Human Target Search and Detection using Autonomous UAV and Deep learning

Abstract

An Unmanned Aerial Vehicle (UAV) is an airborne system or pilotless aircraft which is remotely controlled by a human operator on ground or by an onboard computer such that the vehicle moves autonomously. The range of applications in which UAVs are used is very large. This paper describes the application of developing an autonomous surveillance system using an UAV to identify a given target and/or objects of interest in the terrain over which it flies. Such a system can be used in rescue operations, especially in remote areas where physical access is difficult. It can also be used for military operations, farming or any field where surveillance of a given land area is required. The UAV developed in this work is capable of object detection. A mounted camera is used to give visual feedback, and an onboard processing unit runs image recognition software to identify the target in real time. Optimal algorithms are used to search and find the target from the given search area. After recognition of the target, the UAV can either be used to hold its position so as to have a video feed of the target, or return to its base station once the coordinates have been estimated using GPS modules or relay the GPS location to the base station. This paper describes the implementation of the hardware and software components that lead to the realization of the UAV and the application of object detection. The details of a new search algorithm and an example of object detection is presented . The work presented in this paper is the first part in the attempt to develop a cluster of UAVs meant to work in collaboration to be deployed for search and rescue operations.