移动机器人中视觉里程计技术综述

Overview of Visual Odometry Technology in Mobile Robots

视觉里程计(VO)是实现移动机器人自主导航的主要技术之一,不同类型的VO技术在不同应用场景中受环境和硬件计算能力的影响,导致其性能各有优劣。概述VO技术的发展历程,对基于传统几何和基于深度学习的两类VO技术的性能进行对比与分析,重点介绍传统VO技术中特征点法的原理及其改进方法。在此基础上,归纳VO领域常用的公共数据集并对部分现有方法进行对比评测,为VO技术的实际应用提供参考和借鉴,并展望该领域未来的发展方向。

Visual Odometry(VO)is one of the key technologies to realize autonomous navigation of mobile robots.Due to the influence of environment and hardware computing power,different VO technologies display varying performance in different application scenarios.This paper describes the development history of VO technologies,and compares the performance of traditional geometry-based VO technologies with deep learning-based technologies for analysis.Among all the traditional VO technologies,this paper focuses on the feature point method,describing its principles and improvement approaches.On this basis,the paper summarizes the commonly used public datasets for VO studies,and evaluates part of the existing methods in comparison to provide reference for VO applications.Finally,the paper discusses the future development directions of VO.