透過您的圖書館登入
IP:18.223.106.232
  • 學位論文

以小波理論與非線性演算法建構光譜分析法

Spectra Analysis Using Wavelet Theory and Non-Linear Algorithm

指導教授 : 陳世銘

摘要


本研究之主要目的是建立一套以小波理論為基礎之近紅外光分析模式。以往的光譜分析模式,在光譜前處理及參數設定上步驟複雜;本研究利用小波理論擷取光譜之特徵值,可省去以往繁雜的光譜前處理。本研究是以小波理論(Wavelet Transform)為基礎進行非線性模式分析,主要的研究對象物為青脆枝中的喜樹鹼含量、鳳京白菜的氮含量、檸檬果汁酸度、桶柑果汁糖度,經由鳳京白菜乾粉光譜與氮含量相關性的多次測試,證實分析方法能確實建立近紅外光的分析預測模式。分析程式是以Matlab 6.5撰寫,利用小波理論搭配類神經網路以及基因演算法建立模式,應用Matlab的運算能力,可以達到快速分析的效果。目前本研究自行撰寫的分析模式,已具有相當的準確度,WT-ANN (ANN based on Wavelet Transform)對鳳京白菜氮含量分析結果,Rc最佳可達0.99、RSEC為11. 7%、RSEP為13.8%。對青脆枝中喜樹鹼含量之分析,結果雖不如鳳京白菜,但Rc最佳仍可達0.86、RSEC與RSEP則介於22%至45%之間。而對於檸檬酸度以及桶柑糖度的分析,亦有不錯之分析結果,Rc約可達0.9左右,檸檬酸度之RSEC、RSEP則介於1.5%至2.5%之間,桶柑糖度之RSEC、RSEP則介於6%至8%之間。AWTGA-ANN (Approved GA-ANN based on Wavelet Transform)模式,對以上四種物質的分析結果,大致上與WT-ANN相當,但部份分析結果優於WT-ANN。

並列摘要


The main purpose of this study is to establish a near infrared analysis method that based on wavelet theory (WT). Current spectral analysis procedure is complicated due to its spectrum pretreatment. Using wavelet theory to distil characteristic features in spectra can eliminate the tedious procedure in spectrum pretreatment. The specimens under investigation are Camptothecin content of Nothapodytes foetida, the nitrogen content of Pak-Choi cabbage, total acidity amount of lemon juice, and total sugar content of orange juice. In this study, the programs of analysis models were written using Matlab to conduct Artificial Neural Network (ANN) and Genetic Algorithm (GA) analysis with wavelet transformation. The developed methods show a good capability for spectra analysis. As of prediction of nitrogen content in Pak-Choi cabbage, WT-ANN (ANN based on Wavelet Transform) shows the best results as Rc=0.99, RSEC=11.7% and RSEP=13.8%. Regarding the prediction of Camptothecin content in Nothapodytes foetida, WT-ANN gives the best results as Rc=0.86, RSEC and RSEP being between 22% and 45%. The prediction of acidity amount of lemon juice and sugar content of orange juice by using WT-ANN shows that, Rc is about 0.9; RSEC and RSEP between 1.5% to 2.5% for lemon juice, while RSEC and RSEP between 6% to 8% for orange juice. AWTGA-ANN (Approved GA-ANN based on Wavelet Transform) yields similar results as those of WT-ANN, but some of them are better than WT-ANN.

參考文獻


邵泰璋,史天元。2000。類神經網路於多光譜影像分類之應用。航遙及遙測學刊。5(1):1-14。
張嘉麟。1999。近紅外光與核磁共振應用於乳粉脂肪含量之檢測。碩士論文。台北:國立臺灣大學農業機械工程學系研究所。
蔡志成。2004。以近紅外光檢測幾丁類物質去乙醯率之研究。碩士論文。台北:國立臺灣大學生物產業機電工程學系研究所。
陳加增。2001。近紅外光應用於水果糖酸度線上檢測之研究。碩士論文。台北:國立臺灣大學生物產業機電工程學系研究所。
陳加增。2003以多光譜影像與決策支援系統應用於蔬菜作物氮肥管理之研究。博士論文計畫書。台北:國立臺灣大學生物產業機電工程學。系研究所。

被引用紀錄


伍志翔(2006)。近紅外光光譜標準化模式建立之研究〔碩士論文,國立臺灣大學〕。華藝線上圖書館。https://doi.org/10.6342/NTU.2006.02526

延伸閱讀