Detection of Depressions and Holes in 3D Objects

Cilj ovog diplomskog rada je izrada programa za detekciju rupa i udubljenja na scenama snimljenim RGB-D kamerom. Rubovi rupa i udubljenja matematički predstavljaju sedlaste površine. Objašnjena je matematička pozadina geometrijskih struktura koje će se detektirati. Razvijen je program za detekciju sedlastih površina realiziran u okviru programske biblioteke Robot Vision Library (RVL). Izrađena je funkcija koja detektirane površine predočava uz pomoć Visualization Toolkita (VTK). Program je testiran na raznim scenama na kojima su se nalazili objekti s rupama i udubljenjima poput šalica, zdjela, banana, krofni i rola wc papira u raznim položajima.This work addresses the problem of recognizing holes and dents in scenes captured by RGB-D camera. Edges of holes and dents are mathematically considered as a part of saddle surfaces. Saddle surfaces from the mathematical point of view and corresponding algorithm for detection of saddle surfaces are explained. Saddle surface detection program in this work is developed as a part of Robot Vision Library (RVL) developed by the Faculty of Electrical Engineering, Computer Science and Information Technology (FERIT) in Osijek. RVL will be complemented with a function which detects saddle surfaces implementing explained algorithm and a function for visualization of those detected surfaces. The program was tested on multiple scenes containing objects like donut, mug, bowl, banana and toilet paper in different position