Develop a self-sufficient system capable of detecting people’s physiological parameters during Search and Rescue (SAR) missions

The main objective of this thesis is to develop a self-sufficient system capable of detecting people’s physiological parameters during Search and Rescue (SAR) missions. Specifically, we focused on creating a battery-powered device capable of counting the number of people in an image and extracting physiological parameters rate using the remote photoplethys mography (RPPG) technique. The goal is to integrate this system into a rover or a drone, allowing an operator to control it remotely. In the first part of the study, we focused on hardware selection. It was necessary to identify a sufficiently powerful board capable of handling the computational load of the program while being lightweight, compact, and with low power to allow it to operate on battery. For this reason, we chose the Raspberry Pi 5. After carefully selecting all device components, with particular attention to the cameras, an in-depth analysis was conducted on depth of field and the minimum face size required for the software to func tion correctly, we then concentrated on the power supply. By analyzing the power and current requirements of each component, we determined the need for a 10,000 mAh power bank capable of delivering 5V/5A from a single port. However, since no commercially available power banks met these specifications, we opted for a model with the required ca pacity but with a maximum output of 12V/3A from a single port. We adjusted the power supply using a trigger decoy board and a step-down converter to obtain the necessary input values for the system. In the second part of the study, we focused on developing an application, based on open-source frameworks, to implement the aforementioned functionalities. For body counting, we used YOLO (You Only Look Once), a computer vision algorithm for real time body and object detection. For remote photoplethysmography, the script was devel oped using MediaPipe and HeartPy. MediaPipe allowed us to extract, through FaceMesh, the forehead and cheek regions of the face, which serve as input for HeartPy, an open source library for heart rate analysis. Finally, during the testing phase, we verified that, under standard conditions (di rect natural light and both direct and diffuse artificial light), the obtained results were promising, with heart rate estimation errors below 10 bpm. The main goal of this project is to enable an operator to view in real-time the images captured by the device, so potential future developments include integrating a stabilized camera (particularly useful for drone applications), implementing motors to allow the operator to adjust the camera’s orientation, and using an infrared light source with a dedicated camera to improve performance in low-light conditions.