Single-pixel hyperspectral imaging method and device based on time domain stretching

Abstract

The invention relates to the technical field of hyperspectral imaging, in particular to a single-pixel hyperspectral imaging method and device based on time domain stretching. The method comprises the following steps: broadband signal light passes through a sample to be detected to obtain two-dimensional image information and one-dimensional spectral information, a mask pattern is generated by using a spatial light modulation module, and spatial coding modulation is performed on the signal light carrying the sample information; the broadband signal light subjected to space coding is effectively collected through a multimode optical fiber, different space frequency components are homogenized, then time stretching is conducted through a time stretching module to achieve separation of different wavelengths in the time domain, and time domain information of signal pulses is recorded through a single-pixel detector; time is mapped back to a frequency spectrum through the accurate corresponding relation between the signal light frequency and the time delay amount, and a hyperspectral image is obtained according to the known mask sequence and the signal intensity corresponding to different wavelengths in combination with an association algorithm. According to the technical scheme, high-spectral imaging with high speed, high spectral resolution and high sensitivity can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of hyperspectral imaging, and in particular to a single-pixel hyperspectral imaging method and device based on time domain stretching. Background Art

[0002] Hyperspectral imaging technology can detect and obtain two-dimensional spatial information and one-dimensional spectral information of the target being measured. It is a multi-dimensional information analysis technology that combines imaging technology and spectral technology. Each material has its own unique spectral characteristics in different wavelength regions of the electromagnetic spectrum. Hyperspectral imaging contains detailed spectral responses of target characteristics, has more spectral bands and high spectral resolution, and has the potential to capture spectral differences under different environmental conditions. Therefore, hyperspectral imaging technology has become an indispensable tool for spatial analysis of different chemical components.

[0003] Given the excellent spectral resolution and wide spectral coverage of hyperspectral imaging technology, it covers a wide range of application scenarios, such as: in agricultural research, to retrieve various crop and soil properties, timely and effectively detect early signals such as nutrient deficiency and crop diseases; in remote environmental remote sensing and mineral content mapping, it is usually necessary to draw detailed chemical composition of the earth's crust; in biomedical testing, for pathological examination of tissues and organs, it is necessary to conduct feedback analysis on multiple targets at the same time. Therefore, the development of hyperspectral imaging technology with both fast speed, high spectral resolution and high sensitivity has always been an international research hotspot. Its advantages such as non-contact, non-destructive and detection specificity based on images provide strong support for applications such as precision agriculture, remote sensing and medical diagnosis and analysis.

[0004] However, traditional hyperspectral imaging methods are limited by the spectral scanning method and imaging detection means, and usually have complex optical systems and sophisticated mechanical devices, making it difficult to combine the characteristics of fast speed, high spectral resolution and high sensitivity. Existing dispersive spectral imaging systems rely on gratings or prisms for dispersive spectroscopy, and require scanning devices and array detectors to obtain hyperspectral data cubes. The spectral resolution is limited by the number of grating lines and the pixel scale of the detector, and the imaging speed is also relatively slow. Spectral imaging systems based on narrowband filters obtain spectral information of specific bands through bandpass filters. The number of filters generally determines the number of spectral channels of the system. Fourier transform spectroscopic imagers based on the interference principle usually require focal plane array detectors, and the refresh rate of the spectrometer is inherently affected by the mechanical scanning required to record the interference pattern, which results in a total acquisition time of up to several minutes. Therefore, the development of new spectral methods and imaging means is a very challenging problem for improving the spectral resolution and acquisition speed of spectral imaging systems.

[0005] In recent years, single-pixel imaging technology, as a new type of computational imaging method, uses a spatial light modulation module to generate a series of mask patterns to encode and modulate the light field. The modulated light intensity information is collected by a bucket detector that does not have spatial resolution capabilities. After multiple samplings, the spatial information of the object is reconstructed by the association algorithm. However, the single-pixel imaging architecture used in the field of hyperspectral imaging usually requires spatial dispersion elements or narrow-band filter devices, which limits the spectral resolution and spectral channels and reduces the light flux. Time-stretched spectral detection technology provides a new type of spectral detection method, which uses group velocity dispersion to map spectral information to a time domain waveform, realizes real-time mapping between the signal's time domain information and the spectrum, and uses a single-pixel detector to measure the spectrum at a high rate. There is no need for a spatial time stretch module and mechanical scanning, and it has the advantages of high throughput, wide band, real-time and high resolution. However, in existing technologies such as a system for improving the resolution of spatial imaging optical signals using wavelength division multiplexing technology (CN115494004A) and an ultrafast time-domain stretching imaging device and method (CN110411954A), expanding the time stretching technology to the imaging field also requires the use of a spatial dispersion device, so that different frequency components of the signal light correspond to different spatial coordinates on the target, and the target image is encoded into a temporal broadband pulse, and complete image information at different wavelengths cannot be obtained. Summary of the invention

[0006] The purpose of the present invention is to propose a single-pixel hyperspectral imaging method and device based on time domain stretching. This technical solution can achieve hyperspectral imaging with fast speed, high spectral resolution and high sensitivity, so as to solve the related problems of traditional hyperspectral imaging in spectral scanning mode and imaging detection means.

[0007] To achieve the above-mentioned objectives, in a first aspect, an embodiment of the present disclosure provides a single-pixel hyperspectral imaging method based on time domain stretching, comprising: a wide-band signal light passes through a sample to be tested to obtain two-dimensional image information and one-dimensional spectral information, a mask pattern is generated by a spatial light modulation module, and the signal light carrying the sample information is spatially coded and modulated; the wide-band signal light after spatial encoding is time-stretched by a dispersive medium to achieve separation of different wavelengths in the time domain, and the time domain information of the signal pulse is recorded by a single-pixel detector; the time is mapped back to the spectrum through the precise correspondence between the signal light frequency and the time delay, and a hyperspectral image is obtained according to a known mask sequence and the signal intensities corresponding to different wavelengths in combination with a correlation algorithm.

