Study on Multi Resolution and Depth Information in Partial Stereo Image Registration Scene Using Outer Box Method

Authors

  • Alex Tan Choo Kwok, Rizauddin Ramli,Zuliani Zulkoffli

Abstract

AbstractThis paper presents a stereo vision image registration for multi resolution image. Two PRI images (partial stereo) is captured and a database was constructed by storing multiple view data of CAD model. CAD model shall rotate on an axis at 3 degrees interval and the view is stored into the database. First, PRI shall undergo preprocessing stage to remove background noise and lens distortion. Then, both PRI and CAD images are scaled by using Outer Box Method and transformed into binary image. Correlation and Matched Percentage between CAD and PRI is going through in this stage. PRI produces correlation graph that shows a fluctuate graph in range of 0.98 and 0.99. Its matched percentage graph also shows a fluctuate graph in range of 96.53 to 99.12 percentage. Since the correlation value over 0.90 and match percentage is over 90 percent, it is determined that PRI and CAD are similar. By further analyze the relationship between correlation and the match percentage, the angel of object can be determined. Stereo image matching is faster and more convenient than individual image matching.

Published

2020-12-01

Issue

Section

Articles