請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/51392
標題: | 以高光譜影像技術檢測稻米新鮮度之研究 Evaluation of Rice Freshness Using Hyperspectral Imaging Techniques |
作者: | Chih-Chung Liou 劉執中 |
指導教授: | 廖國基(Kuo-Chi Liao) |
共同指導教授: | 陳世銘(Suming Chen) |
關鍵字: | 高光譜影像,液晶可調式濾鏡,稻米新鮮度, Hyperspectral Imaging,Liquid Crystal Tunable Filter,Rice Freshness, |
出版年 : | 2016 |
學位: | 碩士 |
摘要: | 高光譜影像系統能將空間和光譜資訊做結合,取得高辨識且連續性的影像數據,利用影像像素之概念,來描述高光譜影像上每個像素的光譜反應與待側物的含量關係,本研究開發了高光譜影像系統,並且以此系統進行稻米新鮮度的檢測。
系統硬體可分成影像擷取系統以及取像系統兩大部分。取像系統有封閉、全黑的暗箱,搭配全波段的鹵素光源,建立高光強度、高均勻度的光譜影像量測平面。影像擷取系統主要是以液晶可調式濾鏡、CCD、主鏡頭、relay lens、延伸環、調焦環所組成,經過光學路徑之計算,設計出影像擷取系統。本系統的主鏡頭裝於待測物與液晶可調式濾鏡之間,這能大幅增加系統的可視範圍。本研究開發系統軟體以LabVIEW來控制液晶可調式濾鏡與CCD,可以完成450nm至1100nm的高光譜影像影像之擷取,並進行光場校正、空間校正,使擷取到每粒稻米樣本都具有分析化學成份能力的光譜資訊。 本研究使用修正型部分最小平方迴歸(modified partial least linear regression, MPLSR)的分析方法建立與稻米新鮮度的高光譜影像檢量模式。實驗樣本為多期作之台南11號稻米,在酸鹼值的分析結果中,最佳檢量線的相關係數r值可達0.804,SEC為0.239,而脂肪酸度值的分析結果中,最佳的檢量線,其相關係數r為0.786,SEC為0.362 (mg KOH/100g)。pH値比起脂肪酸度值,有較好的預測能力,而代入檢量線的成份值顯示之色彩分佈影像則具有快速辨識檢測的功能。 本研究設計之高光譜系統配合MPLSR分析建立稻米pH值以及脂肪酸度之檢量模式,藉此預測稻米之新鮮度,以符合快速、非破壞性的檢測目標。 Hyperspectral imaging system can integrate spatial and spectral information to obtain image data with high identifiability and continuity using the concept of image pixels to describe the relationship between the spectral response of each pixel and the ingredient of samples. Hyperspectral imaging system (HSIS) was developed in this study to analyze the freshness of rice. The hardware of HSIS system could be divided into two parts: image acquisition system and measurement chamber. Measurement chamber including a closed-dark chamber and a light source with halogen lamps to provide an even lighting environment; image acquisition system was equipped with the Liquid Crystal Tunable Filter (LCTF), CCD, lens, relay lens, extension tube and focus ring. The main lens was mounted between the LCTF and the sample; this arrangement could significantly increase the visual range of the system. The LabVIEW programs were developed in this study to control both the LCTF and CCD to obtain the hyperspectral images (450-1100nm) of each rice kernel with spatial correction and flat field correction. Modified partial least linear regression (MPLSR) was employed to establish the calibration models for the pH value and fat acidity value of the rice. The variety of rice for this experiment was Tainan No.11. The analysis results of MPLSR model for pH value of r, SEC were 0.804, 0.239 respectively. The analysis results of MPLSR model for fat acidity value of r, SEC were 0.786, 0.362 respectively. A pseudo color diagram could be provided to exhibit the spatial distribution of pH or fat acidity of rice samples. The hyperspectral image system has been successfully developed in this study, and calibration models for the determination of pH values and fat acidity value of rice were well established. The system could be used to predict the freshness of rice in a rapid and non-destructive way. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/51392 |
全文授權: | 有償授權 |
顯示於系所單位: | 生物機電工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-105-1.pdf 目前未授權公開取用 | 2.8 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。