2012 | OriginalPaper | Chapter
Online Topological Mapping of a Sparse Camera Network
Authors : Paulo Freitas, Paulo Menezes, Jorge Dias
Published in: Technological Innovation for Value Creation
Publisher: Springer Berlin Heidelberg
Activate our intelligent search to find suitable subject content or patents.
Select sections of text to find matching patents with Artificial Intelligence. powered by
Select sections of text to find additional relevant content using AI-assisted search. powered by
This article presents a method that without any prior knowledge of a network of cameras, in an automatic and in an on-line mode, estimates their positional relation. The developed method is also able to adapt the estimation of the topology automatically, if some change is made on the network, as adding or removing cameras, and changing the position relative to the adjacent ones. To compute the topological map of a sparse network of cameras, our approach registers events, generated by the entrance and exit of agents (persons or movable objects) from the visible area of the cameras. These events are classified as IN and OUT respectively, and they are stored and associated according to a predefined logic based on the measured entropy level of the overall of events, and joined in relation groups. These groups are then analyzed to infer the topological map. However, it is necessary to develop approaches to avoid and eliminate redundancy of data and remove false relations between the sensors. By this, to achieve reliable results and to prevent loss of data, a method is proposed to adjust the computation of the acquired data based on the entropy evaluation. Then, two algorithms are used to combine the events and to create the relation groups at predefined levels of entropy. Based on the categorization of the relation of the events and groups, is possible to estimate the topological information of sparse networks of cameras. Our experimental results show that, even in situations where there is a significant disorder in the registered events, our method is robust and performs valid estimations.