應用彩色機器視覺技術選別蔬菜種子之研究

Vegetable Seeds Sorting with Color Machine Vision

黃正霖(Cheng-Lin Huang) ； 李芳繁(Fang-Fan Lee) ； 林慶福(Ching-Fu Lin)

《農業機械學刊》 3卷4期 (1994/12) Pp. 1-14

https://doi.org/10.30062/JAM.199412.0001

摘要

本研究是利用機器視覺技術並配合真空吸力式種子選別機、選別電路與噴氣式出料處理，所發展出的一套蔬菜種子選別系統。在此系統中使用兩種顏色系統來作選別的模式，即 R 模式與 I 模式，當種子的平均灰度值大於標準值，則判定為不良種子，反之，則為良好的種子。
選別時，將種子倒入種子室內，利用真空吸力之方式使種子吸附在吸盤上，並隨吸盤的轉動在最高點時進行取像，取像後將分析結果之判斷位元組送入佇列區內，當種子到達最低點被刮離後，在佇列區之判斷位元組會輸出，若該種子被判定為良好的種子，它會直接落入出料口；若該種子被判定為不良的種子時，它將被噴離到另一個出口。
由實驗結果顯示，結球白菜種子在 R 模式下，其準確率為 98.75%, I模式為 97.4%。甘藍種子在 R 模式下，其準確率為 97.4%, I 模式為 96.5%。每處理一粒種子從取像到影像分析結束約需0.16秒，亦即若馬達轉速夠快選別機每分鐘約可處理375粒種子，而種子選別機目前受限於馬達轉速最快可處理約83粒種子。

關鍵字

機器視覺 ； 選別 ； 蔬菜種子

並列摘要

This vegetable seeds sorting machine is equipped with a vacuum-type seeds sucking disk, an electric sorting circuit, an air jet discharger, and a color machine vision system. Two color models, namely, the R model and the I model, were used in this study. When the gray-level value of R or I image of a seed was greater than the threshold, this seed was considered as a bad seed; otherwise, it was judged as a good seed.
During the sorting process, the seeds were poured into the seed room first, then singularized by the sucking disk. As the seed reached the highest position on the sucking disk, the CCD camera grabbed the image of the seed and the computer analyzed this image immediately. Then, when the seed reached the lowest position, it was scraped off from the disk and falling down into a container if it was a good seed. If the seed was bad, it would be blown into another container by air jet while it was falling down. The sorting results obtained after image analysis were stored in an array in byte type to control the action of the air jet.
Experimental results show that for Chinese cabbage seeds sorting accuracy rate is 98.75% for the R model, and 97.4% for the I model. For cabbage seeds, the accuracy rate is 97.4% and 96.5% for the R model and I model, respectively. It took 0.16 seconds to sort a seed including image acquisition; namely, 375 seeds could be sorted per minute if the motor speed was fast enough. Due to the limit of the motor speed, the actual sorting capacity is 83 seeds per minute.

並列關鍵字

Machine vision ； Sorting ； Vegetable seeds