Abstract
A problem of failure in target tracking with occlusion or out of frame has been considered. We propose a method being able to continue the tracking by taking into account road information. We deal with a scene that Unmanned Aerial Vehicle(UAV) is tracking a car. Boundary of the road is detected based on color information. Occlusion or out of the frame of the road are extrapolated with cubic splines. State of the car consists of velocity and posture of the car. Time evolution of the state is represented by random walk in system model. Posture of the car is set to a direction where the road continues. Observation model represents a relative positional of the car and UAV. Likelihood becomes joint distribution of the feature of the car and the road. The tracking is sequentially accomplished by state estimation using particle filter. Effectiveness of proposed method has been shown through tracking experiments with simulation and actual image.