[0008] Beneficial effects of the basic scheme: This technical scheme detects the spectral dimension of the imaging object by time stretching, which overcomes the difficulties of spatial dispersion elements and narrow-band filter devices in improving spectral resolution. Without the need for spatial slits, the light flux of the hyperspectral imaging system has been significantly improved. Combined with a high-speed data acquisition system, it can achieve sub-wavenumber spectral resolution, more than 1000 spectral channels and high-sensitivity hyperspectral imaging in a wide band.

[0009] This technical solution utilizes a spatial light modulator to spatially encode the signal light field, and combines it with time-stretched spectral detection technology to achieve hyperspectral image data acquisition with only a single-pixel detector, avoiding dependence on high-precision mechanical scanning devices and large-array detection devices. At the same time, it avoids the trade-off between imaging spatial resolution and the number of spectral channels, and increases the stability, robustness and applicability of the hyperspectral imaging system.

[0010] Since the single-pixel imaging architecture used requires multiple iterative sampling, a series of known specific coding masks and the intensity information of different wavelengths corresponding to them are correlated and solved in the subsequent hyperspectral image data processing. This technology can be combined with compressed sensing algorithms or reconstruction methods based on deep learning to significantly reduce the number of measurements and data processing time, and improve the imaging frame rate without affecting the quality of hyperspectral imaging.

[0011] As an implementable preferred solution, the spatial light modulation module generates m specific coding patterns The signal light carrying the sample information is spatially intensity encoded. The process is expressed as:

[0012] in, is the total intensity value of the signal light after being encoded by the i-th mask pattern. The broadband signal light passing through the sample to be tested is expressed as , where r is the spatial position, is the wavelength of the incident light.

[0013] As an implementable preferred solution, an adjustable filter is added before the time stretching module to calibrate the relationship between the signal light frequency and the time delay. By setting different filtering frequencies of the adjustable filter and measuring the corresponding time, the corresponding relationship between the signal light frequency and the time delay is obtained.

[0014] As an implementable preferred solution, the time is mapped back to the spectrum through the precise correspondence between the signal light frequency and the time delay. According to the known mask sequence and the signal intensity corresponding to different wavelengths, the hyperspectral image is obtained by combining the correlation algorithm, including the following: For wavelength There are m intensity values , the target to be measured is at wavelength The two-dimensional image at is , solved by: .

[0015] In the second aspect, the embodiments of the present disclosure also provide a single-pixel hyperspectral imaging device based on time domain stretching, including a light source module, a spatial modulation module, a time stretching module, a data acquisition module and an image processing module; the light source module is used to output a wide-band signal light to illuminate a sample to be tested to obtain two-dimensional image information and one-dimensional spectral information; the spatial modulation module is used to encode and modulate the signal light carrying the sample information; the time stretching module is used to disperse the encoded wide-band signal light in the time domain, and the data acquisition module records the time domain information of the signal pulse; the image processing module is used to map the time back to the spectrum through the precise correspondence between the signal light frequency and the time delay, and obtain a hyperspectral image based on a known mask sequence and the signal intensities corresponding to different wavelengths in combination with a correlation algorithm.

[0016] As an implementable preferred solution, the light source module includes a broadband near-infrared light source that can be extended to the visible light band, the medium-wave and long-wave infrared bands, and the terahertz region.

[0017] As an implementable preferred solution, the spatial modulation module includes a liquid crystal spatial light modulator, a digital micromirror array modulator and a light emitting diode array, and is capable of controlling intensity parameters, phase parameters and modulation speed parameters.

[0018] As an implementable preferred solution, the data acquisition module includes a single-pixel detector and an oscilloscope; the single-pixel detector is used to measure the time-domain stretched pulse containing spatial encoding information; the oscilloscope is used to sample and quantize the collected analog electrical signals and convert them into digital signals.

[0019] As an implementable preferred scheme, the data acquisition module includes an InGaAs detector, a single photon detector and a time-correlated single photon counter; the InGaAs processor is used to detect the light source emitted by the light source module as a start trigger signal; the single photon detector is used to realize single photon detection of the signal light after time stretching, and measure the time domain stretching pulse containing spatial coding information as a stop signal; the time-correlated single photon counter is used to record the time difference between the start signal and the end signal, and the cumulative measurement obtains a statistical histogram.

[0020] As an implementable preferred solution, it also includes a first plano-convex lens and a second plano-convex lens for expanding the wide-band signal light; it also includes a third plano-convex lens and a fourth plano-convex lens for adjusting the size of the light beam carrying the sample information signal light and projecting the image information onto the spatial light modulation module. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the principle of single-pixel hyperspectral imaging method based on time domain stretching.

[0022] Figure 2 Schematic diagram of hyperspectral imaging results based on time domain stretching.

[0023] Figure 3 Schematic diagram of a single-pixel hyperspectral imaging device based on time domain stretching in Example 2.

[0024] Figure 4 Schematic diagram of a single-pixel hyperspectral imaging device based on time domain stretching in Example 3. DETAILED DESCRIPTION

[0025] In order to make the technical solution and advantages of the present application clearer, the technical solution of the present invention will be further described in detail below in conjunction with the accompanying drawings. It will be understood that the specific embodiments described herein are only partial embodiments of the present invention, which are only used to explain the present application, rather than to limit the present application. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered isolated, and they can be combined with each other to achieve better technical effects. The same reference numerals appearing in the drawings of the following embodiments represent the same features or components, which can be applied to different embodiments.

