In the last few years, the decrease of hardware costs and the contemporary increase of processing capabilities make possible the development of more and more complex surveillance systems with hundreds of sensors deployed in large areas. New functionalities are now available such as scene understanding, context-dependent processing and real-time user-driven functionality selection. The efficient use of these tools in architecturally complex systems for advanced scene analysis needs the development of specific data fusion algorithms able to merge multi-source information provided by a large number of homogeneous or heterogeneous sensors. In this context, the possibility of distributing the intelligence is one of the more innovative and interesting research fields for such systems. Therefore, several studies are focused on how the logical tasks can be partitioned between smart sensors, intermediate processing nodes and control centers. Typical tasks of surveillance systems (context analysis, object detection, tracking...) are organized into hierarchical chains, where lower levels in the architectures provide input data for higher level tasks. Each element of the architecture is capable of autonomous data processing. The complexity of such systems remarks the importance of good design choices for both logical and physical architecture. The main objective of this book chapter is to present possible solutions to evaluate the overall performance and technical feasibility as well as the interactions of the subparts of distributed multi-sensor surveillance systems. Performance evaluation of a multi-level hierarchical architecture does not pertain only to the accuracy of involved algorithms but several other aspects must be considered as the data communication between smart sensors and higher level nodes, the computational complexity and the memory used. Then, in order to define a procedure to assess the performances of this kind of systems a general model of smart sensors, intermediate processing nodes and control centers has been studied taking into account the elements involved in a multi-sensor distributed surveillance tasks. The advantage of the proposed architecture analysis method are: 1) to allow quantitative comparison of different surveillance structures through the evaluation of performance metrics and 2) to validate the algorithm choice with respect to the physical structure (communication links, computational load...) available. Finally example structures are compared in order to find the best solution to some benchmark problems. © 2011 Springer-Verlag Berlin Heidelberg.
Distributed intelligent surveillance systems modeling for performance evaluation
MALUDROTTU, STEFANO;REGAZZONI, CARLO
2011-01-01
Abstract
In the last few years, the decrease of hardware costs and the contemporary increase of processing capabilities make possible the development of more and more complex surveillance systems with hundreds of sensors deployed in large areas. New functionalities are now available such as scene understanding, context-dependent processing and real-time user-driven functionality selection. The efficient use of these tools in architecturally complex systems for advanced scene analysis needs the development of specific data fusion algorithms able to merge multi-source information provided by a large number of homogeneous or heterogeneous sensors. In this context, the possibility of distributing the intelligence is one of the more innovative and interesting research fields for such systems. Therefore, several studies are focused on how the logical tasks can be partitioned between smart sensors, intermediate processing nodes and control centers. Typical tasks of surveillance systems (context analysis, object detection, tracking...) are organized into hierarchical chains, where lower levels in the architectures provide input data for higher level tasks. Each element of the architecture is capable of autonomous data processing. The complexity of such systems remarks the importance of good design choices for both logical and physical architecture. The main objective of this book chapter is to present possible solutions to evaluate the overall performance and technical feasibility as well as the interactions of the subparts of distributed multi-sensor surveillance systems. Performance evaluation of a multi-level hierarchical architecture does not pertain only to the accuracy of involved algorithms but several other aspects must be considered as the data communication between smart sensors and higher level nodes, the computational complexity and the memory used. Then, in order to define a procedure to assess the performances of this kind of systems a general model of smart sensors, intermediate processing nodes and control centers has been studied taking into account the elements involved in a multi-sensor distributed surveillance tasks. The advantage of the proposed architecture analysis method are: 1) to allow quantitative comparison of different surveillance structures through the evaluation of performance metrics and 2) to validate the algorithm choice with respect to the physical structure (communication links, computational load...) available. Finally example structures are compared in order to find the best solution to some benchmark problems. © 2011 Springer-Verlag Berlin Heidelberg.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
