一种基于粒子滤波的双目视觉惯导SLAM技术

A visual inertial SLAM technology based on particle filter

摘要:

针对智能无人系统的定位与地图构建问题，提出一种基于粒子滤波的双目视觉惯导SLAM（即时定位与地图构建）方法。算法基于传统粒子滤波思想设计实现，后端处理时仅对位姿状态量进行滤波，有效解决了传统算法的维度爆炸问题，减少计算量的同时保证一定的精确度，兼顾SLAM算法精确性和即时性的要求。实验结果表明，方法整体定位精度相对误差低于5%，在光照条件适宜的小型场景效果更佳，误差低于3%，能够达到分米级精度，证明了SLAM方法能够完成SLAM系统的要求，实现即时定位与地图构建功能，具有一定的准确性、稳定性和鲁棒性。粒子滤波能够应用于视觉惯导SLAM领域并达到较高的精度要求。

Abstract:

Aiming at the problem of positioning and map construction in the intelligent unmanned system, a visual inertial SLAM method based on particle filter is proposed. This method is designed and implemented based on the traditional particle filter idea. In the back-end processing, only the pose state is filtered, which effectively solves the dimensional explosion problem of the traditional algorithm, reduces the amount of calculation while ensuring a certain degree of accuracy, and takes into account the accuracy and immediacy requirements of SLAM. The experimental results show that the relative error of the overall positioning accuracy of this method is less than 5%, and the effect is better in small scenes with suitable lighting conditions, which can achieve decimeter-level accuracy. It is proved that the SLAM method can fulfill the requirements of SLAM, realize the functions of simultaneous localization and mapping. The method has certain accuracy, stability and robustness. Particle filtering can be applied to the field of visual inertial SLAM and achieve higher accuracy requirements.