Video Event Recognition by Combining HDP and Gaussian Process

In this paper, we present a framework for automatically analyzing activities and interactions, and recognizing traffic states from surveillance video. Activities and interactions are firstly learned by Hierarchical Dirichlet Process (HDP) models based on low-level visual features. Based on the learning results, a Gaussian Process (GP) classifier is trained to classify the traffic states in online video. Furthermore, the temporal dependencies between video events learned by HDP-Hidden Markov Models (HMM) are effectively integrated into GP classifier to enhance the accuracy of the classification. Our framework couples the benefits of the generative models-HDP with the discriminant models-GP. We validate the proposed model by applying it to the analysis of the three standard video datasets over crowded traffic scenes and compare it with other baseline models. Experimental results demonstrate that our model is effective and efficient.