Real-Time Visual SLAM Using FastSLAM and the Microsoft Kinect Camera

The use of visual sensors (i.e. cameras) in Simultaneous Localization and Mapping (SLAM) as a replacement for the traditional laser range finder may, depending on the application, have several key advantages. These include low price, size, and weight as well as the amount of information that can be gathered about the environment. In this paper, measurements are assumed to be taken from low cost cameras, e.g. the recently launched Microsoft Kinect sensor. Information about the environment will be extracted using sophisticated feature extraction and data association algorithms. Here, the challenge lies in the computational complexity of the image processing algorithms. Therefore, a feature extraction algorithm based on the Oriented FAST and Rotated BRIEF (ORB) algorithm will be discussed. Further, a novel efficient and robust data association algorithm for Visual SLAM features will be presented. Initial experiments were conducted in an office environment, showing that the proposed Visual SLAM algorithm achieves precise localization while approaching real-time performance. Topics: Components & Technologies / Mobile Robotics Keywords: SLAM, Vision, Microsoft Kinect, Feature Extraction, Data Association