Cascaded attention and grouping for object recognition from video

Object detection is an enabling technology that plays a key role in many application areas, such as content based media retrieval. Cognitive vision systems are here proposed where the focus of attention is directed towards most informative processing. The attentive detection system uses a cascade of increasingly complex classifiers of radial basis functions (RBF) networks for the stepwise identification of regions of interest (ROI) and refined object hypotheses. While the coarse classifier is used to determine first approximations on the ROI, more complex classifiers are used to give sufficiently accurate and consistent pose estimates. Objects are modelled by local appearances and in terms of posterior distributions in eigenspace. The experimental results were led for the automatic detection of brand objects in Formula One broadcasts and clearly illustrate the benefit in applying decision making on attention and probabilistic grouping.