diplomsko delo
Tjaž Vračko (Author), Božidar Potočnik (Mentor)

Abstract

V diplomskem delu smo raziskali metode računalniškega vida za detekcijo in razpoznavanje evrskih kovancev v digitalnih slikah. Preučili in analizirali smo obstoječe metode za detekcijo kovancev ter predstavili njihove prednosti in slabosti. V delu predstavimo lasten algoritem za detekcijo in razpoznavo evrskih kovancev, ki temelji na Houghovi transformaciji, barvnih in teksturnih značilnicah, postopku na osnovi vreče besed in strojih podpornih vektorjev; za popravljanje rezultatov razpoznave uporablja informacije o velikostnih razmerjih kovancev. Algoritem smo implementirali in testirali na štirih testnih množicah slik. Ugotovili smo, da naš algoritem deluje najbolje na slikah z belim ozadjem, na katerih se nahaja veliko različnih tipov kovancev. Na takšnih slikah smo v povprečju dosegli 81,53-odstotno uspešnost pravilnega razpoznavanja kovancev. Izkazalo se je tudi, da je 20 cm tista oddaljenost kamere od kovancev, pri kateri dosežemo najvišjo uspešnost razpoznave in ustreza razdalji, ki bi jo tudi sicer izbrali za slikanje od 10 do 30 kovancev, položenih na mizo.

Keywords

razpoznavanje kovancev;Houghova transformacija;stroji podpornih valjev;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher: T. Vračko
UDC: 004.921(043.2)
COBISS: 21412630 Link will open in a new window
Views: 1053
Downloads: 124
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: Recognition of coins in digital images by using computer vision and machine learning
Secondary abstract: In this paper we present our research on computer vision methods pertaining to the recognition of euro coins in digital images. We studied and analysed existing coin recognition methods and present their strengths and weaknesses. We introduce our own algorithm for recognising euro coins, which is based on the Hough transform, different colour and texture descriptors, as well as the bag of words principle and support vector machines. It uses knowledge about coin size ratios to later correct the results of the classification process. We have implemented the proposed algorithm and tested it on four testing sets of images. Our algorithm works best on images with a white background, on which multiple kinds of coins are located. On such images we have achieved an average correct coin recognition ratio of 81.53%. The results also show that a camera distance of 20 cm from the coins produces the highest recognition rate. This distance matches one naturally chosen when taking a picture of roughly 10 to 30 coins, laid flat onto a table.
Secondary keywords: coin recognition;HOG;SIFT;Hough transform;support vector machines;
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, 49 str.
ID: 10924736