Title :
Ego-Action Analysis for First-Person Sports Videos
Abstract :
A new algorithm enables a fully automatic real-time video segmentation solution for dynamic first-person sports videos. The proposed approach leverages the latest in robust vision-based ego-motion estimation and unsupervised learning using nonparametric Bayesian modeling.
Keywords :
Bayes methods; image segmentation; sport; unsupervised learning; ego action analysis; ego motion estimation; nonparametric Bayesian modeling; robust vision; sports videos; unsupervised learning; video segmentation; Algorithm design and analysis; Cameras; Computer vision; Inference algorithms; Pervasive computing; Real time systems; Videos; ego-action; ego-motion; first-person point-of-view video; pervasive computing; sports video analysis; wearable computing;
Journal_Title :
Pervasive Computing, IEEE
DOI :
10.1109/MPRV.2012.28
