BEV-DWPVO: BEV-based Differentiable Weighted Procrustes for Low Scale-drift Monocular Visual Odometry on Ground

TL;DR

This paper introduces BEV-DWPVO, a ground-plane assumption-based monocular visual odometry system that uses Bird's-Eye View features and a differentiable Procrustes solver to reduce scale drift and improve pose estimation accuracy.

Contribution

It proposes a novel BEV-based approach with a differentiable weighted Procrustes method, simplifying pose estimation and enabling end-to-end training without auxiliary tasks.

Findings

Outperforms existing MVO methods on NCLT, Oxford, and KITTI datasets.

Effectively reduces scale drift in long-sequence visual odometry.

Simplifies pose estimation by reducing DoF from 6 to 3.

Abstract

Monocular Visual Odometry (MVO) provides a cost-effective, real-time positioning solution for autonomous vehicles. However, MVO systems face the common issue of lacking inherent scale information from monocular cameras. Traditional methods have good interpretability but can only obtain relative scale and suffer from severe scale drift in long-distance tasks. Learning-based methods under perspective view leverage large amounts of training data to acquire prior knowledge and estimate absolute scale by predicting depth values. However, their generalization ability is limited due to the need to accurately estimate the depth of each point. In contrast, we propose a novel MVO system called BEV-DWPVO. Our approach leverages the common assumption of a ground plane, using Bird's-Eye View (BEV) feature maps to represent the environment in a grid-based structure with a unified scale. This enables us to reduce the complexity of pose estimation from 6 Degrees of Freedom (DoF) to 3-DoF. Keypoints are extracted and matched within the BEV space, followed by pose estimation through a differentiable weighted Procrustes solver. The entire system is fully differentiable, supporting end-to-end training with only pose supervision and no auxiliary tasks. We validate BEV-DWPVO on the challenging long-sequence datasets NCLT, Oxford, and KITTI, achieving superior results over existing MVO methods on most evaluation metrics.