Estimacija poze čovjeka temeljena na fuziji podataka s inercijalnih senzora i kamere

undergraduate thesis

• Title (english): Human pose estimation based on fusion of data from inertial sensors and a camera
• Author: Ivan Gusić
• Mentor: Tamara Petrović (mentor)
• Mentor: Goran Vasiljević (sumentor)
• Committee member: Tamara Petrović (član povjerenstva)
• Committee member: Goran Vasiljević (član povjerenstva)
• Committee member: Dean Martinović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-07-15, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Estimacija ljudske poze ima za cilj određivanje položaja dijelova ljudskog tijela i zglobova, čime se omogućava razumijevanje ponašanja i kretanja čovjeka u okolini. Estimaciju poze čovjeka moguće je odrediti različitim metodama u kojima se vrlo često primjenjuju barem dva različita izvora podataka kako bi procjena bila što preciznija. U ovom završnom radu korištena je kamera koja snima čovjeka kao jedan izvor podataka, te inercijalni senzori postavljeni na tijelo čovjeka (ruke i trup) kao drugi izvor podataka. Nijedan od tih izvora samostalno ne daje zadovoljavajuće rezultate estimacije te se stoga koriste informacije iz oba izvora kako bi se, uz primjenu Kalmanovog filtra, dobio model koji je bolji i stvarnije prikazuje pozu čovjeka. Budući da se pomoću inercijalnih senzora dobiva 3D poza čovjeka, a pomoću kamere 2D poza, napravljena je odgovarajuća pretvorba 3D u 2D pozu primjenom tzv. pinhole modela kamere. Analizom na testnom skupu podataka zaključeno je da je fuzijska metoda bolja za određivanje poze čovjeka u odnosu na metode za određivanje poze samo iz slike ili IMU senzora.

Abstract (english)

Human pose estimation aims to predict the position of human body parts and joints, which enables the understanding of human behavior and movement in the environment. It is possible to calculate the human pose estimation by using different approaches, in which very often at least two different data sources are applied in order to make the estimation as precise as possible. In this bachelor’s thesis, a camera that records a person is used as one data source, and inertial sensors placed on the human body (arms and torso) as another data source. None of these sources independently provide satisfactory estimation results, and therefore information from both sources is used in order to obtain a model that better and more realistically depicts the human pose, with the application of the Kalman filter. Since a 3D pose of a person is obtained using inertial sensors, and a 2D pose is obtained using a camera, a suitable conversion of 3D into a 2D pose was made using the so-called pinhole camera model. The analysis on the test data set concluded that the fusion method is better for determining the pose of a person compared to the methods for determining the pose only from the image or IMU sensor.

• Keywords: pinhole model; Kalmanov filter; fuzija senzora; estimacija poze; ROS; kamera; IMU
• Keywords (english): pinhole model; Kalman filter; sensor fusion; pose estimation; ROS; camera; IMU
• Language: croatian
• URN:NBN: urn:nbn:hr:168:121588
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-15 10:15:59