Structure Aware SLAM using Quadrics and Planes

  • 2018-08-13 16:09:18
  • Mehdi Hosseinzadeh, Yasir Latif, Trung Pham, Niko Suenderhauf, Ian Reid
  • 0

Abstract

Simultaneous Localization And Mapping (SLAM) is a fundamental problem inmobile robotics. While point-based SLAM methods provide accurate cameralocalization, the generated maps lack semantic information. On the other hand,state of the art object detection methods provide rich information aboutentities present in the scene from a single image. This work marries the twoand proposes a method for representing generic objects as quadrics which allowsobject detections to be seamlessly integrated in a SLAM framework. For scenecoverage, additional dominant planar structures are modeled as infinite planes.Experiments show that the proposed points-planes-quadrics representation caneasily incorporate Manhattan and object affordance constraints, greatlyimproving camera localization and leading to semantically meaningful maps. Theperformance of our SLAM system is demonstrated in https://youtu.be/dR-rB9keF8M .

 

Quick Read (beta)

loading the full paper ...