Open Access
Add to BibSonomy
Abstract
We design and construct a broadband integrated multi-channel imaging spectrometer (MCIS) from visible light to near-infrared. This system can directly obtain spectral images that conform to the consistent visual habits of the human eyes through a single exposure of the detector. The genetic algorithm is used to calculate system parameters to minimize pixel waste between spectral channels, achieving nearly 100% utilization of detector pixels. The field stop suppresses stray light in the system. This device is used for imaging an optical-resolution target, an object, and a furnace to verify the basic principles of the system. The results indicate that the system can effectively utilize detectors to monitor high-temperature objects in the visible to near-infrared wavelength range.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Combustion flame monitoring, an important field in today’s world engine combustion diagnosis, refers to determining the stage of combustion reaction by obtaining information such as flame intensity, spectrum, shape and other information, and then analyzing the composition of fuel to control the combustion process. Combustion flame monitoring is of great significance for protecting the safety of combustion equipment, reducing the greenhouse effect, and improving fuel combustion efficiency [1–3]. As an important part of combustion flame monitoring, optical sensing cameras play an irreplaceable role in providing numerical data for flame state measurement, providing visual perception for scientific research, and helping to study fuel composition [4–7].
As technology advances, the functions of optical cameras used for combustion flame monitoring continue to expand, and imaging spectrometers that can simultaneously obtain two-dimensional (2D) spatial and spectral information of objects are used for combustion flame monitoring [8–10]. For imaging spectrometers, the main evaluation indicators for their performance include imaging resolution, time resolution, spectral resolution, etc. [11–13]. Monitoring combustion flames is essentially real-time monitoring of dynamic targets, which requires imaging spectrometers to have high temporal resolution.
Imaging spectrometers can be divided into scanning imaging spectrometers and snapshot imaging spectrometers based on their data collection methods [14]. Scanning imaging spectrometers generally need to scan the scene in spatial or spectral dimensions to obtain complete spatial and spectral information about the target scene [15–18]. Due to motion artifacts, it is difficult for scanning imaging spectrometers to obtain accurate information about dynamic targets. The snapshot imaging spectrometer can obtain all two-dimensional (2D) spatial information and one-dimensional (1D) spectral information through a single exposure, which meets the requirements of real-time monitoring of dynamic targets.
Researchers have proposed various snapshot imaging spectrometers for dynamic target monitoring. Among them, indirect measurement snapshot imaging spectrometers can effectively balance the contradiction between spatial resolution and spectral resolution, and are easy to achieve miniaturization. However, their processing or calculation principles are usually complex [19]. For example, Jian Xiong et al. designed an imaging spectrometer based on the principle of compressed encoded aperture (CASSI) using reconfigurable metasurfaces, which has high time and spectral resolution. However, due to the need to directly integrate the metasurfaces onto the detector surface, this imaging spectrometer has high requirements for alignment accuracy [20]. Pengwei Zhou et al. designed a computer tomography spectrometer based on metamaterial surfaces, which has high energy utilization efficiency. But its time resolution is limited by the time complexity of the algorithm being solved [21].
The algorithm principle of direct measurement snapshot spectral imaging is simple, and the time resolution of the system is also less limited by the time complexity of the algorithm [20]. For example, a Fourier transform spectrometer designed by Yupeng Chen has good real-time performance and a small error. However, due to the need for two paths of light to form an interference image, the system's structure is not compact [10]. Yiqun JI et al. studied the integrated field spectrum of adding a microlens array at the first image plane. This structure has a compact imaging structure and can select a variety of structural forms for light splitting after the microlens array. However, the gaps between microlens arrays inevitably lead to information loss [22,23]. Feng Huang et al. studied a multi-aperture imaging spectrometer using an array of nine cameras, which improved the light efficiency of traditional imaging spectrometers based on bandpass filters. However, the use of camera arrays significantly increases the volume and weight of the system, making the imaging spectrometer less portable [24].
