PhD forum: Investigating the performance of a multi-modal approach to unusual event detection

Abstract

Abstract-In this paper, we investigate the parameters underpinning our previously presented system for detecting unusual events in surveillance applications [1]. The system identifies anomalous events using an unsupervised data-driven approach. During a training period, typical activities within a surveilled environment are modeled using multi-modal sensor readings. Significant deviations from the established model of regular activity can then be flagged as anomalous at run-time. Using this approach, the system can be deployed and automatically adapt for use in any environment without any manual adjustment. Experiments carried out on two days of audio-visual data were performed and evaluated using a manually annotated groundtruth. We investigate sensor fusion and quantitatively evaluate the performance gains over single modality models. We also investigate different formulations of our cluster-based model of usual scenes as well as the impact of dynamic thresholding on identifying anomalous events. Experimental results are promising, even when modeling is performed using very simple audio and visual features.

Learning Algorithms for Anomaly Detection from Images

“…As was concluded in (Ahmed, 2013), "the field of automated visual surveillance has attracted a lot of recent attention, with many large research projects such as an European Erasmus Mundus project (Erasmus Mundus, n.d.) granted on the subject. Some of the most recent research works on outlier detection in image sequences are being performed at the Computer Vision Laboratory at the Swiss Federal Institute of Technology (ETH) Zurich, Switzerland (Breitenstein et al, 2009;Schuster et al, 2010), and at the Center for Sensor Web Technologies at Dublin City University in Dublin, Ireland (Kuklyte et al, 2009;Kuklyte et al, 2011). "…”

Efficient and effective automated surveillance agents using kernel tricks

“…Some of the most recent research on outlier detection in image sequences is being carried out at the Computer Vision Laboratory at the Swiss Federal Institute of Technology (ETH) Zurich in Zurich, Switzerland, 31,32 and at the Center for Sensor Web Technologies at Dublin City University in Dublin, Ireland. 33,34 Gowsikhaa et al have added the supplemental capabilities of face, hand and gesture recognition, to determine suspicious specific actions such as looking over a neighbour’s script during an examination. 35 Most of these researchers, however, only present their own algorithms, without providing any quantitative performance comparisons on labelled data.…”

“…Kuklyte et al conducted experiments using an audio-visual camera placed in the corridor outside their research laboratory, and the images presented look similar to the images obtained from our real-life outdoor setting. Kuklyte et al 34 also investigated the parameters involved with the system presented in their earlier work. 33 They claimed in later work 34 that their proposed system can be easily deployed and automatically adapt to any environment without any manual adjustment.…”

“…Kuklyte et al 34 also investigated the parameters involved with the system presented in their earlier work. 33 They claimed in later work 34 that their proposed system can be easily deployed and automatically adapt to any environment without any manual adjustment. This condition is also met by the agents proposed in this paper, and is experimentally demonstrated by application to a simple, already-deployed system.…”