Visual simultaneous localization and mapping using direct-based method for unmanned aerial vehicle (UAV)

Badzli, Muhammad Harith (2023) Visual simultaneous localization and mapping using direct-based method for unmanned aerial vehicle (UAV). Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

	Text (24 Pages) Visual simultaneous localization and mapping using direct-based method for unmanned aerial vehicle (UAV).pdf - Submitted Version Download (575kB)
	Text (Full text) Visual simultaneous localization and mapping using direct-based method for unmanned aerial vehicle (UAV).pdf - Submitted Version Restricted to Registered users only Download (5MB)

Abstract

The Direct Sparse Odometry (DSO) technique is a new form of visual odometry that makes use of a direct and sparse structure and motion formulation to achieve unprecedented precision. In this project, the objective is to apply the DSO algorithm on the Unmanned Aerial Vehicle (UAV) application. The main studies in this project are focusing on the DSO Algorithm parameter setting to achieve an optimum setting to be applied on real-time. Another objective is to evaluate the parameter and performance of DSO algorithm. The data evaluation was based on three different environments near the university campus. In this project, the Realsense D435i Camera was applied to the RDDRONE-FMUK66 with interface of the Raspberry Pi 3 B+ model to capture the data. This project managed to analyze suitable point values on the active points and gradient parameter setting. Three environments also have been experimented with the same parameter which concerns on point density, keyframe management and sparsity of the environment. Conclusion from this project is DSO on UAV can be improved in order to gain a stable data processing to apply on the algorithm.

Actions (login required)

View Item