Exploration of The Vector Fusion Method for Basic Behavior Unit Segmentation from Visual Data

It becomes an increasingly important research area to automatically analyze object behaviors from visually captured data (e.g., motion) or video recordings. Among this research, the automatic basic behavior unit (BBU) discovery is very important. In this paper, we explore the applicability of the vector fusion (SBP) method, a multi-variate vector visualization technique, in BBU segmentation. This technique is also inherently a data dimension reduction technique: it reduces the multiple dimensional data into two dimensional (SBP)space, and the spatial and temporal analysis in SBP space can help discover the underlying data groups. We present results on a physical system and a synthetic mouse-in-a-cage scenario. The vector fusion method provides a good distinction and interpretation for the bouncing ball example and the analytical data from the synthetic video simulation upon certain selected features. Our experiments show that several factors influence the effectiveness of the vector fusion method in BBU segmentation. The temporal analysis in SBP space seems to be very effective to detect periodic BBUs. Overall, this method is simple and effective for grouping BBUs with periodic motion