蕈菇產瓶栽培走菌檢測系統之研製

Abstract

蕈菇產業在世界各國均屬高產值的精緻產業，綜合鮮銷菇類及衍生產品在全球的銷售值約可達600億美元以上，其生產潛力和市場價值均不容忽視。在台灣，食用菇類生產同樣是重要的精緻產業之一，近年來以自動化太空包栽培技術和環控菇舍管理系統，漸漸發展為工廠化、企業化和國際化的產業。目前食用菇類栽培的作業流程仍需大量的勞力才得以進行，但因社會人口結構的變遷，逐漸面臨雇工不易的窘境，影響生產活動的進行，因此改善作業環境與開發省工設備成為了目前面臨的重要課題。 本研究著重於蕈菇栽培瓶栽培作業中，對人工目測栽培瓶走菌程度和感染雜菌的檢驗進行省工改善，目標是建立一個蕈菇栽培瓶檢測系統，使用機器視覺檢測雜菌汙染和生長異常的栽培瓶。研究栽培瓶側面展開圖的影像處理方法，計算正常走菌區域的百分比，作為選別依據，目前一次處理只需10秒。並參考廠商的目視檢測標準，設定檢測標準為上方走菌90%以上且側向走菌60%以上。檢測走菌1/2和2/3的實驗模擬樣本，百分比分別為50.02%和66.59%，可知得到的影像資訊完整且清晰。以實際栽培瓶樣本進行檢測，得到以正常走菌區域的百分比做為選別依據，可成功選別不良品。 依此設計由氣動機構、打光室、輸送帶和可程式控制器組成的檢測系統，同樣以上方走菌60%、75%和100%，側向走菌1/3、1/2和2/3的實驗模擬樣本進行測試，檢測結果上方走菌58.52%、73.98%和99.514%，側向走菌32.02%、49.94%和66.10%，皆十分貼近樣本設計值，準確率高。再取得32瓶實際栽培瓶樣本，待系統調整完成後，進行三次檢測，並與實驗打光室結果進行比較。除了好菌走菌不完全樣本，因為仍有走菌，部分樣本被判斷錯誤，但其他樣本皆被檢測系統成功選別。經測試得到系統選別正確率為91.07%，並可自動將不良品排除，目前完成一次半籃8瓶的檢測約需費時6~7分鐘。

The mushroom industry is a quality agriculture with high output value in every country. In all kinds of fresh mushroom and its derivatives, the sales value can reach more than 60 billion US dollars worldwide, it shows the importance of its production potential and market value. The edible mushroom cultivation is also one of the important quality agriculture in Taiwan. By using the automatic bag cultivation and fruiting room environment control system, recently the production develops into an enterprise and international industry. The edible mushroom cultivation process still need a large amount of labor, which become a predicament because of the changing of the population structure and the shortage of labor. Therefore, the improvement of working environment and development of labor saving device become the most important topic facing nowadays. This research focused on the improvement on manual fault inspection for spawn level and bacterial infection during the mushroom bottle cultivation process. The aim of this research is to establish a mushroom bottle inspection system using machine vision to recognize the bacterial infected and abnormal bottles. This research used image processing method to get bottle’s development (expansion). Counting the percentage of normal spawning area to recognize the defective bottle. The image processing takes 10 seconds. According to the manual inspection standard, the inspection standard is set to reach over 90% in top view and 60% in side view. When inspecting the simulated samples with 1/2 and 2/3 spawn level, the percentage counting results were 50.02% and 66.59%. When inspecting the real samples, the percentage of normal area was proved to be used as the grading standard to recognize the defective bottles. According to the results above, an inspection system was designed and established. The system consists of pneumatic mechanism, lighting chamber, conveyor and PLC. When inspecting the simulated samples with 60%, 75% and 100% in up view and 1/3, 1/2 and 2/3 in side view spawn level. The average results were 58.52%, 73.98% and 99.514% in up view and 32.02%, 49.94% and 66.10% in side view. Also picking 32 real samples to adjust the system, then inspect these bottles by inspection system three times repeatedly. Except part of the mushroom spawning incomplete samples were misjudged because of the spawning between system adjustment, rest of the samples were all recognized successfully. The accuracy of the inspection system is 91.07%, and the defective bottles were removed automatically. An inspection of 8 bottles per process needs 6-7 minutes.