Human intention inference based on visual sensor data

Prisutnost robota i drugih autonomnih sustava u svakodnevnim okruženjima poput kućanstava, radnih okruženja, škola i bolnica iziskuje sigurnost u upravljanju takvim sustavima s naglaskom na očuvanju sigurnosti čovjeka. Precizno i pouzdano predviđanje namjere kretanja osobe koje potiče sigurnu i učinkovitu suradnju čovjeka i robota može biti temeljem autonomnih sustava, posebice ako se može postići lako dostupnima i financijski pristupačnim uređajima poput RGB-D kamerom. Kinect senzorom mjereni su položaji kostura osobe pri radnji dohvaćanja jednog od 2, 3 ili 4 predmeta postavljenih na radnu površinu ispred osobe. Predloženi model korišten je za predviđanje predmeta koji će biti dohvaćen, što predstavlja namjeru tog kretanja.The increasing presence of robots and other autonomous systems in everyday environments, such as households, workplaces, schools and hospitals, requires their safe cohabitation and collaboration with humans. Accurate and reliable human intention recognition that fosters safe and efficient human-robot collaboration is thus a crucial component of these autonomous systems, especially if attainable from relatively cheap and widely available sensors such as RGB-D cameras. Skeleton tracking of a person reaching for one of two, three or four objects placed on a table in front of a person is obtained using Kinect camera. Proposed model was used for predicting which object is being reached, i.e. predicting the intention of reaching motions