Abstract

Understanding and considering refraction effects are important parts of the demanding task of multimedia photogrammetry, especially with planar interfaces, so-called ”flat ports”. Yet, it remains challenging to determine reliable calibration results that are both quickly acquired and physically interpretable. In this contribution, a novel object-based optimization algorithm, relying on ray tracing methods, is introduced. It enables calibrating physical parameters of all involved refractive properties with reduced computational effort, compared to other standard algorithms in ray tracing. We show that this solution produces equally accurate results as other ray tracing approaches while improving processing speed by a factor of approximately ten and providing a statistical metric in object space. Furthermore, we show in a laboratory investigation that explicit calibration of refractive properties is crucial even with orthogonally aligned bundle-invariant interfaces for highest accuracy, as accuracy in object space is decreased by about 10% with implicit calibration. With deviation from orthogonality by about ten degrees this decreases even further to almost no useful results and accuracy loss of more than 50% compared to explicit calibration results.

Highlights

  • Image acquisition in or through multiple media, especially water, suffers from many quality-degrading and geometry-altering influences, compared to single-media image acquisition

  • Flat ports are widely used and it remains of high interest to fully understand and calibrate the imaging geometry for photogrammetric analyses, especially when requiring high accuracy

  • Algorithms for optical real-time orientation, such as simultaneous localization and mapping (SLAM) require high performance. This applies to large structure from motion (SfM) networks which are impracticable with many models due to long computation times

Read more

Summary

Introduction

Image acquisition in or through multiple media, especially water, suffers from many quality-degrading and geometry-altering influences, compared to single-media image acquisition. The ray path is altered at refracting interfaces according to Snell’s law, rendering the standard pinhole model with additional distortion parameters invalid (Treibitz et al 2012). This is often compensated by either calibrating under water with standard parameters of photogrammetric models (e.g. Brown (1971)), designed for air-applications to implicitly calibrate the imaging geometry or by explicit calibration of the refractive imaging properties. Algorithms for optical real-time orientation, such as simultaneous localization and mapping (SLAM) require high performance This applies to large structure from motion (SfM) networks which are impracticable with many models due to long computation times. This improves these disadvantages by introducing a different error function to reduce computational complexity significantly

Methods
Results
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call