the spatiotemporal evolution process of X-rays.

[0033] Furthermore, the semiconductor response medium 12 includes a semiconductor response medium body made of semiconductor material, an antireflection coating deposited on the side of the semiconductor response medium body near the X-ray imaging structure 5, and a metal reflective layer deposited on the side of the semiconductor response medium body near the first semi-transparent mirror 11. Specifically, the semiconductor response medium body is the substrate that carries the photo-object interaction, and it is typically made of semiconductor material with ultra-high carrier mobility and fast photoelectric response characteristics, such as the CdSe material mentioned above. On the side of the body near the X-ray imaging structure 5, that is, the front surface facing the X-ray incident direction, an antireflection coating is deposited. The design of this antireflection coating is crucial; its refractive index is precisely matched, and its main function is to reduce interface reflection loss when X-rays penetrate the semiconductor material, ensuring that as much X-ray photon energy as possible can enter the interior of the semiconductor body, thereby efficiently exciting carriers and improving detection sensitivity. On the side of the body near the first semi-transparent mirror 11, that is, the rear surface facing the incident direction of the detection light, a metal reflective layer is deposited. The function of this metallic reflective layer is to efficiently reflect the chirped pulse probe light incident from the rear surface back, preventing transmission loss, while ensuring that the reflected probe light can carry the information on carrier concentration changes caused by X-ray excitation at the front surface and be successfully returned to the imaging recording module. It should be understood that although this embodiment describes the specific layout of the antireflection coating on the front surface and the metallic reflective layer on the rear surface, in other embodiments, the specific position and type of the coating can be adaptively adjusted according to the specific requirements of the optical path design, as long as the core functions of antireflection and reflection are met.

[0034] Furthermore, the metal reflective layer is made of gold or copper, and its thickness is 10 nm. For example, the metal reflective layer can be made of gold, which has excellent chemical stability and high reflectivity in the visible to near-infrared bands, maintaining its reflective performance without degradation under complex experimental environments over a long period. Alternatively, the metal reflective layer can be made of copper, which also possesses excellent reflective properties and is relatively cheaper. Regardless of whether gold or copper is used, its thickness is set to 10 nm. This specific value of 10 nm is not arbitrarily chosen, but determined based on the balance mechanism between reflectivity and transmitted X-rays. On the one hand, if the metal layer is too thick, for example, exceeding 50 nm, although the reflectivity of the probe light is extremely high, it will cause a large amount of X-rays incident on the front surface to be absorbed or blocked by the metal layer before penetrating to the semiconductor body, severely weakening the X-ray transmission and reducing the carrier excitation efficiency. On the other hand, if the metal layer is too thin, for example, less than 5 nm, a continuous and dense reflective film layer cannot be formed, the reflectivity of the probe light drops sharply, resulting in insufficient intensity of the reflected light carrying information, affecting the imaging signal-to-noise ratio. The 10nm thickness is just right to ensure sufficient reflectivity for the probe light while allowing X-rays to pass through the thin layer and enter the semiconductor body with high transmittance, achieving the best balance between the dual requirements of reflection and transmission.

[0035] Furthermore, the X-ray imaging structure 5 is preferably a small aperture, approximately 15 μm in diameter, formed in the vacuum target chamber 3, thus constituting pinhole imaging, which is simple and reliable. Alternatively, the X-ray imaging structure 5 is preferably a KB microscope mounted on the shell of the vacuum target chamber 3, thereby achieving higher spatial resolution than pinhole imaging. These two parallel examples each have their advantages and disadvantages and are suitable for different application scenarios. When the X-ray imaging structure 5 is a small aperture, for example, approximately 15 μm in diameter, formed in the vacuum target chamber 3, it constitutes a simple pinhole imaging structure. The significant advantage of this approach is its extremely simple and reliable structure, low manufacturing cost, and absence of aberrations that may be introduced by complex optical components, making it very suitable for routine diagnostic experiments where spatial resolution requirements are not extreme but system robustness requirements are extremely high. However, the disadvantage of pinhole imaging is that spatial resolution is limited by the aperture size; too small an aperture will result in insufficient X-ray flux, while too large an aperture will lead to a decrease in resolution. As an alternative, when the X-ray imaging structure 5 is a KB microscope mounted on the vacuum target chamber 3, the KB microscope utilizes a combination of multiple reflecting mirrors to focus X-rays with a high numerical aperture, thus achieving a spatial resolution several times higher than that of pinhole imaging, and clearly distinguishing the fine structural features of the plasma. However, its disadvantages lie in its relatively complex structure, extremely high requirements for assembly and adjustment, and high cost. Therefore, experimenters can flexibly choose between these two alternatives based on specific resolution requirements and budget constraints.

[0036] In this embodiment, the probe light generation module includes a probe laser 6, an adjustable attenuator 7, a dispersive medium 9, a beam expander 10, and a first semi-transparent and semi-reflective mirror 11. The probe laser emitted by the probe laser 6 is attenuated by the adjustable attenuator 7, and then the spectrum is stretched in time by the dispersive medium 9 and converted into a chirped pulse as the probe light. The probe light is expanded by the beam expander 10 and then reflected by the first semi-transparent and semi-reflective mirror 11 to the rear surface of the semiconductor response medium 12.

[0037] Specifically, the probe laser 6 emits an ultrashort pulse laser with a specific spectral width. This laser first passes through an adjustable attenuator 7 for intensity modulation to prevent excessively strong probe light from causing additional carrier excitation interference to the semiconductor response medium 12. The modulated probe laser then enters the dispersive medium 9, a key device for generating chirped pulses. Utilizing the physical property of different refractive indices and propagation speeds for different wavelengths (i.e., dispersion effect), the spectral components that were originally overlapping in time are separated after passing through the medium, with longer and shorter wavelength components emitted sequentially. This elongates the ultrashort pulse in time, converting it into a chirped pulse with spectral components arranged chronologically as the probe light. This chirping mechanism establishes a precise mapping between spectrum and time, allowing subsequent time information to be deduced by resolving the spectrum. The chirped pulse probe light then passes through a beam expander 10 for beam widening, ensuring its cross-sectional size completely covers the imaging area of ​​the semiconductor response medium 12. The expanded probe light is directed towards the first semi-transparent mirror 11, which reflects the probe light so that it enters from the rear surface of the semiconductor response medium 12. It should be understood that although the figure shows the optical path layout where the probe light enters the dispersive medium 9 after being oriented by the reflector 8, in other embodiments, depending on spatial layout requirements, the reflector 8 can be omitted or more reflectors can be used for a multi-reflection optical path design, as long as the probe light passes smoothly through each device sequentially.

[0038] Furthermore, the adjustable attenuator 7 is preferably an adjustable attenuator plate, which is used to adjust the intensity of the probe laser.

