Fourier transform spectral imaging apparatus and spectral imaging method

Abstract

The application provides a Fourier transform spectral imaging device and a spectral imaging method, and relates to the field of spectral imaging.The Fourier transform spectral imaging device comprises a Michelson interferometer, a digital micromirror, a photoelectric detector and a control host.The Michelson interferometer comprises a scanning galvanometer fixedly arranged on a rotation shaft, and is used to generate first interference light and guide the first interference light to a sample to be measured in a state that the scanning galvanometer reciprocally rotates along the rotation shaft.The digital micromirror is used to display a plurality of modulation patterns, and modulate the intensity of second interference light by using the plurality of modulation patterns to obtain a plurality of third interference light beams, wherein the second interference light is obtained by the first interference light after the first interference light interacts with the sample to be measured.The photoelectric detector is used to receive the plurality of third interference light beams to obtain a plurality of modulation interference patterns.The control host is used to perform compressed sensing imaging based on the plurality of modulation patterns and the plurality of modulation interference patterns to obtain a spectral image.

Description

Technical Field

[0001] This application relates to the field of spectral imaging, specifically to a Fourier transform spectral imaging device and a spectral imaging method. Background Technology

[0002] Fourier Transform Spectroscopic Imaging (FTSI) is an optical detection technology that uses a Fourier transform imaging spectrometer to achieve spectral analysis and imaging. It can simultaneously acquire spatial and spectral information of a target object and has advantages such as high spectral resolution, high signal-to-noise ratio, and wide spectral range. It is widely used in many fields.

[0003] Traditional Fourier transform imaging spectrometers utilize a Michelson interferometer to generate interference light, which is then collected by an array detector. Finally, the host computer executes a Fourier transform algorithm to convert the interference light signal into a frequency domain spectrum, generating a three-dimensional data cube containing two spatial dimensions and one spectral dimension. This allows for the simultaneous analysis of the sample's spectral information and spatial distribution. However, due to the low frame rate of the array detector, the scanning speed of the Michelson interferometer must be significantly reduced to prevent aliasing in the final frequency domain spectrum. This reduces the overall imaging speed of the Fourier transform imaging spectrometer, thus limiting its application in dynamic real-time spectral imaging. Summary of the Invention

[0004] To address the aforementioned issues, this application provides a Fourier transform spectral imaging device and a spectral imaging method.

[0005] This application proposes a Fourier transform spectral imaging device, comprising: a Michelson interferometer, including a scanning galvanometer fixedly mounted on a rotating axis, wherein the Michelson interferometer generates a first interference beam while the scanning galvanometer reciprocates along the rotating axis, and guides the first interference beam to the sample under test; a digital micromirror, used to display multiple modulation patterns, and using the multiple modulation patterns to modulate the intensity of a second interference beam to obtain multiple beams of third interference beams, wherein the second interference beam is obtained by the interaction of the first interference beam with the sample under test; a photodetector, used to receive the multiple beams of third interference beams to obtain multiple modulation interferograms; and a control host, used to perform compressed sensing imaging based on the multiple modulation patterns and multiple modulation interferograms to obtain a spectral image.

[0006] According to an embodiment of this application, the control host is further configured to provide a drive signal to the Michelson interferometer, wherein the drive signal is composed of a plurality of identical drive sub-signals; the Michelson interferometer is configured to drive the scanning galvanometer to perform reciprocating rotational motion based on the drive signal.

[0007] According to an embodiment of this application, the control host is further configured to provide trigger signals to the digital micromirror and the photodetector at a first time and a second time, respectively, wherein the time difference between the first time and the second time is half of the total duration of the driving sub-signal; the digital micromirror is configured to switch the displayed modulation pattern in response to the trigger signal, so as to use the switched modulation pattern to modulate the intensity of the second interference light to obtain the third interference light; the photodetector is configured to perform photoelectric conversion on the third interference light in response to the trigger signal to obtain a modulation interferogram.

[0008] According to an embodiment of this application, the driving sub-signal is a triangular wave signal, and the duration of the rising edge of the triangular wave signal is equal to the duration of the falling edge.

