A framework for a video analysis tool for suspicious event detection

This paper proposes a framework to aid video analysts in detecting suspicious activity within the tremendous amounts of video data that exists in today’s world of omnipresent surveillance video. Ideas and techniques for closing the semantic gap between low-level machine readable features of video data and high-level events seen by a human observer are discussed. An evaluation of the event classification and detection technique is presented and a future experiment to refine this technique is proposed. These experiments are used as a lead to a discussion on the most optimal machine learning algorithm to learn the event representation scheme proposed in this paper.

A resource-efficient event algebra

Events play many roles in computer systems, ranging from hardware interrupts, over event-based software architecture, to monitoring and managing of complex systems. In many applications, however, individual event occurrences are not the main point of concern, but rather the occurrences of certain event patterns. Such event patterns can be defined by means of an event algebra, i.e., expressions representing the patterns of interest are built from simple events and operators such as disjunction, sequence, etc.We propose a novel event algebra with intuitive operators (a claim which is supported by a number of algebraic properties). We also present an efficient detection algorithm that correctly detects any expression with bounded memory, which makes this algebra particularly suitable for resource-constrained applications such as embedded systems.     

Information assimilation framework for event detection in multimedia surveillance systems

Most multimedia surveillance and monitoring systems nowadays utilize multiple types of sensors to detect events of interest as and when they occur in the environment. However, due to the asynchrony among and diversity of sensors, information assimilation – how to combine the information obtained from asynchronous and multifarious sources is an important and challenging research problem. In this paper, we propose a framework for information assimilation that addresses the issues – “when”, “what” and “how” to assimilate the information obtained from different media sources in order to detect events in multimedia surveillance systems. The proposed framework adopts a hierarchical probabilistic assimilation approach to detect atomic and compound events. To detect an event, our framework uses not only the media streams available at the current instant but it also utilizes their two important properties – first, accumulated past history of whether they have been providing concurring or contradictory evidences, and – second, the system designer’s confidence in them. The experimental results show the utility of the proposed framework.     

Automatic real-time event detection for seismic networks

A low-power microprocessor-based seismic signal detection system has been developed for monitoring and recording earthquakes detected from multichannel seismic network data. The system operates on-line to the network and may be battery powered for field applications or as a precaution against power failures. The algorithm is designed to cope with varying noise conditions at the network outstations and detects local, regional, and teleseismic events. The sampling rate for each channel is 100 samples/sec. with 10-bit accuracy. The system is operating with the LOWNET network of BGS and is suitable for large or small aperture networks because 20 sec of data for each channel are stored in memory. The system is designed to detect seismic events for recording on either analogue or digital recorders depending on the requirements and preferences of the network management.     

Sensor virtualization for underwater event detection

Distributed event detection is a popular application in Underwater Wireless Sensor Networks (UWSNs). The Base Station (BS) collects the measurements from multiple sensor nodes, and makes a decision based on the sensors’ reports. However, due to the unpredictable moving of underwater sensor nodes and interference among multiple events, it is difficult to guarantee the accuracy of event detection. In this paper, we propose a sensor virtualization approach to deal with the event detection problem in UWSNs. The final decision making at the BS will be implemented with the reports of multiple virtual sensors. Although the events may happen in a large scale, the locations where the events happen are relatively sparse in the underwater environment. Consider the sparse property of events, we employ the technique of compressive sensing to recover the original signal from the correlated sensors’ measurements. Through a proper signal reconstruction, the accurate event detection can be reached with a remarkable low sensing overhead. We implement the sensor virtualization based on the compressive sensing technique. Our approach is suitable for the high dynamic topology of UWSN, and it can improve the accuracy of event detection and reduce energy consumption in UWSNs.     

Optimal sensor activation for diagnosing discrete event systems

The problem of dynamic sensor activation for event diagnosis in partially observed discrete event systems is considered. Diagnostic agents are able to activate sensors dynamically during the evolution of the system. Sensor activation policies for diagnostic agents are functions that determine which sensors are to be activated after the occurrence of a trace of events. The sensor activation policy must satisfy the property of diagnosability of centralized systems or codiagnosability of decentralized systems. A policy is said to be minimal if there is no other policy, with strictly less sensor activation, that achieves diagnosability or codiagnosability. To compute minimal policies, we propose language partition methods that lead to efficient computational algorithms. Specifically, we define “window-based” language partitions for scalable algorithms to compute minimal policies. By refining partitions, one is able to refine the solution space over which minimal solutions are computed at the expense of more computation. Thus a compromise can be achieved between fineness of solution and complexity of computation.     

