基於影像辨識之鳳梨苗種植機構之控制系統模擬分析

Abstract

現在農業正朝向智慧型農業發展，由於勞動力短缺以及農業人口高齡化的關 係，欲以農業機械化以及農業自動化來解決並且改善上述問題。而現在本系與桃 園區及高雄區農業改良場正在研發一型鳳梨苗種植機，用機械來取代以前傳統人 工種植鳳梨的方式，在不降低產量的情形下減少作業的時間以提高產值。此種植 機主要功能為鋪塑膠布及種植鳳梨苗，而本研究主要針對種植鳳梨苗時株距不一 的問題做改善，用影像的方式來做自動控制之模擬，並且調適 PID 參數設計控制 器，藉此方法來有效解決問題。 而實驗主要利用簡易的自製自走車進行模擬，用自行設計的蓋印機構代替複 雜的種植機構，再加裝網路攝影機且將 RGB 色彩空間轉化成 HSV 並進行一些影 像處理來去除雜訊及正確辨識，最後根據目標輸出所需調整 PID 參數，找到最適 合的數值設計出 PID 控制器。結果顯示，在本研究環境設定下，當自走車欲達到 每一次蓋印的距離為 150mm 時，以 KP 為 0.08、KI 為 0.001、KD 為 0.005 有最好 的效果，可以使誤差很小且響應時間很短，同時也示意了模擬中用影像的方式配 合 PID 控制器可以有效的解決問題。

Nowadays, agriculture is developing towards smart agriculture. Due to the shortage of labor and the aging of the agricultural population, it is necessary to solve and improve the above problems with agricultural mechanization and agricultural automation. Now, the Department and the Taoyuan District and the Kaohsiung District Agricultural Research and Extension Station are developing a type of pineapple seedling planter, which would replace the traditional method of artificially planting methods, and increase the output without reducing the yields. The main function of this planter is to lay plastic sheets and plant pineapple seedlings. This study is mainly aimed at improving the problem of plant distance when planting pineapple seedlings, using images recognition to simulate, and adapting PID parameter design controller. The experiment mainly uses a simple self-propelled car model to simulate, replaces the complex planting mechanism with a self-designed stamping mechanism, adds a network camera and converts the RGB color space into HSV color space and performs some image processing to remove noise and correct identification. Finally, adjust the PID parameters according to the target output, then find the most suitable value to design the PID controller. The results show that under the setting of this research environment, the best effect is obtained with KP of 0.08, KI of 0.001 and KD of 0.005, which can make the error small when the self-propelled vehicle wants to reach the distance of 150 mm for each stamping. The response time is very short, and it also indicates that the image can be used in the simulation with the PID controller to solve the problem effectively.