Imaging spectrometers can improve the information acquisition capability of combustion flame monitoring systems. By clearly identifying the gaps in existing publications, existing research has not proposed an integrated multi-channel imaging spectrometer with a wide spectrum from visible light to near-infrared. To realize real-time monitoring of combustion flame, we designed and constructed a multi-channel imaging spectrometer (MCIS) that can simultaneously obtain spatial and spectral information about targets. To improve the integration of the instrument, we use an aperture array (ATAR) to achieve multiple spectral channels. We designed an image plane and added a field stop before the ATAR. The purpose of doing so is to improve the efficiency of the detector and effectively suppress stray light in the system. The system has a full field of view of 26.5 ° and a single channel F-number of 4.23. After fabrication and alignment of the lenses and optical axis, a test experience with an optical resolution target and an actual target was performed.
The remainder of this paper is organized as follows: Design principles are presented in Section 2. Section 3 elaborates on optical simulations and analysis. The prototype and experimental results are presented in Section 4. Section 5 provides a brief summary of the study.
2. Method
The MCIS is composed of a front common aperture lens group (FCALG), four sub-aperture lens groups (SALG) and a rear common aperture lens group (RCALG). Each SALG contains a filter to form four spectral imaging sub-channels of MCIS. The schematic diagram of MCIS imaging principle is shown in Fig. 1. The radiation from the target is adjusted by FCALG to adjust the peripheral contour of the beam. After passing through SALG and its built-in filters, it is divided into four different spectral imaging sub-channels. Finally, the images of the four spectral imaging sub-channels are imaged on the detector through RCALG.
The initial structure was built first. A genetic algorithm is used to determine the initial structural parameters of the system. The optical parameters of the four SALGs except the filter transmittance are set to the same, so only a single channel is analyzed in the analysis process. The effective use of the detector target surface is achieved by constraining the objective function, which requires that the image formed by each channel after splicing needs to fill the whole detector target surface. The meaning of each symbol used in the analysis is shown in Fig. 2.
The filter in MCIS can be considered as a thin flat glass. The influence of this thin flat glass on the direction of system light can be ignored. Thin flat glass will not be analyzed during the initial structural analysis process. In addition, RCALG plays a role in adjusting the overall size and position of the image after various spectral imaging channels are connected in series. Therefore, we first analyzed FCALG and SCALG [25,26]. Then we analyzed the entire MCIS.
In the same imaging medium, according to the ideal imaging relationship:
where ${f_1}$ is the focal length of FCALG, ${l_1}^\prime$ is the image distance of FCALG, ${l_1}$ is the object distance of FCALG, ${l_2}^\prime$ is the image distance of SALG, and ${l_2}$ is the object distance of SALG.The relationship between the offset of SALG and the offset of the image center of the eccentric structure is:
where ${\Delta _1}$ is the offset of the image center of the eccentric structure, $\Delta$ is the offset of SALG, and ${\Delta _2}$ is the difference between $\Delta$ and ${\Delta _1}$.Then, based on the analysis of the imaging process of RCALG and the relationship between object and image size, as both the object and image of RCALG are located in the air, the vertical magnification of the RCALG is:
According to the ideal imaging relationship:
According to Eq. (1)–(6) and the geometric relationship in Fig. 2, the offset of the image center of the MCIS system can be expressed as:
In order to effectively use the detector, the offset of the image center is equal to half of the image height. Because the target size is 7.68 mm, the image size of a single spectral imaging sub-channel is 3.84 mm, so the half image height is 1.92 mm. According to the use requirements of the detector, ${l_3}^\prime$=17.5 mm. Therefore, the objective function constructed for genetic algorithm analysis is expressed as:
The objective function ${F_m}$ is optimized by genetic algorithm [27]. The specific process is:
- (1) Encode and initialize the population. Firstly, we set the parameters $\Delta ,{f_1},{f_2},{f_3},{d_{12}}$ as a gene. The random values of genes within a certain range forms a coding string called a chromosome. Each coding string represents a potential solution. Then, multiple chromosomes are generated randomly. Construct an individual so that each individual carries a chromosome, and all individuals form the initial population M.