[0009] According to an embodiment of this application, the control host is further configured to train a convolutional neural network using multiple spectral image samples, and generate multiple modulation patterns based on the model parameters of the multiple convolutional kernels included in the trained convolutional neural network, so as to send multiple modulation patterns to the digital micromirror during spectral imaging.

[0010] According to an embodiment of this application, the control host is used to input spectral image samples into a convolutional neural network to obtain multiple single-pixel measurement values, perform compressed sensing imaging based on the model parameters of the multiple convolutional kernels included in the convolutional neural network and the multiple single-pixel measurement values ​​to obtain a reconstructed spectral image, and update the model parameters of the convolutional neural network based on the loss value between the reconstructed spectral image and the spectral image samples to obtain a trained convolutional neural network.

[0011] According to an embodiment of this application, the first interference light includes a first scanning light and a first reference light. The Michelson interferometer further includes a light source, a beam splitter, a grating, a lens module, and a reflector. The light source is used to provide detection light. The beam splitter is used to split the detection light into a second scanning light and a second reference light, and guides the second scanning light to the grating and the second reference light to the reflector. The grating is used to diffract and split the second scanning light to obtain a first beam. The lens module is used to converge the first beam to the rotation axis. The scanning galvanometer is used to introduce a phase delay into the first beam when it reciprocates along the rotation axis to obtain a second beam, and uses the second beam as the first scanning light, outputting the first scanning light through the lens module, the grating, and the beam splitter. The reflector is used to reflect the second reference light to obtain the first reference light, so that the first reference light can be output through the beam splitter.

[0012] According to an embodiment of this application, the Fourier transform spectral imaging device further includes:

[0013] A beam expander is used to expand the beam of the first interference beam in order to increase the area of ​​the light spot formed by the first interference beam illuminating the sample under test.

[0014] A focusing lens is used to converge the third interference light onto the photodetector.

[0015] Another aspect of this application proposes a Fourier transform spectral imaging method, comprising: acquiring a first interference beam while the scanning galvanometer of the Fourier transform spectral imaging device is in a state of reciprocating rotation along the axis of rotation; guiding the first interference beam to the sample to be tested so that the first interference beam interacts with the sample to be tested to obtain a second interference beam; using multiple modulation patterns to modulate the intensity of the second interference beam to obtain multiple beams of third interference beam; performing photoelectric detection on the multiple beams of third interference beam to obtain multiple modulation interferograms; and performing compressed sensing imaging based on the multiple modulation patterns and multiple modulation interferograms to obtain a spectral image.

[0016] This embodiment of the application sets up a digital micromirror in a Fourier transform spectrometer to display the modulation pattern required by the compressed sensing algorithm. This allows the interference light interacting with the sample to be modulated directly by the modulation pattern outside the control host. The resulting modulation interferogram is then received by a photodetector and transmitted to the control host for spectral reconstruction, ultimately achieving the acquisition of sample information. The digital micromirror allows the interference light modulation module in the imaging process to be independent, no longer relying on the control host to perform this operation. The control host only needs to execute the Fourier transform and compressed sensing algorithm spectral reconstruction process, thus simplifying the control host's execution procedure. Compared to technical solutions that use the control host to execute the entire compressed sensing algorithm process, imaging efficiency is improved. Attached Figure Description

[0017] The above and other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0018] Figure 1 A schematic diagram of a Fourier transform spectral imaging device according to an embodiment of this application is shown.

[0019] Figure 2 A schematic diagram of the process of reconstructing a spectral image from a modulation interferogram according to an embodiment of this application is shown.

[0020] Figure 3 A schematic diagram of a Michelson interferometer according to an embodiment of this application is shown.

[0021] Figure 4 A schematic diagram of the modulation interferogram acquisition process according to an embodiment of this application is shown.

[0022] Figure 5 A flowchart of a modulation pattern customization method based on a convolutional neural network according to an embodiment of this application is shown.

[0023] Figure 6A schematic diagram of the imaging principle of a Fourier transform spectrometer according to an embodiment of this application is shown.

