Remote Controlled Unmanned Water Vehicle with Human Detection and GPS Using Yolov4 for Flood Search Operations

Abstract

During flooding, search and rescue operations (SAR) is immediately conducted. Thus, it must be performed with an assessment and well-defined procedure. Search and rescue operations rescue those stranded or needing immediate medical attention. However, SAR volunteers are not guaranteed safe operations, especially in hazardous environments like flooding. To lessen environmental risks and assessment time when searching flooded areas, the researchers developed a remote-controlled unmanned water vehicle (UWV) prototype with a real-time video that can project 180 degrees in its surroundings and captures images after detecting humans. The purpose of this study is to use the You Only Look Once version 4 (YOLOv4) algorithm for human detection in an unmanned water vehicle application for SAR operations. The paper proposes an alternative way of SAR applications by using UWV aside from the most common application, an unmanned aerial vehicle (UAV). The overall methodology of the study starts with the connection of the prototype to the internet and the detection of location using Global Positioning System (GPS). Then, a live feed is projected in the laptop from the prototype’s perspective using Logitech Camera. With the YOLOv4 algorithm, the prototype detects humans, captures an image, and sends it as a notification through the user's email. It tells how many people are detected and delivers the prototype's location using Global Positioning System (GPS). The accuracy of the human detection was tested in various settings that could hinder the detection, such as obscure settings, detection with animals, and in an uncontrolled setting. Based on the data gathered from the tests, the overall accuracy is 91.49%.