Optimal view path planning for visual SLAM

Abstract

In experimental design and 3D reconstruction it is desirable to minimize the number of observations required to reach a prescribed estimation accuracy. Many approaches in the literature attempt to find the next best view from which to measure, and iterate this procedure. This paper discusses a continuous optimization method for finding a whole set of future imaging locations which minimize the reconstruction error of observed geometry along with the distance traveled by the camera between these locations. A computationally efficient iterative algorithm targeted toward application within real-time SLAM systems is presented and tested on simulated data. © 2011 Springer-Verlag.