[0024] Figure 7 A flowchart of a Fourier transform spectral imaging method according to an embodiment of this application is shown. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0026] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0027] All terms used herein, including technical and scientific terms, have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0028] Existing improved Fourier transform imaging spectrometers propose using compressed sensing algorithms to overcome the scanning speed limitation of Michelson interferometers caused by the sampling theorem. Compressed sensing algorithms are not limited by the sampling theorem; they do not rely on dense, uniform sampling of the signal. Instead, they reconstruct the complete signal by collecting a small amount of core information and combining it with the algorithm. Therefore, they do not require sampling at twice the highest frequency of the signal. Consequently, this algorithm can significantly reduce the amount of data during reconstruction, saving storage resources.

[0029] However, in the process of spectral imaging using compressed sensing algorithms, a control unit is required to generate and display the modulation pattern needed for the modulation process, modulate the interference light passing through the sample, and execute the calculations of the compressed sensing algorithm to obtain the final measurement results. This results in the control unit performing too many tasks, making the program complex and slowing down its operation. Therefore, it is urgent to adopt technical means to offload the tasks performed by the control unit, enabling some tasks to be completed by hardware, thereby simplifying the control unit's execution program and improving imaging efficiency.

[0030] Figure 1 A schematic diagram of a Fourier transform spectral imaging device according to an embodiment of this application is shown.

[0031] like Figure 1 As shown, the Fourier transform spectral imaging device may include: a Michelson interferometer 1, a digital micromirror 2, a photodetector 3, and a control host 4.

[0032] The Michelson interferometer 1 includes a scanning galvanometer fixedly mounted on a rotating axis. The Michelson interferometer 1 generates a first interference beam while the scanning galvanometer reciprocates along the axis, and guides this first interference beam to the sample under test. A digital micromirror 2 displays multiple modulation patterns and uses these patterns to modulate the intensity of a second interference beam, resulting in multiple beams of third interference beams. The second interference beams are obtained by the interaction of the first interference beam with the sample under test. A photodetector 3 receives these multiple beams of third interference beams, generating multiple modulation interferograms. A control unit 4 performs compressed sensing imaging based on the multiple modulation patterns and interferograms to obtain a spectral image.

[0033] In a Michelson interferometer, the light emitted from the source is split into two beams with different propagation directions. Since these two beams originate from the same source, they have the same frequency, a stable phase relationship, and a consistent vibration direction, satisfying the interference conditions. After propagating along different paths, these two beams will meet again and superimpose. When the two beams rejoin, if the optical path difference is an integer multiple of the wavelength, the superposition results in increased light intensity, forming bright fringes; if the optical path difference is an odd multiple of half the wavelength, the superposition results in decreased light intensity, forming dark fringes. By changing the propagation path length of one beam, the optical path difference between the two beams will change, and the interference fringes will move or deform accordingly. The propagation path length of one beam can be changed by the reciprocating rotation of the scanning mirror. When the scanning mirror deflects at an angle, the reflection angle of the incident beam on the scanning mirror also changes, thus changing the path length of the reflected beam in the interferometer's optical path.

[0034] The interference light generated by the Michelson interferometer 1 passes through the sample and is modulated by the digital micromirror 2. The digital micromirror 2 can be an optical modulation device composed of an array of numerous tiny mirrors (hereinafter referred to as micromirrors), each of which can independently rotate a small angle around its own axis. When light shines on the digital micromirror 2, each micromirror corresponds to a pixel in the image. The modulation pattern displayed by the digital micromirror 2 is essentially a specific combination of the micromirror's on / off states, which can be controlled by the control host 4 using binary signals. A micromirror receiving a "1" signal rotates to the open state, and a micromirror receiving a "0" signal rotates to the closed state. When the incident light to be modulated reaches a micromirror in the open state, it will precisely enter the imaging optical path such as the projection lens, while when it reaches a micromirror in the closed state, the incident light will deviate from the imaging optical path. Based on the above principle, the spatial intensity distribution of the incident light perfectly matches the brightness distribution of the modulation pattern, thereby modulating the light signal transmitted through the modulation pattern.

[0035] The modulated interferogram obtained after modulation is further detected and converted into a signal by photodetector 3. Photodetector 3 can be composed of a photosensitive element, a photoelectric signal conversion structure, and auxiliary circuitry, which can convert optical signals into electrical signals. When light radiation irradiates the photosensitive material of photodetector 3, the electrons in the material absorb the energy of photons and undergo state changes, thereby causing changes in electrical properties, such as changes in current, voltage, or resistance. The strength of these electrical signals is related to the intensity and wavelength of the incident light. Through subsequent amplification and processing by the circuitry, the detection and signal conversion of the optical signals can be achieved.

