Feature Matching for Indoor-Oriented Visual Odometry

Abstract

The simultaneous localization and mapping technology [1] [2] refers to the moving object positioning itself according to the characteristics of the environment and constructing the map incrementally [3] [4]. This technology can realize the trajectory tracking of the robot without temporary positioning infrastructure. Visual odometry [5] based on camera sensors has developed rapidly in recent years. As a front end of visual SLAM, it can replace lidar to calculate mileage, thereby reducing system cost and enriching map information. This paper briefly describes the concept and development of visual odometry for mobile robots, proposes a visual odometry method suitable for indoor environments, and compares different existing visual odometry methods. The results show that the proposed scheme achieves faster and more accurate mileage calculation in predictable scenarios, which can be used in the navigation of indoor mobile robots.