Principles and practice of real-time visual tracking for navigation and mapping

Abstract

We present examples of real-time mobile navigation systems and approaches to vision-based Simultaneous Localization and Mapping (SLAM) based on the second generation of our image processing library, XVision. Although the application-field of XVision is not limited to mobile navigation, the real-time requirements and limited computational resources on mobile systems provide a good testbed for it. We discuss the general architecture of the library and its modular processing pipeline, and we show how multiple tracking and low-level image processing primitives for color, texture and disparity can be combined to produce vision-guided navigation systems. The applications we discuss make use of XVision capabilities to solve the temporal correspondence problem by tracking an image feature in a given image domain.