EXPLOITING INERTIAL SENSING IN MOSAICING AND VISUAL NAVIGATION

In vision systems used in robotics, inertial and earth field magnetic sensors can provide valuable data about the observer ego-motion, as well as an absolute orientation reference. This article exploits the inertial orientation measurements to compensate the rotational degrees of freedom, in two different domains. First, inertial data is used to project images on a leveled plane, relaxing the demands on interest point matching algorithms when performing image mosaicing. Second, in the rotation-compensated, pure translation case, full homographies are reduced to planar homologies, and the ratio of heights over the ground plane on two views are calculated more accurately. Both techniques are validated over outdoor image sequences including aerial images from an remotely piloted blimp.