magistrsko delo
Matej Kristjan Mestek (Author), Janez Perš (Mentor), Klemen Grm (Co-mentor)

Abstract

V sledečem delu je predstavljen sistem računalniškega vida za optimizacijo odreza vroče valjanih jeklenih plošč na vročih škarjah znotraj obrata vroče valjarne. Izdelan sistem služi kot predstavitev delovanja takšnega sistema in kot orodje za pomoč pri odločanju operaterjev. V delu so najprej predstavljeni osnovni problem in zahteve, ki jim mora sistem zadostiti. Opisana so tudi sorodna dela in nekaj osnovnih segmentacijskih metod. Sledi pregled teoretičnih znanj računalniškega vida in konvolucijskih nevronskih mrež, ki so predpogoj za razumevanje izdelanega sistema. Podrobno so prikazane tudi uporabljene metode in algoritmi. Nato predstavimo uporabljeno strojno opremo, izdelan sistem in učenje sistema na realnih podatkih. Na koncu sta opisana delovanje sistema in uspešnost delovanja. Uspešnost delovanja je prikazano v dveh delih. V prvem predstavimo kvaliteto delovanja segmentacijskega modela, ki je ključnega pomena za dobro delovanje celotnega sistema, v drugem delu pa določimo končno kvaliteto delovanja celotnega sistema.

Keywords

računalniški vid;konvolucijske nevronske mreže;U-net;optimizacija;magisteriji;

Data

Language: Slovenian
Year of publishing:
Typology: 2.09 - Master's Thesis
Organization: UL FE - Faculty of Electrical Engineering
Publisher: [M. K. Mestek]
UDC: 004.93(043.3)
COBISS: 156557571 Link will open in a new window
Views: 32
Downloads: 11
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Shape detection and cut optimization of hot-rolled plates
Secondary abstract: In this thesis we present a computer vision system for the crop optimization of hot rolled steel plates in a hot rolling mill. The system is meant to demonstrate the workings of such a system and is used as a tool for operators in the decision making process. In the thesis, we first define the problem and working requirements of the system. We also show an overview of similar works in the field and describe some segmentation methods. Next, we give the basic theoretical knowledge of the used computer vision and convolutional neural network methods, necessary for understanding the workings of the system. The hardware, finished system and its training on real process data is shown next. In the last part, we present the final operation of the system and its performance. The performance is shown in two parts. In the first part, we present the performance of the segmentation model that is vital for the overall good operation of the whole system. In the second part, we show the final performance of the whole system.
Secondary keywords: computer vision;convolutional neural networks;U-net;crop optimization;
Type (COBISS): Master's thesis/paper
Study programme: 1000316
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za elektrotehniko
Pages: XVI, 63 str.
ID: 19325631