[0036] The photodetector 3 transmits the received modulation interferogram to the control host 4, which then uses a compressed sensing algorithm to reconstruct the spectrum based on the modulation pattern and its corresponding modulation interferogram.

[0037] Compressed sensing algorithms are not limited by the sampling theorem, allowing them to acquire small amounts of measurement data at rates far lower than traditional sampling rates. Then, mathematical optimization algorithms are used to accurately reconstruct the original high-dimensional signal from this low-dimensional data. The compressed sensing workflow consists of two steps: First, signal modulation, where a measurement matrix is ​​designed to modulate the original signal, transforming the high-dimensional signal into low-dimensional measurements with far fewer dimensions than the original. One form of the measurement matrix can be a modulation pattern displayed by a digital micromirror. Second, signal reconstruction, where algorithms are used to infer the original signal from the low-dimensional measurements.

[0038] Figure 2 A schematic diagram of the process of reconstructing a spectral image from a modulation interferogram according to an embodiment of this application is shown.

[0039] like Figure 2 As shown, the reconstruction process consists of two steps: Fourier transform and compressed sensing reconstruction. First, the modulation interferograms I1, I2, ..., I... n Perform a Fourier transform to obtain the modulation spectrum. For Michelson interferometer 1, the modulation spectrum at optical frequency ν is... The theoretical expression for is shown in formula (1). Wherein, Indicates the sequence number of the modulation pattern; This indicates the frequency of the light incident on the axis of the scanning galvanometer; Indicates modulation pattern The corresponding modulation coefficient; This represents the collected interference light signal; The group delay is expressed as shown in formula (2), where F represents the focal length of the 4F system and N represents the number of grating lines. t represents the rotational speed of the scanning galvanometer, and t represents time.

[0040]

[0041] Secondly, the control host 4 runs the compressed sensing algorithm to reconstruct the spectral image based on the modulation spectrum at the optical frequency ν and the corresponding modulation pattern. The general solution principle of compressed sensing is shown in formula (3), where, φ represents the reconstructed spectral image; φ represents the measurement matrix. The principle of solving the measurement matrix is ​​shown in formula (4). First, the modulation pattern is... Flattened into row vectors Then, it is obtained by combining the rows; The measured value is obtained by synthesizing n modulation spectra at optical frequency ν.

[0042]

[0043] Compressed sensing reconstruction is performed on the modulation spectrum at each ν point to obtain a spectral image, which is a three-dimensional data cube.

[0044] This application proposes a Fourier transform spectral imaging device and method. The device consists of a Michelson interferometer 1, a digital micromirror 2, a photodetector 3, and a control host 4. The interference light generated by the Michelson interferometer 1 passes through the sample and is modulated by a modulation pattern displayed on the digital micromirror 2. The modulated interference light is received by the photodetector 3, and finally, the control host 4 runs a compressed sensing algorithm to reconstruct the spectral image. This application simplifies the control host's execution process by setting a digital micromirror within the Fourier transform spectrometer, using the digital micromirror to display the modulation pattern required for the compressed sensing algorithm, and modulating the interference light. The control host's execution tasks are partially implemented through external hardware, thus improving the imaging speed compared to technical solutions that use the control host to perform the entire compressed sensing algorithm process.

[0045] Figure 3 A schematic diagram of the structure of a Michelson interferometer according to an embodiment of this application is shown.

[0046] like Figure 3As shown in this embodiment, the Michelson interferometer 1 includes a mirror 101, a beam splitter 102, a light source 103, a grating 104, a lens module 105, and a scanning mirror 106. The light source 103 provides detection light, and the beam splitter 102 splits the detection light into a second scanning light and a second reference light, guiding the second scanning light to the grating 104 and the second reference light to the mirror 101. The grating 104 diffracts the second scanning light to obtain a first beam; the lens module 105 converges the first beam to the rotation axis; the scanning mirror 106 introduces a phase delay into the first beam during reciprocating rotation along the rotation axis to obtain a second beam, and uses the second beam as the first scanning light, outputting the first scanning light through the lens module 105, the grating 104, and the beam splitter 102. The mirror 101 reflects the second reference light to obtain the first reference light, which is then output through the beam splitter 102.

