Abstract
Without neuromorphic hardware, artificial stereo vision suffers from high resource demands and processing times impeding real-time capability. This is mainly caused by high frame rates, a quality feature for conventional cameras, generating large amounts of redundant data. Neuromorphic visual sensors generate less redundant and more relevant data solving the issue of over- and undersampling at the same time. However, they require a rethinking of processing as established techniques in conventional stereo vision do not exploit the potential of their event-based operation principle. Many alternatives have been recently proposed which have yet to be evaluated on a common data basis. We propose a benchmark environment offering the methods and tools to compare different algorithms for depth reconstruction from two event-based sensors. To this end, an experimental setup consisting of two event-based and one depth sensor as well as a framework enabling synchronized, calibrated data recording is presented. Furthermore, we define metrics enabling a meaningful comparison of the examined algorithms, covering aspects such as performance, precision and applicability. To evaluate the benchmark, a stereo matching algorithm was implemented as a testing candidate and multiple experiments with different settings and camera parameters have been carried out. This work is a foundation for a robust and flexible evaluation of the multitude of new techniques for event-based stereo vision, allowing a meaningful comparison.
Highlights
Frame-based visual sensors produce large amounts of redundant data while operating with a limited temporal resolution, which leads to over and undersampling at the same time
It can not be applied to neuromorphic hardware. While this can be a disadvantage when compared to an algorithm that is executed on neuromorphic hardware, it can be seen as an advantage: The algorithm is sufficiently efficient without the need for dedicated hardware
This work provides a robust and flexible way to evaluate this multitude of new algorithms by introducing a comprehensive benchmark environment for depth reconstruction
Summary
Frame-based visual sensors produce large amounts of redundant data while operating with a limited temporal resolution, which leads to over and undersampling at the same time. As artificial stereo vision has always been especially suffering from high resource demands and computation times, researchers developed many event-based techniques for this problem Steffen et al (2019). Beside profiting from less redundant data, event-based methods offer an additional matching criterion: Time. A profound and reasonable evaluation of any algorithm greatly benefits from open benchmark datasets. While they have been done intensively for frame-based stereo vision Seitz et al (2006); Scharstein and Szeliski (2002), comparable studies are still missing for respective event-based techniques
Sign-in/Register to view highlights & summary 
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