[0039] Furthermore, the probe laser 6 is a femtosecond laser with an output spectral coverage of 40 nm ± 5 nm, and the dispersive medium 9 can temporally broaden the femtosecond pulse to 10 ps-200 ps. Specifically, the probe laser 6 must be a femtosecond laser because only extremely short pulses on the femtosecond scale possess a sufficiently wide initial spectral bandwidth to provide the basic spectral resources for subsequent dispersion broadening. Its output spectral coverage is set at 40 nm ± 5 nm, a range that includes multiple parallel endpoint examples, such as a spectral coverage of 35 nm, 40 nm, or 45 nm. The width of the spectral coverage directly determines the total time window width after the chirped pulse is temporally stretched, thus having a decisive impact on the final temporal resolution. The role of the dispersive medium 9 is precisely to temporally broaden this wide-spectrum femtosecond pulse, with the broadening range set at 10 ps-200 ps. For example, when using a high-dispersion photonic crystal fiber with a length of approximately 80 cm as the dispersion medium 9, the femtosecond pulse can be broadened to a shorter time window of approximately 10 ps, ​​suitable for detecting physical processes with extremely fast evolution, requiring extremely high time resolution but short total duration. When using a high-dispersion photonic crystal fiber with a length of approximately 16 m, the pulse can be broadened to a longer time window of approximately 200 ps, ​​suitable for detecting physical processes with longer duration but relatively relaxed time resolution requirements. This mapping relationship between the spectral range and the broadened time window is the core source of the time resolution capability of this invention: the wider the spectral range, the longer the time dimension that can be separated by dispersion, and the longer the evolutionary process that can be recorded; within a given spectral range, the greater the dispersion of the dispersion medium, the higher the temporal separation of spectral components, and the larger the time interval between adjacent spectral components, thus ensuring the accuracy of inferring time information through spectral resolution, i.e., the time resolution. It should be understood that although this embodiment provides specific examples of glass rod length and broadening time, in actual operation, experimenters can flexibly change media of different lengths or with different dispersion coefficients according to specific detection timescale requirements in order to adjust the time window and resolution to the best matching state.

[0040] Furthermore, a reflector 8 is provided between the adjustable attenuator 7 and the dispersive medium 9 to adjust the propagation direction of the probe laser and accurately guide the probe laser into the incident end of the dispersive medium 9.

[0041] Furthermore, the beam expander 10 preferably employs a Galilean or Keplerian beam expander system to enlarge the probe spot to the required size, covering the imaging area of ​​the semiconductor response medium 12.

[0042] In this embodiment, the synchronous control system 1 is used to control the relative delay between the target laser emitted by the target laser 2 and the probe laser emitted by the probe laser 6, so that the X-rays and chirped pulses interact within the semiconductor response medium 12, and finally the probe light reflecting back from the semiconductor response medium 12 with information on the change in carrier concentration is introduced into the imaging recording module.

[0043] Specifically, the synchronous control system 1 establishes signal connections with the target laser 2 and the probe laser 6, respectively. Its internal high-precision delay generator sends trigger control signals to the two lasers according to the set relative delay parameters. This timing coordination mechanism is crucial, ensuring that at the same instant the X-rays excited by the target laser reach the front surface of the semiconductor response medium 12, the chirped pulse probe light, after dispersion broadening, also reaches the rear surface of the semiconductor response medium 12. At this moment, the X-rays and probe light undergo a spatiotemporal encounter and interaction within the semiconductor response medium 12: the transient change in carrier concentration generated by the X-rays excited on the front surface instantaneously alters the refractive index and absorptivity of the semiconductor response medium 12 for the specific wavelength probe light incident at that moment, thus causing the probe light reflected back from the rear surface to carry information about the carrier concentration change at the corresponding spatial location. This interaction principle cleverly modulates the invisible spatiotemporal evolution of ultrafast X-rays into the intensity and spectral changes of the chirped pulse in the visible light band.

[0044] In this embodiment, the imaging recording module includes an imaging system 13 and a hyperspectral resolution camera 14. The probe light reflecting back from the semiconductor response medium 12, carrying information on the change in carrier concentration, passes through the first semi-transparent mirror 11 and is imaged onto the hyperspectral resolution camera 14 by the imaging system 13, recording a two-dimensional image containing multispectral information in a single exposure.

[0045] Specifically, the probe light reflected from the back surface of the semiconductor response medium 12 already carries information about the carrier concentration changes caused by the spatiotemporal evolution of X-rays, and its propagation direction is again directed towards the first semi-transparent mirror 11. Due to the semi-transparent and semi-reflective properties of this mirror, the reflected probe light is no longer reflected, but passes directly through the first semi-transparent and semi-reflective mirror 11 and then enters the imaging system 13. The imaging system 13 is usually composed of multiple lenses, such as a 4F imaging system composed of the first lens 13-1 and the second lens 13-2, which clearly images the effective area of ​​the semiconductor response medium 12 onto the sensing surface of the hyperspectral resolution camera 14 at a set magnification. The hyperspectral resolution camera 14 is a snapshot imaging device that can simultaneously acquire spatial and spectral information in a single exposure. Its core idea is that since the probe light has been broadened into a chirped pulse by the dispersive medium 9, different wavelengths correspond to different times. The distribution of different spectral channels in the two-dimensional image recorded by the hyperspectral resolution camera 14 essentially reflects the X-ray evolution state at different times. Therefore, by recording a two-dimensional image containing multispectral information in a single exposure, the spatiotemporal evolution data of the entire ultrafast process can be completely captured without relying on high-speed electronic scanning, breaking through the time resolution limit of traditional detectors.

[0046] Furthermore, the imaging system 13 preferably employs an image magnification system (i.e., a 4F imaging system) composed of a first lens 13-1 and a second lens 13-2, and the magnification can be adjusted according to the target area and the size of the camera's sensitive surface (e.g., 2x, 5x, 10x). Therefore, the probe light carrying information on carrier concentration changes passes through the first semi-transparent mirror 11 and is then imaged onto the hyperspectral resolution camera 14 via the first lens 13-1 and the second lens 13-2.

[0047] Furthermore, in order to isolate stray light and reduce its impact on the imaging recording module, the first semi-transparent mirror 11, the semiconductor response medium 12, the imaging system 13 and the hyperspectral resolution camera 14 are all housed in the light shield 15. Only X-rays are allowed to irradiate the semiconductor response medium 12 and the probe light is allowed to strike the first semi-transparent mirror 11. Other stray light will not enter the subsequent optical path.

[0048] In this embodiment, the calibration optical path includes a spectrometer 17, a streak camera 18, and a second semi-transparent mirror 16 disposed between the dispersive medium 9 and the beam expander 10. The probe light emitted from the dispersive medium 9 is split into two after hitting the second semi-transparent mirror 16. One beam of probe light that passes through the second semi-transparent mirror 16 is hit by the beam expander 10, and the other beam of probe light reflected by the second semi-transparent mirror 16 is hit by the spectrometer 17 and then by the streak camera 18.