- (2) Calculate fitness values and natural selection. In order to evaluate the advantages and disadvantages of each individual, we calculate the fitness value of each chromosome so that the fitness function fit meets:
The smaller the objective function value, the larger the individual fitness value. Then, all individuals are selected using the turntable method. When the turntable method is used for selection, the individuals with high fitness function fit have a higher probability of being selected. At the same time, in order to avoid the exclusion of good genes, individuals with low fitness function value also have a certain probability of being selected.
- (3) Crossover and variation. The selected individuals carry chromosomes for mating, that is, the corresponding genes on the chromosome are randomly exchanged. Then, the genes of some individuals are randomly selected for mutation to produce offspring chromosomes. After that, the individuals carrying sub-chromosomes form a new population, and the next generation population is generated according to steps (2) and (3).
- (4) Termination conditions and outputs. The termination condition is that the maximum evolution algebra or fitness function value set in the algorithm has not changed in several generations. If the termination condition is not met, return to the second step. If the condition is met, the output parameter is the initial structure. The flow chart of the whole algorithm is shown in Fig. 3.
3. Simulation and analysis
3.1 Optical system design and simulation
The main system parameters are shown in Table 1. The operating wavelength of the system is 510 nm to 1650 nm, the field angle is 26.5 °, and the optical length is only 235 mm. The F number of single spectral imaging sub-channel is 4.25, and the focal length is 12.8 mm.
Table 1. Specifications of the proposed optical system
We constructed the initial structure of the MCIS using the method described in section 2. To suppress stray light, a primary image plane is designed in the FCALG system. Using the primary image plane as the dividing point in FCALG, the lens in front of the primary image in FCALG is classified as FCALG-F, and the lens behind the primary image plane in FCALG is classified as FCALG-B, as shown in Fig. 4. At this point, FCALG-F, FCALG-B, four SALGs, and RCALG are set to the paraxial plane. The lens data is shown in Table 2.
Table 2. Initial structure lens data
Then the actual MCIS is designed based on the initial structure of MCIS. The MCIS is composed of FCALG, four SALGs and RCALG, as shown in Fig. 5. MCIS can synchronously acquire an object's spectral and spatial features from four different spectra. Firstly, the light entering the whole system is adjusted to a parallel light with an appropriate aperture through FCALG, providing enough space for the mechanical installation of the four SALGs. There is a primary image plane in FCALG, and there is a field aperture at the primary image plane, which not only improves the utilization of the detector target surface, but also effectively suppresses the stray light of the system. Then, the light enters the SALG. Each SALG contains a filter to form four spectral imaging sub-channels symmetrically distributed in the four quadrants of the XOY plane. The pressure ring in front of SALG is ATAR. The addition of ATAR divides the aperture of the system and achieves detector multiplexing. At the same time, the four SALGs adjust the images of each spectral imaging sub-channel so that the images of each channel form a rectangle without overlap. Finally, RCALG adjusts the size and position of the image after each channel splicing, so that the image falls on the detector and fills it.
The optical system exhibits good imaging performance. The modulation transfer function (MTF) of the system in different fields of view is shown in Fig. 6. At a Nyquist frequency of 33lp/mm, the MTF value for each working field of view is greater than 0.5. Figure 7 shows the spot diagram of the system. The root-mean-square (RMS) diameters over the full FOV are smaller than 15μm, which is smaller than 1 pixel. Figure 8 illustrates that the distortion of the full FOV in the MCIS is less than 1.47%.
3.2 Tolerance analysis
Through tolerance analysis, the feasibility of the actual processing and assembly of the optical system was evaluated. We study the influence of small machining and assembly errors on the imaging quality of the MCIS. The evaluation standard is MTF with an average of 33lp/mm. Compensation is the distance from the last side of the last lens of the RCALG to the image plane. The tolerance distribution of the design system is shown in Table 3.
