3D reconstruction from un-calibrated images using optical flow magnitude based method

• Published in: Advanced Science Letters
• Main Author: 2-s2.0-85023766191
• Format: Article
• Language: English
• Published: American Scientific Publishers 2017
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85023766191&doi=10.1166%2fasl.2017.8307&partnerID=40&md5=f4aba5f9d0f6ba2b04440167eb0546d9

3D reconstructions from image sequences captured by single moving camera have increased in demand as it is low in cost, light weight, and easy to be mounted on flying robot to capture the large scale of image. However, the main challenge of this approach is to reconstruct 3D model from un-calibrated image sequences which are captured from various translation and rotation of camera position. Moreover, existing methods which require camera calibration process based on Structure from Motion (SfM) and optimization are cumbersome, affects the generality and accuracy of 3D model. Therefore, this research proposed a 3D reconstruction from un-calibrated images based on optical flow magnitude equation. This approach involves feature extraction and matching, magnitude calculation from optical flow and 3D point reconstruction. An equation developed from optical flow magnitude is used to recover the camera rotation and translation. Therefore, this approach simplifies the process to recover camera rotation and translation parameter which is essential to provide an accurate 3D reconstruction from un-calibrated images. © 2017 American Scientific Publishers All rights reserved.