Uncertainty Estimation of Dense Optical-Flow for Robust Visual Navigation

TL;DR

This paper introduces a dense optical-flow algorithm with full uncertainty estimation for improved robustness in visual navigation, enhancing SLAM performance for ground and aerial robots.

Contribution

It presents a novel method to estimate full optical flow uncertainty and integrates it into SLAM, improving robustness and accuracy over existing approaches.

Findings

Enhanced robustness and accuracy demonstrated on KITTI dataset

Full uncertainty estimation improves collision avoidance

Effective for both ground and aerial robotic navigation

Abstract

This paper presents a novel dense optical-flow algorithm to solve the monocular simultaneous localization and mapping (SLAM) problem for ground or aerial robots. Dense optical flow can effectively provide the ego-motion of the vehicle while enabling collision avoidance with the potential obstacles. Existing work has not fully utilized the uncertainty of the optical flow -- at most an isotropic Gaussian density model. We estimate the full uncertainty of the optical flow and propose a new eight-point algorithm based on the statistical Mahalanobis distance. Combined with the pose-graph optimization, the proposed method demonstrates enhanced robustness and accuracy for the public autonomous car dataset (KITTI) and aerial monocular dataset.