Optical fiber hyperspectral imaging system and method based on single-pixel calculation imaging

Abstract

The invention discloses an optical fiber hyperspectral imaging system and method based on single-pixel calculation imaging. The system comprises a structured light projection module, an imaging object, a compound eye type optical fiber probe, an optical fiber spectrum detection module and a data acquisition and processing module. Single-pixel calculation imaging and optical fiber spectrum detection are combined, joint acquisition of spatial information and spectrum information is achieved on the premise that the internal structure of the optical fiber spectrum detection module is not changed, and therefore a traditional optical fiber spectrometer has the hyperspectral imaging capacity. By introducing a compound eye type optical fiber probe structure, a plurality of optical fiber sub-units are combined according to a predetermined spatial layout, so that different optical fiber sub-units have differences in spatial orientation, target optical signals from different spatial directions are received respectively, and multi-direction optical signals are converged and received under a single-point spectrum detection condition, so that the detection accuracy is improved. Therefore, the equivalent optical signal receiving field of view of the system is expanded. Meanwhile, in combination with a single-pixel calculation imaging mechanism, spectrum distinguishing and spatial reconstruction are realized at a sampling rate lower than a full-space sampling condition, so that the data acquisition amount is reduced, and the data redundancy is reduced.

Description

Technical Field

[0001] This invention belongs to the field of hyperspectral imaging technology, and particularly relates to a fiber optic hyperspectral imaging system and method based on single-pixel computational imaging. Background Technology

[0002] Hyperspectral imaging is an imaging technique that integrates imaging science and spectroscopy, enabling the simultaneous acquisition of spectral features and spatial images, offering the advantage of combined image and spectrum. By capturing the reflection or radiation characteristics of a target object at different wavelengths, hyperspectral imaging technology can reveal information such as the target object's chemical composition and physiological state, thus finding wide application in fields such as remote sensing, agriculture, and medical imaging.

[0003] Fiber optic spectrometers offer advantages such as small size, high spectral resolution, and wide spectral coverage. They typically use optical fibers to introduce the light signal to be measured into a spectral analysis module to detect the spectral information of the target. However, fiber optic spectrometers can only acquire single-point spectral information and lack spatial resolution, making it difficult to meet the imaging requirements of simultaneously acquiring the spatial distribution and spectral characteristics of a target.

[0004] To achieve the joint acquisition of spatial and spectral information, traditional hyperspectral imaging systems typically employ pushbroom, spot-scan, or snapshot imaging structures, constructing hyperspectral data cubes through area array detectors, scanning mechanisms, or spectral encoding devices. However, these systems generally suffer from complex structures, large sizes, and high manufacturing and maintenance costs. Furthermore, they require the acquisition of a large amount of spectral data during imaging, which can easily lead to data redundancy and low imaging efficiency. Under low-light or specific wavelength conditions, their imaging efficiency and signal-to-noise ratio are also limited.

[0005] In recent years, computational imaging techniques based on single-point detectors have gradually attracted attention. These techniques do not rely on spatially resolved area array detectors; imaging can be achieved simply by combining structured modulation and computational reconstruction algorithms, offering potential advantages in system simplification and low-light imaging. However, existing related solutions mostly focus on building independent imaging systems and are still difficult to integrate directly with mature commercial fiber optic spectrometers. Furthermore, when using a single optical fiber as the optical signal receiving unit, the effective imaging field of view remains limited by the fiber's numerical aperture and receiving angle, making it difficult to meet the application requirements of large-field-of-view hyperspectral imaging. Summary of the Invention

