Carriers equipped with high-frame-rate sensors face challenges such as high computational load and decreased positioning accuracy when moving rapidly indoors.This paper proposes an enhanced MSCKF-based VIO algorithm tailored for indoor environments to address these issues.Firstly,by applying constraints based on gradients and eigenvalues to the feature point detection results,the quality of feature point extraction and consequently the pose estimation accuracy of the VIO algorithm are enhanced.Then,one-dimensional inverse depth parameterization for map points is employed to augment the state,reducing the computational complexity and thereby increasing system processing speed.Finally,comprehensive evaluations of the proposed algorithm are conducted using both the public EuRoC datasets and real-world scenarios,assessing the algorithm's trajectory estimation accuracy,processing time,and CPU utilization.The experimental results demonstrate that,compared to three mainstream VIO methods—S-MSCKF,VINS-Mono,and PL-VIO—the proposed algorithm achieves an improvement in positioning accuracy of at least 19.18%,while also ensuring lower processing times and CPU usage,thus maintaining system real-time performance.
维普
万方数据