[0049] Specifically, in order to accurately convert the spectral information recorded by the hyperspectral resolution camera 14 into temporal information, the spectral-temporal mapping relationship of the chirped pulse must be calibrated beforehand. The calibration optical path is designed for this purpose. The second semi-transparent mirror 16 is placed in the optical path between the dispersive medium 9 and the beam expander 10, and proportionally splits the chirped pulse probe light emitted after being broadened by the dispersive medium 9. One beam of light passes through the second semi-transparent mirror 16 and continues to be directed along the main optical path to the beam expander 10 for subsequent detection and imaging; while the other beam of light is reflected by the second semi-transparent mirror 16 and directed to the spectrometer 17 and the streak camera 18. The spectrometer 17 is responsible for collecting the spectral component distribution of the chirped pulse, while the streak camera 18 simultaneously measures the temporal information corresponding to each spectral component. The combination of the two allows for the accurate calibration and recording of the spectral-temporal mapping relationship data, providing an indispensable reference support for subsequent algorithms to reverse-engineer the spectral dimension into the temporal dimension. It should be understood that although the calibration optical path uses the second semi-transparent and semi-reflective mirror 16 for beam splitting in this embodiment, other beam splitting devices such as polarizing beam splitters can also be used in other embodiments, as long as a portion of the probe light can be stably directed to the spectrometer and streak camera.

[0050] Example 2:

[0051] A high spatiotemporal resolution detection method based on the high spatiotemporal resolution detection device of Embodiment 1 is performed according to the following steps:

[0052] S1. The probe laser emitted from the probe laser 6 is adjusted in intensity by the adjustable attenuator 7. The spectrum is then stretched in time by the dispersive medium 9 and converted into a chirped pulse as the probe light. The probe light is guided to the spectrometer 17 by the second semi-transparent and semi-reflective mirror 16 and finally recorded by the streak camera 18. The mapping relationship between the spectrum and time is then calibrated.

[0053] Specifically, the core of this step lies in establishing a precise mapping reference between the spectral and temporal dimensions of the chirped pulse. This is a prerequisite for subsequently reversing the spectral image into a temporal evolution sequence. The ultrashort pulse emitted from the probe laser 6 is adjusted to a suitable intensity by the adjustable attenuator 7 before entering the dispersive medium 9. The dispersive medium 9 utilizes its material dispersion effect to sequentially separate the spectral components of different wavelengths along the time axis, forming a long-pulse chirped probe beam. Subsequently, this chirped pulse reaches the second semi-transparent mirror 16 and is split, with a portion reflected and directed to the calibration optical path. The spectrometer 17 is responsible for accurately acquiring the intensity distribution of each wavelength component in the chirped pulse, while the streak camera 18 simultaneously records the arrival time information of these wavelength components. By comparing the spectral data acquired by the spectrometer 17 with the time data recorded by the streak camera 18 point by point, the mapping relationship between spectrum and time can be accurately calibrated and established, clarifying which wavelength corresponds to which moment, providing an absolute time reference coordinate for subsequent time inversion.

[0054] S2. Using a spectrally calibrated tungsten halogen lamp as the calibration light source, the light is irradiated onto the diffuse scattering plate to produce uniformly distributed light. After filtering out unwanted spectra and adjusting the light intensity, the spectrum of the light is recorded using a hyperspectral resolution camera 14 to establish the mapping relationship between gray values ​​and spectral response.

[0055] Specifically, this step aims to eliminate the inconsistencies in spectral response among pixels of the hyperspectral resolution camera 14 caused by differences in manufacturing processes, ensuring that the grayscale of subsequently acquired images accurately reflects the spectral intensity of the incident light. A spectrally calibrated tungsten halogen lamp is used as a standard calibration light source with a known spectral distribution. The light emitted from this lamp illuminates a diffuser plate, which converts point light sources or beams with a certain divergence angle into spatially uniform diffuse reflected light, ensuring that the entire sensing surface of the hyperspectral resolution camera 14 receives uniform illumination. In the optical path, unwanted stray spectra are filtered out by setting filters, and the luminous intensity of the tungsten halogen lamp is adjusted or an attenuator is added to ensure that the incident light intensity is precisely within the linear response range of the hyperspectral resolution camera 14, avoiding overexposure or underexposure. The hyperspectral resolution camera 14 records the exposure of this uniform calibration light, extracts the grayscale value of each pixel, and compares it with the known standard spectral distribution to establish a mapping relationship between the grayscale value of each pixel and its corresponding spectral response characteristics, thus completing the spectral calibration of the camera.

[0056] S3. The target is placed in the set position in the vacuum target chamber 3. The synchronous control system 1 controls the target laser 2 and the probe laser 6 to emit the target laser and probe laser respectively according to the set relative delay. The X-rays generated by the target laser are focused on the target and the probe laser is converted into chirped probe light through the dispersive medium 9. They arrive at the front surface and the back surface of the semiconductor response medium 12 at the same time. That is, when the X-rays arrive at the front surface of the semiconductor response medium 12, the probe light arrives at the back surface of the semiconductor response medium 12. The X-rays and the probe light interact within the semiconductor response medium 12. That is, the spatiotemporal evolution of the X-rays is converted into a change in the carrier concentration in the semiconductor response medium 12. The probe light reflected back from the back surface of the semiconductor response medium 12 carries information about the change in carrier concentration.

[0057] Specifically, this step is the most crucial step in the entire detection method in terms of timing coordination. Its physical essence is to ensure the precise spatial and temporal encounter between the X-ray event and the probe light readout. After the target 4 is installed at the designated position in the vacuum target chamber 3, the synchronous control system 1 sends trigger signals to the target laser 2 and the probe laser 6 according to the preset relative delay parameters. Since it takes a certain amount of time for the target laser to excite the target 4 to generate X-rays and propagate to the front surface of the semiconductor response medium 12, and it also takes a certain amount of time for the probe laser to broaden through the dispersive medium 9 and propagate to the rear surface of the semiconductor response medium 12, the physical paths and propagation speeds of the two are different. Therefore, precise delay compensation must be performed by the synchronous control system 1. Only when the delay is set precisely can it be ensured that the X-rays arrive at the front surface of the semiconductor response medium 12 at the same instant, and the chirped pulse probe light also arrives at the rear surface of the semiconductor response medium 12 at the same instant. At this moment, the transient change in carrier concentration excited by X-rays on the front surface and the probe light of a specific wavelength corresponding to a specific moment incident on the rear surface meet and interact in the medium, so that the reflected probe light is modulated at the corresponding spatial position, thereby recording the ultrafast spatiotemporal evolution information of X-rays into the spectrum and intensity changes of the probe light in real time.

[0058] S4. The probe light reflecting back from the back surface of the semiconductor response medium 12, carrying information on the change in carrier concentration, passes through the first semi-transparent mirror 11 and is imaged onto the hyperspectral resolution camera 14 by the imaging system 13. The hyperspectral resolution camera 14 records a snapshot spectral image containing multispectral information in a single exposure.

