LiDAR Reflection Recovery via Shadow Boxing

초록

Light detection and ranging (LiDAR) sensor is a device that can effectively collect 3D spatial information using lasers. However, reflective surfaces like mirrors pose a challenge for LiDAR, making it difficult to distinguish between objects in front and behind a mirror and their reflected counterparts. In this paper, we propose the Shadow Boxing Method, a technique for effectively classifying reflected points that appear to be behind the mirror. We will differentiate reflected points from points behind the mirror, and further restore these identified points to their original 3D locations by accurately detecting the relative position of the mirror and the angle of reflections. This will overcome the existing limitations of LiDAR sensors, enabling more accurate sensing. We envision that our idea can potentially improve the safety of LiDAR-based autonomous vehicles even when confronted with highly reflective surfaces such as traffic safety mirrors.

키워드