Simultaneous aerial vehicle localization and Human tracking

摘要

In this paper, we propose a robotic video surveillance system. A prototype system is implemented on an aerial vehicle, a quadcopter. It is capable of following a person or any moving object, while simultaneously localizing it, i.e., measuring the coordinates of the quadcopter on a scaled map. The system can operate well, even if it is flying in an unknown and GPS-silent environment. The prototype system consists of 3 major modules. First is an `Image-based Visual Servoing (IBVS)' module for tracking the desired object or person, and sending the error parameters to the actuators. The second module is `Parallel Tracking and Mapping (PTAM)' module. The third module is an `Extended Kalman Filter (EKF)' module. The PTAM and EKF modules help in estimating the location of the quadcopter at every time-instant. The proposed system enables the quadcopter to follow the desired object smoothly, while keeping the tracked object in the center of the frame and maintaining a constant distance from the object. The effectiveness of the system is validated by computing the precise measurements of the aerial movements, while ensuring the minimal error. The proposed method can be helpful for quadcopters used in diverse applications such as surveillance, monitoring and tracking a desired object.