Primjena računalnog vida u održavanju trupa putničkih zrakoplova

master's thesis

• Title (english): Application of computer vision in the maintenance of the fuselage of passenger aircraft
• Author: Damjan Konjevod
• Mentor: Dragutin Lisjak (mentor)
• Committee member: Darko Ivančević (predsjednik povjerenstva)
• Committee member: Milan Vrdoljak (član povjerenstva)
• Committee member: Ivica Smojver (član povjerenstva)
• Granter: University of Zagreb; Faculty of Mechanical Engineering and Naval Architecture; Zagreb
• Defense date and country: 2022-05-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Aviation, Rocket and Space Technology; Aviation Technology and Maintenance

Abstract

U ovome radu razvijena je aplikacija za inspekciju trupa putničkih zrakoplova primjenom tehnologije računalnog vida. Objašnjeni su problemi trenutnog procesa održavanja i prikazana su tipična oštećenja trupa zrakoplova. U trećem je poglavlju dan uvod u tehnologiju računalnog vida, prikazane su njezine tipične zadaće i ukratko su objašnjene različite vrste detektora. Prikazana je primjena konvolucijskih neuronskih mreža, njihova struktura i način na koji funkcioniraju. Za detekciju anomalija izabran je YOLO algoritam, objašnjena je njegova struktura i razlike između različith verzija modela. Detaljno je prikazan proces kreiranja i treniranja konvolucijske neuronske mreže kao i različith verzija YOLO detektora anomalija u programskom paketu MATLAB. Na kraju rada provedena je usporedba rezultata te je izveden zaključak.

Abstract (english)

In this paper, an application for the inspection of the fuselage of passenger aircraft using computer vision technology was developed. The problems of the current maintenance process are explained and typical damage to the fuselage is shown. The third chapter introduces computer vision technology, presents its typical tasks, and briefly explains the different types of detectors. The application of convolutional neural networks, their structure and the way they function are presented. The YOLO algorithm was chosen to detect anomalies, its structure and differences between different versions of the model were explained. The process of creating and training a convolutional neural network as well as different versions of YOLO anomaly detectors in the MATLAB software package is presented in detail. At the end of the paper, a comparison of the results was made and a conclusion was drawn.

• Keywords: inspekcija zrakoplova; tehnologija računalnog vida; konvolucijska neuronska mreža; YOLO algoritam
• Keywords (english): aircraft inspection; computer vision technology; convolutional neural network; YOLO algorithm
• Language: croatian
• URN:NBN: urn:nbn:hr:235:742147
• Study programme: Title: Aeronautical Studies; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka zrakoplovnog inženjerstva (magistar/magistra inženjer/inženjerka zrakoplovnog inženjerstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Embargoed access; Embargo expiration date: 2024-05-11
• Created on: 2022-05-11 15:29:10