[0026] In addition, unless otherwise defined, the technical terms or scientific terms used in the description of the present invention should have the common meanings understood by those skilled in the art in the art to which the present invention belongs.

[0027] The present invention is further described in detail below in conjunction with the accompanying drawings: Figure numerals: light source module 1, filter 2, first plano-convex lens 3, second plano-convex lens 4, sample to be tested 5, third plano-convex lens 6, fourth plano-convex lens 7, spatial light modulation module 8, fifth plano-convex lens 9, reflecting silver mirror 10, adjustable filter 11, first beam collimator 12, multimode optical fiber 13, time stretching module 14, single-pixel detector 15, high-speed oscilloscope 16, image processing module 17, time-correlated single-photon counter 18, second beam collimator 19, neutral density filter 20, single-photon detector 21, detector 22.

[0028] Embodiment 1 Reference Figure 1 The embodiment of the present disclosure provides a single-pixel hyperspectral imaging method based on time domain stretching, comprising the following steps.

[0029] Step S100: After the broadband signal light passes through the sample to be tested, it carries the two-dimensional spatial intensity information and one-dimensional spectrum information of the sample.

[0030] Step S200, using a spatial light modulation module to generate a series of specific mask patterns, and performing spatial coding modulation on the signal light field carrying sample information. The spatial intensity and phase of the signal light field can be precisely controlled to carry more information; The spatial light modulation module can be placed before or after the sample to be tested, corresponding to the structured light illumination and structured light detection in the single-pixel imaging architecture, and performs spatial coding modulation through a specific mask sequence.

[0031] For spatial coding modulation operation, it is assumed that the broadband signal light passes through the sample to be tested and is expressed as , where r is the spatial position, is the wavelength of the incident light, and the spatial light modulation module generates m specific coding patterns The signal light carrying the sample information is spatially intensity encoded. The above process can be expressed as:

[0032] in, It is the total intensity value of the signal light after being encoded by the i-th mask pattern, which contains the intensity information of different wavelengths.

[0033] In step S300, the spatially encoded signal light is collected through a multimode optical fiber and the different spatial frequency components are homogenized. The signal light is then time-stretched by a time-stretching module to achieve separation of different wavelengths in the time domain. A single-pixel detector is used to measure and record the time-domain stretched pulse containing the spatial coding information. This embodiment does not require spatial dispersion elements and narrowband filter devices.

[0034] The use of multimode optical fiber with a large core diameter can not only improve the coupling efficiency of the wide-band signal light after spatial encoding, but also make the energy distribution of different spatial frequency components more uniform in the multimode optical fiber, so that the encoded information of different wavelengths can be effectively collected by the time stretching module and then time stretched. The multimode optical fiber can also be replaced by a spatial light guide to homogenize the light beam.

[0035] Using optical time stretching to resolve spatially encoded total intensity values An adjustable filter is added before the time stretching module to calibrate the relationship between the signal light frequency and the time delay. By setting different filter frequencies of the adjustable filter and measuring the corresponding time with the data acquisition module, the corresponding relationship between the signal light frequency and the delay can be obtained.

[0036] For example, using a dispersion-shifted fiber with a length of L to perform time-domain stretching, for an incident wavelength , whose frequency is , the group velocity dispersion is The corresponding dispersion parameter D is related to the group velocity and wavelength, that is, , where c is the speed of light, then the delay can be obtained theoretically , is the deviation of the filter frequency from the center frequency of the wide-band signal light. The signal light frequency will correspond to the delay amount one by one. Since the adjustable filter itself will introduce a delay amount a' and the delay amount a introduced by other components in the system, the delay amount at this time is By setting different filter frequencies of the adjustable filter, the data acquisition system measures the corresponding delay τ, and the experimental parameters a and k can be obtained by fitting according to the measured data. Here, k should be consistent with the theoretical value DL. When the delay and frequency relationship calibration is completed, the adjustable filter can be removed.

[0037] Step S400, mapping the time back to the spectrum through the precise correspondence between the signal light frequency and the time delay, and reconstructing the hyperspectral image of the sample in combination with the correlation algorithm according to the known mask sequence and the signal intensity corresponding to different wavelengths.

[0038] The main parameters that determine the spectral resolution of the hyperspectral imaging system include: the dispersion parameter D of the dispersion element, the dispersion propagation distance L, the sampling rate f of the electronic digitizer, and the analog bandwidth B of the detection system (including the photodetector and the electronic digitizer). In an ideal electronic digitizer, B and f are determined by the Nyquist criterion f = 2B, although in reality B <f / 2。

[0039] Number of sampling points for stretched pulses in the time domain , as follows:

[0040] in, is the spectral width of the incident light.

[0041] Therefore, the sampling rate of the electronic digitizer places a limit on the spectral resolution of the system: .

[0042] The analog bandwidth of the detection system limits the system spectral resolution to: .

[0043] In addition, the group velocity dispersion (GVD) of the dispersive element limits the system spectral resolution to:

[0044] GVD can be viewed as the temporal equivalent of the far-field spectral resolution in spatial diffraction, where is the central wavelength of the incident light.

[0045] Finally, combining all the spectral resolution limitations, the spectral resolution limit of hyperspectral images is:

[0046] At this time, the number of spectral channels K of the hyperspectral imaging system can be calculated as:

[0047] A trade-off is made between the spectral width, laser repetition frequency and the dispersion of the time stretching module to avoid overlap between adjacent pulses after stretching. Assuming the repetition frequency of the pulse is R, the relationship should be satisfied: In one embodiment, a quantum correlated photon time stretching method is used to separate the overlap between adjacent pulses through the correlation between photon pairs, so that there is no longer a trade-off between high laser repetition frequency and spectral bandwidth.

