Brown Planthopper Detection System

Abstract

This research is to revolutionize the pest control methods in a paddy field farming industry. This project develops a computer vision and Artificial Intelligence (AI) system for accurate and efficient way of detecting pest attacks in paddy field. In this paper, the focus is on Nilaparvata Lugens or more commonly known as Brown Planthopper a high-risk pest in the paddy farming industry. This pest has been the main contribute to paddy plant diseases in Malaysia. A recoded loss as much as 12% was recorded in the time span of 2015-2021 and this percentage has been growing as year passes. Currently, the process of detecting this pest in paddy field is only by manual labour which is labour intensive and a time-consuming process. Moreover, this pest infests the paddy field without any early warning signs. Therefore, this project was conducted to help farmers detect this pest for early pest control. This project uses image recognition to identify brown planthoppers and the YOLOv8 algorithm to train a model to detect and count the number of brown planthoppers on sticky traps. A camera-type C-mount lens has been used to get a clearer image. The images that have been collected for this project are as many as 800 images. In addition, the Internet of Things (IoT) system also successfully sends messages to its users. This project has achieved an accuracy of 85% for the detection and counting of the Brown Planthopper with the used of 6000 images for the learning process. Which is considered 23% more accurate compared to other experiments conducted.