Robust monocular visual odometry for road vehicles using uncertain perspective projection

David Van Hamme,Wilfried Philips,Werner Goeman,Peter Veelaert

doi:10.1186/s13640-015-0065-6

David Van Hamme, Wilfried Philips + Show 2 more

Open Access

https://doi.org/10.1186/s13640-015-0065-6

Copy DOI

Abstract

Many emerging applications in the field of assisted and autonomous driving rely on accurate position information. Satellite-based positioning is not always sufficiently reliable and accurate for these tasks. Visual odometry can provide a solution to some of these shortcomings. Current systems mainly focus on the use of stereo cameras, which are impractical for large-scale application in consumer vehicles due to their reliance on accurate calibration. Existing monocular solutions on the other hand have significantly lower accuracy. In this paper, we present a novel monocular visual odometry method based on the robust tracking of features in the ground plane. The key concepts behind the method are the modeling of the uncertainty associated with the inverse perspective projection of image features and a parameter space voting scheme to find a consensus on the vehicle state among tracked features. Our approach differs from traditional visual odometry methods by applying 2D scene and motion constraints at the lowest level instead of solving for the 3D pose change. Evaluation both on the public KITTI benchmark and our own dataset show that this is a viable approach for visual odometry which outperforms basic 3D pose estimation due to the exploitation of the largely planar structure of road environments.

Highlights

Visual odometry is an increasingly important research domain in the field of intelligent transportation systems
The aim of this research is to prove that our 2D approach is a viable alternative to traditional 3D pose estimation for visual odometry
The potential improvements afforded by bundle adjustment and more precise ground plane estimation for the proposed method are to be explored in future work

Summary

Introduction

Visual odometry is an increasingly important research domain in the field of intelligent transportation systems. Many emerging and existing applications related to consumer vehicles rely on accurate position estimation. Some examples of these applications are navigation, lane assistance, collision warning, and avoidance. The positioning data for these applications is provided by satellite-based systems such as GPS, GLONASS, or GALILEO, sometimes augmented by closer-range communication (as in DGPS) or additional sensors scanning the local environment (e.g., a lane assist camera). The reliance of these applications on satellite navigation is a threat to their full-time availability. Due to the four-dimensional nature of the problem (3D positioning and time synchronization), signals from at least four satellites must be received in order to obtain a positional

Objectives

Results

Conclusion