[0048] Through time-stretched spectral detection, the time information is mapped back to the spectrum to obtain the intensity information corresponding to different wavelengths. Since the spatial light modulation module generates m specific mask patterns, it is necessary to perform iterative sampling m times. At this time, for the wavelength There are m intensity values , the target to be measured is at wavelength The two-dimensional image at is , can be solved by: .

[0049] Reference Figure 2 , the spectral information corresponding to a single spatial pixel in a hyperspectral image is continuous in time. By mapping the time information and the spectrum, the two-dimensional images corresponding to different wavelengths are solved with the help of the association algorithm, and finally a hyperspectral image based on time domain stretching can be obtained. Obviously, in order to obtain faster or even real-time hyperspectral imaging, it is necessary to significantly reduce the required mask patterns and the number of measurements. Choosing a suitable reconstruction method helps to improve the measurement speed and image quality. Compressed sensing has been proven to be a way to utilize the redundant information of most natural signals or images. By finding the sparse solution of the underdetermined linear system to reconstruct the measured signal, a compressed sampling rate far below the Nyquist sampling frequency can be achieved. Another common reconstruction method is deep learning, which includes two stages: training and reconstruction. In the training stage, the weight coefficients of the network are trained using the data set, and in the reconstruction stage, the measurement domain information is mapped to the image domain, which can also maintain good imaging speed and imaging quality at low sampling rates. Selecting a suitable reconstruction algorithm based on indicators such as running speed, algorithm performance, sampling compatibility, noise robustness, and reconstruction quality can meet different application requirements.

[0050] The disclosed embodiment also provides a single-pixel hyperspectral imaging device based on time domain stretching, including a light source module 1, a spatial light modulation module 8, a time stretching module 14, a data acquisition module and an image processing module 17.

[0051] The wide-band signal light output by the light source module 1 illuminates the sample 5 to obtain two-dimensional image information and one-dimensional spectral information. The signal light carrying the sample information is encoded and modulated by the spatial light modulation module 8, and then the encoded wide-band signal light is dispersed in the time domain by the time stretching module 14. The time domain information of the signal pulse is recorded by the data acquisition module. Finally, the image processing module 17 maps the time back to the spectrum through the precise correspondence between the signal light frequency and the time delay. According to the known mask sequence and the signal intensities corresponding to different wavelengths, the correlation algorithm is used to solve the hyperspectral image containing the two-dimensional spatial intensity information and one-dimensional spectral information of the sample 5 to be tested.

[0052] Embodiment 2 In an embodiment of the present disclosure, a single-pixel hyperspectral imaging device based on time-domain stretching is capable of detecting near-infrared signal light in a wide band to achieve hyperspectral imaging.

[0053] Reference Figure 3 A single-pixel hyperspectral imaging device based on time domain stretching includes a light source module 1, a filter 2, a first plano-convex lens 3, a second plano-convex lens 4, a sample to be tested 5, a third plano-convex lens 6, a fourth plano-convex lens 7, a spatial light modulation module 8, a fifth plano-convex lens 9, a reflective silver mirror 10, an adjustable filter 11, a beam collimator 12, a multimode optical fiber 13, a time stretching module 14, a single-pixel detector 15, an oscilloscope 16, and an image processing module 17.

[0054] The light source module 1 is used to output broadband near-infrared light, and the nonlinear medium that generates the supercontinuum includes but is not limited to waveguides, soft glass optical fibers, photonic crystal optical fibers and other media. The light source module 1 includes but is not limited to: broadband near-infrared light sources, and can be extended to the visible light band, medium-wave and long-wave infrared bands, and even the terahertz region, with a wide range of detection bands to meet the spectral analysis in different application scenarios. In this embodiment, the light source module uses a near-infrared supercontinuum laser.

[0055] The filter 2 is used to perform spectrum filtering on the wide-band signal, filter out unnecessary spectral components, and retain the sample absorption spectrum. In this embodiment, a bandpass filter is used.

[0056] The first plano-convex lens 3 and the second plano-convex lens 4 are used to expand the wide-band signal light to illuminate more sample areas and improve the imaging field of view. In this embodiment, the first plano-convex lens 3 and the second plano-convex lens 4 are C-band plano-convex lenses with focal lengths of 50 mm and 100 mm, respectively.

[0057] The sample 5 to be tested includes but is not limited to: liquid, biological tissue cells, chemical materials, etc. The detection target has different light absorption degrees for different wavelengths. The experimental system can obtain the absorption rate of each pixel of the detection target for each wavelength by measuring the imaging results with and without samples.

[0058] The third plano-convex lens 6 and the fourth plano-convex lens 7 are used to adjust the size of the light beam carrying the sample information signal light and project the image information onto the spatial light modulation module 8. In this embodiment, the third plano-convex lens 6 and the fourth plano-convex lens 7 are C-band plano-convex lenses with focal lengths of 75 mm and 50 mm, respectively.

[0059] The spatial light modulation module 8 includes, but is not limited to, modulation devices such as liquid crystal spatial light modulators, digital micromirror array modulators, and light-emitting diode arrays, which can control parameters such as intensity, phase, and modulation speed. In this embodiment, the spatial light modulation module 8 uses a digital micromirror array modulator DMD, which is composed of many small aluminum reflective mirrors, each of which is called a pixel. The digital micromirror refresh rate is 10 kHz, the size of the micromirror is 10.8 μm, the resolution is 1280×800, and each mirror can deflect ±12° around the diagonal of the micromirror. The DMD can switch a series of specific mask patterns to perform spatial coding modulation on the image.

