一种场景稳健的单目视觉里程计算法

A Scene Robust Monocular Visual Odometry Algorithm

针对利用单目相机采集的图像序列进行实时车载平台位姿估计问题,对比了不同单目半直接典型算法的原理和试验结果以及不同场景、运动状态、光照和耦合因素下的同名点跟踪算法、长时间场景稳健的高精度位姿估计方法、位姿优化方法的试验结果。通过与两个典型半直接MVO算法进行了计算过程多个阶段和计算结果多个方面的对比,得出每个阶段和整体结果更好的计算方法;最终总结提出了一种场景稳健的单目半直接视觉里程计算法并利用序列真实数据进行了试验验证。试验结果表明,该算法的长时间位姿估计的场景稳健性和计算精度均显著优于目前典型的半直接MVO算法,位姿估计精度比ERL算法提升10%以上,计算效率与典型的ERL算法相当,能够满足各类单目视觉里程计应用场景需求。

For solving the real-time pose estimation problem utilizing sequential images sampled by monocular camera on vehicle platform,this paper compared various semi-direct MVO theories and their experiment results,together with experiment results of homonymous point tracking algorithms,accurate pose estimation methods for long-time and scene robust computation,and pose optimization methods in different scenes,motion states,illumination conditions and coupled factors.Through comparing the stages in calculation process and aspects of calculation results with two typical semi-direct MVO algorithm,better algorithms were derived for each stage and whole result.At last,a scene robust semi-direct MVO algorithm was proposed and validated using sequential real dataset.Experiment results illustrate that,the robustness to scenes and accuracy of long-time pose estimation of the proposed algorithm are much better than current typical semi-direct MVO algorithms.The proposed algorithm enhances 10%of ERL accuracy and the time cost is fairly the same as ERL.This will satisfy multiple MVO application requirement.