Composing the feature map retrieval process for robust and ready-to-use monocular tracking

► Offline based feature map retrieval method for camera-based tracking in augmented reality applications. ► Using constrained Bundle Adjustment for higher reconstruction accuracy after incremental map building with SLAM. ► Applying Randomized Trees based feature learning to reconstructed feature maps to handle environments with arbitrary geometries. ► Performing a statistical analysis of feature visibility for higher performance in self-occluding scenes. ► Supporting frame-to-frame tracking with inertial sensor measurements (sensor fusion).