Table 3. Tolerance allocations results
Figure 9 exhibits the relationship between image quality and attainable probability under the proposed tolerance allocation. The MTF is eventually better than 0.52 at 33lp/mm with a probability of 90%. This shows that the optical system has a good feasibility of practical processing and adjustment.
3.3 Stray light analysis
Stray light is a type of light that can affect imaging quality and should not be detected by detectors. For MCIS, the formation of stray light is mainly due to the circular shape of the image periphery of each channel. If the images of each spectral imaging channel are pieced together without any gaps, it will cause the images of each channel to overlap and form crosstalk. This kind of crosstalk can be solved by designing a primary image plane in FCALG and inserting a field stop at the primary image plane. Stray light analysis results are shown in Fig. 10.
Fig. 10. Suppression of stray light in the system. (a) Image irradiance distribution before adding a field stop; (b) Image irradiance distribution after adding a field stop
Figure 10(a) and Fig. 10(b) respectively show the irradiance distribution in the image plane of MCIS without and with a field stop. From Fig. 10(a), it can be seen that when there is no field stop in MCIS, due to the crosstalk of light between SALGs, there is a significantly stronger irradiance at the image plane intersection of each spectral imaging sub-channel compared to other regions. From Fig. 10(b), it can be seen that when there is a field stop in MCIS, the crosstalk between SALGs is suppressed. However, due to the blocking effect of the field stop, black edges with weaker irradiance than other areas will be generated at the intersection of the image planes of each spectral imaging sub-channel. This black edge can be reduced or even eliminated by accurately determining the size of the square hole in the middle of the field stop based on the coordinate values of the edge coordinates at the primary image plane in FCALG. In addition, the energy received by the detector is slightly reduced, which may cause the reduction of signal-to-noise ratio (SNR). Although it will not affect the current prototype's detection performance, the issue can be addressed in the future by increasing the gain or exposure time of the detector.
4. Prototype and experiment
4.1 Optomechanical structure of the device
The optical structure of the device was designed and fabricated, as shown in Fig. 11. The overall structure options a modular design, which is divided into four parts and is integrated. These four parts include the three lens groups mentioned in section 3 and a filter wheel for switching between visible and near-infrared bands. Each part is connected by the flange structure. The introduction of filter wheels increases the number of spectral channels from 4 to 8 (Near-infrared bands: (900-1300) nm, (1420-1510) nm, (1510-1590) nm, (1590-1650) nm); Visible light bands: (510-620) nm, (620-700) nm, (700-750) nm, (750-780) nm).
Fig. 11. Optical of the device contains FCALG, filter wheels, SALG, and RCALG The arrow indicators the direction of motion of the parts
4.2 Adjustment and experimental results of the device
We use the resolution of the optical system to measure the imaging performance of MCIS. Because filters have a strong attenuation effect on light and have little impact on imaging quality, resolution testing was conducted on MCIS without filters. To test the imaging performance of the system, we established an experimental setup. All optical instruments were installed on a separate optical platform, where the light source was a halogen lamp and an optical resolution target (1951 US Air Force) was placed on the focal plane of a collimator. The device was placed in the image space of a parallel light tube. After the light from the light source hits the target, it enters the MCIS through a parallel light tube and ultimately forms an image on the detector. The spatial resolution of MCIS can be calculated by the similarity relationship between objects and images of parallel light tubes.
Figure 12 shows the obtained resolution target image. In the image, the third row in the fourth group can be distinguished, indicating that the spatial resolution of the image reaches 15.2μm.
In the testing experiment of image plane splicing in four spectral imaging sub-channels, the structure of the experimental device is shown in Fig. 13. We placed the first target object 0.6 m away from MCIS and illuminated the first target with a halogen lamp. Due to the same reasons as in the resolution testing process, filters weren’t placed in the MCIS during this process. Figure 14(a) shows the imaging results of the first target object. There is a slight gap between the four channel images. To further evaluate the size of the gap, the second target object with more uniform brightness was placed at the same position as the first target for the experiment. The collected image is shown in Fig. 14(b). As depicted in Fig. 14(b), the maximum gap between the upper and lower channel images is about 6 pixels, and the maximum gap between the left and right channel images is 8 pixels.
