Exploiting Feature Confidence for Forward Motion Estimation
This work addresses a specific difficulty in VIO for vehicle navigation, representing an incremental improvement in domain-specific applications.
The paper tackles the challenge of estimating forward motion in Visual-Inertial Odometry (VIO) for vehicles in large-scale outdoor environments by proposing a method that analyzes feature confidence using IMU data and incorporates it into a Bayesian framework, resulting in improved performance over baseline VIO on the KITTI dataset.
Visual-Inertial Odometry (VIO) utilizes an Inertial Measurement Unit (IMU) to overcome the limitations of Visual Odometry (VO). However, the VIO for vehicles in large-scale outdoor environments still has some difficulties in estimating forward motion with distant features. To solve these difficulties, we propose a robust VIO method based on the analysis of feature confidence in forward motion estimation using an IMU. We first formulate the VIO problem by using effective trifocal tensor geometry. Then, we infer the feature confidence by using the motion information obtained from an IMU and incorporate the confidence into the Bayesian estimation framework. Experimental results on the public KITTI dataset show that the proposed VIO outperforms the baseline VIO, and it also demonstrates the effectiveness of the proposed feature confidence analysis and confidence-incorporated egomotion estimation framework.