Gaussian elliptical fitting based skin color modeling for human detection

Abstract

Human detection has now become an important and integral part of the maximum of computer vision applications. Surveillance is one such significant application. Surveillance with automatic human detection and event understanding is a faster and smart solution for today's surveillance. There have been a number of techniques which can be utilized for achieving human detection in an image or video. Each of these uses different types of features to exploit different characteristics of human being. So is the case that every of these techniques has some or other pitfall. Seeing all those, a novel technique based on skin color modeling is proposed in this paper for human detection in real-time surveillance videos. The proposed approach outperforms the other popular approaches by a significant rate on common measures of True Positive Rate, False Positive Rate and Accuracy.