An Automatic Monitoring System for Dragon Fruit Using Convolutional Neural Networks (CNN) and Internet of Things (IoT)

Abstract

Plant diseases and pests have led to a decline in the quality of dragon fruit produced in Banyuwangi regency, Indonesia. Infections in dragon fruit cause rot, and farmers struggle to identify the pests responsible. To address this problem, this work proposed two concepts for the classification and monitoring systems of dragon fruit. The classification was done by processing some images of dragon fruits captured by DLSR camera and utilizes a convolutional neural network with three layers for training and testing. The monitoring system is based on the Internet of Things to tracks the status of ripe, raw, and rotten fruits. The application of the dragon fruit classification system to ripe, rotten, and raw fruits has yielded results that increase fold accuracy by 0.976, 0.981, and 0.986, respectively, with 200 training data in each of the three training and testing phases. There is a decrease of 0.024, 0.019, and 0.014 in fold loss accuracy. Meanwhile, the monitoring system's platform integrates the classification of dragon fruit to monitor the condition of ripe, raw, and rotting fruit in real time. With the implementation of the classification and monitoring system, farmers will be better equipped to predict when dragon fruit will ripen and prevent the spread of rot to other fruits.