ABSTRACT
Floods are a lethal force of nature that not only cause heavy economical and infrastructural damage but most importantly take away invaluable lives. These floods can last days to weeks during which people in the affected regions get cut off from essential supplies. Rescue teams are deployed to save affected people, and in the process it is highly probable that the team members may lose their lives. We need to have a mode of transport for supplies to reach the flood victims that does not involve sending in human beings to the affected areas. Unmanned surface vehicles (USVs) are used in a wide range of areas where there is a risk to mankind. Since World War II, advancement has been seen in control systems and navigation for USVs, but we are still lacking in terms of the object detection prowess of these vehicles. A USV, armed with object detection capability interfaced with sensor data, has the potential to be the solution to the rescue ops problem. The challenge in this approach, however, would be to deploy computer vision algorithms on resource-constrained devices which would require a lot of optimization for accurate performance. Hence, we propose an analysis to identify the best algorithms for collision detection and design a cost-effective prototyping method for USVs.
