| 摘要: | 紅外光光譜影像系統可提供物質中分子組成的二維空間分布資訊，因此這樣的系統已被應用於許多領域中，如食物品質分析。為了拍攝物體於紅外光波段的影像，常見的做法是以二維陣列偵測器或單點掃描的干涉儀架構取得物體影像，但二維陣列偵測器在紅外光波段的價格通常昂貴，且干涉儀架構通常對系統穩定度的要求較高，因此紅外光光譜影像系統較難以普及於日常生活中的應用。
為了突破上述限制，本研究架設一結合散射介質與單像素成像技術的光譜影像系統，以單像素成像還原帶有物體空間與光譜資訊的光斑影像後，再依靠反摺積重建物體光譜影像。為了測試系統架構的可行性，目前以可見光波段進行實驗。當點光源通過散射介質後，於成像面上產生的光斑圖案具有記憶效應與光譜去相關性，因此在取得物體於成像面形成的光斑影像後，便可利用反摺積還原物體光譜影像。本研究成功以此作法還原單色物體影像與三色物體的光譜影像。接著模擬系統以單像素成像還原物體光斑影像的過程，並以反摺積成功重建單色物體影像。但實際以系統進行單像素成像與反摺積的影像重建流程後，發現系統無法重建單色物體影像。在先前模擬的流程中加入雜訊後，我們確認影像還原過程中的雜訊影響物體影像重建。未來若能提高入射系統的光強度，或提升偵測器的靈敏度，便有機會提高訊雜比，減少雜訊對物體影像重建的影響。
;An infrared (IR) spectral imaging system is able to provide information on the two-dimensional spatial distribution of the molecular composition of an object, and there has been many applications of such a system, for example food quality analysis. An image of the object in the IR region is commonly taken by a two-dimensional focal plane array (FPA) or a single-point scanning interferometer, but an IR FPA is often expensive, and an interferometer often requires stable environmental conditions; thus, it is harder for an IR spectral imaging system to be commonly applied to applications in daily life.
To overcome these limitations, a system combining a scattering medium and single-pixel imaging technique is set up in our research. After recovering the speckle pattern containing the spatial and spectral information of an object, the spectral image of the object can be reconstructed by deconvolution. To test the feasibility of the system setup, the experiment is carried out in the visible region. After a point light source passes through a scattering medium, the speckle pattern formed in the image plane has the property of memory effect and spectral decorrelation; thus, the spectral image of an object can be recovered by deconvolution after obtaining the speckle image formed by an object in the image plane. The image of a monochromatic object and the spectral image of an object with red, green, and blue colors are successfully recovered by the proposed method in our research. By simulating the process of recovering the speckle image of an object by a system using single-pixel imaging, an image of the monochromatic object is reconstructed by the deconvolution. However, after performing the process of image reconstruction using single-pixel imaging and deconvolution by a real system, it was discovered that the system is unable to recover the image of a monochromatic object. After adding noises into the previous simulation process, we confirmed that the reconstruction of an object’s image is affected by noises in the process of recovering an image. If the intensity of light illuminating the system is higher, or the sensitivity of the detector is improved, it might be possible to improve the signal-to-noise ratio; thus, the effect of noises on the reconstruction of an object’s image can be reduced. |
