NORB: A Stream-Based and Non-Blocking FPGA Accelerator for ORB Feature Extraction

Qixing Zhang, Hao Sun, Qi Deng, Heng Yu, Yajun Ha

Research output: Chapter in Book/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

Oriented FAST and Rotating BRIEF (ORB) is a key technique for visual feature extraction and matching, which forms the foundation of the state-of-the-art ORB-SLAM systems. Due to the computational complexity, FPGA accelerators for ORB computing are usually used when running ORB-SLAM on low-power platforms. Previous implementations of ORB accelerators need to block the input stream when computing the rBRIEF descriptors, and cannot achieve high throughput on all pixels. In this paper, we propose a stream-based accelerator of ORB feature extraction which achieves non-blocking computation by caching the columns of the corresponding window buffer. The proposed accelerator is implemented on a Zynq UltraScale SoC, and the experimental result shows it achieves an average latency of 1.4ms, which is 44% faster than the state-of-the-art approach, with similar output accuracy. The system has a low resource utilization with power consumption being only 1.5W.

Original languageEnglish
Title of host publicationICECS 2023 - 2023 30th IEEE International Conference on Electronics, Circuits and Systems
Subtitle of host publicationTechnosapiens for Saving Humanity
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9798350326499
DOIs
Publication statusPublished - 2023
Event30th IEEE International Conference on Electronics, Circuits and Systems, ICECS 2023 - Istanbul, Turkey
Duration: 4 Dec 20237 Dec 2023

Publication series

NameICECS 2023 - 2023 30th IEEE International Conference on Electronics, Circuits and Systems: Technosapiens for Saving Humanity

Conference

Conference30th IEEE International Conference on Electronics, Circuits and Systems, ICECS 2023
Country/TerritoryTurkey
CityIstanbul
Period4/12/237/12/23

Keywords

  • FPGA Accelerator
  • Feature Extraction
  • ORB
  • ORB-SLAM

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Information Systems
  • Biomedical Engineering
  • Electrical and Electronic Engineering
  • Instrumentation
  • Artificial Intelligence

Fingerprint

Dive into the research topics of 'NORB: A Stream-Based and Non-Blocking FPGA Accelerator for ORB Feature Extraction'. Together they form a unique fingerprint.

Cite this