Online Pedestrian Group Walking Event Detection Using Spectral Analysis of Motion Similarity Graph

TL;DR

This paper introduces an online spectral clustering method to detect pedestrian groups in video sequences, effectively identifying group events in real-time for surveillance applications.

Contribution

It presents a novel spectral analysis approach for real-time group detection in videos, specifically tailored for online processing and PETS2015 challenge requirements.

Findings

Effective group detection on PETS2015 dataset

Real-time identification of moving pedestrian groups

Improved accuracy over previous methods

Abstract

A method for online identification of group of moving objects in the video is proposed in this paper. This method at each frame identifies group of tracked objects with similar local instantaneous motion pattern using spectral clustering on motion similarity graph. Then, the output of the algorithm is used to detect the event of more than two object moving together as required by PETS2015 challenge. The performance of the algorithm is evaluated on the PETS2015 dataset.