Crowd Violence Detection Using Global Motion-Compensated Lagrangian Features and Scale-Sensitive Video-Level Representation

Abstract

Lagrangian theory provides a rich set of tools for analyzing non-local, long-term motion information in computer vision applications. Based on this theory, we present a specialized Lagrangian technique for the automated detection of violent scenes in video footage. We present a novel feature using Lagrangian direction fields that is based on a spatio-temporal model and uses appearance, background motion compensation, and long-term motion information. To ensure appropriate spatial and temporal feature scales, we apply an extended bag-of-words procedure in a late-fusion manner as classification scheme on a per-video basis. We demonstrate that the temporal scale, captured by the Lagrangian integration time parameter, is crucial for violence detection and show how it correlates to the spatial scale of characteristic events in the scene. The proposed system is validated on multiple public benchmarks and non-public, real-world data from the London Metropolitan Police. Our experiments confirm that the inclusion of Lagrangian measures is a valuable cue for automated violence detection and increases the classification performance considerably compared to stateof-the-art methods.

Timed-image based deep learning for action recognition in video sequences

“…Technicolor [23] and Mediaeval [24], [25] on the basis of evaluation performed by a consortium of experts. • the proposal of feature extraction and learning for certain specific violence facts, for example the violence: induced by the crowd [26], [27], [28], [29] , the following Section IV-B proposes two experimental benchmarking frameworks: the first considers, in a joint framework, both VSD@Mediaeval and VSD@Technicolor. The second proposes upgrading the database obtained by fusing VSD@Mediaeval and VSD@Technicolor in order to derive a 3-level violence category database.…”

A Novel Low Processing Time System for Criminal Activities Detection Applied to Command and Control Citizen Security Centers

“…There are already several approaches to detect crimes and violence in video analysis, as shown by [8][9][10][11]. However, the Colombian National Police does not implement any method for the specific case of the detection of criminal events.…”

Video-based Bottleneck Detection utilizing Lagrangian Dynamics in Crowded Scenes

“…Instead of considering individual trajectories only, this information can be compactly represented within so called Lagrangian fields. Examples of Lagrangian fields that have been applied for video analysis are the arc length field [8] for segmentation or the direction field for violence detection [12] or action recognition [1]. One specifically popular type of Lagrangian fields are Finite-Time Lyapunov Exponents (FTLE) which quantify the amount of separation between neighboring path lines.…”