Real-time3-D Reconstruction Algorithm Based On The Kinect And MonoSLAM

The3D reconstruction is refers to create a mathematical model of the3D objectssuitable for computer representation and processing, in the computer environment,that is the fundation of processing, operating and analysis the objects nature for themathematical models, but also the key technologies of virtual reality. It’s widely usedin autonomous navigation of the mobile robot, aviation and remote sensing, industrialautomation systems, medical imaging, virtual reality and other fields. As one of themost popular research direction of the computer vision,3D reconstruction technologygets more and more attentions of the researchers.This article launched a study on the algorithm of the SLAM(SimultaneousLocalization and Mapping) by introducing the RGB-D camera Kinect to obtain thedepth information of the3D scene, an algorithm of the3-D reconstruction based onthe Kinect and monocular vision SLAM achieved. We mainly completed the followingworks:(1) This paper talks about several kinds of methods and techniques for pointextraction based on the visual characteristics: Harris, SIFT, SURF and also FAST areintroduced respectively. Considering the real-time3D reconstruction, In this paper,we choose the FAST algorihem for corner detection and also correct the cumulativeerror in the SLAM system.(2)We study on the monocular vision SLAM algorithm based on the ExtendedKalman Filter (EKF), including the principle of the EKF, SLAM system model. Weconstructed the monocular vision SLAM framework based on the EKF, on this basis,we did the experiment,and gives the experimental analysis.(3)This paper studied on the visual landmark initialization in the monocularvision SLAM, introducing the depth camera Kinect to obtain the depth map, analyzingthe basic principles of the Kinect, establish a framework of the EKF-SLAM based onthe Kinect. Realize the function of real-time3D reconstruction in the indoorenvironment.(4)This paper also studied on the real-time3D reconstruction based on theKinect and optical flow tracking and structure from motion algorithm, compared withreal-time3D reconstruction based on the Kinect and Monocular vision SLAM, to verify the reliability and stability of the proposed method.