Aspects of optimal viewpoint selection and viewpoint fusion

摘要: In the past decades, most object recognition systems were based on passive approaches. But in last few years a lot of research was done field active recognition. this context, there are several unique problems to be solved, such as fusion views and selection an optimal next viewpoint. In paper we present approach solve problem choosing (viewpoint selection) these for 3D fusion). We formally define additional optimization show how use reinforcement learning viewpoint training continuous state spaces without user interaction. context focus modeling reward. also multiple density propagation, discuss advantages disadvantages two approaches practical evaluation densities, namely Parzen estimation trees.