Fig. 14. Experimental results of spectral imaging sub-channel image plane stitching test. (a) Imaging results of the first target object; (b) Imaging results of the second target object
In order to simulate combustion flame temperature monitoring, we established an experimental optical path, as shown in Fig. 15. The furnace was placed 0.6 m from the lens. We adjusted the height of MCIS so that the light spots in each spectral imaging sub-channel can be fully displayed in the image plane. The integration time of the detector is set to 2 ms. At this time, the temperature of the blackbody furnace is 1000 ℃. Figure 16(a) and Fig. 16(b) show the imaging effects in the visible and infrared bands, respectively, indicating that MCIS has the ability to monitor high-temperature objects.
Fig. 16. The images of furnace obtained in different wavebands. (a) Visible light; (b) Near infrared
5. Conclusion
In this study, we propose MCIS to obtain spatial and spectral information from visible light to near-infrared for real-time monitoring of combustion flames. The MCIS uses ATAR to achieve snapshot spectral imaging. The device not only suppresses stray light but also minimizes pixel waste between spectral channels without aliasing. The field of view angle of the system is 26.5 °, with a volume of 300 mm × 130 mm × 150 mm. Based on the tolerance and straight light analysis results, we designed the optical and coating requirements of the device. Finally, we conducted experimental research. The results have demonstrated the performance of MCIS and demonstrated its reliability in combustion flame monitoring. Compared to traditional temperature measurement systems (such as colorimetric temperature measurement systems), this device only requires one detector to obtain images of multiple spectral bands, greatly saving system costs and improving system compactness.
Funding
National Key Research and Development Program of China (2021YFC2202100).
Disclosures
The authors declare that there are no conflicts of interest related to this article.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. Y. Xia, J. Zhang, C. Tang, et al., “Research and application of online monitoring of coal and biomass co-combustion and biomass combustion characteristics based on combustion flame,” J. Energy Inst. 108, 101191 (2023). [CrossRef]
2. Z. Omiotek and A. Smolarz, “Combustion process monitoring based on flame intensity time series,” Proc. Inst. Mech. Eng. Pt. I: J. Syst. Contr. Eng. 235(6), 809–822 (2021). [CrossRef]
3. Y. Zhou, C. Zhang, X. Han, et al., “Monitoring combustion instabilities of stratified swirl flames by feature extractions of time-averaged flame images using deep learning method,” Aerosp. Sci. Technol. 109, 106443 (2021). [CrossRef]
4. J. Matthes, P. Waibel, M. Kollmer, et al., “Camera based flame stability monitoring and control of multi-burner systems using deep learning based flame detection,” Therm. Sci. Eng. Prog. 41, 101859 (2023). [CrossRef]
5. J. Matthes, M. Kollmer, M. Eberhard, et al., “A High-Speed Camera-Based Measurement System for the High-Pressure Entrained-Flow Gasification,” Chem. Eng. Technol. 45(12), 2313–2322 (2022). [CrossRef]
6. C. Zhou, S. Wang, Y. Cao, et al., “Online Pyrometry Calibration for Industrial Combustion Process Monitoring,” Processes 10(9), 1694 (2022). [CrossRef]
7. L. Li, F. Gubarev, and A. Mostovshchikov, “Synchronized Two-Camera Laser Monitor for Studying Combusting Powder Systems,” Symmetry 14(4), 656 (2022). [CrossRef]
8. F. Castillo, L. Arias, and H. O. Garcés, “Estimation of temperature, local and global radiation of flames, using retrieved hyperspectral imaging,” Measurement 208, 112459 (2023). [CrossRef]
9. T. Hong, M. Zhao, T. Qin, et al., “Reconstructing temperature field based on imaging spectrum,” Journal of Harbin Engineering University 43(7), 7 (2022).
