High-Level Features Extraction for Video Surveillance Indexing and Search

A. Wali and A.M. Alimi (Tunisia)


Feature fusion, Classifier fusion, Optical Flow, Active Contour, Support Vector Machines.


In this paper, we present an overview of a new approach to indexing and search the video sequence by the content that has been developed within the REGIMVid project. A large part of our system has been developed as part of TRECVID’2008 evaluation. The platform termed REGIMVidToolbox provides Hight-level feature extraction from audio-visual content and video search. Our subsystem is based on the MPEG-7 eXperimental Model (XM), with extensions to provide descriptor extraction from shaped image segments. Thereby, it supports local descriptors reflecting real image content. We describe the architecture of the toolbox as well as providing an overview of the descriptors supported to date. We also briefly describe the search task. We then demonstrate the usefulness of the toolbox in the context of feature extraction, objects learning and retrieval in large collections of video surveillance dataset.

Important Links:

Go Back