[0047] According to an embodiment of this application, in the Michelson interferometer 1, a grating 104, a lens module 105, and a scanning mirror 106 are arranged sequentially. The grating 104 and the lens module 105, and the lens module 105 and the scanning mirror 106 form a 4F structure, that is, the distance between the grating 104 and the lens module 105, and the distance between the lens module 105 and the scanning mirror 106 are all F. The incident light travels from the grating 104 through the lens module 105 to the scanning mirror 106, is reflected by the scanning mirror 106, and returns to the grating 104 through the lens module 105. The total optical path is 4F, where F is the focal length of the lens in the lens module 105.

[0048] According to an embodiment of this application, the detection light emitted by the light source 103 first reaches the beam splitter 102. The beam splitter 102 can be a glass plate coated with a semi-transparent and semi-reflective film. When the incident light reaches the beam splitter 102, half of the light is reflected and the other half is transmitted. The light reflected by the beam splitter 102 is the second scanning light. The second scanning light then reaches the grating 104. The grating 104 can be composed of a large number of parallel slits of equal width and spacing, which can diffract and split the second scanning light to obtain a first beam. The first beam is focused by the lens module 105 to the rotation axis of the scanning galvanometer 106. As the scanning galvanometer 106 reciprocates along the rotation axis, the reflected optical path of the first beam continuously changes, thereby introducing a phase delay to obtain a second beam. The second beam is then used as the first scanning light and output through the lens module 105, the grating 104, and the beam splitter 102. The lens module 105 can be a combination of two biconvex lenses. The divergence center of the light emitted from the grating 104 coincides with the object-side focal point of the biconvex lens. According to the imaging principle of a convex lens, the emitted light becomes parallel to the optical axis after passing through the biconvex lens. The parallel light is reflected by the scanning galvanometer 106 and returns along the original path. When it re-enters the biconvex lens, the parallel light is focused back to the original object-side focal point position by the biconvex lens. The detection light emitted from the light source 103 is transmitted through the beam splitter 102 to obtain the second reference light. The second reference light is reflected by the reflecting mirror 101 to obtain the first reference light. The first reference light interferes with the first scanning light to obtain the first interference light.

[0049] According to embodiments of this application, the aforementioned Fourier transform spectral imaging device may further include a beam expander. The beam expander is used to expand the beam of the first interference light to increase the area of ​​the light spot formed by the first interference light illuminating the sample under test. The beam expander may consist of an input lens and an output lens. The input lens may be a short focal length concave lens, responsible for receiving the incident original beam (e.g., the first interference light) and diverging it. The output lens may be a long focal length convex lens, which re-collimates and magnifies the diverged beam, ultimately outputting a parallel beam with a larger diameter.

[0050] According to embodiments of this application, the aforementioned Fourier transform imaging spectrometer may further include a focusing lens. The focusing lens is used to converge the third interference light to the photodetector 3. The focusing lens may be a cemented doublet achromatic focusing lens, consisting of a convex lens and a concave lens. The convex lens deflects the incident light along the optical axis, while the concave lens compensates for the deflection angle of the convex lens for different wavelengths of light, ultimately enabling light of different wavelengths to converge to the same focal point after the combined action of the two lenses.

[0051] According to an embodiment of this application, the control host 4 is further configured to provide a drive signal to the Michelson interferometer, wherein the drive signal is composed of multiple identical drive sub-signals. The Michelson interferometer 1 is used to drive the scanning galvanometer 106 to perform reciprocating rotational motion based on the drive signal. The drive sub-signals can be triangular wave signals, where the duration of the rising edge of the triangular wave signal is equal to the duration of the falling edge.

