Navigation for Performing Visual Surveys of Non-Planar Surfaces

This paper addresses the problem of navigating a robotic vehicle over non-planar surfaces for the purpose of performing visual surveys. The key issue lies in generating a spatiallysmooth camera orientation command that will point the camera toward the surface during surface traversal. The proposed solution filters high spatial frequency local surface measurements using a low spatial frequency spline-based model of the surface. The filtering occurs by updating the spline surface model with local measurements using a linear Kalman filter. Subsequently, the spline point projection algorithm generates the camera orientation command from the updated spline surface. The results show that when the spline-based surface model is used to smooth the high frequency measurements, the resulting camera orientation command resembles a spatially-smoothed version of the orientation command produced by local surface orientation measurements. Though the results of this paper focus on underwater visual survey applications, the proposed solution could be applied to any visual survey of a high spatial frequency surface, such as spacecraft inspection.