Semi-automated CCTV surveillance: The effects of system confidence,system accuracy and task complexity on operator vigilance,reliance and workload

Recent advances in computer vision technology have lead to the development of various automatic surveillance systems, however their effectiveness is adversely affected by many factors and they are not completely reliable. This study investigated the potential of a semi-automated surveillance system to reduce CCTV operator workload in both detection and tracking activities. A further focus of interest was the degree of user reliance on the automated system. A simulated prototype was developed which mimicked an automated system that provided different levels of system confidence information. Dependent variable measures were taken for secondary task performance, reliance and subjective workload. When the automatic component of a semi-automatic CCTV surveillance system provided reliable system confidence information to operators, workload significantly decreased and spare mental capacity significantly increased. Providing feedback about system confidence and accuracy appears to be one important way of making the status of the automated component of the surveillance system more ‘visible’ to users and hence more effective to use.     

DSNotify – A solution for event detection and link maintenance in dynamic datasets

The dynamics of linked datasets may lead to broken links if data providers do not react to changes appropriately. Such broken links denote interrupted navigational paths between resources and may lead to unavailability of data. As a possible solution, we developed DSNotify, an event-detection framework that informs actors about various types of changes and allows them to maintain links to resources in distributed linked data sets. For representing changes we developed the DSNotify Eventset Vocabulary. Different from other vocabularies it applies a resource-centric perspective and preserves the timely order of changes. We further describe our reusable evaluation infrastructure, which can be extended for extracting change sets from arbitrary linked datasets.     

Motion analysis for event detection and tracking with a mobile omnidirectional camera

A mobile platform mounted with omnidirectional vision sensor (ODVS) can be used to monitor large areas and detect interesting events such as independently moving persons and vehicles. To avoid false alarms due to extraneous features, the image motion induced by the moving platform should be compensated. This paper describes a formulation and application of parametric egomotion compensation for an ODVS. Omni images give 360 view of surroundings but undergo considerable image distortion. To account for these distortions, the parametric planar motion model is integrated with the transformations into omni image space. Prior knowledge of approximate camera calibration and camera speed is integrated with the estimation process using a Bayesian approach. Iterative, coarse-to-fine, gradient-based estimation is used to correct the motion parameters for vibrations and other inaccuracies in prior knowledge. Experiments with a camera mounted on various types of mobile platforms demonstrate successful detection of moving persons and vehicles.Published online: 11 October 2004     

On-line fault detection in discrete event systems by Petri nets and integer linear programming

The paper addresses the fault detection problem for discrete event systems in a Petri Net (PN) framework. Assuming that the structure of the PN model and the initial marking are known, faults are modelled by unobservable transitions. Moreover, we assume that there may be additional unobservable transitions associated with the system legal behaviour and that the marking reached after the firing of any transition is unknown. The proposed diagnoser works on-line: it waits for the firing of an observable transition and employs an algorithm based on the definition and solution of some integer linear programming problems to decide whether the system behaviour is normal or exhibits some possible faults. The results characterize the properties that the PN modelling the system fault behaviour has to fulfill in order to reduce the on-line computational effort.     

Event rates and aggregation in hierarchical discrete event systems

A discrete event system (DES) is a dynamical system whose evolution in time develops as the result of the occurrence of physical events at possibly irregular time intervals. Although many DES's operation is asynchronous, others have dynamics which depend on a clock or some other complex timing schedule. Here we provide a formal representation of the advancement of time for logical DES via interpretations of time. We show that the interpretations of time along with a timing structure provide a framework to study principles of the advancement of time for hierarchical DES (HDES). In particular, it is shown that for a wide class of HDES the event rate is higher for DES at the lower levels of the hierarchy than at the higher levels of the hierarchy. Relationships between event rate and event aggregation are shown. We define a measure for event aggregation and show that there exists an inverse relationship between the amount of event aggregation and the event rate at any two successive levels in a class of HDES. Next, we study how to design the timing structure to ensure that there will be a decrease in the event rate (by some constant factor) between any two levels of a wide class of HDES. It is shown that if the communications between the various DES in the HDES satisfy a certain admissibility condition then there will be a decrease in the event rate. These results for HDES constitute the main results of this paper, since they provide the first mathematical characterization of the relationship between event aggregation and event rates of the HDES and show how to design the interconnections in a HDES to achieve event rate reduction. Several examples are provided to illustrate the results.The authors gratefully acknowledge the partial support of the Jet Propulsion Laboratory. Please address all correspondence to K. Passino (email: passino@eagle.eng.ohio-state.edu).     