[0059] Specifically, this step achieves a single capture conversion from dynamic physical processes to static two-dimensional image data. The reflected probe light carrying information on carrier concentration changes returns along its original path, passing through the first semi-transparent mirror 11 and entering the imaging system 13. The imaging system 13 clearly images the active area of ​​the semiconductor response medium 12 onto the sensing surface of the hyperspectral resolution camera 14 at a specific magnification. Since the probe light has been broadened into a chirped pulse by the dispersive medium 9, its different wavelength components correspond to different time points. The hyperspectral resolution camera 14 has the ability to resolve different wavelength spectra at the pixel level. Therefore, when this probe light containing time-series information is projected onto the camera's sensing surface, the camera can record all spectral channel information at all spatial locations in a single exposure within a very short time, forming a two-dimensional snapshot spectral image containing multispectral information. This physical process completely eliminates the time bottleneck of traditional electronic detectors that require frame-by-frame scanning and reading, achieving complete freezing and capture of irreversible ultrafast processes.

[0060] S5. First, the snapshot spectral image recorded by the hyperspectral resolution camera 14 is processed to obtain a two-dimensional multispectral image. Then, the two-dimensional multispectral image is processed to obtain an image sequence of X-ray evolution over time.

[0061] Specifically, this step represents the macroscopic transformation concept of the entire detection method. Its core logic lies in reversibly expanding the static spectral dimension information obtained in step S4 into dynamic temporal dimension information based on the spectral-temporal mapping relationship defined in step S1. Since different spectral channels in the snapshot spectral image essentially record the X-ray state at different times, by processing the snapshot spectral image and converting the spectral dimension distribution into a temporal dimension distribution according to the spectral-temporal mapping relationship, multiple two-dimensional spatial images representing the X-ray state at different times can be extracted and reconstructed from this static two-dimensional spectral image, thus forming an image sequence of X-ray evolution over time. It should be understood that this only describes the macroscopic transformation concept from spectral image to temporal evolution sequence. Details regarding specific band separation, interpolation reconstruction algorithms, and neural network optimization involved in data processing will be further elaborated in subsequent embodiments, and this embodiment does not limit this aspect.

[0062] Through the steps S1 to S5 described above, this embodiment establishes a rigorous method flow and timing logic. From the establishment of calibration benchmarks, camera response calibration, and precise synchronization control of spatiotemporal encounters, to snapshot capture of a single exposure, and then to the macroscopic inversion conversion from spectrum to time, each step is interconnected and jointly supports the core technical concept of acquiring high spatiotemporal resolution evolution data of irreversible ultrafast physical processes under a single exposure.

[0063] Example 3:

[0064] Based on Example 3, this example further details the underlying hardware structure of the hyperspectral resolution camera 14 and the preliminary reconstruction process in step S5, providing specific physical structure and algorithm steps to support the functional limitations.

[0065] The hyperspectral resolution camera 14 is a snapshot-type spectral imaging camera. The snapshot spectral image recorded by the hyperspectral resolution camera 14 has a pixel size of 2040×2040, consisting of 340×340 spectral imaging units. Each unit contains 6×6 pixels, and the size of a single pixel is 9μm×9μm. Each pixel surface forms a 6×6 Fabry-Perot cavity structure. Specifically, the hyperspectral resolution camera 14, as the core hardware of this invention for recording multispectral information in a single exposure, adopts a design that combines a large-area pixel arrangement with micro-nano optical structures. The entire sensing surface consists of a large area array of 2040×2040 pixels to ensure that it can accommodate sufficiently rich spatial detail information. These pixels are logically and physically divided into 340×340 spectral imaging units, and each spectral imaging unit is a microarray consisting of 6×6, or 36 pixels. The size of a single pixel is 9μm×9μm, which ensures both sufficient photoelectric signal collection area to improve the signal-to-noise ratio and maintains a high spatial sampling density. Most importantly, a 6×6 Fabry-Perot cavity (FP cavity) structure is stacked on the surface of each pixel using ultra-precision interference coating and photolithography. It should be understood that although this embodiment illustrates a 6×6 FP cavity arrangement, in other implementations, depending on the required number of capture frames and spectral channels, other array sizes such as 4×4 or 8×8 can be used, as long as the requirement to acquire multispectral channel information in a single exposure is met.

[0066] The physical principle behind pixel-level spectral selection achieved by the Fabry-Perot cavity structure lies in the fact that a Fabry-Perot cavity is essentially an optical resonant cavity composed of two parallel high-reflectivity films and an intermediate dielectric layer. When incident light enters the cavity, the light wave undergoes multiple reflections and interferences between the two films. Only light with wavelengths that satisfy the resonance condition (i.e., the cavity length is exactly an integer multiple of half the wavelength) can form a standing wave within the cavity and ultimately be transmitted and received by the photoelectric sensor at the bottom of the pixel. Other wavelengths that do not satisfy the resonance condition are reflected or absorbed. By fabricating Fabry-Perot cavities with different cavity lengths or dielectric layer thicknesses on the 36 pixels of the same spectral imaging unit, these 36 pixels each have the highest transmittance response to 36 different center wavelengths, thus achieving pixel-level spectral selection and dispersion at the physical level. This design cleverly avoids the slits or prisms required by traditional spectrometers, allowing two-dimensional spatial information and one-dimensional spectral information to be simultaneously compressed and recorded on a single two-dimensional image in a single exposure, providing the original data foundation for subsequent inference of the time dimension from the spectral dimension.

[0067] Combination Figure 2In step S5, firstly, after band separation, the snapshot spectral image is separated into 36 sparse spectral images. A convolutional filter dependent on the probability of spectral bands in the snapshot spectral image is designed. By interpolating and reconstructing the image using the weighted values ​​of neighboring known pixels and normalizing the results, a preliminary complete multispectral image is generated. Specifically, after the hyperspectral camera 14 completes a single exposure, the output snapshot spectral image is a two-dimensional grayscale image with a size of 2040×2040, which interweaves spatial and spectral information. Since the 36 pixels in each spectral imaging unit correspond to 36 different spectral channels, during the band separation process, the algorithm extracts pixels corresponding to the same spectral channel one by one from the entire snapshot spectral image according to the 6×6 periodic arrangement of the spectral imaging units and rearranges them. Since each spectral channel occupies only 1 / 36 of the pixel position in the original image, each extracted single-spectral channel image is spatially discontinuous, with a large number of missing pixel holes, hence the name 36 sparse spectral images.

[0068] Example 4:

[0069] Based on Example 3, this example further describes the symmetric design and interpolation rules of the convolution filter in step S5 in a sinking defense manner. By explaining the physical and mathematical meaning of each parameter in the formula and establishing strict interpolation boundary conditions, the core algorithm is prevented from being easily replaced or circumvented.

[0070] Since the spectral image is arranged in a repeating pattern in both the horizontal and vertical directions, the convolution filter is designed symmetrically, and its expression is:

[0071] ;

[0072] In the above formula, and These represent the coordinates of the pixels. Indicates the coordinate position as pixel values, Indicates the coordinate position as The pixel value.

