| In recent times, intelligent, unmanned vehicles with onboard computers and sensors have been used in many ways for better serving humanity. An important use case of such vehicles is the deployment in the field of Disaster Rescue and Relief. Many remote controlled systems have been tested in real life disaster situations, with dramatic increase in productivity of Rescue and Relief teams and a huge decrease in loss of precious lives. These systems, however, are not cost effective and thus, out of reach of most organizations involved in these activities. Based on these observations, I felt that a generic and robust system which is also affordable and easily deployable/manageable is the need of the hour. These factors, along with the availability of affordable technology, motivated me to focus my research on the use of thermal imagery for person detection from Unmanned Aerial Vehicles (UAVs) in disaster situations. The person detection system works in a hierarchical, multi-phase deployment, with each step having its own significance. The onboard thermal and Raspberry Pi cameras record images at a pre-determined interval, which are processed for detection onboard the UAV's computer. These images are compressed and wirelessly sent to the ground control station, along with the UAV flight status information (location co-ordinates, airspeed, ground-speed, and altitude) at a near real-time speed. The ground control system organizes the data and is responsible for alerting users when successful detections are made. The system's use of mesh network architecture makes it highly scalable and flexible, with various multi-nodal deployment options. |
