A Weighing Euler Pre-integration Method in the Visual-Inertial Odometry

Abstract

Visual Inertial Odometry (VIO) has been widely used in many fields. Inertial Measurement Unit (IMU) pre-integration is an important part of data preprocessing throughout the VIO process. And improving the accuracy of pre-integration is of great significance for the subsequent joint initialization, motion estimation and optimization. This paper proposed a weighing Euler pre-integration method in a VIO, which is based on monocular camera and IMU. It weights the contiguous IMU data. Meanwhile, the influence of noise and biases are both discussed. In the monocular vision and inertial tightly coupled system, the experimental results proved to be ideal with the EuRoC MAV datasets, which can make better use of the angular velocity and acceleration measurements, and show a great improvement compared to the existing pre-integration technology.