[0052] According to an embodiment of this application, the control host 4 is further configured to provide trigger signals to the digital micromirror 2 and the photodetector 3 at the first and second moments of the driving sub-signal, respectively, wherein the time difference between the first and second moments is half the total duration of the driving sub-signal; the digital micromirror 2 is configured to switch the displayed modulation pattern in response to the trigger signal, so as to use the switched modulation pattern to modulate the intensity of the second interference light to obtain the third interference light; the photodetector 3 is configured to perform photoelectric conversion on the third interference light in response to the trigger signal to obtain a modulation interferogram.

[0053] Figure 4 A schematic diagram of the modulation interferogram acquisition process according to an embodiment of this application is shown.

[0054] like Figure 4 As shown, in order to acquire the data required for a single spectral imaging, the control host 4 outputs a triangular wave drive signal to the scanning mirror 106 of the Michelson interferometer 1, and ensures that the duration of its rising edge and falling edge are equal, so that it performs a uniform reciprocating symmetrical scanning motion to generate stable interference light.

[0055] At the beginning (first moment) and halfway through (second moment) of each scanning cycle of the scanning galvanometer 106, the control host 4 outputs a synchronous trigger signal to the digital micromirror 2 and the photodetector 3. Upon detecting the trigger signal, the digital micromirror 2 immediately switches the modulation pattern, and simultaneously, the photodetector 3 immediately begins acquiring the modulation interferogram under this modulation pattern. Thus, in each scanning cycle of the scanning galvanometer 106, two modulation patterns are switched, two intensity modulations are performed, and two modulation interferograms are measured. Finally, after... In one scan cycle, all n modulation patterns are switched, and n modulation interferograms I1, I2, ..., I3 are acquired. n Since the scanning of the scanning galvanometer 106 and the switching of the modulation pattern are synchronized, and the scanning speed of the galvanometer can usually reach several hundred scanning cycles per second, the above scheme can effectively improve the signal acquisition speed.

[0056] According to an embodiment of this application, the control host 4 is also used to train a convolutional neural network using multiple spectral image samples, and generate multiple modulation patterns based on the model parameters of the multiple convolutional kernels included in the trained convolutional neural network, so as to send multiple modulation patterns to the digital micromirror 2 during spectral imaging.

[0057] According to an embodiment of this application, the control host 4 is used to input spectral image samples into a convolutional neural network to obtain multiple single-pixel measurement values, perform compressed sensing imaging based on the model parameters of the multiple convolutional kernels included in the convolutional neural network and the multiple single-pixel measurement values ​​to obtain a reconstructed spectral image, and update the model parameters of the convolutional neural network based on the loss value between the reconstructed spectral image and the spectral image samples to obtain a trained convolutional neural network.

[0058] Figure 5 A flowchart of a modulation pattern customization method based on a convolutional neural network according to an embodiment of this application is shown.

[0059] like Figure 5 As shown, the method includes operations S501 to S504.

[0060] In operation S501, multiple spectral image samples are input into a convolutional neural network, and n single-pixel measurement values ​​of multiple spectral image samples are obtained by performing n convolution operations on the multiple spectral image samples.

[0061] In this operation, the model parameters of n convolutional kernels can be converted into the data format supported by the Digital Microscope 2. Therefore, n convolutional kernels correspond to n modulation patterns.

[0062] In operation S502, n single-pixel measurements and corresponding n convolution kernels are input into the compressed sensing imaging algorithm to obtain a reconstructed spectral image.

[0063] In operation S503, the loss value between the reconstructed spectral image and multiple spectral image samples is calculated.

[0064] In operation S504, the model parameters of the convolution kernel of the convolutional neural network are updated.

[0065] This process is iterated until the loss value converges and falls below the expected level. After iteration, the model parameters of the n convolutional kernels are converted into a data format supported by the digital micromirror 2, which can then be uploaded to the digital micromirror 2 for Fourier transform spectral imaging. The control host 4 inputs spectral image samples into the convolutional neural network to obtain multiple single-pixel measurements. Based on the loss value between the input and measured values, it updates and optimizes the model parameters of the convolutional kernels, customizing the optimal modulation pattern for the compressed sensing imaging algorithm in the Fourier transform spectral imaging system. This pattern is then displayed through the digital micromirror 2 to achieve modulation of the interference light signal.

