Abstract
Abstract. Complex activity modeling and identification of anomaly is one of the most interesting and desired capabilities for automated video behavior analysis. A number of different approaches have been proposed in the past to tackle this problem. There are two main challenges for activity modeling and anomaly detection: 1) most existing approaches require sufficient data and supervision for learning; 2) the most interesting abnormal activities arise rarely and are ambiguous among typical activities, i.e. hard to be precisely defined. In this paper, we propose a novel approach to model complex activities and detect anomalies by using non-parametric Gaussian Process (GP) models in a crowded and complicated traffic scene. In comparison with parametric models such as HMM, GP models are nonparametric and have their advantages. Our GP models exploit implicit spatial-temporal dependence among local activity patterns. The learned GP regression models give a probabilistic prediction of regional activities at next time interval based on observations at present. An anomaly will be detected by comparing the actual observations with the prediction at real time. We verify the effectiveness and robustness of the proposed model on the QMUL Junction Dataset. Furthermore, we provide a publicly available manually labeled ground truth of this data set.
Highlights
Activity modeling and automatic anomaly detection in videos have become active fields in computer vision and machine learning because of the wide deployments of the surveillance cameras
We propose a novel approach to model activities and detect anomalies using non-parametric Gaussian Process (GP) regression models (Rasmussen and Williams, 2006)
We propose a novel method based on non-parametric GP models to exploit spacial-temporal dependencies among ac
Summary
Activity modeling and automatic anomaly detection in videos have become active fields in computer vision and machine learning because of the wide deployments of the surveillance cameras. Feature-based (descriptor) approaches are useful to model and recognize activities They in general use an appropriate classifier to classify the learned activities and detect anomalies. The typical classifier includes GP classifier and SVM (Chathuramali and Rodrigo, 2012; Althloothi et al, 2014; Hasan and Roy-Chowdhury, 2014) are widely adopted because of their advantage in terms of high classification accuracy and relative simpler learning process They are supervised models and a training data set with manually assigned labels is necessary in advance. We propose a novel approach to model activities and detect anomalies using non-parametric Gaussian Process (GP) regression models (Rasmussen and Williams, 2006). We propose a novel method based on non-parametric GP models to exploit spacial-temporal dependencies among ac- The relationships between different activities in different regions are learned by GP regression model using the Automatic Relevance Determination (ARD) kernel (Chu and Ghahramani, 2005)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have