Multi and single epipolar geometry-based filters vs. affine and conformal 2D transformation-based filters
Department of Civil Engineering, Benghazi University, Benghazi, Libya.
Research Article
Global Journal of Engineering and Technology Advances, 2022, 10(03), 032–051.
Article DOI: 10.30574/gjeta.2022.10.3.0047
Publication history:
Received on 04 February 2022; revised on 11 March 2022; accepted on 13 March 2022
Abstract:
The accuracy of Optical Robot Navigation (ORN) depends mainly on the quality of the Automatic Image Matching (AIM) results. Conformal 2D transformation-based Filter (C-2DF), affine 2D transformation-based Filter (A-2DF) and Single Epipolar Geometry-based Filter (S-EGF) are common filters used in ORN. Multi-Epipolar Geometry-based Filter (M-EGF) has been introduced, tested and evaluated by the author and compared by each filter of those mentioned above in separate research papers. This paper comes to include all these individually comparisons, with additional details, showing the advantages and limitations of all filters comparing to each other. Tests show that C-2DF and A-2DF have failed to deal with AIM results in areas with open, narrow, and confused DOF. Also, they have failed to find out the right mathematical model in data with high rate of mismatched points and data obtained from images with difficult view angles. With limited DOF and low rate of errors, C-2DF and A-2DF have provided relatively sufficient results, which can be used for ORN applications that do not require precision. A-2DF is relatively better than C-2DF due to its flexibility to deal with figures including different scales, which is the case when dealing with different levels of DOF and different capturing angles between the cameras. The processing time is another disadvantage of C-2DF and A-2DF, where these are based on iterative estimation methods. Tests display how A-2DF is slowest, which keeps it away from using with real time ORN applications.
Tests show that S-EGF and M-EGF are timesaving and able to deal with any AIM results, regardless the DOF, view angle and errors rate in observations. S-EGF is affected in areas including lines parallel to the cameras base line. M-EGF has offered the best results in terms of providing error-free filtered matched points in all tests. This can be attributed to the high restriction level of this filter, where the probability for the mismatched point to pass throughput the three co-planarity equations is nearly zero. M-EGF and S-EGF are affected by the quality of the Interior Orientation Elements (IOEs) and Exterior Orientation Parameters (EOPs) of the three cameras, leading to rejection a small number of corrected matched points, which can be avoided with professionally manufactured ORN systems. Tests illustrated that S-EGF and M-EGF are extremely high-speed and S-EGF is the faster and M-EGF, C-2DF, and A-2DF comes after, respectively. All results indicated that M-EGF is the best, as it is fast, restricted, reliable, and error-free technique and is suitable for real-time precise ORN applications.
Keywords:
Multi and Single Epipolar Geometry; Affine and Conformal 2D Transformation; Automatic Image Matching; SLAM; Optical Robot Navigation
Full text article in PDF:
Copyright information:
Copyright © 2022 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution Liscense 4.0