A Review of Machine Learning Approaches for Human Detection through Feature Based Classification

Abstract

Human detection has always been a task of sincere importance for many automation activities under computer vision. The problem concentrates on identifying regions of human presence in an image or frames of running video. The application areas may have the varied role of human detection objectives ranging from normal to serious and very critical/sensitive. Vision based attendance, traffic flow analysis, driver assistance, etc., are examples of applications with a normal role. Simultaneously, it plays a serious role in vision based theft identification system and a critical role in intruder detection in the border or sensitive places. This paper presents and explores various machine learning based human detection techniques. These techniques include feature learning based and deep learning based human detection paradigms. Through experiments, it has been found that Deep Neural Network based human detection techniques provide more efficient results than feature learning based techniques in terms of detection accuracy.