Goal-based trajectory analysis for unusual behaviour detection in intelligent surveillance

Abstract

Video surveillance systems are playing an increasing role in preventing and investigating crime, protecting public safety, and safeguarding national security. In a typical surveillance installation, a human operator has to constantly monitor a large array of video feeds for suspicious behaviour. As the number of cameras increases, information overload makes manual surveillance increasingly difficult, adding to other confounding factors like human fatigue and boredom.The objective of an intelligent vision-based surveillance system is to automate the monitoring and event detection components of surveillance, alerting the operator only when unusual behaviour or other events of interest are detected. While most traditional methods for trajectory-based unusual behaviour detection rely on low-level trajectory features, this thesis improves a recently introduced approach that makes use of higher-level features of intentionality. Individuals in a scene are modelled as intentional agents instead of simply objects. Unusual behaviour detection then becomes a task of determining whether an agent's trajectory is explicable in terms of learned spatial goals. The proposed method extends the original goal-based approach in three ways: first, the spatial scene structure is learned in a training phase; second, a region transition model is learned to describe normal movement patterns between spatial regions; and third, classification of trajectories in progress is performed in a probabilistic framework using particle filtering. Experimental validation on three published third-party datasets demonstrates the validity of the proposed approach.iii

Exploring Behavior Analysis in Video Surveillance Applications

“…Different activity representations can be grouped into three categories as shown in Fig. 1: object based representations [9,14], pixel based representations [15-17, 20, 21], and other feature representations [22,24]. In the following subsections, we review the work presented in each category.…”

“…These features include trajectory or blob-level descriptors such as bounding box and shape. A trajectory-based feature is prevalently utilized to represent the motion history of an object in a scene [14]. Usually, a trajectory is formed by associating a set of attributes of detected object, such as appearance features and velocity over successive frames using motion tracking algorithms (Fig.…”

Systematic Review and Classification on Video Surveillance Systems

“…Video detecting category addresses issues regarding to automatic detection of anomalous, forbidden, dangerous events or abandoned object (counting moving people, ship detection, after-the-fact event, intruder detection, trajectory-based unusual behavior detection, motion detection, mult iple moving object detection, face detection, pedestrian detection, vehicle detection, unattended object detection, etc) . Video encoding [91], [92], [93], [94], [95], [96], [97], [98], [99], [100], [101], [102], [103], [16], [104] [105], [106], [107], [108], [109], [110], [7], [111] [112], [113], [114], [115] [116],[8], [117] Object detection is performed by co mmon statistical learning techniques with dynamically learning background model of the scene and applies the reference model to find out which section of the scene match with mov ing object. Reasoning refers to generating new explanations, facts and knowledge of dynamic scenes by applying inference engine and method (rule and case based reasoning, Bayesian network, decision tree).…”

Comparison of genetic algorithm components and selection variants in unlawful behavior detection of hand movement

“…The purpose of a VSS is to observe and detect any unlawful behavior occurred in an authorized venue or any area of interest. It helps to prevent and investigating crime and also to ensure individual's privacy, safety and security [1][2].…”

“…The manual VSS detection is no longer practical since the possibility of crime acts to be foreseen and misinterpreted is very high due to size and load of the images that need to be scan through. The number of Video Surveillance Data (VSD) has increased exponentially which made it increasingly difficult for human to observe all channels continuously, yet it is money and time consuming [2][3][4][5].…”