Fuzzy classification system for outdoor video databases retrieval

Summary form only given. We propose a content-based video database retrieval system. The aim is to extract moving objects from video sequences in order to classify them into predefined categories according to image-based properties, and then robustly index them. The principal idea is to exploit useful properties of fuzzy geometry classification in order to provide a robust method to classify motion regions into three generic categories: pedestrian, vehicle or no identified object. The classified motion regions are used as templates for metrical training algorithms and as keys for tree indexing technique.