[0006] To overcome the limitations of existing fiber optic spectrometers, such as their ability to acquire only single-point spectral information, limited imaging field of view, and the complex structure, high cost, and significant data redundancy of traditional hyperspectral imaging systems, this invention proposes a fiber optic hyperspectral imaging system and method based on single-pixel computational imaging. This invention utilizes the principle of single-pixel computational imaging, combined with a non-area array single-point spectral detection method, to achieve the joint acquisition of spatial and spectral information without altering the internal optical structure of the fiber optic spectral detection module, thereby enabling traditional fiber optic spectrometers to possess hyperspectral imaging capabilities. By employing a combination of single-point detection and computational reconstruction, this invention avoids dependence on high-density array detectors and complex optical beam splitting components, effectively reducing system complexity and implementation cost. Simultaneously, the use of spectral integration improves the efficiency of light signal utilization under low-light conditions, which is beneficial for improving the imaging signal-to-noise ratio. Furthermore, by leveraging the sparsity of the target spectral information, spectral differentiation and spatial reconstruction that meet imaging requirements can be achieved at sampling rates lower than those used for full-space sampling, thereby reducing the amount of data acquisition and data redundancy generated during hyperspectral imaging. Further, this invention introduces a compound-eye fiber optic probe structure, combining multiple fiber subunits according to a predetermined spatial layout, ensuring that different fiber subunits have different spatial orientations, thus corresponding to different incident angle receiving directions. By converging and receiving target light signals from different spatial directions through multiple fibers, the effective optical signal receiving field of view of the system is expanded from the receiving angle of a single fiber to the combined range of receiving angles from multiple fibers, overcoming the technical bottleneck of limited field of view in single-fiber imaging. While maintaining the system's miniaturization, structural simplicity, and single-point spectral detection characteristics, this invention achieves an effective expansion of the imaging field of view, meeting the application requirements of large-field-of-view hyperspectral imaging. This invention has the advantages of simple system structure, low cost, small size, low data redundancy, strong low-light imaging capability, and large imaging field of view, demonstrating good engineering feasibility and application prospects.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A fiber optic hyperspectral imaging system based on single-pixel computational imaging includes: Structured light projection module 1: used to project a time-varying structured light field onto the imaging object 2 to spatially modulate the target scene; Compound eye fiber optic probe 3: includes multiple fiber sub-units (4), each fiber sub-unit (4) is combined according to a predetermined spatial layout to form a compound eye-like structure. Different fiber sub-units have different spatial orientations, thus corresponding to different incident angle receiving directions. It is used to perform non-imaging convergence reception of target light signals from different spatial directions under single-point spectral detection conditions, so as to expand the optical signal receiving field of view of the system. Fiber optic spectral detection module 5: connected to the compound eye fiber optic probe 3, used to perform spectral integration and measurement on the received target light signal, and output spectral intensity information as a single-point spectral detection unit; Data acquisition and processing module 6: It is used to control the structured light projection module 1 to load and output a preset spatial modulation sequence, and to synchronously process the spectral intensity information corresponding to different spatial modulation states at a sampling rate lower than the full space sampling condition. Combined with the spatial modulation sequence, the spectral differentiation and spatial reconstruction of the imaging object 2 are realized through a single-pixel computational imaging algorithm, thereby reducing the data redundancy generated during hyperspectral imaging.

[0008] The structured light projection module 1 is a projection-type structured light projection device, such as a digital micromirror array modulator or a liquid crystal spatial light modulator, and the spatial modulation sequence loaded in the structured light projection module 1 is a random matrix, a Hadamard matrix, or a Fourier matrix.

[0009] The receiving angles of different fiber subunits 4 in the compound eye fiber optic probe 3 are at least partially different or partially overlapping, so as to achieve converged reception of target optical signals from multiple spatial directions.

[0010] The fiber subunit 4 is a multimode fiber, and the fiber subunit 4 is combined in a concentric, array, or irregular distribution manner.

[0011] The fiber optic spectral detection module 5 is a commercial fiber optic spectrometer, and its internal optical structure does not need to be changed.

[0012] The spectral resolution of the fiber optic spectral detection module 5 is adjustable to meet the hyperspectral imaging requirements of different imaging objects.

[0013] The system can cover a wide spectral range from ultraviolet to near-infrared, and is suitable for hyperspectral imaging scenarios such as plant leaf classification, medical tissue detection, and industrial material analysis.

[0014] A fiber optic hyperspectral imaging method based on single-pixel computational imaging includes: Step 1: Use the structured light projection module (1) to project a time-varying structured light field onto the imaging object (2) to spatially modulate the imaging object (2); Step 2: Receive the reflected or transmitted light generated by the imaging object (2) under the structured light field through the compound eye fiber optic probe (3), and transmit the received light signal to the fiber optic spectral detection module (5). Step 3: Pre-set the sampling rate according to the number of colors or the number of spectral regions of the imaging object (2) to complete the acquisition of spectral intensity information corresponding to different spatial modulation states under low sampling rate; Step 4: Combine the collected spectral intensity information with the corresponding spatial modulation sequence, and use a single-pixel computational imaging algorithm to perform spectral differentiation and spatial reconstruction of the imaged object (2).

