Output details

An EM/E-MRF algorithm for adaptive model based tracking in extremely poor visibility

This paper presents an algorithm for visual segmentation and tracking of moving objects in image sequences. The work is unusual, in that it tackles the problem of vision in extremely poor visibility conditions – in contrast the vast majority of machine vision is done in comparatively clear visibility conditions. Poor visibility is tackled by a novel machine learning approach, where models of both the tracked object and the background are continuously relearned online during tracking. Additionally, poor visibility is tackled by presenting a novel extension of Markov Random Field theory which combines observations with predictions within a principled probabilistic framework.