Automated position estimation of target using view extrapolation

Abstract

Position Estimation of target has always remained a critical task in defense applications. A variety of techniques exists for evaluation of position of objects over time. This paper discusses the idea of view morphing to generate future images of scenes having moving objects. It is highlighted that the concepts of view interpolation may be extended to synthesize new views that are NOT present between the given views with reference to time and/or position. This problem is addressed using View Extrapolation. It is based on the assumption that present development of non-stationary objects will continue in the same direction and with unvarying speed. The problem has been solved by dividing it into the three steps of Prewarping, View Extrapolation and Postwarping. It is pointed out that due consideration be given to the time passed between the capture of original views and time at which the new view is to be generated. This will help in finding the motion related parameters of the scene. This paper outlines an algorithm and highlights the issues to be considered to generate future images using the information in existing ones.