[0073] Specifically, this symmetrical design expression has profound mathematical and physical significance. Mathematically, H(x,y)=H(-x,-y) indicates that the function is evenly symmetric about the origin, meaning that the system's response value at a spatial coordinate is exactly equal to its response value at a centrally symmetric position. Physically, this spatial symmetry characteristic stems from the 6×6 Fabry-Perot cavity filter combination of the spectral imaging unit in the snapshot spectral imaging module of the hyperspectral resolution camera 14, which is periodically repeated in both the horizontal and vertical directions. Due to the periodicity and symmetry of this hardware arrangement, the point spread function or system response function of the imaging recording module must also exhibit a symmetrical distribution in space. Therefore, designing the convolution filter to match its response characteristics with the symmetrical arrangement of the physical hardware ensures that the weight allocation in each spatial direction is fair and conforms to the actual light field distribution during interpolation reconstruction, avoiding orientational deviations or artifacts introduced by filter asymmetry. It should be understood that although this embodiment illustrates the mechanism of even-symmetric design based on a 6×6 FP cavity arrangement, in other embodiments, if the spectral imaging unit adopts an asymmetric arrangement logic, the symmetry design of the convolution filter can also be adjusted to a form with other symmetric characteristics, as long as the spatial response of the filter is consistent with the physical arrangement characteristics of the hardware.

[0074] When performing interpolation, pixels that are farther away are assigned smaller weights, that is: when At that time, This distance weighting logic is the core of the interpolation distortion prevention mechanism. In the formula... and Let represent the spatial Euclidean distances from two different pixel locations to the target location to be interpolated. This logic explicitly stipulates that pixels farther away have smaller weights, while pixels closer to each other have larger weights. The physical and mathematical basis for this setting is that, in the fields of optical imaging and spatial signal propagation, the correlation and continuity between neighboring pixels are usually much stronger than those between distant pixels. The true spectral intensity of the location to be interpolated is most affected by its spatially closest known pixel. However, as the spatial distance increases, due to medium scattering, aberrations, and the locality of physical processes, the statistical correlation between distant pixels and the true value of the location to be interpolated rapidly decreases. Therefore, assigning higher weights to nearby pixels can maximize the use of local continuity constraints to approximate the true value; while limiting the weights of distant pixels can effectively suppress interference from irrelevant noise or discontinuous textures from distant locations, thereby significantly reducing interpolation errors and preventing image edge blurring and high-frequency detail distortion.

[0075] During interpolation, the positions of known pixel values ​​are not modified. This principle is the bottom-line constraint for preserving the authenticity of the original data. The known pixel values ​​directly acquired by the hyperspectral resolution camera 14 in a single exposure are the only physically measured data that have not been estimated by any algorithm, representing the most authentic spectral response intensity at that spatial location. Modifying or smoothing these known pixel values ​​during interpolation may make the entire image appear more uniform, but it actually undermines the objectivity and accuracy of the original measurement data, introducing uncontrollable algorithmic errors, which is unacceptable in laser fusion diagnostics that require high-precision spatiotemporal inversion. Therefore, adhering to the principle of not modifying known pixels ensures that the reference anchor points of the reconstructed image are absolutely reliable, and any interpolation and reconstruction operations can only be performed in unknown regions, thus preserving the hard data skeleton of the original spectral image.

[0076] When interpolating a pixel value at any location within any spectral band, the result is obtained from at least two known pixel values ​​in the vicinity of that location. Specifically, this boundary condition sets the minimum data support scale for the interpolation operation. In two-dimensional space, interpolation using only one known pixel cannot determine the direction, leading to significant randomness and directional ambiguity. However, relying on at least two known pixel values ​​allows the establishment of a preliminary gradient and directional constraint on the two-dimensional plane, ensuring that the interpolation result is no longer a random extension determined by a single point, but rather a reasonable estimate constrained by the combined spatial distribution trends of multiple points. This method of obtaining interpolation from at least two nearby known pixel values ​​mitigates distortion by offsetting the amplification effect of single-point errors through multi-point constraints. When the local trends of these two or more known pixels are consistent, the interpolation result smoothly transitions with the trend; when the trends of the known pixels differ, the interpolation algorithm uses distance weights for compromise and reconciliation, avoiding drastic jumps or isolated noise points in the interpolation region. It should be understood that although this embodiment emphasizes the bottom line requirement of at least 2 known pixels, in actual algorithm execution, in order to obtain a smoother and more accurate reconstruction effect, weighted interpolation calculation is usually performed using as many known pixels as possible around the location to be interpolated (e.g., all known points in a 6×6 neighborhood). As long as the number of known pixels involved in the calculation is not less than 2, the best balance between computational efficiency and distortion prevention can be achieved.

[0077] Then, based on the convolutional filter and the convolution-based weighted interpolation method, a preliminary complete multispectral image is generated.

[0078] Finally, a neural network is used to reconstruct the preliminary complete multispectral image to obtain a hyperspectral image that is close to the real spectral image.

[0079] The residual module concatenates all preceding features and passes them to each subsequent layer; its expression is as follows:

[0080] ;

[0081] In the above formula, Indicates the first Cascaded features of layer output, Indicates from 0 to Feature maps generated by the layer This refers to cascading. Specifically, in traditional convolutional neural networks, as the network depth increases, each layer only receives the output of the previous layer as input. This serial transmission mechanism easily leads to the gradual attenuation or even loss of low-frequency basic features extracted from shallow layers during the layer-by-layer nonlinear transformation. Therefore, although deep networks possess stronger expressive power, they often struggle to effectively utilize early basic information. This invention employs a cascading operation... This breaks down the serial barrier; its dimensional splicing mechanism lies in: connecting the network from the initial layer 0 to the current layer... All feature maps generated by the layer The features are spliced ​​and fused along the channel dimension to form an ultra-high-dimensional feature matrix containing all historical feature information, which serves as the first feature matrix. The core of this feature transfer and loss prevention principle lies in the fact that each layer can directly access and utilize the multi-scale features extracted from all previous layers, ranging from coarse-grained to fine-grained. This ensures that the basic spatial structure and spectral contour information of shallow layers are not submerged during the deep abstraction process, thus providing extremely rich multi-dimensional data support for subsequent high-resolution reconstruction. It should be understood that although this embodiment describes a cascading method that stitches along the channel dimension, other cascading strategies such as feature addition or weighted fusion can also be used in other implementations, as long as it ensures that the preceding features can be transferred to the subsequent deep network without loss or with low loss.

