- 期刊
- OpenAccess
使用影像處理進行棗子分級之研究
Jujubes Grading Using Image Processing
摘要
本研究使用彩色影像處理系統對棗子進行選別、大小量測及顏色分級。本研究同時量測棗子之糖度，並探討其與表皮顏色之關係。本研究之目的為建立棗子選別與分級之方法，以供研製分級機之參考。 在分級前，先將表皮顏色不正常或被小鳥啄食之樣品加以剔除，剔除方法是使用Sobel運算子對樣品影像處理，再以臨界值法分割，即可偵測出不良樣品而加以剔除。在大小量測方面，使用Hotelling轉換法計算樣品之長軸及短軸。使用RGB和HSI顏色座標系統量測樣品之顏色。樣品顏色之分級則使用最小距離分類器、Fisher's線性分類器及Bayes分類器。實驗結果顯示，長軸誤差約小於8%，短軸誤差約小於6%；樣品之重量與投影面積間之相關係數為0.99；顏色愈黃之樣品其糖度愈高。使用H和S做為顏色分級參數時，最小距離分類器之分級正確率為79%，Fisher's線性分類器之正確率為91%，而Bayes分類器之正確率則為95%。
並列摘要
This study utilized a color image processing system to detect bad jujubes, to measure the size and to grade the color for good samples. The soluble solids content of samples juice was measured and its relation with the skin color was investigated. The objective of this study is to establish the sorting and grading method for jujubes. The method established could be used in designing a grading machine. Before grading, the samples with abnormal skin color or bird-pecked holes were removed. These culls were detected by using the Sobel operator and thresholding segmentation. The Hotelling transform method was employed in calculating the major and minor axes. The RGB and HSI color systems were used to measure the color of the samples. The minimum distance classifier, the Fisher's linear classifier, and the Bayes classifier were utilized to grade samples color. Experimental results show that the errors are less than 8% and 6% for the major axis and minor axis, respectively. The correlation coefficient between the weights and the projected areas is 0.99. The soluble solids content of yellow samples is greater than that of the green ones. Using the H and S values as the parameters for grading color, the grading accuracy rates are 79% for the minimum distance classifier, 91% for the Fisher's classifier, and 95% for the Bayes classifier.