Multiple human detection in images based on differential evolution and HOG-LBP

Abstract

In this paper a method for multiple human detection in the image has been presented. This method uses differential evolution (DE) algorithm to improve window position detection speed and HOG-LBP algorithm for feature extraction. Fitness function for DE algorithm is SVM and in the final state, a postprocessing on detected windows by DE algorithm is performed. This method has been tested on INRIA datasets and its precision for detecting humans in the image is 92% which is better than state of the art methods.