Sustav za detekciju prometnih traka u slikama snimljenim iz vozila

undergraduate thesis

• Title (english): System for Detection of Traffic Lanes in Images Recorded from a Vehicle
• Author: Marino Voćanec
• Mentor: Tomislav Hrkać (mentor)
• Committee member: Tomislav Hrkać (predsjednik povjerenstva)
• Committee member: Marin Oršić (član povjerenstva)
• Committee member: Josip Šarić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U ovom radu opisana je implementacija sustava za detekciju prometnih
traka. Sustav je implementiran primjenom dubokog učenja odnosno pomoću
umjetne potpuno konvolucijske neuronske mreže. Arhitektura mreže korištena
za segmentaciju slika je inspirirana modelom U-Net. Nadalje, u radu se opisuje
postupak odabira i pripreme skupa slika za učenje, validaciju i ispitivanje.
Također, razmotren je i problem odabira modela s odgovarajućom razinom
ekspresivnosti. Nakon odabira opisuje se postupak učenja na većem skupu
slika, dobiveni rezultati se interpretiraju te se ukazuje na moguće probleme,
nedostatke i ograničenja trenutne inačice sustava. Na koncu, predlažu se
moguće buduće nadogradnje sustava kako bi postao dovoljno kvalitetan za
ugradnju u cestovna vozila.

Abstract (english)

This paper desribes implementation of a system used for road lane
detection. System is implemented with the application of deep learning,
convolutional neural networks to be exact. Architecture of a network used for
image segmentation is inspired by the model U-Net. Furthermore, this paper
describes the process of choosing and preparing data for training, validation
and testing. The problem of choosing adequate level of expressiveness for a
model is also considered. After choosing the right model, process of training a
model on a larger dataset is described, results are interpreted and possible
problems, imperfections and limitations of a current version of a system are
indicated. At last, paper proposes possible future upgrades which could
improve the system in a way to make it possible to build in road vehicles.

• Keywords: prometne trake; autonomna vožnja; detekcija; segmentacija; duboko učenje; konvolucijske neuronske mreže; U-Net
• Keywords (english): road lanes; autonomous driving; detection; segmentation; deep learning; convolutional neural networks; U-Net
• Language: croatian
• URN:NBN: urn:nbn:hr:168:251921
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2020-01-07 22:55:02