Gradnja 3D modela s pomočjo oblaka točk in barvnih slik

Blatnik, Blažka (author), Skočaj, Danijel (mentor)

Na področju umetnega zaznavanja, računalniškega vida in robotike se pogosto uporabljajo različni viri podatkov o okolju. Edinstven in informativen vir je tudi oblak točk, ki predstavlja in opisuje 3D prostor okoli nas. Področja uporabe pa segajo precej preko robotike, saj je mogoče oblak točk zajeti in združiti v reprezentativen model objekta iz realnega prostora. S kasnejšo obdelavo lahko dosežemo, da je pridobljeni model dovolj natančno reprezentativen. Primerno obdelan objekt je zato mogoče tudi natisniti. Cilj diplomske naloge je bil izboljšati algoritem ICP z uporabo podatkov z barvne slike za začetno oceno pozicije. Ocenjena pozicija je bila rezultat prilagojenega algoritma SfM.

• Language: Slovenian
• Keywords: 3D, oblak točk, globinska kamera, sledenje, rekonstrukcija
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2018
• PID: 20.500.12556/RUL-102758
• Publication date in RUL: 07.09.2018

Secondary language

• Language: English
• Title: Building a 3D model with point cloud and color images

Abstract

In the field of artificial recognition, computer vision and robotics there are usually different means of getting representation of surrounding place. One of a kind and unique source of data is a point cloud representation which gives us 3D description of the surroundings. Use cases of point cloud far surpass robotics as it can be used to reconstruct a 3D model from the real world. With post processing we can obtain accurate models that can be later on printed with a 3D printer. The goal of this thesis has been to improve ICP algorithm with the use of 2D data from colored images, to approximate camera positions. Approximate camera position is the end result of SfM algorithm.

• Keywords: 3D, point cloud, depth camera, tracking, reconstruction