Abstract:ABSTRACTEvent detection plays an essential role in video content analysis and remains a challenging open problem. In particular, the study on detecting human-related video events in complex scenes with both a crowd of people and dynamic motion is still limited. In this paper, we investigate detecting video events that involve elementary human actions, e.g. making cellphone call, putting an object down, and pointing to something, in complex scenes using a novel spatio-temporal descriptor based approach. A new spatio-temporal descriptor, which temporally integrates the statistics of a set of response maps of low-level features, e.g. image gradients and optical flows, in a space-time cube, is proposed to capture the characteristics of actions in terms of their appearance and motion patterns. Based on this kind of descriptors, the bag-of-words method is utilized to describe a human figure as a concise feature vector. Then, these features are employed to train SVM classifiers at multiple spatial pyramid levels to distinguish different actions. Finally, a Gaussian kernel based temporal filtering is conducted to segment the sequences of events from a video stream taking account of the temporal consistency of actions. The proposed approach is capable of tolerating spatial layout variations and local deformations of human actions due to diverse view angles and rough human figure alignment in complex scenes. Extensive experiments on the 50-hour video dataset of TRECVid 2008 event detection task demonstrate that our approach outperforms the well-known SIFT descriptor based methods and effectively detects video events in challenging real-world conditions.

Group Event Detection with a Varying Number of Group Members for Video Surveillance

Group Event Detection for Video Surveillance

Intelligent Video Surveillance for Checking Attendance of Traffic Controllers in Level Crossing

Detecting Group Activities with Multi-Camera Context

Human activity recognition for video surveillance

Detecting Abnormal Activities in Video Surveillance with Multi-Models

A new package-group-transmission-based algorithm for human activity recognition in videos.

A New Network-Based Algorithm for Human Group Activity Recognition in Videos.

Automated Classification of Two Persons' Interactive Activities.

Deep Structured Models For Group Activity Recognition

Activity Recognition Using A Combination of Category Components And Local Models for Video Surveillance.

Online Recognition of Group Actions in Intelligent Meeting Scenario

Multi-object Events Recognition from Video Sequences Using Extended Finite State Machine

Recognizing Human Group Activities with Localized Causalities

Detecting video events based on action recognition in complex scenes using spatio-temporal descriptor.

A Network-Based Algorithm for Recognizing Group Activities with Multiple People and Crowded Scenes

Associating Groups of People

Unsupervised group-based crowd dynamic behavior detection and tracking in online video sequences

Robust Event Detection Scheme for Complex Scenes in Video Surveillance

Multi-target Tracking Identification System under Multi-Camera Surveillance System

Small Group People Behavior Analysis Based on Temporal Recursive Trajectory Identification