Maritime Human Drone Teaming For Search and Rescue Operations

Drowning detection is a critical safety measure aimed at identifying and preventing drowning incidents in various environments, such as swimming pools, beaches, rivers, and other bodies of water. Drowning is a leading cause of accidental death worldwide, and early detection is essential for saving lives and ensuring water-related activities remain safe. The goal of drowning detection systems is to promptly identify when a person is in distress or submerged in the water, triggering an immediate response to prevent potential drowning incidents. In this proposed work, a monitoring system using UAV (drones) to detect a drowning person is proposed. The system includes an operator, camera gear, Raspberry Pi3, and rescue UAV. The real-time image is assisted by the drowning detection algorithm, and judges whether there are targets to be rescued, so the performance of detection algorithm used in this process is very important. The drowning activity detection is identified using deep learning algorithm. After comparison with other models/methods, the deep learning model chosen for drowning detection is Convolutional Neural Networks (CNN). Conv2dLstm is the convolution layer used for developing the model, since the proposed work deals with 3D input data instead of 1D vector. Deep learning model is successfully built using CNN architecture which has an accuracy of 0.9341. The processing time taken for the model to detect the drowning activity is 5 seconds where the rescue operation can be done immediately to save the person. Real time implementation of the proposed system has been achieved using Raspberry Pi 3. For the rescue operation, human-drone communication is established. The controller is mounted on a carrier device and is tested in real time environment. The drowning detection system has successfully identified a person drowning in a swimming pool.