[0060] The fifth plano-convex lens 9 is used to collimate the signal light after being encoded by the digital micromirror array DMD. In this embodiment, a C-band plano-convex lens with a focal length of 75 mm is used.

[0061] The reflective silver mirror 10 is used to reflect the encoded signal light.

[0062] The adjustable filter 11 is used to calibrate the relationship between the signal light frequency and the time delay in the measurement. The delay introduced by the adjustable filter 11 is a fixed value. After calibration, the device will be removed from the system.

[0063] The first beam collimator 12 is used to couple the spatial signal light into the multimode optical fiber 13 .

[0064] The multimode optical fiber 13 is used to improve the collection efficiency of the spatial frequency components of the signal light (incoherent light) and to shape the signal light mode propagating in the optical fiber so that the energy distribution between different modes of the signal light tends to be uniform. The multimode optical fiber can also be replaced by a spatial light guide to homogenize the light beam. In this embodiment, the core diameter of the multimode optical fiber is preferably 105 μm and the length is preferably 2 m.

[0065] The time stretch module 14 is used to achieve temporal separation of different frequency components by utilizing group velocity dispersion. The time stretch module 14 includes but is not limited to: single-mode optical fiber, arrayed waveguide grating, multimode waveguide or chirped fiber Bragg grating and other dispersive media. In this embodiment, a 20 km communication band optical fiber is used as the time stretch module, and its zero dispersion point is at 1310 nm, and there is a large positive dispersion near 1550 nm, the dispersion parameter is 17.4 ps / (nm∙km), and the loss is 0.2 dB / km.

[0066] The data acquisition module can record the intensity information of different wavelengths separated in the time domain. In this embodiment, the data acquisition module includes a single pixel detector 15 and an oscilloscope 16 .

[0067] The single pixel detector 15 is used to quickly detect the time-stretched signal light and convert the optical signal into an electrical signal. In this embodiment, a high-speed InGaAs detector is used, and its detection wavelength range covers 900-1700 nm and the detection bandwidth is 30 GHz.

[0068] The high-speed oscilloscope 16 has a bandwidth of 33 GHz and a sampling rate of 80 Gs / s, and is used to sample and quantize the collected analog electrical signals and convert them into digital signals.

[0069] The image processing module 17 is used to map the collected time information back to the spectrum, and solve the hyperspectral image of the sample through a correlation algorithm based on the known mask sequence and the intensity information corresponding to different wavelengths. In this embodiment, the image processing module 17 is built into the processing computer.

[0070] This embodiment is implemented as follows: First, after the wide-band near-infrared signal light passes through the sample 5 to be tested, it carries the two-dimensional spatial intensity information and one-dimensional spectral information of the sample. Then, the spatial light modulation module 8 generates a series of specific mask patterns to perform spatial coding modulation on the signal light carrying the sample information.

[0071] Specifically, the spectral range output by the light source module 1 with a repetition frequency of 10 MHz covers 1-2 μm, and the signal light in the 1.4-1.65 μm band is filtered out by the filter 2. The 4F system composed of the first plano-convex lens 3 and the second plano-convex lens 4 is used for beam expansion, thereby covering a larger area of ​​the sample 5 to be tested and obtaining the spatial intensity and spectral information of the sample. Subsequently, the 4F system composed of the third plano-convex lens 6 and the fourth plano-convex lens 7 is imaged on the spatial light modulation module 8 to realize spatial coding modulation.

[0072] Subsequently, the spatially encoded signal light is collected by the multimode optical fiber 13, and is time-stretched by the time-stretching module 14 to achieve separation of different wavelengths in the time domain, and the time-domain stretched pulse containing the spatially encoded information is measured using a single-pixel detector.

[0073] Specifically, the encoded signal light is collimated by the fifth plano-convex lens 9, and then collected into the multimode optical fiber 13 through the reflective silver mirror 10 and the beam collimator 12. The large-core multimode optical fiber 13 can collect more spatial frequency components of the signal light and shape the signal light mode propagating in the optical fiber so that the energy distribution between different modes in the optical fiber tends to be uniform. It is then time-stretched by the 20 km single-mode optical fiber time-stretching module 14, detected by the high-speed InGaAs single-pixel detector 15, and the collected analog electrical signal is converted into a digital signal using a high-speed oscilloscope 16.

[0074] Finally, the time is mapped back to the spectrum through the precise correspondence between the signal light frequency and the delay amount, and the hyperspectral image is reconstructed based on the known mask sequence and the signal intensity corresponding to different wavelengths combined with the correlation algorithm.

[0075] Specifically, an adjustable filter 11 is placed in the optical path to calibrate the correspondence between the signal light frequency and the time delay. The adjustable filter 11 is then taken out, and the collected time-domain stretched pulse time information is mapped back to the spectrum through the image processing module 17. The intensity information of different wavelengths corresponding to each coding mask can be obtained, and the hyperspectral image of the sample 5 to be tested can be obtained by combining the correlation algorithm.

[0076] This embodiment adopts a single-pixel imaging architecture combined with time-stretched spectral detection technology to separate the encoded signal light in the time domain. Although it is hoped to stretch the pulse as much as possible in order to obtain higher spectral resolution and more spectral channels and reduce the requirements for high detection bandwidth and sampling rate of the data acquisition system, this may cause overlap between the significantly stretched pulses and adjacent pulses. Therefore, there is a mutually restrictive relationship between the spectral bandwidth, laser repetition frequency and dispersion amount.