[0066] It should be noted that there are many types of compressed sensing imaging algorithms, and the above is only one example according to the embodiments of this application. However, regardless of the algorithm used to train the customized modulation pattern, the compressed sensing imaging algorithm mentioned here should be consistent with the compressed sensing imaging algorithm used to reconstruct the spectral image above. Similarly, the modulation pattern obtained from training should also be consistent with the modulation pattern uploaded to the digital micromirror 2 during the imaging process. In addition, the number of convolution kernels n needs to be adjusted according to the actual spectral image reconstruction effect, and can usually be taken as 5% to 20% of the total number of pixels in the target spectral image. This method of customizing the modulation pattern improves the spectral imaging quality of the compressed sensing algorithm when there are fewer modulation patterns, thereby reducing the number of modulation interferogram measurements required to reconstruct a high-quality spectral image, which helps to improve the speed of Fourier transform spectral imaging.

[0067] Figure 6 A schematic diagram of the imaging principle of a Fourier transform spectrometer according to an embodiment of this application is shown.

[0068] like Figure 6 As shown in the embodiment of this application, the light emitted from the light source 103 in the Michelson interferometer 1 is reflected and transmitted after reaching the beam splitter 102. The reflected light, as the second scanning light, is diffracted and split after reaching the grating 104, and then exits in parallel through the lens module 105. Finally, it reaches the reciprocating rotating scanning mirror 106 and is reflected to obtain the second beam. The second beam, as the first scanning light, is output through the lens module 105, the grating 104, and the beam splitter 102. The transmitted light, as the second reference light, is reflected after reaching the reflecting mirror 101 to obtain the first reference light. The first reference light interferes with the first scanning light to obtain the first interference light. The first interference light is expanded by the beam expander 5 and interacts with the sample 6 to obtain the second interference light. The second interference light is modulated by the modulation pattern displayed in the digital micromirror 2 to obtain the third interference light. The third interference light is focused by the focusing lens 7 to the photodetector 3 and received by the photodetector 3 to obtain the modulation interferogram. After receiving the modulation interferogram from the photodetector 3, the control host 4 reconstructs the spectral image based on the modulation interferogram and the modulation pattern through the compressed sensing algorithm, and finally obtains the spatial and spectral information of the sample 6 to be tested.

[0069] Figure 7 A flowchart of a Fourier transform spectral imaging method according to an embodiment of this application is shown.

[0070] like Figure 7 As shown, the method includes operations S701 to S705.

[0071] When operating S701, the scanning galvanometer of the Fourier transform spectral imaging device is in a state of reciprocating rotation along the axis of rotation, and the first interference light is acquired.

[0072] In operation S702, the first interference light is directed to the sample to be tested so that the first interference light interacts with the sample to obtain the second interference light.

[0073] By operating S703, multiple modulation patterns are used to modulate the intensity of the second interference light to obtain multiple beams of third interference light.

[0074] By operating the S704, photoelectric detection is performed on multiple beams of third interference light to obtain multiple modulation interferograms.

[0075] When operating the S705, compressed sensing imaging is performed based on multiple modulation patterns and multiple modulation interferograms to obtain spectral images.

[0076] This application utilizes a digital micromirror within a Fourier transform spectral imaging device to display the modulation pattern required for the compressed sensing algorithm, thereby modulating the interference light signal. After acquiring the modulated interferogram using a photodetector, the host computer executes a spectral reconstruction process to obtain the spatial location and spectral information of the target sample. The digital micromirror in this application allows the interference light modulation step in the compressed sensing algorithm to be independent of the host computer, enabling external hardware to perform some tasks of the compressed sensing algorithm. This simplifies the host computer's calculation program and improves the spectral imaging speed compared to schemes where the entire compressed sensing algorithm is executed by the host computer.

[0077] The embodiments of this application have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of this application. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. The scope of this application is defined by the appended claims and their equivalents. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this application, and all such substitutions and modifications should fall within the scope of this application.

