以機器視覺引導機器人選別水果

Robotic Sorting of Fruits Using Machine Vision

張文宏(Wen-Hung Chang) ； 陳世銘(Suming Chen)

《農業機械學刊》 2卷3期 (1993/09) Pp. 11-24

https://doi.org/10.30062/JAM.199309.0002

摘要

水果選別作業爲農產品加工工程中極爲重要之一環。本研究主要是利用黑白影像處理卡，針對水果之大小及顏色的選別，探討打光的條件及背景顏色的選擇，並應用基本的影像處理技巧，尋找最佳的處理條件，再以機器手臂來完成出料的動作，配合輸送帶的使用，以建立一套連貫的選別系統作業流程。
大小選別是利用攝影機所攝取之二維影像，以水果之投影面積所佔的點數多寡，作爲判斷大小等級的指標；顏色選別則是以水果之投影面積的平均灰度値作爲判斷顏色等級的標準。以檸檬爲例，經本系統之大小選別量測試驗後，所得之投影面積點數與實際體積有R^(2)=0.97之線性相關，選別正確率達93.3％；經顏色選別試驗之後，所量測之平均灰度值與以色差計量測之明暗度（即L值），有R^(2)=0.91之線性相關，選別正確率與人工分級結果比較則達93.3％。本系統亦可同時考慮大小及顏色兩項因素，作綜合性之選別，其選別正確率可達91.1％。對檸檬而言，機器手臂之爪具以挾持式之效果最好。

關鍵字

選別 ； 大小 ； 顏色 ； 機器人 ； 機器視覺

並列摘要

Fruit sorting is one of the important agricultural processing operations. The sorting of size and color of fruits was investigated by using image processing techniques. The optimum sorting conditions and methods were established through proper design of an illumination chamber and well examining of background colors. A robot was then used to integrate the entire sorting system.
The sizing was based on calculating the total number of pixels by thresholding of fruit's image. The color grading was decided by the averaged gray level of image. Lemons were used for experiments. The result of sizing experiments showed that the measured area had a linear relationship (R-Square=0.97) with the actual volume and the sorting accuracy was above 93.3%. The result of color sorting tests indicated that the calculated averaged gray level had a linear relationship (R-Square=0.91) with the L value recorded by the color difference meter, and the sorting accuracy was above 93.3%.The system can also be used for combined sorting of both size and color, in which, thefruits were divided into 9 grades, and a sorting accuracy of 91.1% was reported. The grasp-type end effector works best with lemons.

並列關鍵字

Sorting ； Size ； Color ； Robot ； Machine vision