부산대학교 도서관

The aim of this research is mainly to develop a pesticide sprayer than can detect the health of the plant and spray pesticide, if necessary, in an agricultural setting. The proposed method was demonstrated in a GUI in MATLAB and Blynk that allows the user to analyze the health of the plant and control the spraying application using a mobile phone. The spraying system was implemented using Arduino Mega 2560 and GSMsim808. The performance of the system was evaluated in terms of the accuracy and response time of the disease identification system and the specific spraying pressure test. It is observed that CNN proves to 90% accurate for plant disease classification whereas SVM is 85%. However, it is also noted that CNN takes an average of 9.6 seconds to respond while SVM takes 0.9 seconds. This is due to the large training database using 50-layer neural network that CNN runs every time. It is noted that the user can control the spraying system and the movement of the robot smoothly with a runtime of about 3 hours and 45 minutes. All in all, the developed protype proved to be accurate and effective in spraying pesticide on crops that need pesticide with optimum user control.