Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Acceleration on a reconfigurable logic platform of the ORB-SLAM2 algorithm for autonomous underwater vehicles

Maragkaki Maria

Full record


URI: http://purl.tuc.gr/dl/dias/A6B1708C-30D4-4617-AD8C-BB1E19EE7CB5
Year 2022
Type of Item Diploma Work
License
Details
Bibliographic Citation Maria Maragkaki, "Acceleration on a reconfigurable logic platform of the ORB-SLAM2 algorithm for autonomous underwater vehicles", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2022 https://doi.org/10.26233/heallink.tuc.94051
Appears in Collections

Summary

Over the last few years the use of visual Simultaneous Localization and Mapping(vSLAM) algorithms gained widespread development and use in allareas, e.g., self-driving cars, robots, aerial drones, autonomous underwatervehicles and more. Autonomous underwater vehicles have various applicationsranging from garbage collection in shallow ports and port mapping, tofinding holes in fishery nets. Underwater scenarios are complex and costlydue to the large amount of sensors needed such as Doppler Velocity Log(DVL) sensors, depth sensors etc. The use of vSLAM algorithms in these applicationsis important, leading to a need for real time implementation onlow-power platforms. In this case either a platform with a fast processorbut with high power consumption is used in order to have the real time implementation,or a low-power consumption processor with lower processingpower in frames per second is used, resulting to undesirably slow systemperformance. Field Programmable Gate Arrays (FPGAs) and Graphics ProcessingUnits (GPUs) can offer real time implementation with low energycost. In this thesis we have developed an FPGA-based architecture to acceleratethe most time consuming part of the ORB-SLAM2 algorithm, i.e. theOriented FAST and Rotated BRIEF (ORB) feature extraction part. The proposedarchitecture requires per image 60% less energy vs. the software implementationof the ORB part of ORB-SLAM2 algorithm, while maintainingcompetitive performance vs. a high-end processor.

Available Files

Services

Statistics