Building 2D Maps with Integrated 3D and Visual Information using Kinect Sensor

Abstract

RGB-D sensors can be used as a cost effective alternative to expensive laser scanners for small scale indoor mapping applications. Since RGB-D cameras provide both color and depth data, it is possible to fuse these two data frames in 2D mapping to exploit the advantages over laser scanners. An improved technique is to capture the navigation environment in an application for goal-oriented navigation and situation assessment using a mobile robot. The proposed technique is based on incorporating RGB images and 3D information to existing 2D Simultaneous Localization And Mapping (SLAM); which has not been clearly investigated before. By integrating visual and 3D information using an inexpensive RGB-D camera, the mobile robot is better equipped to perform tasks such as sign detection, object detection, and text understanding. The motivation for such an approach is to use 3D and visual information with reduced complexity. The performance of 2D mapping is exploited here by replacing the laser scanner with a RGB-D camera. In addition, memory usage is reduced by selecting key RGB-D frames from its sequence by comparing the number of overlapped RGB features. The proposed approach is evaluated for accuracy and consistency using experimental data gathered from a real environment.