[0082] The final output of the residual module Represented as:

[0083] ;

[0084] In the above formula, Represents module functions, This represents a residual connection. This expression reveals the core mechanism by which residual connections solve the vanishing gradient problem in deep networks. In extremely deep neural networks, the gradient needs to undergo multiple multiplication operations during error backpropagation. If the absolute value of the derivative of each layer is less than 1, the gradient will rapidly decay to zero, causing the parameters of shallower network layers closer to the input to be almost impossible to update, i.e., the vanishing gradient phenomenon. This invention introduces residual connections... , the module's input The gradient is directly transferred to the output via a jump, and added to the features after multiple nonlinear transformations. This allows the gradient to be directly and losslessly propagated back to the shallow layers during backpropagation via this short jump path, essentially opening a high-speed bypass channel for gradient flow that is unaffected by multi-layer decay. (Module function) Typically, convolutional operations and activation functions are used to extract incremental features from the residuals, while residual connections ensure the identity mapping of the underlying information. The combination of these two allows the network to learn only the differences between the input and output, significantly reducing the optimization difficulty and making it possible to build ultra-deep reconstruction networks to extract extremely fine spectral textures.

[0085] To further clarify the physical architecture of the neural network, this embodiment supplements the specific network topology: The entire reconstruction network uses a convolutional layer as the front-end feature extraction entry point, responsible for mapping the preliminary complete multispectral image to a high-dimensional feature space; then, three cascaded residual modules are connected in series. As mentioned above, these three residual modules not only contain residual skip connections internally, but also pass historical features to each other through cascading operations, forming a dense feature reuse chain, ensuring the completeness of deep feature extraction; after the three residual modules, two deconvolutional layers are set to progressively upsample the high-dimensional abstract feature map to restore it to the original spatial resolution size; each deconvolutional layer is followed by a modified linear unit activation function, introducing nonlinear correction capability, accelerating network convergence and enhancing the sparse representation of features; finally, the output layer outputs the reconstructed hyperspectral image. It should be understood that although this embodiment lists three cascaded residual modules and a specific number of convolutional and deconvolutional layers, in practical applications, the number of residual modules can be increased to four, five, or more, depending on the degree of blurring in the initial multispectral image and the reconstruction quality requirements. The upsampling rate of the deconvolutional layers can also be flexibly adjusted, as long as the network as a whole can achieve a nonlinear mapping from a low-resolution interpolated image to a high-resolution true spectral image. Through this deep topological architecture of cascaded and residual interweaving, the neural network can mine and restore high-frequency details beyond the interpolation limit from the initial complete multispectral image, obtaining a hyperspectral image close to the original spectral image, providing the most reliable data source for the final accurate spatiotemporal transformation.

[0086] Based on the spectral and temporal calibration data of the chirped pulses, the 36 spectral images were converted into a sequence of X-ray images at 36 time points, with time intervals of approximately 0.5 ps to 5 ps (depending on the length of the dispersive medium).

[0087] Example 5:

[0088] Within the framework of step S5 in Embodiment 2, this embodiment further elaborates on the mathematical modeling and spatiotemporal transformation operators of snapshot spectral images, providing the most basic mathematical support for image processing and establishing a rigorous mapping logic from the spectral dimension to the time dimension.

[0089] First, the snapshot spectral image Represented as:

[0090] ;

[0091] This mathematical model is a precise abstraction of the physical process of single-exposure recording by a hyperspectral resolution camera. In the above formula, It represents the original light intensity distribution, which reflects the true two-dimensional spatial intensity characteristics of X-rays when they excite charge carriers on the front surface of the semiconductor response medium 12. It is the initial physical quantity without any spectral modulation or coordinate transformation. This represents the spectral selection operator, whose physical and mathematical significance lies in realizing the transformation from two-dimensional data to a three-dimensional matrix. Specifically, it represents the original light intensity. While spectral density is merely a function of two-dimensional spatial coordinates, the hyperspectral camera 14, through the Fabry-Perot cavity structure on its pixel surface, introduces a third dimension of spectral information into the two-dimensional spatial plane, allowing light intensity at different spatial locations to be allocated to different spectral channels. (Spectral selection operator) It is the mathematical expression describing this allocation mechanism that discretizes and maps the originally continuous two-dimensional spatial light intensity into a three-dimensional data cube containing both space and spectrum, according to the periodic spectral sampling rules of the camera hardware. This represents the correction coefficient operator, which is used to correct spectral differences between different sensors. Due to manufacturing limitations, the Fabry-Perot cavities corresponding to each pixel of the hyperspectral resolution camera may have slight shifts in center wavelength or inconsistencies in transmittance on the actual response curves. Without correction, directly extracted spectral data will introduce systematic errors. The operator calculates a set of compensation coefficients by comparing the measured response characteristics of each pixel with the standard ideal response, ensuring that the final extracted spectral intensity can truly restore the original energy distribution of the incident light. and These represent the spectral selection operators respectively. and correction coefficient operator The calculated new coordinates This represents the wavelength. It should be understood that the coordinate transformation logic here reflects the spatial remapping of the image after spectral sampling and rearrangement. Due to the periodic extraction of spectral channels, the spatial continuity of the original image is broken and reassembled, resulting in new coordinates. and It is defined precisely to describe this repositioning in the spectral-spatial joint domain.

[0092] Furthermore, the correction coefficient operator The expression is:

[0093] ;

[0094] This summation logic reveals the macroscopic statistical mechanism of spectral correction. In the above equation, This indicates the number of spectra, for example, in the aforementioned 6×6 Fabry-Perot cavity structure. That is, 36 spectral channels; This represents the correction coefficients for different spectra. This expression indicates that, for a specific spectral band... The final correction coefficient is not determined by a single pixel, but is obtained by summing and aggregating the correction coefficients corresponding to that band at all spatial locations within the imaging area. The advantage of this summation logic is that it utilizes the statistical averaging characteristics of the entire array space to smooth out random manufacturing errors of local pixels, thereby obtaining a more robust and accurate global spectral correction benchmark and effectively suppressing the interference of single-point anomalous responses on the overall spectral reconstruction.

[0095] Then, the snapshot spectral image Converted into an image sequence of X-rays evolving over time Its expression is:

[0096] ;

[0097] This expression completes the most crucial mathematical leap of this invention, from the spectral dimension to the temporal dimension. In the above formula, Indicates time, Operators for converting spectra to time. Conversion Operators It is not a simple constant matrix; its physical meaning is determined by the spectral-time mapping characteristics of the chirped pulse. As emphasized in the previous embodiments, after the probe light is broadened by the dispersive medium 9, different wavelength components are precisely separated on the time axis, forming a specific spectral-time correspondence. Operator This mathematical encapsulation of the physical mapping relationship precisely measured in step S1 through optical path calibration defines which specific time slice on the time axis corresponds to each wavelength slice in the three-dimensional spectral data cube. Furthermore, the operator... The process also includes the mechanism of intensity normalization. Because the intensity distribution of different wavelength components is not uniform during the broadening process of the chirped pulse, and the spectrum output by the detector laser 6 may also have intensity fluctuations, without normalization, the directly converted time-series image will not accurately reflect the proportional relationship of X-ray intensity evolution over time. Therefore, While performing coordinate mapping, the operator also corrects the light intensity ratio of the image data for each time slice based on the calibrated spectral intensity distribution curve, eliminating time intensity distortion caused by the spectral inhomogeneity of the probe light source itself. Through the operator... The function of this method is to precisely solve and project the static snapshot image, which was originally unfolded in the spectral dimension, onto the temporal dimension, thereby outputting a dynamic image sequence that can realistically reproduce the ultrafast spatiotemporal evolution of X-rays. It should be understood that although this embodiment provides specific mathematical expressions and operator definitions, in actual programming implementation, operators... The specific form may be a lookup table, interpolation function or matrix transformation, etc. As long as its core logic follows the spectral-time mapping and light intensity normalization rules of chirped pulses, the spatiotemporal conversion purpose of this invention can be achieved.