[0077] In this embodiment, the temporal separation of different frequency components is 0.35 ns / nm, and the maximum delay difference is 100 ns. By using a high-speed InGaAs single-pixel detector 15 and a high-speed oscilloscope 16, high-spectral imaging with a detection wavelength range of 1.4-1.65 μm is achieved, and the system spectral resolution is about 0.8 cm -1 , the number of hyperspectral acquisition channels is about 1350. Since the current switching speed of the spatial light modulation module is 10 kHz, when 64×64 pixels are selected for encoding and reconstruction, it takes 0.5 s to obtain the hyperspectral data cube.

[0078] If a faster 20 kHz spatial light modulation module and compressed sensing algorithm are adopted, the near-infrared hyperspectral data cube acquisition time in this scheme can be as low as ms. Finally, the hyperspectral imaging method and device based on this technical scheme can have the advantages of high acquisition speed, wide spectral range, high spectral resolution, high spectral channel and high sensitivity.

[0079] Embodiment 3 In an embodiment of the present disclosure, a single-pixel hyperspectral imaging device based on time-domain stretching is capable of performing coincidence counting detection on wide-band signal light in a low-light scene to achieve highly sensitive hyperspectral imaging.

[0080] Reference Figure 4 A single-pixel hyperspectral imaging device based on time domain stretching includes: a light source module 1, a filter 2, a first plano-convex lens 3, a second plano-convex lens 4, a sample to be tested 5, a third plano-convex lens 6, a fourth plano-convex lens 7, a spatial light modulation module 8, a fifth plano-convex lens 9, a reflective silver mirror 10, an adjustable filter 11, a first beam collimator 12, a multimode optical fiber 13, a time stretching module 14, an image processing module 17, a time-correlated single-photon counter 18, a second beam collimator 19, a neutral density filter 20, a single-photon detector 21, and a detector 22.

[0081] The light source module 1 is used to output broadband near-infrared light, and the nonlinear medium that generates the supercontinuum includes but is not limited to waveguides, soft glass optical fibers, photonic crystal optical fibers and other media. The light source module 1 includes but is not limited to: broadband near-infrared light sources, and can be expanded to the visible light band, medium-wave and long-wave infrared bands, and even the terahertz region, with a wide range of detection bands to meet the spectral analysis in different application scenarios. In this embodiment, the light source module adopts a near-infrared supercontinuum laser. In this embodiment, the light source module 1 adopts a near-infrared supercontinuum laser.

[0082] The filter 2 is used to transmit the required spectral components in the wide-band signal light and reflect the unnecessary spectral components. In this embodiment, a reflective bandpass filter is used.

[0083] The first plano-convex lens 3 and the second plano-convex lens 4 are used to expand the wide-band signal light to illuminate more sample areas and improve the imaging field of view. In this embodiment, the first plano-convex lens 3 and the second plano-convex lens 4 are C-band plano-convex lenses with focal lengths of 50 mm and 100 mm, respectively.

[0084] The sample 5 to be tested includes but is not limited to: liquid, biological tissue cells, chemical materials, etc. The detection target has different light absorption degrees for different wavelengths. By measuring the imaging results with and without samples, the absorption rate of each pixel of the detection target for each wavelength can be obtained.

[0085] The third plano-convex lens 6 and the fourth plano-convex lens 7 are used to adjust the size of the light beam carrying the sample information signal light and project the image information onto the spatial light modulation module 8. In this embodiment, the third plano-convex lens 6 and the fourth plano-convex lens 7 are C-band plano-convex lenses with focal lengths of 75 mm and 50 mm, respectively.

[0086] The spatial light modulation module 8 includes, but is not limited to, modulation devices such as liquid crystal spatial light modulators, digital micromirror array modulators, and light-emitting diode arrays, which can control parameters such as intensity, phase, and modulation speed. In this embodiment, the spatial light modulation module 8 uses a digital micromirror array modulator DMD, which is composed of many small aluminum reflective mirrors, each of which is called a pixel. The digital micromirror refresh rate is 10 kHz, the size of the micromirror is 10.8 μm, the resolution is 1280×800, and each mirror can deflect ±12° around the diagonal of the micromirror. The DMD can switch a series of specific mask patterns to perform spatial coding modulation on the image.

[0087] The fifth plano-convex lens 9 is used to collimate the signal light after being encoded by the digital micromirror array DMD. In this embodiment, a C-band plano-convex lens with a focal length of 75 mm is used.

[0088] The reflective silver mirror 10 is used to reflect the encoded signal light.

[0089] The adjustable filter 11 is used to calibrate the relationship between the signal light frequency and the time delay in the measurement. The delay introduced by the adjustable filter 11 is a fixed value. After calibration, the device will be removed from the apparatus.

[0090] The first beam collimator 12 is used to couple the spatial signal light into the multimode optical fiber 13 .

[0091] The multimode optical fiber 13 is used to improve the collection efficiency of the spatial frequency components of the signal light (incoherent light) and to shape the signal light mode propagating in the optical fiber so that the energy distribution between different modes of the signal light tends to be uniform. The multimode optical fiber can also be replaced by a spatial light guide to homogenize the light beam. In this embodiment, the core diameter of the multimode optical fiber 13 is preferably 105 μm and the length is preferably 2 m.

[0092] The time stretch module 14 is used to achieve temporal separation of different frequency components by utilizing group velocity dispersion. The time stretch module 14 includes but is not limited to: single-mode optical fiber, arrayed waveguide grating, multimode waveguide or chirped fiber Bragg grating and other dispersive media. In this embodiment, a 20 km communication band optical fiber is used as the time stretch module, and its zero dispersion point is at 1310 nm, and there is a large positive dispersion near 1550 nm, the dispersion parameter is 17.4 ps / (nm∙km), and the loss is 0.2 dB / km.

