Pose estimation algorithm for mobile Augmented Reality based on inertial sensor fusion
Augmented Reality (AR) applications have become increasingly ubiquitous as it integrates virtual information such as images, 3D objects, video and more to the real world, which further enhances the real environment. Many researchers have investigated the augmentation of the 3D object on the digital...
Saved in:
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English English |
Published: |
2021
|
Subjects: | |
Online Access: | http://irep.iium.edu.my/96221/1/02-Alam_PoseEstimationv11.pdf http://irep.iium.edu.my/96221/7/96221_Pose_Estimation_Algorithmv11.pdf http://irep.iium.edu.my/96221/ |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | Augmented Reality (AR) applications have become increasingly ubiquitous as it integrates virtual information such as images, 3D objects, video and more to the real world, which further enhances the real environment. Many researchers have investigated the augmentation of the 3D object on the digital screen. However, certain loopholes exist in the existing system while estimating the object's pose, making it inaccurate for Mobile Augmented Reality (MAR) applications. Objects augmented in the current system have much jitter due to frame illumination changes, affecting the accuracy of vision-based pose estimation. This paper proposes to estimate the pose of an object by blending both vision-based techniques and MEMS sensor (gyroscope) to minimize the jitter problem in MAR. The algorithm used for feature detection and description is Oriented-FAST Rotated-BRIEF (ORB), whereas to evaluate the homography for pose estimation, Random Sample Consensus (RANSAC) is used. Furthermore, gyroscope sensor data is incorporated with the vision-based pose estimation. We evaluated the performance of augmenting the 3D object using the techniques, vision-based, and incorporating the sensor data using the video data. After extensive experiments, the validity of the proposed method was superior to the existing vision-based pose estimation algorithms. After incorporating the sensor (gyroscope) data with the vision-based pose estimation, the result shows improved pose estimation performance and augmentation of the 3D object in MAR applications. The proposed method has proven to be successful in overcoming the problem of jitter in the existing system. |
---|