Human detection for a video surveillance applied in the ‘SmartMonitor’ system

Abstract

Human detection is one of the key and crucial tasks in video surveillance systems and is important for the purpose of object tracking, fall detection, human gait analysis or abnormal event detection. This paper concerns the application of two classifiers for human detection in the ‘SmartMonitor’ system — an intelligent security system based on image analysis. The classifiers are based on the Histogram of Oriented Gradients (HOG) descriptor and simple Haar-like features. The paper provides a brief description of the main system characteristics, discusses the problem of human detection and includes some results of the experiments performed using various parameters of HOG and Haar classifiers that were trained using benchmark databases and tested using appropriate video sequences. The paper aims at investigating the effectiveness and performance of both methods applied separately before incorporating them into the ‘SmartMonitor’ system’s video processing model.