[0093] The data acquisition module can record the intensity information of different wavelengths separated in the time domain. In this embodiment, the data acquisition module includes a single photon detector 21 and a time-correlated single photon counter 18 .

[0094] The single-photon detector 21 is a low time jitter single-photon detector. In this embodiment, an InGaAs single-photon detector is used, and its detection wavelength range covers 900-1700 nm, the time jitter is 70 ps, ​​and the dark count is 2 kcps. It is used to realize ultra-sensitive single-photon detection of the signal light after time stretching as a stop signal.

[0095] The second beam collimator 19 is used to collect the light reflected by the bandpass filter.

[0096] The detector 22 is used to detect the light source reflected by the bandpass filter as a start trigger signal. In this embodiment, an InGaAs detector is used, and its detection wavelength range covers 900-1700 nm.

[0097] The time-correlated single-photon counter 18 records the time difference between the start signal and the end signal, and obtains a statistical histogram by cumulative measurement. Its digital resolution is 1 ps, the time jitter is <10 ps, ​​and the detection time window can reach 10 μs.

[0098] The image processing module 17 is used to map the collected time information back to the spectrum, and solve the hyperspectral image of the sample through a correlation algorithm based on the known mask sequence and the intensity information corresponding to different wavelengths. In this embodiment, the image processing module 17 is built into the processing computer.

[0099] The neutral density filter 20 is used to attenuate the wide-band signal light irradiated on the sample to a single-photon level, thereby simulating a weak-light detection scenario.

[0100] This embodiment is implemented as follows: First, the wide-band near-infrared signal light is attenuated to the single-photon level by the neutral density filter 20. After passing through the sample to be tested 5, it carries the two-dimensional spatial intensity information and one-dimensional spectral information of the sample. Subsequently, the spatial light modulation module 8 is used to generate a series of specific mask patterns to perform spatial coding modulation on the signal light carrying the sample information.

[0101] Specifically, the spectral range output by the light source module 1 with a repetition frequency of 10 MHz covers 1-2 μm, the filter 2 transmits the signal light in the 1.4-1.65 μm band and reflects the signal light outside the required band, the 4F system composed of the first plano-convex lens 3 and the second plano-convex lens 4 expands the transmitted signal light, attenuates it to the single-photon level through the neutral density filter 20, and then passes through the sample to be tested 5 to obtain the spatial intensity and spectral information of the sample, and the 4F system composed of the third plano-convex lens 6 and the fourth plano-convex lens 7 is imaged on the spatial light modulation module 8 to realize spatial coding modulation.

[0102] Subsequently, the spatially encoded signal light is collected by the multimode optical fiber 13, and time-stretched by the time stretching module 14 to achieve separation of different wavelengths in the time domain. The indium gallium arsenide single-photon detector 21 is used to measure the time-domain stretched pulse containing the spatial encoding information as a stop signal. In addition, the reflected near-infrared light is detected by the detector 22 as a start signal, and the correlated spectrum measurement is achieved through the time-correlated single-photon counting module.

[0103] Specifically, the encoded signal light is collimated by the plano-convex lens 9 and then collected into the multimode optical fiber 13 through the reflective silver mirror 10 and the beam collimator 12. The large-core multimode optical fiber 13 can collect more spatial frequency components of the signal light and shape the signal light mode propagating in the optical fiber so that the energy distribution between different modes in the optical fiber tends to be uniform. Then, it is time-stretched by the 20km single-mode optical fiber time-stretching module 14 and detected by the InGaAs single-photon detector 21 as a stop signal. At the same time, the near-infrared light reflected by the filter 2 is collected by the light collimator 16 and detected by the InGaAs detector 22. The time-correlated single-photon counter 18 is used for correlation detection. Through multiple measurements, a statistical histogram of the correlation spectrum detection delay and the photon count is obtained in the image processing module 17.

[0104] Finally, the time is mapped back to the spectrum through the precise correspondence between the signal light frequency and the delay amount, and the hyperspectral image is reconstructed based on the known mask sequence and the signal intensity corresponding to different wavelengths combined with the correlation algorithm.

[0105] Specifically, an adjustable filter 11 is placed in the optical path to calibrate the correspondence between the signal light frequency and the delay amount, and then the adjustable filter 11 is taken out, and the collected time-domain stretched pulse time information is mapped back to the spectrum through the image processing module 17, so that the intensity information of different wavelengths corresponding to each coding mask can be obtained, and the hyperspectral image of the sample to be tested can be obtained by combining the correlation algorithm.

[0106] This example demonstrates the high sensitivity advantage of this technology in low-light scenarios. Through time-correlated single-photon counting, the arrival time of the upconverted photons is accurately recorded by a low-timing jitter photon counter, and the separation of different frequency components in time is 0.35 ns / nm, which can obtain about 0.8 cm under single-photon illumination. -1 The high spectral resolution and the number of hyperspectral acquisition channels are about 1350. The time to obtain the hyperspectral data cube depends on the integration time of the photon coincidence counting of the time-correlated single photon counter. The light loss in the time stretch module can be compensated by optically pumping it with one or more high-power lasers and amplifying the time-stretched pulses by stimulated Raman scattering, thereby significantly reducing the integration time of photon detection.

[0107] This implementation can also introduce efficient and low-noise frequency up-conversion technology, which can up-convert wide-band signal light to the visible / near-infrared band with superior detection devices and low-loss stretching media, and the selection of nonlinear crystals of different materials can extend the detection band to the long-wave infrared or terahertz region to meet the needs of sensitive hyperspectral imaging in this spectral band.

