Real-Time Intensity-Based Mirror Detection for Optimizing Point Cloud Data and Occupancy Grid Map

Abstract

In recent years, indoor mobile robots have played an increasingly important role in various home, medical, commercial, and industrial applications. However, mirror surfaces commonly found in indoor environments pose challenges to the localization and navigation of indoor mobile robots. In environments with mirror surfaces, robots may misjudge the location of obstacles owing to laser reflections, leading to accidental collisions or mission failure. In this study, a single 2D LiDAR was used to identify the location of the mirrors in the environment, and to optimize the point cloud data and occupancy grid map using the mirror locations. The main innovation is the use of the intensity information of the reflected laser beam and the inherent symmetry of mirrors for real-time detection, including polygonal mirrors and mirrors without contours. Currently, the method has been experimented with in several complex environments, the accuracy of mirror identification exceeds 97%, and effectively modified the erroneous occupancy grid map.