Structured-light Depth-sensing Camera for SLAM

Abstract

Structured light depth cameras offer a fast, accurate, and affordable means of depth-estimation for augmented reality applications using Simultaneous Localization and Mapping (SLAM) that traditional LIDAR scanners cannot. By projecting light-encoded information into a scene, structured light methods solve the traditional stereo correspondence problem more efficiently, allowing real-time, high resolution 3D reconstruction. Although temporal projection schemes produce accurate depth maps, they require heavy thresholding and averaging. In this paper we present a temporal projection scheme using Walsh codes that is self-referencing, fast, and accurate.