[0108] This embodiment avoids the difficulties of traditional hyperspectral imaging technology in spectral scanning methods and imaging detection means. It does not require mechanical scanning devices, spatial dispersion elements and large-array detection devices. By combining time-stretched spectral detection and single-pixel imaging technology, this technical solution can have the advantages of high acquisition speed, wide spectral range, high spectral resolution, high spectral channels and high sensitivity, and will provide a new hyperspectral analysis method for application scenarios such as product sorting, precision agriculture, environmental monitoring, cultural relics protection, criminal investigation, camouflage identification, etc.

[0109] The above contents are only embodiments of the present invention. The common sense such as the known specific structures and characteristics in the scheme is not described in detail here. The ordinary technicians in the relevant field know all the common technical knowledge in the technical field to which the invention belongs before the application date or the priority date, can obtain all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for the ordinary technicians in the relevant field to implement this application. It should be pointed out that for the technicians in this field, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent. The protection scope required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to explain the content of the claims.

Claims

1. A single-pixel hyperspectral imaging method based on time domain stretching, characterized in that: include: The wide-band signal light passes through the sample to be tested to obtain two-dimensional image information and one-dimensional spectrum information. The spatial light modulation module is used to generate a mask pattern to perform spatial coding modulation on the signal light carrying the sample information. The spatially encoded wide-band signal light is collected through a multimode optical fiber and the different spatial frequency components are homogenized. The time stretching module then performs time stretching to achieve the separation of different wavelengths in the time domain, and a single-pixel detector is used to record the time domain information of the signal pulse. The time is mapped back to the spectrum through the precise correspondence between the signal light frequency and the time delay, and the hyperspectral image is obtained based on the known mask sequence and the signal intensities corresponding to different wavelengths combined with the correlation algorithm.

2. The single-pixel hyperspectral imaging method based on time domain stretching according to claim 1, characterized in that: The spatial light modulation module generates m specific coding patterns The signal light carrying the sample information is spatially intensity encoded. The process is expressed as: in, is the total intensity value of the signal light after being encoded by the i-th mask pattern. The broadband signal light passing through the sample to be tested is expressed as , where r is the spatial position, is the wavelength of the incident light.

3. The single-pixel hyperspectral imaging method based on time domain stretching according to claim 1, characterized in that: An adjustable filter is added before the time stretching module to calibrate the relationship between the signal light frequency and the time delay. By setting different filter frequencies of the adjustable filter and measuring the corresponding time, the corresponding relationship between the signal light frequency and the time delay is obtained.

4. The single-pixel hyperspectral imaging method based on time domain stretching according to claim 1, characterized in that: Through the precise correspondence between the signal light frequency and the time delay, the time is mapped back to the spectrum. According to the known mask sequence and the signal intensity corresponding to different wavelengths, the hyperspectral image is obtained by combining the correlation algorithm. Includes the following: For wavelength There are m intensity values , the target to be measured is at wavelength The two-dimensional image at is , solved by: 。 5. A single-pixel hyperspectral imaging device based on time domain stretching, characterized in that: It includes a light source module, a spatial modulation module, a time stretching module, a data acquisition module and an image processing module; the light source module is used to output a wide-band signal light to illuminate the sample to be tested to obtain two-dimensional image information and one-dimensional spectrum information; the spatial modulation module is used to encode and modulate the signal light carrying the sample information; the time stretching module is used to disperse the encoded wide-band signal light in the time domain, and the data acquisition module records the time domain information of the signal pulse; the image processing module is used to map the time back to the spectrum through the precise correspondence between the signal light frequency and the time delay, and obtain the hyperspectral image according to the known mask sequence and the signal intensity corresponding to different wavelengths in combination with the correlation algorithm.

6. The single-pixel hyperspectral imaging device based on time domain stretching according to claim 5, characterized in that: The light source module includes a broadband near-infrared light source that can be extended to the visible light band, the medium-wave and long-wave infrared bands, and the terahertz region.

7. The single-pixel hyperspectral imaging device based on time domain stretching according to claim 5, characterized in that: The spatial modulation module includes a liquid crystal spatial light modulator, a digital micromirror array modulator and a light emitting diode array, and can control intensity parameters, phase parameters and modulation speed parameters.

8. The single-pixel hyperspectral imaging device based on time domain stretching according to claim 5, characterized in that: The data acquisition module includes a single-pixel detector and an oscilloscope; the single-pixel detector is used to measure the time-domain stretched pulse containing spatial coding information; the oscilloscope is used to sample and quantize the collected analog electrical signals and convert them into digital signals.

9. The single-pixel hyperspectral imaging device based on time domain stretching according to claim 5, characterized in that: The data acquisition module includes an InGaAs detector, a single-photon detector and a time-correlated single-photon counter; the InGaAs processor is used to detect the light source emitted by the light source module as a start trigger signal; the single-photon detector is used to realize single-photon detection of the signal light after time stretching, and measure the time-domain stretching pulse containing spatial coding information as a stop signal; the time-correlated single-photon counter is used to record the time difference between the start signal and the end signal, and the cumulative measurement obtains a statistical histogram.

10. A single-pixel hyperspectral imaging device based on time domain stretching according to any one of claims 5 to 9, characterized in that: It also includes a first plano-convex lens and a second plano-convex lens for expanding the wide-band signal light; it also includes a third plano-convex lens and a fourth plano-convex lens for adjusting the size of the light beam carrying the sample information signal light and projecting the image information onto the spatial light modulation module.

Images