20.500.12556/RUL-173708 Razvoj sistema za samodejno prepoznavanje ustreznosti nastavitev podvozja gokart vozila na osnovi slik tekalnih površin pnevmatik Development of a system for automatic recognition of the suitability of go-kart chassis settings based on images of tire treads Nastavitve podvozja gokart vozila bistveno vplivajo na oprijem z voziščem. Vsaka nastavitev in vsaka komponenta na gokartu drugače vpliva na oprijem in s tem na vožnjo. Čeprav lahko različne kombinacije nastavitev vodijo do podobnih rezultatov, je obnašanje vozila na stezi drugačno. Vsak voznik ima svojo kombinacijo nastavitev, ki mu najbolj ustreza. Medtem ko izkušeni vozniki prek občutkov in opažanj ob vožnji ter videza tekalnih površin pnevmatik zlahka prepoznajo, ali so nastavitve podvozja ustrezne ter kako jih v primeru neustreznosti popraviti, imajo manj izkušeni vozniki težave že pri ugotavljanju, ali so nastavitve ustrezne ter ali je v primeru neustreznosti težava v tem, da je preveč ali premalo oprijema. V magistrskem delu smo razvili sistem za samodejno prepoznavanje ustreznosti nastavitev podvozja gokart vozila na osnovi slik tekalnih površin pnevmatik. Sistem smo zasnovali na osnovi konvolucijskih nevronskih mrež in metode semantične segmentacije. Rezultati kažejo obetavnost uporabe metode semantične segmentacije. The chassis settings of a go-kart vehicle significantly affect the grip on the track. Each setting and each component on a go-kart affects the grip and driving in a different way. Although different combinations of settings can lead to similar results, the vehicle's behaviour on the track is different. Each driver has their own combination of settings that suits them best. While experienced drivers can easily recognize whether the chassis settings are appropriate and how to correct them in case of inadequacy through their feelings and observations while driving and the appearance of the tire treads, less experienced drivers have difficulty even in determining whether the settings are appropriate and whether, in case of inadequacy, the problem is too much or not enough grip. In this master's thesis, we developed a system for automatically recognizing the adequacy of the chassis settings of a go-kart vehicle based on images of tire treads. We designed the system based on convolutional neural networks and the semantic segmentation method. The results show the promise of using the semantic segmentation method. karting nastavitve podvozja pnevmatike avtomatsko prepoznavanje oprijema slikovna analiza tekalne površine konvolucijske nevronske mreže semantična segmentacija karting chassis settings tires automatic grip recognition image analysis of tire tread convolutional neural networks semantic segmentation true false false Slovenski jezik Angleški jezik Magistrsko delo/naloga 2025-09-20 08:30:25 2025-09-20 08:30:27 2025-10-11 03:50:14 0000-00-00 00:00:00 2025 0 0 XVIII, 77 str. 2025 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 62-52:004.932.2:629.371.26(043.2) 276687 250201091 43937.pdf 43937.pdf 1 7199CBA75025E94E301934BFBCD98EAA 3c554d732721d89ffbf54c87f7e14e263071d5ce4cbc964b70b74efda74026f2 1e8a4b29-95eb-11f0-9328-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=218358 Fakulteta za strojništvo 0 0 0