20.500.12556/DKUM-47284 ANALIZA VPLIVA BARVNEGA SPEKTRA NA UČINKOVITOST PREPOZNAVANJA OBJEKTOV V DIGITALNIH SLIKAH ANALYSIS OF THE IMPACT OF THE COLOUR SPECTRUM ON OBJECT RECOGNITION EFFICIENCY IN DIGITAL IMAGES V diplomskem delu se ukvarjamo s študijem vpliva izbire barvnega spektra, v katerem so zajete digitalne slike, na uspešnost prepoznavanja objektov. Kot testno aplikacijo smo izbrali problem prepoznavanja prometnih znakov v slabih osvetlitvenih pogojih. S pomočjo sodobnih tehnologij smo razvili algoritem, ki na digitalnih posnetkih prepoznava prometne znake stop. Algoritem smo preizkusili na posnetkih, zajetih v različnih barvnih spektrih, in sicer v realnem okolju ter ob slabih osvetlitvenih pogojih. Najprej smo algoritem preizkusili na posnetkih, zajetih v vidnem spektru, nato pa še na posnetkih, zajetih v infrardečem (IR) spektru. Na podlagi dobljenih rezultatov smo proučili, kako izbira barvnega spektra vpliva na učinkovitost prepoznavanja objektov. Ugotovili smo, da je ob slabih osvetlitvenih pogojih bolje izbrati spekter IR. The diploma thesis deals with the effect of the selected colour spectrum in which the digital images are captured, on the success of object identification. The problem of identifying traffic signs in poor lighting conditions was selected as the test application. Using state-of-the-art technology, an algorithm for identification of stop traffic signs in digital images was developed. The algorithm was tested by using digital images which were taken in different colour spectra, in the real environment and in poor lighting conditions. First, the algorithm was tested using images which were taken in the visible spectrum and later with images which were taken in the infrared spectrum (IR). Based on the obtained results, the effect of the colour spectrum selection on the efficacy of object identification was studied. It was discovered that the IR spectrum is more suitable for poor lighting conditions. računalniški vid prepoznavanje objektov barvni spekter prometni znaki computer vision object recognition colour spectrum traffic signs true true false U. Žižek Slovenski jezik Angleški jezik Diplomsko delo/naloga 2015-01-28 12:06:55 2015-05-22 17:19:42 2022-05-20 03:07:32 2021-05-06 03:22:30 2015 0 [Maribor 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.932.75(043.2) 19026710 URN:SI:UM:DK:P2URUS0C UN_Zizek_Uros_2015.pdf UN_Zizek_Uros_2015.pdf 1 2725CB07E544EF07CC79F401E19D5336 cc839114feac90d01c0891538fc763b666c83590c2428a1c17736c851c6d6e67 9809d518-7c0b-11eb-bb7a-00155d0001ca https://dk.um.si/Dokument.php?lang=slv&id=70920 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0