Home // VISUAL 2018, The Third International Conference on Applications and Systems of Visual Paradigms // View article

K-Nearest Neighbours Based Classifiers for Moving Object Trajectories Reconstruction

Authors:
Muhammad Afzal
Karim Ouazzane
Vassil Vassilev

Keywords: Video surveillance; Real-time video analytics; Model-driven motion description; Moving objects tracking; Trajectory reconstruction; Incremental algorithms; Machine learning

Abstract:
This article presents an exemplary prototype implementation of an Application Programming Interface (API) for incremental reconstruction of the trajectories of moving objects captured by Closed-Circuit Television (CCTV) and High-Definition Television (HDTV) cameras based on K-Nearest Neighbor (KNN) classifiers. This paper proposes a model-driven approach for trajectory reconstruction based on machine learning algorithms which is more efficient than other approaches for dynamic tracking, such as RGB-D (Red, Green and Red Color model with Depth) images or scale or rotation approaches. The existing approaches typically need a low-level information from the input video stream but the environment factors (indoor light, outdoor light) affect the results. The use of a predefined model allows to avoid this since the data is naturally filtered. Experiments on different input video streams demonstrate that the proposed approach is efficient for solving the tracking of moving objects in input streams in real time because it needs less granular information from the input stream. The research reported here is part of a research program of the Cyber Security Research Centre of London Metropolitan University for real-time video analytics with applicability to surveillance in security, disaster recovery and safety management, and customer insight.

Pages: 7 to 12

Copyright: Copyright (c) IARIA, 2018

Publication date: June 24, 2018

Published in: conference

ISSN: 2519-8645

ISBN: 978-1-61208-647-7

Location: Venice, Italy

Dates: from June 24, 2018 to June 28, 2018