[0015] In step 3, the low sampling rate is set according to the number of colors or the number of spectral regions of the imaging object (2) to reduce the amount of data acquisition required during hyperspectral imaging, thereby reducing data redundancy and improving imaging efficiency.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Compared with existing hyperspectral imaging technologies that rely on area array detectors or multi-channel imaging devices, this invention introduces a system architecture that combines single-pixel computational imaging with fiber optic spectral detection, thereby effectively decoupling the spatial modulation process from the spectral detection process. Without changing the internal structure of the fiber optic spectral detection module, it achieves the joint acquisition of spatial and spectral information of the target scene, thus simplifying the system composition, reducing hardware complexity, and improving the system's versatility and scalability.

[0017] 2. This invention employs a compound-eye fiber optic probe structure, combining multiple fiber optic subunits according to a predetermined spatial layout. Different fiber optic subunits exhibit spatial orientation differences, corresponding to different incident angles and receiving directions. This enables the system to perform non-imaging convergence reception of target light signals from different spatial directions under single-point spectral detection conditions. While maintaining system compactness and structural simplicity, it expands the effective optical signal receiving field of view from the single fiber to a combined range of multiple fiber receiving angles, thereby overcoming the problem of limited field of view for single-fiber reception and improving the system's applicability in large field-of-view imaging and low-light conditions.

[0018] 3. This invention employs a single-pixel computational imaging mechanism to acquire and reconstruct spectral signals corresponding to different spatial modulation states at a sampling rate lower than that of full-space sampling conditions. While ensuring spectral discrimination capability, it reduces the amount of data acquisition and lowers the data redundancy generated during hyperspectral imaging, thereby improving imaging efficiency. It is suitable for applications requiring rapid detection and real-time performance.

[0019] In summary, this invention combines single-pixel computational imaging with a fiber optic spectrometer, achieving joint imaging of spatial and spectral information without the need for complex array detectors and additional beam-splitting optical structures. Furthermore, it realizes imaging field of view expansion based on multi-directional light signal convergence and spectral discrimination capability under low sampling rate conditions. It has advantages such as simple system structure, low data volume, high imaging efficiency, and strong engineering feasibility. Attached Figure Description

[0020] Figure 1This is a schematic diagram of the fiber optic hyperspectral imaging system for single-pixel computational imaging provided by the present invention.

[0021] Figure 2 This is a schematic diagram of the compound eye fiber optic probe module provided by the present invention.

[0022] Figure 3 This is a schematic diagram of the data acquisition and processing module.

[0023] In the diagram: 1-structured light projection module, 2-imaging object, 3-compound eye fiber optic probe, 4-fiber optic subunit, 5-fiber optic spectral detection module, 6-data acquisition and processing module. Detailed Implementation

[0024] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0025] like Figure 1 As shown, a fiber optic hyperspectral imaging system and method based on single-pixel computational imaging includes: Structured light projection module 1: used to project a time-varying structured light field onto the imaging object 2 to achieve spatial modulation of the target scene. It can use a digital micromirror array modulator or a liquid crystal spatial light modulator to achieve the spatial light modulation function. In this embodiment, the structured light projection module 1 uses a commercial projector, model EPSON CB-U05. This projector is based on 3LCD projection technology and can stably output a structured light field in the visible light band, meeting the requirements of single-pixel computational imaging for spatial modulation stability and brightness. The projector can project a pre-loaded spatial modulation sequence onto the surface of the imaging object 2 in the form of structured light, realizing temporal spatial encoding of the target scene; The spatial modulation sequence can be a random matrix, a Hadamard matrix, or a Fourier matrix; In this embodiment, the spatial modulation sequence adopts a Fourier matrix with a size of 128*128; Imaging object 2: Located on the projection light path of structured light projection module 1, used to reflect or transmit structured light field. The imaging object 2 is a target sample with spatial structure and spectral differences. Compound-eye fiber optic probe 3: used to receive light signals reflected or transmitted by the imaging object 2, and transmit the light signals to the fiber optic spectral detection module 5 through optical fiber; like Figure 2 As shown, the compound eye fiber optic probe 3 is composed of multiple fiber sub-units 4. Each fiber sub-unit 4 is combined according to a predetermined spatial layout to form a compound eye-like structure, so that different fiber sub-units correspond to different spatial orientation directions. This is used to converge and receive target light signals from multiple spatial directions under single-point spectral detection conditions, thereby expanding the optical signal receiving field of view of the system.

