Vision Based SLAM in Dynamic Scenes

Abstract

In this project, we studied the vision-based simultaneous localization and mapping (SLAM) problem from a novel perspective. We employed multiple independently moving cameras, while conventional studies are limited with a single camera (or a multi-camera rig where the relative positions between cameras are fixed). Our flexible configuration of cameras makes this algorithm applicable to robot teams, which also makes this study the world's first vision based SLAM algorithm for robot teams. Furthermore, the collaboration among multiple cameras allows us to deal with challenging dynamic scenes which make most of previous algorithms fail. This work was accepted for publication at the IEEE Transaction of Pattern Analysis and Machine Intelligence (TPAMI), with impact factor 5.9, #1 in all engineering and AI.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 20, 2012
Accession Number
ADA577304

Entities

People

  • Ping Tan

Organizations

  • National University of Singapore

Tags

Communities of Interest

  • Autonomy
  • Materials and Manufacturing Processes
  • Sensors

DTIC Thesaurus Topics

  • Algorithms
  • Artificial Intelligence
  • Autonomous Navigation
  • Cartography
  • Computations
  • Computer Graphics
  • Computer Science
  • Detection
  • Detectors
  • Electronic Mail
  • Engineering
  • Guidance
  • Kalman Filters
  • Maps
  • Micro Air Vehicles
  • Robots
  • Simultaneous Localization And Mapping

Fields of Study

  • Computer science

Readers

  • Computer Vision.
  • Technical Research and Report Writing.

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • Autonomy