Detection of Human Bodies from the Background in an Image Using Piecewise Linear-Support Vector Machine

Detection of the human bodies from the background in an image is challenged by the view and posture variation problem. In this paper, a piecewise linear support vector machine (PL-SVM) method is proposed for solving the identification problems of humans from an image, challenged by the view and posture variation. The aim is to use the piecewise discriminative function for constructuction of a non-linear classification boundary that can differentiate multi-view and multi-posture human bodies from the backgrounds in a high dimensional feature space. PL-SVM training is designed as an iterative procedure of feature space division and linear SVM training, aiming at the maximized marginality of local linear SVM's. Each section SVM model is responsible for subspace, corresponding to a human cluster of a special view or posture. In the PL-SVM, a cascaded detector is proposed with block orientation features and a histogram of oriented gradient features. This identification of humans in an image is implemented by using Matlab. Using this method, presence of humans can also be detected in videos also.