Sistem Peringatan Dini Banjir Rob Secara Real-Time Menggunakan Algoritma You Only Look Once Berbasis Internet of Things

Abstract

Tidal flooding is a serious threat to coastal areas, potentially causing infrastructure damage, property loss, and health disturbances. This study aims to develop an early warning system for tidal flooding based on the Internet of Things (IoT) and Computer Vision to provide early alerts about potential flooding, allowing communities to take preventive measures. The system has two alert statuses: warning, triggered when wind speed reached or exceeded the threshold and is directed toward land, and danger, triggered when tidal water levels exceed the safe limit. Anemometer, Wind Vane, DHT11, and a USB camera with the You Only Look Once (YOLO) algorithm are used to detect these conditions, with Raspberry Pi 5 as the main processing unit. Testing results indicate that the IoT devices, consisting of an Anemometer, Wind Vane, and DHT11, successfully collected data, despite minor delays in data acquisition, which remained within acceptable limits (±1s). Data from the three sensors were also successfully transmitted via serial communication for further processing. The YOLOv8 model demonstrated excellent performance in detecting tidal water objects, with Precision of 1.000, Recall of 1.000, mAP50 of 0.995, and mAP50-95 of 0.984, accurately detecting all test data. Furthermore, testing of the early warning system showed that it successfully sent a warning alert to the telegram channel when wind speed reached or exceeded the threshold and wind direction was toward land. The system also successfully sent a danger alert when tidal water levels exceeded the predetermined safe threshold.