臺灣博碩士論文加值系統

本研究使用機器視覺代替人眼進行落花生仁選別。選別機之主要設備包括
兩部攝影機、影像擷取卡、影像處理加速卡以及微電腦；並以微處理器介
面技術控制取像及吹氣時間。研究過程是，首先分析靜態落花生仁影像，
建立去除不良落花生之方法。然後分析動態影像，探討靜態與動態選別分
劃值之差異。最後使用兩部攝影機擷取與分析落花生仁不同兩面之影像，
進行動態選別試驗。由靜態影像分析之結果得知，重度顏色不正常落花生
仁可以損壞比率值做為選別參數；輕微顏色不正常落花生仁可以損壞比率
值及平均色相值二條件做為選別參數；皺皮落花生仁可以平均一次差分值
做為選別參數；發芽落花生仁可以細密度值做為選別參數；破粒落花生仁
可以平均色相值及損壞比率值二條件做為選別參數。動態選別時，不良落
花生仁之判別方法與靜態選別者相同，惟動態選別時之樣品不包括皺皮及
發芽落花生仁。每一顆落花生仁動態選別所需之時間為0.44秒；動態選別
結果，選別準確率為94.8%。

The purpose of this study is to sort peanut kernels by
machine vision insted of eyes. The main equiment of the sorting
machine are two CCD cameras,a color frame grabber, a high speed
frame processor, and a microcomputer.Furthermore,the technology
of microprocessor interface is used to control the image
grabbing time and the air blowing time. In this study,image of
the stationary samples were analyzed first to establish the
methods to remove bad kernels. then, the images of the moving
samples were analyzed to investigate the difference between
the thresolds of the static and dynamic sorting test.
Finally, dynamic sorting test was conducted by
using two cameras to grab two different sides of the sample and
analyzing these two images immediately. The results of the
stationary images analyses reveal that the damage ratio can
be used as the sorting parameter for the seriously abnormal-
colored kernels; the damage ratio and the average hue
have to be used as the sorting parameters for slightly
abnormal-colored kernels; the average first difference can be
used to remove the shrunk kernels;the compactness can be used
to remove the kernels with sprout; and the average hue and the
damage ratio are needed to sort out the broken samples. In
dynamic sorting test, the sorting parameters used were the
same as those used in static sorting test; however, the
shrunk samples and samples with sprout were not included in
dynamic test. In dynamic test, it took about 0.44 seconds to
sort a sample and the sorting accuracy rate is 94.8%。