视觉惯导紧耦合的行人室内自主定位方法

Indoor autonomous positioning method with visual inertial tightly coupled for pedestrians

针对行人在大型复杂建筑环境中的高精度和高可靠性室内定位需求,传统的基于视觉点特征方法易受环境纹理缺失、相机快速运动导致图像模糊而定位失效问题,提出了一种基于视觉点线特征与IMU紧耦合的行人室内自主定位方法。在视觉惯导融合导航系统框架下,前端部分,在点特征基础上引入结构化建筑环境中丰富的线特征,并采取基于梯度密度过滤机制的改进线特征提取策略,剔除局部线特征密集区域;利用点线特征与IMU紧耦合优化机制提高行人位姿估计及定位的准确性和稳定性。通过利用EuRoC数据集和在实际楼道场景下的实验,特别是在弱纹理、光照变化等条件下实验,验证了所提方法进行行人室内定位的准确性和可行性。

Aiming at the demand for high-precision and high-reliability indoor positioning of pedestrians in a large and complex building environment,the traditional visual point feature methods are susceptible to environmental texture loss,fast camera motion and blurry images resulting in positioning failure.This paper proposes a pedestrian indoor autonomous positioning method(IMPL-VINS)based on tightly coupled visual point and line features and IMU.Under the framework of the visual inertial integrated navigation system(VINS),in the front part,the rich line features in the structured building environment are introduced based on the point features,and an improved line feature extraction strategy based on the gradient density filtering mechanism is adopted to eliminate the local line feature dense areas.The back part coupled the point-line features and IMU data.The tight coupling optimization mechanism improves the accuracy and stability of pedestrian pose estimation and positioning.By using the EuRoC data set and experiments in actual corridor scenes,especially under the conditions of weak texture and light changes,the accuracy and feasibility of the proposed method for indoor positioning of pedestrians are verified.