A robust 4.0 dual-classifier for determining the internal condition of watermelons using YOLOv4-tiny and sensory

Abstract

This study presents a robust internet of things (IoT) based approach to solve the challenge of sorting fruit (watermelons) either as a raw material or final product in fruit manufacturing lines. A real-time objection detection technique called you only look once (YOLO) was used in the features detection, extraction, and matching of watermelons. The hardware framework of the system was developed on an Arduino microprocessor which integrates the sensors and camera into the system. The accuracy of the developed classifier is about 88% with a loss of 0.3, with images captured automatically saved on a designated folder which makes the detection and classification of upcoming products in the production line more accurate. The classified watermelons were further categorized into two possible states of ripe or rotten condition with an accuracy rate of 85%-90% with the tested data. These data were sent to the cloud via the Wi-Fi module and can be accessed using the Things Speak website (which is an application programming interface (API) for data retrieval and storage via the internet). An easy download option was incorporated into the system to obtain data from predictions and the cloud to a designated comma separated values (CSV) file locally for documentation and reference.