Human detection in public environment using GHOG: Gaussian of mixtures &Histogram of Oriented Gradients

In this paper, a novel method GHoG is proposed based on Gaussian Mixture Model and Histogram of Oriented Gradients for human detection in a public places. The human region and non-human region feature vectors are extracted using Histogram of Oriented Gradients (HoG) and the feature vectors are trained using SVM classifier to detect and classify the human and non-human region. Now, Gaussian of Mixture model (GMM) is applied to the human detected region. Thus the human alone is segmented from the image, as a result we get the human postures. However, the background or the non-human region is not considered for segmentation and time to model the background is reduced. In our algorithm, we handle illumination variation and dynamic background. Our proposed method provides 96% detection accuracy for our video set and 100% detection accuracy for INRIA dataset.