以机器视觉分的级文心兰切花之研究.pdfVIP

農業機械學刊 第 10 卷第1 期 2001 年 3 月 17 以機器視覺分級文心蘭切花之研究 席友亮 1，李芳繁 2 1. 國立中興大學農業機械工程學系碩士 2. 國立中興大學農業機械工程學系教授，本文通訊作者 摘要 本研究之目的是使用機器視覺技術萃取文心蘭切花之特徵參數，以進行文心 蘭切花之分級。本文使用兩種方法對文心蘭切花進行分級。第一種方法係依據文 心蘭切花分級標準加以分級，所使用之分級參數有花部長、莖部長及分枝數。第 二種方法係使用類神經網路對切花分級，輸入網路的參數有花部投影面積、花部 邊界長度、花部長、莖部長、切花中間部分莖粗以及莖部底端莖粗。第一種分級 方法之分級結果與人工分級結果比較，相符的程度為 72%，而第二種分級方法之 分級結果與人工分級比較，相符程度為 79%。 關鍵詞：文心蘭、分級、機器視覺、類神經網路 ONCIDIUM CUT FLOWER GRADING WITH MACHINE VISION Yu-Liang Hsi 1, Fang-Fan Lee 2 1. Former Graduate Student, Department of Agricultural Machinery Engineering, National Chung Hsing University. 2. Professor, Department of Agricultural Machinery Engineering, National Chung Hsing University, Corresponding Author. The objective of this study is to use digital image processing techniques to extract feature parameters of oncidium cut flowers for grading. Two methods were employed to grade the cut flowers. The first method utilized the length of the flower part, the length of the stem part, and the number of branches to grade the flowers according to the grading criteria of oncidium cut flowers. The second method used an artificial neural network to grade the cut flowers. The projected area of the flower part, the boundary length of the flower part, the length of the flower part, the length of the stem part, and the stem diameters of the cut flower in the middle as well as at the end of the stem were used as the grading input parameters to the neural network. The grading accuracy of the first method was 72% compared wi