[0098] Example 6:

[0099] To more clearly demonstrate the operational effectiveness and irreplaceability of the technical solution of this invention in real extreme environments, this embodiment applies the aforementioned high spatiotemporal resolution detection device and method to a specific scenario of black cavity plasma X-ray spatiotemporal diagnosis in laser fusion experiments for detailed explanation. It should be understood that although this embodiment uses laser fusion black cavity diagnosis as an example, the core concept of this invention is also applicable to the diagnosis of other ultrafast processes in high-energy-density physics research. The embodiment is merely illustrative and not restrictive.

[0100] In this laser fusion experiment scenario, specific operating parameters are introduced: the high-energy laser emitted from the target laser 2 is focused on the target 4 in the vacuum target chamber 3, instantly generating high-temperature, high-density plasma and radiating X-rays; the probe laser 6 is a femtosecond laser with an output center wavelength of 800nm, a spectral coverage of 40nm, and a pulse width of 35fs; the dispersion medium 9 is a high-dispersion photonic crystal fiber with a length of about 4m, which broadens the femtosecond pulse in time to a chirped pulse of about 50ps as the probe light.

[0101] The entire diagnostic process was strictly executed according to the steps of the aforementioned high spatiotemporal resolution detection method. In step S1, the femtosecond pulse emitted by the probe laser 6 was adjusted in intensity by the adjustable attenuator 7 and then passed through a high dispersion photonic crystal fiber of about 4m. The spectrum was stretched in time and converted into a chirped pulse of about 50ps. After being split by the second semi-transparent and semi-reflective mirror 16, a portion of the chirped pulse was directed to the spectrometer 17 and the streak camera 18 to accurately determine the mapping relationship between each wavelength component in the 40nm spectral range and the time within the 50ps time window.

[0102] In step S2, a spectrally calibrated tungsten halogen lamp is used as a calibration light source to calibrate the mapping relationship between grayscale values ​​and spectral response of the hyperspectral resolution camera 14, thereby eliminating the response differences of each pixel in the camera array.

[0103] In step S3, after the target 4 is placed in the vacuum target chamber 3, the synchronous control system 1 controls the target laser 2 and the probe laser 6 to emit the target laser and probe laser respectively according to the set relative delay. At this time, the timing coordination mechanism of the synchronous control system 1 is crucial. It ensures that the target laser is focused on the X-rays generated by the target 4, and that the approximately 50 ps chirped pulse probe light, after passing through a high-dispersion photonic crystal fiber of about 4 m, arrives synchronously at the front and rear surfaces of the semiconductor response medium 12, respectively. The X-rays excite carriers on the front surface, while the chirped probe light corresponding to a specific wavelength at a specific moment enters from the rear surface and interacts with the change in carrier concentration inside the medium, so that the reflected probe light carries the spatiotemporal evolution information of the X-rays.

[0104] In step S4, the probe light carrying information on carrier concentration changes, reflected from the back surface of the semiconductor response medium 12, passes through the first semi-transparent mirror 11 and is then imaged onto the hyperspectral resolution camera 14 by the imaging system 13. The hyperspectral resolution camera 14 records a snapshot spectral image containing multispectral information in a single exposure. Since the camera has a pixel size of 2040×2040 and consists of 340×340 spectral imaging units, each unit containing 6×6 pixels and forming a 6×6 Fabry-Perot cavity structure on its surface, this snapshot spectral image essentially freezes the X-ray spatial state at all times within a 50ps time window.

[0105] In step S5, the snapshot spectral image recorded by the hyperspectral resolution camera 14 is first processed. After band separation, the snapshot spectral image is separated into 36 sparse spectra. A convolution filter dependent on the spectral band probability of 1 / 36 is designed using a symmetric design expression. Based on the interpolation rule that pixels with greater distances have smaller weights, a preliminary complete multispectral image is generated by interpolating and normalizing the weighted values ​​of neighboring known pixels. Subsequently, a neural network is used to reconstruct the preliminary complete multispectral image. The residual module concatenates all the previous features and passes them to each subsequent layer, expressed as follows: The final output of the residual module This yields a hyperspectral image that closely resembles the original spectral image. Finally, the snapshot spectral image is... Represented as and using conversion operators Convert snapshot spectral images into a sequence of X-ray images that evolve over time. .

[0106] Because the dispersive medium 9 broadens the spectrum to a time window of approximately 50 ps and covers a spectral range of 40 nm, combined with the resolution capability of the 36 spectral channels of the hyperspectral camera 14, the algorithm ultimately reconstructs a sequence of 36 X-ray evolution images with time intervals of approximately 1.4 ps. The derivation logic for this time interval is as follows: dividing the total time window of 50 ps by 36 spectral channels yields a time resolution of approximately 1.4 ps, thus clearly reproducing the entire ultrafast evolution dynamics of black cavity plasma X-rays with extremely high time resolution.

[0107] This embodiment fully demonstrates the irreplaceable nature and commercial value of the device and method of this invention in the extreme application scenario of laser fusion. At the laser fusion experimental site, there is extremely strong neutron radiation, gamma rays, and intense electromagnetic noise interference. Traditional electronic-based diagnostic systems, such as streak cameras or framing cameras, are highly susceptible to such interference, leading to distorted measurement data or even component damage. They typically require additional, heavy, and complex protective devices, increasing system cost and complexity, and the protective effect is difficult to guarantee completely. In contrast, the high spatiotemporal resolution detection device based on hyperspectral imaging used in this invention is essentially a non-electronic optical detection method. Its core components are a hyperspectral resolution camera 14 and various optical lenses and media. These optical elements have natural immunity to neutrons, gamma rays, and electromagnetic noise, allowing for stable operation in extremely harsh environments without the need for complex anti-interference protection devices. More importantly, laser fusion target-hitting experiments are irreversible, single-shot ultrafast physical processes that cannot be repeatedly triggered to accumulate data. This invention, through the spectral-time mapping of chirped pulses and single-exposure recording by a hyperspectral camera, can acquire a complete sequence of 36 spatiotemporal evolution images in a single target-hitting operation. This completely overcomes the fatal flaw of traditional pump-probe technology, which can only measure repetitive phenomena. It provides a revolutionary diagnostic tool for laser fusion and high-energy-density physics research that is simple in structure, highly resistant to interference, and has extremely high spatiotemporal resolution.

[0108] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention. Those skilled in the art, under the guidance of the present invention, can make various similar representations without departing from the spirit and claims of the present invention, and such modifications all fall within the protection scope of the present invention.

