Real-time Fire Detection for Video Surveillance Applications using Multi Experts System

In this proposed system the method is able to detect �re by analyzing the Videos acquired by surveillance cameras is introduced. Two main novelties have been introduced �rst, complementary Information, respectively based on color, shape variation and Motion analysis, are combined by a multi expert system. Second, a novel descriptor based on a bag-of-words approach has been proposed for representing motion. The previous systems were developed but the results were not much signi�cant as they were having their own limitations. Some of them were working on color while some on shape or motion hence they had own drawbacks. Our system will try to overcome those as we are proposing the system that will have a separate module for shape variation, a module for color evaluation and a module for motion evaluation. The multi expert based system takes the input of all this modules and then the decisions taken by the experts are combined by a MES (multi expert system) classi�er based on a weighted voting rule, which �nally assigns a class to each blob. The Main advantage deriving from this approach lies in the fact that the overall performance of the system can be signi�cantly increases with a relatively small effort