| 摘要: | 在研究物質分子組成中，紅外光譜影像系統可提供波長與二維空間的資訊，因此已廣泛使用在諸多領域中。現今普遍的紅外光譜影像系統多使用陣列偵測器或空間掃描式干涉架構來獲得光譜影像，但紅外光波段中的陣列偵測器價格通常昂貴，而使用干涉技術的架構易受到環境擾動的影響，這些阻礙將不利於紅外光譜影像系統在生活上的應用和研究上的普及。
為了打破上述的阻礙，因此本研究提出一單像素成像與散射介質的單照式多光譜影像系統，藉由光通過散射介質後的光斑具有空間相關性與光譜去相關性，以單像素成像重建的光斑會帶有物體波長與空間資訊的特性，並利用反摺積來還原物體的光譜影像。在系統第一階段的開發中，以可見光波段進行實驗，成功以陣列相機拍攝光斑並還原物體光譜影像，但使用單像素成像重建光斑並還原後的結果卻不理想。因此在第二階段研究中，對既有的系統提出待改善的問題與優化的方法，並以單像素相機來取代陣列相機並還原物體影像作為研究的核心目標。
優化後的系統除了降低人為操作的誤差、提升還原影像的品質，更結合了一套校正系統，來解決光斑成像在不同相機間放大率的問題。在第二階段的研究成果中，利用陣列相機知悉系統光斑的相關參數以外，並順利使用單像素相機拍攝可見顆粒分布的光斑影像，也成功還原單色光物體和多色光物體的光譜影像，然而卻發現系統對於分辨不同波長時，結果沒有預期來的好。最後也透過模擬和實驗來確信，系統上不同波長的光斑彼此存在過高的關聯性，並且以不同壓縮比率對光斑影像進行調整能影響反摺積後所還原的影像品質。因而在未來的研究中，若能調整系統元件的相關參數，便有機會降低不同波長間光斑的高關聯性，並提升系統對於分離波長的能力。
;In the study of molecular composition of an object, an infrared spectral imaging system can provide the information about the object of wavelength and two-dimensional space, so the system has been widely used in many research fields. In today’s infrared spectral imaging systems mostly use an array detector or a spatial scanning interferometer to acquire spectral images, however, an array detector in infrared region is usually expensive, and an interferometer is easily affected by disturbances of the environment. Thus, these obstacles are disadvantages of an infrared spectral imaging system to commonly apply for daily life and at the research.
To breakthrough these obstacles, a snapshot multispectral imaging system based on single pixel imaging and scattering medium is set up, thanks to the spatial correlation and the spectral de-correlation on the speckle since the lights pass through the scattering medium, the recovered speckle pattern using by the single-pixel imaging will have the wavelength and spatial information of the object, then spectral image of the object can be reconstructed by the deconvolution. The development of the system in the first stage, it is successfully reconstructed the spectral image of the object by using array detector to capture the speckle pattern in the visible light region. Nevertheless, the recovered speckle pattern using by the single-pixel imaging or the reconstruction result is not ideal. Therefore, the primary goal in the second stage, the existing system will be improved and optimized, the single-pixel detector is used to replace the array detector when captures speckle and then reconstructed image.
The optimized system not only reduces the error of human operation and increase the quality of the reconstruction image, also incorporates a calibration system to solve the problem of magnification when the speckle is imaging between different cameras. In the results of the second stage, the array camera is used for knowing the relevant parameters of the speckle pattern on the system, and the speckle pattern is successfully recovered by single-pixel imaging, also well reconstructed the spatial image of the monochromatic and polychromatic objects. However, the system is found that when distinguishing different wavelengths, the results arn′t as good as expected. Finally, through simulation and experiments, it′s confirmed that the speckles of different wavelengths on the system are too highly correlated with each other, and proof that adjusting the speckle image with different compression ratios can affect the quality of the reconstruction image after the deconvolution. Therefore, in the future, if the system components could be adjusted, there will be an opportunity to decrease the high correlation of the speckle image between different wavelengths and improve the ability of the system for separating wavelengths. |