Visualization of time-dependent data with feature tracking and event detection

This paper presents an innovative method to analyze and visualize time-dependent evolution of features. The analysis and visualization of time-dependent data are complicated because of the immense number of data involved. However, if the scientist's main interest is the evolution of certain features, it suffices to show the evolution of these features. The task of the visualization method is to extract the features from all frames, to determine the correspondences between features in successive frames, to detect significant events or stages in the evolution of the features, and, finally, to visualize the results. The method described here performs all these steps, and it is applied to a number of applications.     

The concept of residuals for fault localization in discrete event systems

In this paper an approach for fault localization in closed-loop Discrete Event Systems is proposed. The presented diagnosis method allows fault localization using a fault-free system model to describe the expected system behavior. Via a systematic comparison of the observed and the expected behavior, a fault can be detected and a set of fault candidates is determined. Inspired by residuals known from diagnosis in continuous systems, different set operations are introduced to generate the fault candidate set. After fault detection and a first fault localization, a procedure is given to render the fault localization more precisely by an analysis of the further observed system behavior. Special emphasis is given to the use of identified models for the fault-free system behavior. The approach is explained using a laboratory manufacturing facility.     

Fault detection and isolation in manufacturing systems with an identified discrete event model

In this article a generic method for fault detection and isolation (FDI) in manufacturing systems considered as discrete event systems (DES) is presented. The method uses an identified model of the closed-loop of plant and controller built on the basis of observed fault-free system behaviour. An identification algorithm known from literature is used to determine the fault detection model in form of a non-deterministic automaton. New results of how to parameterise this algorithm are reported. To assess the fault detection capability of an identified automaton, probabilistic measures are proposed. For fault isolation, the concept of residuals adapted for DES is used by defining appropriate set operations representing generic fault symptoms. The method is applied to a case study system.     

Fault detection for discrete event systems using Petri nets with unobservable transitions

In this paper we present a fault detection approach for discrete event systems using Petri nets. We assume that some of the transitions of the net are unobservable, including all those transitions that model faulty behaviors. Our diagnosis approach is based on the notions of basis marking and justification, that allow us to characterize the set of markings that are consistent with the actual observation, and the set of unobservable transitions whose firing enable it. This approach applies to all net systems whose unobservable subnet is acyclic. If the net system is also bounded the proposed approach may be significantly simplified by moving the most burdensome part of the procedure off-line, thanks to the construction of a graph, called the basis reachability graph.     

Discrete time-varying fault detection filter for non-uniformly sampled-data systems

This paper investigates the fault detection problem for non-uniformly sampled-data systems. No periodic assumption is made for the sampling instants. In contrast to most currently available resuks that are limited to strictly proper systems, measurement noises are considered. With the operators introduced to capture the inter-sampling behaviors of disturbances and faults, an omine fault detection algorithm is first derived to optimize the ratio-type design objective. It is then equivalently transformed into a recursive algorithm consisting of a discrete time-varying fault detection filter and the corresponding residual evaluation function. As repeated computation of the parity vectors is avoided, the proposed fault detection filter can help reduce the online computational burden with comparison to the existing parity relation based fault detection method.     

A system design for surveillance systems protecting critical infrastructures

Critical infrastructures are attractive targets for attacks by intruders with different hostile aims. Modern information and sensor technology provides abilities to detect such attacks. The objective of this work is to outline a system design for surveillance systems aimed at protection of critical infrastructures, with the focus on early threat detection at the perimeter of critical infrastructures. The outline of the system design is based on an assessment of stakeholder needs. The needs were identified from interviews with domain experts and system operators. The system design of the surveillance system and the user requirements in terms of capabilities were then determined. The result consists of the system design for surveillance systems, comprising the systems capabilities, the systems structure, and the systems process. The outcome of the work will have an impact on the implementation of the surveillance systems with respect to the sensors utilized, the sensor data algorithms and the fusion techniques.     

Algorithms for anomaly detection of traces in logs of process aware information systems

This paper discusses four algorithms for detecting anomalies in logs of process aware systems. One of the algorithms only marks as potential anomalies traces that are infrequent in the log. The other three algorithms: threshold, iterative and sampling are based on mining a process model from the log, or a subset of it. The algorithms were evaluated on a set of 1500 artificial logs, with different profiles on the number of anomalous traces and the number of times each anomalous traces was present in the log. The sampling algorithm proved to be the most effective solution. We also applied the algorithm to a real log, and compared the resulting detected anomalous traces with the ones detected by a different procedure that relies on manual choices.