Claims

1. A high spatiotemporal resolution detection device based on hyperspectral imaging, characterized in that, include: The X-ray response module includes a target laser, a vacuum target chamber, and a semiconductor response medium. An X-ray imaging structure is provided on the vacuum target chamber. The target laser emitted by the target laser is focused on the target in the vacuum target chamber to generate X-rays. The X-rays are imaged on the front surface of the semiconductor response medium by the X-ray imaging structure. The probe light generation module includes a probe laser, an adjustable attenuator, a dispersive medium, a beam expander, and a first semi-transparent and semi-reflective mirror. The probe laser emitted by the probe laser is attenuated by the adjustable attenuator, and the spectrum is stretched in time by the dispersive medium and converted into a chirped pulse as the probe light. The probe light is expanded by the beam expander and then reflected by the first semi-transparent and semi-reflective mirror to the back surface of the semiconductor response medium. A synchronous control system is used to control the relative delay between the target laser emitted by the target laser and the probe laser emitted by the probe laser, so that X-rays and chirped pulses interact within a semiconductor response medium. The imaging recording module includes an imaging system and a hyperspectral resolution camera. The probe light reflecting back from the semiconductor response medium, carrying information on the change in carrier concentration, passes through the first semi-transparent mirror and is then imaged onto the hyperspectral resolution camera by the imaging system, recording a two-dimensional image containing multispectral information in a single exposure. The calibration optical path includes a spectrometer, a streak camera, and a second semi-transparent mirror positioned between a dispersive medium and a beam expander. The probe light emitted from the dispersive medium is split in two after hitting the second semi-transparent mirror. One beam of probe light passes through the second semi-transparent mirror and hits the beam expander, while the other beam of probe light reflected by the second semi-transparent mirror passes through the spectrometer and hits the streak camera.

2. The high spatiotemporal resolution detection device according to claim 1, characterized in that, The semiconductor response medium includes a semiconductor response medium body made of semiconductor material, an antireflective coating deposited on the side of the semiconductor response medium body near the X-ray imaging structure, and a metal reflective layer deposited on the side of the semiconductor response medium body near the first semi-transparent mirror.

3. The high spatiotemporal resolution detection device according to claim 2, characterized in that, The metal reflective layer is made of gold or copper and has a thickness of 10 nm.

4. The high spatiotemporal resolution detection device according to claim 1, characterized in that, The X-ray imaging structure is a small hole opened in the vacuum target chamber or a KB microscope mounted on the vacuum target chamber.

5. The high spatiotemporal resolution detection device according to claim 1, characterized in that, The probe laser is a femtosecond laser with an output spectrum covering a range of 40 nm ± 5 nm. The dispersive medium can stretch the femtosecond pulse in time to 10 ps-200 ps.

6. A high spatiotemporal resolution detection method based on the high spatiotemporal resolution detection device according to any one of claims 1 to 5, characterized in that, Follow these steps: S1. The probe laser emitted from the probe laser has its intensity adjusted by an adjustable attenuator. The spectrum is then stretched in time by the dispersive medium and converted into a chirped pulse as the probe light. The probe light is guided to the spectrometer by the second semi-transparent and semi-reflective mirror and finally recorded by the streak camera. The mapping relationship between the spectrum and time is then calibrated. S2. Using a spectrally calibrated tungsten halogen lamp as the calibration light source, the light is irradiated onto the diffuse scattering plate to produce uniformly distributed light. After filtering out unwanted spectra and adjusting the light intensity, a hyperspectral resolution camera is used to record the light spectrum and establish the mapping relationship between gray values ​​and spectral response. S3. After the target is placed in the vacuum target chamber, the synchronous control system controls the target laser and the probe laser to emit the target laser and the probe laser respectively according to the set relative delay. The X-ray generated by the target laser is focused on the target and the probe laser is converted into chirped pulse probe light through the dispersive medium. They arrive at the front surface and the back surface of the semiconductor response medium synchronously and interact within the semiconductor response medium. S4. The probe light carrying information on the change in carrier concentration reflected from the back surface of the semiconductor response medium passes through the first semi-transparent mirror and is imaged onto the hyperspectral resolution camera by the imaging system. The hyperspectral resolution camera records a snapshot spectral image containing multispectral information in a single exposure. S5. First, the snapshot spectral images recorded by the hyperspectral resolution camera are processed to obtain an image sequence of X-ray evolution over time.

7. The high spatiotemporal resolution detection method according to claim 6, characterized in that, The snapshot spectral image recorded by the hyperspectral resolution camera has a pixel size of 2040×2040 and is composed of 340×340 spectral imaging units. Each unit contains 6×6 pixels, and the size of a single pixel is 9μm×9μm. A 6×6 Fabry-Perot cavity structure is formed on the surface of each pixel. In step S5, firstly, after band separation, the snapshot spectral image is separated into 36 sparse spectral images, and a convolution filter that depends on the probability of spectral bands in the snapshot spectral image is designed. By interpolating and reconstructing the weighted values ​​of neighboring known pixels and normalizing them, a preliminary complete multispectral image is generated. Then, based on the convolutional filter and the convolution-based weighted interpolation method, a preliminary complete multispectral image is generated; Finally, a neural network is used to reconstruct the preliminary complete multispectral image to obtain a hyperspectral image that is close to the original spectral image.

8. The high spatiotemporal resolution detection method according to claim 7, characterized in that, In step S5, the convolution filter adopts a symmetrical design, and its expression is: ； In the above formula, and These represent the coordinates of the pixels. Indicates the coordinate position as pixel values, Indicates the coordinate position as Pixel values; When performing interpolation, set when At that time, ; When performing interpolation, the position of known pixel values ​​is not modified. When interpolating the pixel value at any position in any spectral band, it is obtained from at least two known pixel values ​​near that position.

9. The high spatiotemporal resolution detection method according to claim 8, characterized in that, In step S5, when reconstructing the preliminary complete multispectral image using a neural network, the residual module concatenates all the preceding features and passes them to each subsequent layer, as expressed in the following expression: ； In the above formula, Indicates the first Cascaded features of layer output, Indicates from 0 to Feature maps generated by the layer Indicates cascading; The final output of the residual module Represented as: ； In the above formula, Represents module functions, This indicates a residual connection.

10. The high spatiotemporal resolution detection method according to claim 6, characterized in that, In step S5, firstly, the snapshot spectral image is... Represented as: ； In the above formula, This represents the original light intensity distribution. Indicates the spectral selection operator, Indicates the correction coefficient operator. and These represent the spectral selection operators respectively. and correction coefficient operator The calculated new coordinates Indicates wavelength; Correction coefficient operator The expression is: ； In the above formula, Indicates the number of spectra. Represents the correction coefficients for different spectra; Then, the snapshot spectral image Converted into an image sequence of X-rays evolving over time Its expression is: ； In the above formula, Indicates time, Operators that represent the conversion between spectrum and time.

Images