Publication Details
Visual Surveillance Metadata Management
Zendulka Jaroslav, doc. Ing., CSc. (UIFS)
Visual surveillance, metadata management, cameras, vision units, movingobjects, data cleaning, integration, persistence, Kalman filter, classification, object model.
The paper deals with a solution forvisual surveillance metadata management. Data coming from many cameras isannotated using computer vision units to produce metadata representing movingobjects in their states. It is assumed that the data is often uncertain, noisyand some states are missing.
The solution consists of the followingthree layers: (a) data cleaning layer - improves quality of the data bysmoothing it and by filling in missing states in short sequences referred to astracks that represent a composite state of a moving object in a spatiotemporalsubspace followed by one camera. (b) Data integration layer - assigns a globalidentity to tracks that represent the same object. (c) Persistence layer -manages the metadata in a database so that it can be used for onlineidentification and offline querying, analyzing and mining. A Kalman filter techniqueis used to solve (a) and a classification based on the moving object's stateand its visual properties is used in (b). An object model for layer (c) ispresented too.
@inproceedings{BUT28830,
author="Petr {Chmelař} and Jaroslav {Zendulka}",
title="Visual Surveillance Metadata Management",
booktitle="Eighteenth International Workshop on Database and Expert Systems Applications",
year="2007",
series="IEEE CPS",
pages="79--83",
publisher="IEEE Computer Society Press",
address="Regensburg",
isbn="978-0-7695-2932-5"
}