Video Semanticanalysis And Event Detection

In the past few years, many researchers focused on intelligent video surveillance system. As its basis, semantic analysis of video and event detection technology has not only theoretical significance but also practical perspective, which extracts strong research interests.In this paper, a set of quaternion Gabor wavelets are constructed to establish a comprehensive analysis tool for local spectral, spatial and temporal characteristics of fire regions. Quaternion Gaussian kernels are utilized to represent the spectral distribution of fire pixel clusters. In addition, a 2D band-pass filter kernel contained in the quaternion Gabor extracts spatial contours of fire regions. Another 1D temporal filter kernel is enforced to capture random flickering behavior in the fire regions, greatly reducing the false alarm rate. Rather than dealing with color channels separately, the color pixels are encoded as quaternion so as to be clustered as whole units in color spaces.A new framework to recognize fire as well as smoke is presented. Smoke region is recognized from its dynamic textures in the proposed fire surveillance system. Experimental results under a variety of conditions show the proposed vision-based surveillance method is capable of detecting flame and smoke reliably.In this paper, we designed a new event detection framework based on HMM, which is used to detect some predefined semantic events in surveillance video. This method depends on our human detection and tracking system, which provides trajectories features for HMM model. In order to detect more complex event such as ElevatorNoEntry, we design a discrete HMM framework to deal with such complex situation. Based on the discrete HMM model, more features can be used and we can detect events more effectively. Our framework is tested in TRECVID evaluation platform, where we got BEST RUN of People-SplitUp event in 2009.