UAV based Human Detection for Search and Rescue Operations in Flood

During disastrous events like floods, the urgency to rescue people facing immediate danger poses a significant challenge. Traditional search operations led by humans often struggle due to dangerous conditions and limited resources. As a response, a new application has emerged: employing Unmanned Aerial Vehicle (UAV) for rescue missions, utilizing image processing to locate individuals in distress. However, existing application relies on deep learning based approach for image analysis suffer from slow human detection, a crucial drawback given the time sensitive nature of such scenarios. To tackle this issue, we have proposed a novel approach using You Look Only Once version 4 (YOLOv4) for rapid human detection. YOLOv4 is renowned for its speed, surpassing traditional methods by swiftly identifying humans within flooded environments. To train and validate our approach, we have used Common Objects in Context (COCO) dataset (pretrained or labeled). The COCO dataset's large size and wide variety (real-world scenarios, including images from natural scenes, indoor environments, and crowded public spaces) of images make it a valuable asset for training and testing human detection models. Through exhaustive experimentation, our method showcases an impressive average accuracy of 79.46% in accurately identifying humans amidst flood-ravaged surroundings. By harnessing the efficiency of YOLOv4 and the diversity of the COCO dataset, our approach holds the potential to significantly enhance the efficiency of UAV-assisted search and rescue operations during critical moments, ultimately resulting in the preservation of more lives in the face of disasters. As well as the human detection information collected from the model facilitates the rescue team, assisting in tasks such as distributing food, providing first aid, and formulating rescue strategies.