Claims

1. A Fourier transform spectral imaging device, characterized in that, include: A Michelson interferometer includes a scanning galvanometer fixedly mounted on a rotating axis. The Michelson interferometer is used to generate a first interference beam while the scanning galvanometer is reciprocating along the rotating axis, and to guide the first interference beam to the sample to be tested. A digital micromirror is used to display multiple modulation patterns and to modulate the intensity of a second interference beam using the multiple modulation patterns to obtain multiple third interference beams, wherein the second interference beam is obtained by the interaction of the first interference beam with the sample under test; A photodetector is used to receive multiple beams of the third interference light to obtain multiple modulation interferograms; and A control host is used to perform compressed sensing imaging based on multiple modulation patterns and multiple modulation interferograms to obtain a spectral image; The control host is also used to provide a drive signal to the Michelson interferometer, wherein the drive signal is composed of a plurality of identical drive sub-signals; The control host is also used to provide trigger signals to the digital micromirror and the photodetector at the first and second moments of the driving sub-signal, respectively, wherein the time difference between the first moment and the second moment is half of the total duration of the driving sub-signal; The digital micromirror is used to switch the displayed modulation pattern in response to the trigger signal, so as to use the switched modulation pattern to modulate the intensity of the second interference light to obtain the third interference light; The photodetector is used to perform photoelectric conversion on the third interference light in response to the trigger signal to obtain the modulation interferogram.

2. The Fourier transform spectral imaging device according to claim 1, characterized in that, The Michelson interferometer is used to drive the scanning galvanometer to perform reciprocating rotational motion based on the driving signal.

3. The Fourier transform spectral imaging device according to claim 2, characterized in that, The driving sub-signal is a triangular wave signal, and the duration of the rising edge of the triangular wave signal is equal to the duration of the falling edge.

4. The Fourier transform spectral imaging device according to claim 1, characterized in that, The control host is also used to train a convolutional neural network using multiple spectral image samples, and generate multiple modulation patterns based on the model parameters of the multiple convolutional kernels included in the trained convolutional neural network, so as to send multiple modulation patterns to the digital micromirror during spectral imaging.

5. The Fourier transform spectral imaging device according to claim 4, characterized in that, The control host is used to input the spectral image samples into the convolutional neural network to obtain multiple single-pixel measurement values. Based on the model parameters of the multiple convolutional kernels included in the convolutional neural network and the multiple single-pixel measurement values, compressed sensing imaging is performed to obtain a reconstructed spectral image. Based on the loss value between the reconstructed spectral image and the spectral image samples, the model parameters of the convolutional neural network are updated to obtain the trained convolutional neural network.

6. The Fourier transform spectral imaging device according to claim 1, characterized in that, The first interference light includes a first scanning light and a first reference light, and the Michelson interferometer further includes a light source, a beam splitter, a grating, a lens module, and a reflector; The light source is used to provide detection light; The beam splitter is used to split the detection light into a second scanning light and a second reference light, and guides the second scanning light to the grating and the second reference light to the reflector; The grating is used to diffract and separate the second scanning light to obtain a first beam. The lens module is used to converge the first beam to the rotating shaft; The scanning galvanometer is used to introduce a phase delay into the first beam to obtain a second beam when it reciprocates along the axis of rotation, and uses the second beam as the first scanning light to output the first scanning light through the lens module, the grating and the beam splitter; The reflector is used to reflect the second reference light to obtain the first reference light, so as to output the first reference light through the beam splitter.

7. The Fourier transform spectral imaging device according to claim 1, characterized in that, Also includes: A beam expander is used to expand the beam of the first interference beam in order to increase the area of ​​the light spot formed by the first interference beam illuminating the sample under test.

8. The Fourier transform spectral imaging device according to claim 1, characterized in that, Also includes: A focusing lens is used to converge the third interference light onto the photodetector.

9. A spectral imaging method, applied to the Fourier transform spectral imaging device as described in any one of claims 1 to 8, characterized in that, The spectral imaging method includes: The first interference light is acquired when the scanning galvanometer of the Fourier transform spectral imaging device is in a state of reciprocating rotation along the axis of rotation; The first interference beam is guided to the sample under test so that the first interference beam interacts with the sample under test to obtain the second interference beam; By using multiple modulation patterns, the intensity of the second interference light is modulated to obtain multiple beams of third interference light; Photoelectric detection is performed on multiple beams of the third interference light to obtain multiple modulation interferograms; and Compressed sensing imaging is performed based on multiple modulation patterns and multiple modulation interferograms to obtain spectral images.

Images