Fuzzy logic based sensor fusion for accurate tracking

Ujwal Koneru, Sangram Redkar, Anshuman Razdan

Research output: Contribution to journalConference article

4 Scopus citations

Abstract

Accuracy and tracking update rates play a vital role in determining the quality of Augmented Reality(AR) and Virtual Reality(VR) applications. Applications like soldier training, gaming, simulations & virtual conferencing need a high accuracy tracking with update frequency above 20Hz for an immersible experience of reality. Current research techniques combine more than one sensor like camera, infrared, magnetometers and Inertial Measurement Units (IMU) to achieve this goal. In this paper, we develop and validate a novel algorithm for accurate positioning and tracking using inertial and vision-based sensing techniques. The inertial sensing utilizes accelerometers and gyroscopes to measure rates and accelerations in the body fixed frame and computes orientations and positions via integration. The vision-based sensing uses camera and image processing techniques to compute the position and orientation. The sensor fusion algorithm proposed in this work uses the complementary characteristics of these two independent systems to compute an accurate tracking solution and minimizes the error due to sensor noise, drift and different update rates of camera and IMU. The algorithm is computationally efficient, implemented on a low cost hardware and is capable of an update rate up to 100 Hz. The position and orientation accuracy of the sensor fusion is within 6mm & 1.5°. By using the fuzzy rule sets and adaptive filtering of data, we reduce the computational requirement less than the conventional methods (such as Kalman filtering). We have compared the accuracy of this sensor fusion algorithm with a commercial infrared tracking system. It can be noted that outcome accuracy of this COTS IMU and camera sensor fusion approach is as good as the commercial tracking system at a fraction of the cost.

Original languageEnglish (US)
Pages (from-to)209-218
Number of pages10
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume6939 LNCS
Issue numberPART 2
DOIs
StatePublished - Oct 5 2011
Event7th International Symposium on Visual Computing, ISVC 2011 - Las Vegas, NV, United States
Duration: Sep 26 2011Sep 28 2011

    Fingerprint

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this