10. Y. Chen, J. Lv, W. Yue, et al., “A Snapshot Infrared Imaging Fourier Transform Spectrometer for Dynamic Target Detection,” Remote Sens. 14(7), 1543 (2022). [CrossRef]
11. Z. Han, W. M. MacCuaig, M. N. Gurcan, et al., “Dynamic 2-deoxy-D-glucose-enhanced multispectral optoacoustic tomography for assessing metabolism and vascular hemodynamics of breast cancer,” Photoacoustics 32, 100531 (2023). [CrossRef]
12. Y. Hu, J. Chang, Z. Ji, et al., “Construction method for designing a spectrometer with variable spectral resolution and wide bandwidth using multiple off-axis convex gratings,” Opt. Express 30(2), 2472–2486 (2022). [CrossRef]
13. F. Qi and P. Zhang, “High-resolution multi-spectral snapshot 3D imaging with a SPAD array camera,” Opt. Express 31(19), 30118–30129 (2023). [CrossRef]
14. N. A. Hagen and M. W. Kudenov, “Review of snapshot spectral imaging technologies,” Opt. Eng. 52(9), 090901 (2013). [CrossRef]
15. Z. Yang, T. Albrow-Owen, H. Cui, et al., “Single-nanowire spectrometers,” Science 365(6457), 1017–1020 (2019). [CrossRef]
16. M. Faraji-Dana, E. Arbabi, A. Arbabi, et al., “Compact folded metasurface spectrometer,” Nat. Commun. 9(1), 4196 (2018). [CrossRef]
17. Y. Xu, A. Giljum, and K. F. Kelly, “A hyperspectral projector for simultaneous 3D spatial and hyperspectral imaging via structured illumination,” Opt. Express 28(20), 29740–29755 (2020). [CrossRef]
18. P. H. Cu-Nguyen, A. Grewe, P. Feßer, et al., “An imaging spectrometer employing tunable hyperchromatic microlenses,” Light: Sci. Appl. 5(4), e16058 (2015). [CrossRef]
19. L. Gao and L. Wang, “A review of snapshot multidimensional optical imaging: measuring photon tags in parallel,” Phys. Rep. 616, 1–37 (2016). [CrossRef]
20. J. Xiong, X. Cai, K. Cui, et al., “Dynamic brain spectrum acquired by a real-time ultraspectral imaging chip with reconfigurable metasurfaces,” Optica 9(5), 461–468 (2022). [CrossRef]
21. P. Zhou, J. Zhou, Y. Wang, et al., “The Computed Tomographic Imaging Spectrometer based on metamaterial surface,” Opt. Mater. 136, 113378 (2023). [CrossRef]
22. S. Zhao, C. Zeng, Y. Ji, et al., “Dual-channel snapshot imaging spectrometer with wide spectrum and high resolution,” Appl. Opt. 62(21), 5830–5838 (2023). [CrossRef]
23. Y. Ji, F. Tan, S. Zhao, et al., “Spatial-spectral resolution tunable snapshot imaging spectrometer: analytical design and implementation,” Appl. Opt. 62(17), 4456–4464 (2023). [CrossRef]
24. F. Huang, P. Lin, R. Cao, et al., “Dictionary Learning-and Total Variation-Based High-Light-Efficiency Snapshot Multi-Aperture Spectral Imaging,” Remote Sens. 14(16), 4115 (2022). [CrossRef]
25. M. Gao, C. Wang, J. Liu, et al., “Design of four-channel array polarized imaging adaptive optical system,” LASER & INFRARED 46(6), 017 (2016).
26. H. He, Y. Ji, J. Zhou, et al., “Optical Design of Decentered Aperture-Divided Polarization Imaging System,” Journal of Optics 33(6), 0622005 (2013). [CrossRef]
27. X. Fan, C. Lin, M. Xu, et al., “Design of the Optical System of Coaxial PGP Imaging Spectrometer with Eliminating Spectral Smile,” Journal of Optics 51(12), 1212005 (2022).
