透過您的圖書館登入
IP:18.224.30.118
  • 學位論文

應用程式語言開發光譜影像分析之圖形介面-以雞隻屠體為例

Development of Graphic User Interface for Spectral Image Analysis by Programming Language- Example of Chicken Carcasses

指導教授 : 謝清祿

摘要


由於光譜影像具有呈現空間及呈現光譜的特色,近年來也逐漸被利用在食品檢測和醫學檢驗等。以國人所食用的雞肉為例,政府為了維護國人食用肉雞上的安全,要求屠宰場雞隻屠體須經獸醫師檢查合格才可上市,但獸醫師在雞隻屠體檢查一般用肉眼去辨識,加上屠宰場環境潮濕悶熱,為了減輕獸醫師的負擔亟需開發一套自動化檢測系統。本研究利用Borland C++ Builder程式語言並結合OpenCV函式庫,開發一套光譜影像分析之圖形介面程式,可將超分光影像系統所拍攝的雞隻屠體光譜影像進行波段光譜組合,並利用特徵波長、顏色及形狀等特徵值組成合格與不合格雞隻屠體的特徵,最後再利用主軸成分分析所獲得的2個主軸分數(score)、4個score及6個score配合線性分類器一次及二次方程式,分別比較判別合格與不合格雞隻樣本的準確度,結果顯示在2個score時配合二次方程式的辨識結果最佳,訓練組的準確率達100%,測試組則為98.7%。期望日後此介面程式可應用於不同農產品或生物樣本光譜影像資料的檢測與分析。

並列摘要


Since the spectral image is capable of presentation the characteristics both in spectrum and spatial, it has been gradually used in food testing and medical inspection. According to wholesome food regulation in Taiwan, every chicken sold in market should be inspected by veterinarian before and after slaughter process. Slaughterhouse is a severe environment and inspection job needs highly mind concentrating. However, arguments between slaughterhouse managers and inspectors occur occasionally. To reduce the burden of veterinarians, it is needed to develop an automated detection system. In this study, a Borland C++ Builder and an OpenCV library were used to develop a graphics user interface (GUI) program for spectral image of chicken carcasses samples. After line spectral images were taken from samples by a hyperspectral imaging system, they were processed by the GUI program to finish band combination, extraction of region of interest (ROI) and features on color and shape. Finally discriminant function was found and tested the classification for wholesome and unwholesome chicken carcasses. Results showed that 2 PCA scores with quadratic equations was the best model and achieved a classification accuracy of 100%, and 98.7% for training and testing, respectively. It is suggested that apply this GUI program to different agricultural or biological samples for detection and analysis.

參考文獻


陳俊榮,2012,建立超分光光譜影像分辨國產合格與不合格雞隻屠體檢測技術之研究,碩士論文,國立屏東科技大學,機械工程系,屏東。
張介政,2013,應用超分光光譜影像線光譜分類不同雞隻病灶之研究,碩士論文,國立屏東科技大學,生物機電工程系,屏東。
吳昭葶,2013,應用光譜及光譜影像檢測雞隻屠體及其病灶之研究,碩士論文,國立屏東科技大學,生物機電工程系,屏東。
王文祺,2006,利用身高及人臉特徵實現遠距環境之數位家庭成員辨識系統,碩士論文,國立成功大學,電機工程學系碩博士班,臺南。
徐百輝,2003,小波轉換應用於高光譜影像光譜特徵萃取之研究,博士論文,國立成功大學,測量工程學系碩博士班,臺南。

延伸閱讀