[0026] In this embodiment, the compound-eye fiber optic probe 3 includes a 1×16 fiber optic interface and a hemispherical outer shell structure. Sixteen multimode fibers are evenly distributed and fixed on the hemispherical shell and connected to the 1×16 fiber optic interface. The fiber optic subunits are arranged in a multi-layer concentric structure, including one central fiber, six intermediate layer fibers, and nine outer layer fibers. Different layers of fibers have different spatial orientations relative to the imaging object, thus corresponding to different receiving direction ranges, as shown in the attached figure. Figure 2 As shown; Fiber optic spectral detection module 5: used to perform spectral measurement on the optical signal transmitted by the compound eye fiber optic probe 3 and output the corresponding spectral intensity information; In this embodiment, the fiber optic spectral detection module 5 is a commercial fiber optic spectrometer, model LBTEK AMOS. The fiber optic spectrometer has 2048 spectral channels, enabling high-resolution spectral sampling of the input optical signal and outputting spectral data as a single-point spectral detection unit.

[0027] Data acquisition and processing module 6: used to acquire, store and process the spectral signals output by fiber optic spectral detection module 5, and realize spectral imaging reconstruction based on single-pixel computational imaging; The data acquisition and processing module 6 is a computer; The processing procedure is as follows: like Figure 3 As shown, after data acquisition is completed, the raw spectral data acquired by the fiber optic spectral detection module is input into the data acquisition and processing module 6. The raw spectral data is a two-dimensional data matrix with dimensions N×M, where N represents the number of spectral channels and M represents the length of the spatial modulation mode sequence.

[0028] In this embodiment, the number of spectral channels N is 2048, corresponding to the number of spectral sampling channels of the fiber optic spectrometer. The spatial modulation mode sequence length M is 8000, corresponding to the number of spatial modulation patterns played at a sampling rate lower than the full-space sampling condition.

[0029] Subsequently, the data processing module 5 performs channel-by-channel separation processing on the two-dimensional spectral data along the spectral channel dimension, sequentially extracting 2048 sets of spectral intensity sequences. Each set of spectral intensity sequences has a length of 8000, and these sequences are used as input data for single-pixel computational imaging algorithms. For the data of each spectral channel, a four-step phase-shift Fourier single-pixel imaging algorithm or other single-pixel computational imaging reconstruction algorithms can be used for processing. Specifically, based on the phase shift relationship between Fourier basis modulation patterns, the spectral intensity signals acquired under different spatial modulation states are calculated to obtain the complex Fourier coefficients corresponding to each spatial frequency component, thereby obtaining the two-dimensional Fourier spectrum of the target scene under that spectral channel.

[0030] After obtaining the complete Fourier spectrum, an inverse Fourier transform is performed to reconstruct the spatial reflectance distribution image corresponding to that spectral channel. The above processing steps are performed on all 2048 spectral channels, and they are arranged and combined according to the spectral channel order to finally obtain a three-dimensional hyperspectral image of the target scene.

[0031] A fiber optic broadband multispectral imaging method based on single-pixel computational imaging includes: Step 1: Project a spatially modulated structured light field onto the imaging object (2) through the structured light projection module (1), and load a preset spatial modulation matrix to generate a time-varying structured light field to spatially modulate the imaging object (2); Step 2: Determine the undersampling ratio of the spatial modulation mode based on the number of spectral regions in the target to be imaged, and play the corresponding number of spatial modulation patterns under conditions lower than full sampling; wherein, the spatial modulation matrix is ​​a random matrix, a Hadamard matrix, a Fourier matrix or a combination thereof, and a Fourier matrix is ​​used in this embodiment; Step 3: During each spatial modulation mode, the optical signal reflected or transmitted by the imaging object (2) is received by the compound eye fiber optic probe (3), and the optical signal is transmitted to the fiber optic spectral detection module (5) for spectral measurement; wherein, the integration time of the fiber optic spectral detection module (5) is set to 100. Step 4: Collect and store the spectral intensity data obtained by the fiber optic spectral detection module (5) in each spatial modulation mode to the data acquisition and processing module (6) to form raw spectral data containing multiple spectral channels and corresponding spatial modulation mode sequences; Step 5: The data acquisition and processing module (6) processes the original spectral data by spectral channel, and under low sampling rate conditions, uses a single-pixel computational imaging reconstruction algorithm to reconstruct the data corresponding to each spectral channel to obtain the spatial image of the target under different spectral channels, thereby obtaining the multispectral image of the target scene.

