Gasiorowski, Pawel, Vassilev, Vassil and Ouazzane, Karim (2018) 3D Simulation-based Analysis of Individual and Group Dynamic Behaviour in Video Surveillance. In: The Third International Conference on Applications and Systems of Visual Paradigms VISUAL 2018, June 24-28, 2018, Venice, Italy.
The visual behaviour analysis of individual and group dynamics is a subject of extensive research in both academia and industry. However, despite the recent technological advancements, the problem remains difficult. Most of the approaches concentrate on direct extraction and classification of graphical features from the video feed, analysing the behaviour directly from the source. The major obstacle, which impacts the real-time performance, is the necessity of combining processing of enormous volume of video data with complex symbolic data analysis. In this paper, we present the results of the experimental validation of a new method for dynamic behaviour analysis in visual analytics framework, which has as a core an agent-based, event-driven simulator. Our method utilizes only limited data extracted from the live video to analyse the activities monitored by surveillance cameras. Through combining the ontology of the visual scene, which accounts for the logical features of the observed world, with the patterns of dynamic behaviour, approximating the visual dynamics of the world, the framework allows recognizing the behaviour patterns on the basis of logical events rather than on physical appearance. This approach has several advantages. Firstly, the simulation reduces the complexity of data processing by eliminating the need of precise graphic data. Secondly, the granularity and precision of the analysed behaviour patterns can be controlled by parameters of the simulation itself. The experiments prove in a convincing manner that the simulation generates rich enough data to analyse the dynamic behaviour in real time with sufficient precision, completely adequate for many applications of video surveillance.
Available under License Creative Commons Attribution Non-commercial No Derivatives.
Download (620kB) | Preview
Downloads
Downloads per month over past year
Downloads each year
View Item |