Towards a Stereo Approach to PTAM

TAIHÚ PIRE; JULIO JACOBO BERLLES

São Carlos

Conferencia; Joint Conference on Robotics and Intelligent Systems 2014, II Workshop on MSc Dissertation and PhD Thesis in Robotics; 2014

Universidad do São Paulo

This work presents the partial results obtained during the development of a system for local Simultaneous Localization and Mapping (SLAM) using a stereo camera. We follow the parallel tracking and mapping spirit that has become a standard in the community, and divide the computation into two threads (PTAM). Our experiments with this system show its real-time performance and the accuracy of the estimated metric maps. The main contribution of this work is the introduction of the stereo constraints in a local non-linear Bundle Adjustment and tracking operations and the release of the code under the name of STAM -standing for Stereo Tracking and Mapping- for its use as a basic local feature-based stereo SLAM.