Claims

1. A fiber optic hyperspectral imaging system based on single-pixel computational imaging, characterized in that, include: Structured light projection module (1): used to project a time-varying structured light field onto the imaging object (2) to spatially modulate the target scene; Compound eye fiber optic probe (3): includes multiple fiber sub-units (4), each fiber sub-unit (4) is combined according to a predetermined spatial layout to form a compound eye-like structure. Different fiber sub-units have different spatial orientations, thus corresponding to different incident angle receiving directions. It is used to perform non-imaging convergence reception of target light signals from different spatial directions under the condition of keeping the single spectral detection channel unchanged, so that the effective light signal receiving field of view of the system is expanded from the receiving angle of a single fiber to the combined range of receiving angles of multiple fibers. Fiber optic spectral detection module (5): connected to the compound eye fiber optic probe (3), used to perform spectral integration and measurement on the received target light signal, and output spectral intensity information as a single-point spectral detection unit; Data acquisition and processing module (6): Used to control the structured light projection module (1) to load and output a preset spatial modulation sequence, and synchronously process the spectral intensity information corresponding to different spatial modulation states at a sampling rate lower than the full space sampling condition. Utilizing the correspondence between the multi-directional light signal superposition result formed by the compound eye fiber probe under the single spectral intensity output condition and the spatial modulation sequence, the spectral differentiation and spatial reconstruction of the imaging object (2) are realized through a single pixel computational imaging algorithm, thereby reducing the data redundancy generated during hyperspectral imaging.

2. The fiber optic hyperspectral imaging system based on single-pixel computational imaging according to claim 1, characterized in that: The receiving angles of different fiber subunits (4) in the compound eye fiber probe (3) are at least partially different or partially overlapping, so that each fiber subunit corresponds to a different spatial receiving direction, thereby realizing the convergence reception of target optical signals from multiple spatial directions.

3. A fiber optic hyperspectral imaging system based on single-pixel computational imaging according to claim 1 or 2, characterized in that: The structured light projection module (1) is a projection-type structured light projection device, such as a digital micromirror array or a liquid crystal spatial light modulator, and the spatial modulation sequence loaded in the structured light projection module (1) is a random matrix, a Hadamard matrix, or a Fourier matrix.

4. The fiber optic hyperspectral imaging system based on single-pixel computational imaging according to claim 1, characterized in that: The fiber optic spectral detection module (5) is a commercial fiber optic spectrometer, and its internal optical structure does not need to be changed.

5. A fiber optic hyperspectral imaging system based on single-pixel computational imaging according to claim 1, characterized in that: The fiber subunit (4) is a multimode fiber, and is combined in a concentric, array, or irregular distribution manner to achieve convergence and reception of multi-angle optical signals while keeping the single-point spectral detection structure unchanged.

6. The fiber optic hyperspectral imaging system based on single-pixel computational imaging according to claim 1, characterized in that: The spectral resolution of the fiber optic spectral detection module (5) is adjustable to meet the hyperspectral imaging requirements of different imaging objects.

7. The fiber optic hyperspectral imaging system based on single-pixel computational imaging according to claim 1, characterized in that: The system can cover a wide spectral range from ultraviolet to near-infrared, and is suitable for hyperspectral imaging scenarios such as plant leaf classification, medical tissue detection, and industrial material analysis.

8. A fiber optic hyperspectral imaging method based on single-pixel computational imaging, characterized in that: Step 1: Use the structured light projection module (1) to project a time-varying structured light field onto the imaging object (2) to spatially modulate the imaging object (2); Step 2: Receive the reflected or transmitted light generated by the imaging object (2) under the structured light field illumination through a compound-eye fiber optic probe (3), and transmit the received light signal to the fiber optic spectral detection module (5). This is in accordance with the fiber optic hyperspectral imaging method based on single-pixel computational imaging as described in claim 8. The low sampling rate is set according to the number of colors or spectral regions of the imaging object (2) to reduce the amount of data acquisition required during hyperspectral imaging, thereby reducing data redundancy and improving imaging efficiency. Step 3: Set the sampling rate according to the number of colors or the number of spectral regions of the imaged object (2); Step 4: Combine the collected spectral intensity information with the corresponding spatial modulation sequence, and use a single-pixel computational imaging algorithm to perform spectral differentiation and spatial reconstruction of the imaged object (2).

9. The fiber optic hyperspectral imaging method based on single-pixel computational imaging according to claim 8, characterized in that: The low sampling rate is set according to the number of colors or spectral regions of the imaging object (2) to reduce the amount of data acquisition required during hyperspectral imaging, thereby reducing data redundancy and improving imaging efficiency.

Images