Zaznavanje trkov na daljnovodih z optičnim interferometrom

diplomsko delo

Jan Mikolič (Author), Aleš Holobar (Mentor), Denis Đonlagić (Co-mentor)

Abstract

V diplomski nalogi skušamo ugotoviti, v kolikšni meri je možno zaznavati in klasificirati trke na jeklenicah daljnovodov z optičnim interferometrom. Na začetku predstavimo osnovne pojme interferometrije in opišemo uporabljen optični interferometer. V jedru diplomske naloge natančneje opišemo eksperimentalni protokol in obdelavo signalov. Nadaljujemo z implementacijo algoritmov za segmentacijo in klasifikacijo zajetih signalov ter predstavimo dobljene rezultate. Segmentacijo izvedemo v domeni števila prehodov signala skozi ničlo, za klasifikacijo pa uporabimo večplastno nevronsko mrežo z algoritmom vzvratnega učenja. Rezultati študije nakazujejo, da sta implementirani segmentacija in klasifikacija uspešni v 77 % izvedenih trkov različnih predmetov.

Keywords

interferometrija;obdelava signalov;klasifikacija;detekcija trkov;računalniški algoritmi;diplomske naloge;

Data

Language:	Slovenian
Year of publishing:	2017
Typology:	2.11 - Undergraduate Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	J. Mikolič
UDC:	004.421:528.872(043.2)
COBISS:	20869398
Views:	1651
Downloads:	184
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Collision detection on transmission lines with optical interferometer
Secondary abstract:	We analyse feasibility of collision detection on transmission lines with optical interferometer. We first provide a brief introduction into interferometry, along with a description of the optical interferometer used for measurements in this study. Afterwards, we describe the conducted experimental protocol and signal processing methodology. The focus is on implementation of algorithms for signal segmentation and collision classification. We used zero-crossing algorithm to transform signals into segmentation domain. Classification of collisions is done with a multilayer neural network trained by the backpropagation algorithm. The results demonstrate an average success rate of 77% for segmentation and classification of collision with five different objects.
Secondary keywords:	interferometry;signal processing;classification;collision detection;
URN:	URN:SI:UM:
Type (COBISS):	Bachelor thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Računalništvo in informacijske tehnologije
Pages:	XII, 34 str.
ID:	10862120