Abstract:
Human detection has enormous application area in
autonomous video surveillance and human-computer interaction.
Detecting suspicious event has become a very crucial issue in the
current circumstance of our society. As a pioneer, a framework is
proposed for detecting human in the sterile zone in this paper.
Since, in the case of the sterile zone, we have to deal with lowresolution
video, that’s why initially input video frames are
enhanced by using local histogram equalization. Then a
background model is created by using the Gaussian Mixture
Model (GMM) where each pixel is represented by a mixture of a
number of Gaussian based on probabilistic method. This modeled
background is then compared with a new frame to detect the
foreground object. After that, the morphological operation is
performed to remove discontinuities and to get the region of
interest (ROI). Then shape and texture features from ROI are
extracted for classification. Finally, combined features from
Histogram of Oriented Gradient (HOG) and Local Binary Pattern
(LBP) are fed into SVM classifier to detect human. In this paper
to achieve better performance in the sterile zone, human shape is
analyzed with HOG along with enumerating local features by
LBP. Moreover, this proposed framework is tested using various
video in different conditions and the outcome demonstrates
remarkable efficiency comparative to other alternatives.
