20.500.12556/DKUM-20665 HITRO RAZPOZNAVANJE OBRAZOV S KNJIŽNICO OPENCV RAPID FACE RECOGNITION BY USING OPENCV LIBRARY V diplomskem delu raziskujemo možnosti uporabe paralelizacije pri zaznavanju obrazov s pomočjo knjižnice OpenCV. V uvodu predstavimo digitalno zaznavo obrazov z metodo iz OpenCV ter možnosti, ki se ponujajo za njeno pohitritev. Naš cilj je natančno preučiti obstoječo metodo ter jo pohitriti z različnimi postopki. Najprej se lotimo analize možnih pristopov pohitritve te metode, študiramo pa tudi njihove omejitve v praksi. Osrednji del diplome je posvečen razvoju in analizi lastne rešitve, ki temelji na pohitritvi osnovne metode za razpoznavanje obrazov iz OpenCV. Razvili smo jo kot množico različnih pristopov pohitritve. Ugotavljamo, da so implementirani postopki precej hitrejši od obstoječe rešitve iz OpenCV, ter da so nekateri bolj primerni za statične slike, nekateri pa za realno-časovno zaznavo. Z našim pristopom smo dosegli pohitritev do devet krat, glede na osnovno metodo zaznavanja obraza. The thesis investigates the use of parallelization in face detection using OpenCV library. In the introduction, we present a method for face detection in OpenCV and the opportunities offered to improve its performance. Our goal is to carefully examine the existing method, and speed up by different procedures. First, we start the analysis of possible approaches to performance improvements of this method as well as their limitations in practice. The main part of the diploma is dedicated to developing and analyzing our own solution, based on the performance improvements of the basic methods for face detection from OpenCV. It was developed as a multitude of different approaches for performance improvements. We find that procedures that are implemented are much faster than existing solutions in OpenCV, and that some are more suitable for static images, but some for real-time detection. With our approach we have achieved a speed up of nine times, compared to the basic face detection method. zaznavanje obrazov računalniški vid paralelizacija digitalne slike face detection computer vision parallelization digital images true true false [M. Jerot] Slovenski jezik Angleški jezik Diplomsko delo/naloga 2011-09-27 10:05:52 2011-09-27 12:13:24 2022-04-14 08:08:32 2023-12-27 03:20:44 2011 0 Maribor 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.921(043.2) 15587862 URN:SI:UM:DK:BXSSFXH4 UNI_Jerot_Miha_2011.pdf UNI_Jerot_Miha_2011.pdf 1 B609F729CE87B7CB65A173DA9F5434BF 362c26e31aaa8326cdcb7610aca5e644da5da656b3d21a82f6144749918986bf ac3a382e-7c05-11eb-bb7a-00155d0001ca 20.500.12556/dkum/7e21cc57-c245-462e-9bf7-47d42b347a8b https://dk.um.si/Dokument.php?lang=slv&id=25730 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0