Date of Publication

12-6-2023

Document Type

Master's Thesis

Degree Name

Bachelor of Science in Mechanical Engineering (Honors) - Ladderized

Subject Categories

Engineering | Mechanical Engineering

College

Gokongwei College of Engineering

Department/Unit

Mechanical Engineering

Thesis Advisor

Alvin Y. Chua
Timothy Scott C. Chu

Defense Panel Chair

Gerardo Augusto

Defense Panel Member

Edwin Sybingco
Isidro Marfori

Abstract/Summary

Modern inspections on concrete pavements and buildings are done using unmanned aerial vehicles (UAV) and machine vision to detect cracks remotely. While drones are efficient in inspections, they mostly rely on global navigation satellite system (GNSS) for positioning. UAV missions on GNSS-denied areas use either manual or semi-automatic pathing due to the lack of a positioning system. This study proposes the use of an ultrawide-band positioning system as an alternative to control and locate the drone for autonomous flight. When combined with convolutional neural networks, it would be able to detect and approximate the location of the crack. Tests were done using the alternative positioning system and classification and segmentation network to detect and locate the various cracks placed on a rectangular test wall of different sizes. The results show that the inspection system was able to autonomously navigate around the test areas and collect images of the wall along with its cracks. The system was able to detect and segment the cracks with 94.85% accuracy and approximate its location through a visualization of the crack location within the test area. This gives proof that ultrawide-band and convolutional neural networks could be used as a solution for autonomous inspection on GNSS-denied areas. If developed and improved, it would be possible to use this for regular autonomous inspection on important structures without satellite coverage.

Abstract Format

html

Language

English

Format

Electronic

Keywords

Neural networks (Computer science); Building inspection

Upload Full Text

wf_yes

Embargo Period

12-6-2024

Download

Share

COinS