Wearable cameras, such as Google Glass and Go Pro, enable video datacollection over larger areas and from different views. In this paper, we tacklea new problem of locating the co-interest person (CIP), i.e., the one who drawsattention from most camera wearers, from temporally synchronized videos takenby multiple wearable cameras. Our basic idea is to exploit the motion patternsof people and use them to correlate the persons across different videos,instead of performing appearance-based matching as in traditional videoco-segmentation/localization. This way, we can identify CIP even if a group ofpeople with similar appearance are present in the view. More specifically, wedetect a set of persons on each frame as the candidates of the CIP and thenbuild a Conditional Random Field (CRF) model to select the one with consistentmotion patterns in different videos and high spacial-temporal consistency ineach video. We collect three sets of wearable-camera videos for testing theproposed algorithm. All the involved people have similar appearances in thecollected videos and the experiments demonstrate the effectiveness of theproposed algorithm.
展开▼
