Obogatena resničnost gibanja uporabnika v realnem času na mobilni napravi s pomočjo podatkov globinskega senzorja

ŠKERJANC, NEJC (author), Peer, Peter (mentor)

Ljudje težimo k vedno bolj natančni rekonstrukciji realnega sveta. Uporaba samo RGB podatkovega toka ne daje zadovoljivih rezultatov pri detekciji in rekonstrukciji človeškega telesa. Z vpeljavo globinskega podatkovnega toka, ki nam ga posreduje globinski senzor, lahko rekonstruriramo okolico. Algoritem z uporabo sinhronega RGB in globinskega podatkovnega toka izvaja detekcijo s prepletom obeh tokov. S poznavanjem fizikalnih zakonitosti človeškega telesa lahko detektiramo posamezne dele človeškega telesa. Na podlagi detekcije sledi izris animiranega lika (z uporabo slikovnega atlasa), ki kar najbolje prekriva detektiranega človeka in človeško telo. Rezultat je knjižnica, napisana v C++ jeziku za čimboljšo prehodnost med platformami. Kvalitativna in kvantitativna analiza je pokazala, da je z uporabo sinhronega RGB in globinskega podatkovnega toka detekcija človeškega telesa bolj natančna in učinkovita, kot samo uporaba posameznega podatkovnega toka.

• Language: Slovenian
• Keywords: globinski senzor, detekcija okostja, animacija, obogatena resničnost, realni čas, mobilne naprave
• Work type: Master's thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2016
• PID: 20.500.12556/RUL-87583
• Publication date in RUL: 02.12.2016

Secondary language

• Language: English
• Title: Real-time augmented reality of user motion on mobile device using depth sensor

Abstract

People constantly aim to enhance the precision of the digital reconstruction of the real world. The detection and reconstruction of the human body does not provide comprehensive results with only RGB data stream. By introducing depth data flow from depth sensor we can create the reconstruction of our surroundings. The algorithm performs the detection of the human body with the intertwine of synchronous RGB and depth data stream. With the knowledge of physical rules of the human body proportions we can detect human body parts. With the results of the detection the application then draws an animated character (using sprite sheet) that covers as much as possible of the previously detected human body. The result is a library, which is written in the C++ programming language for a good transferability between platforms. Qualitative and quantitative analysis show that using synchronous RGB and depth data stream make the reconstruction more accurate and effective than using only RGB or depth data stream.

• Keywords: depth sensor, skeleton detection, animation, augmented reality, real-time, mobile devices