Abstract: Accurately detecting humans under occlusion in images plays a critically important role in many computer vision applications. Extraction of effective features is the key to this task. Promising features should be discriminative, robust to various variations and easy to compute. In this paper, we present center-symmetric local binary patterns (CS-LBP) and Histogram of oriented gradients (HOG), for human detection with occlusion. HOG feature calculates the gradient magnitude and the gradient direction of the local image. The main drawback of HOG feature extraction is that, it does not work well with different texture and pose of human and accuracy of detection is less. Rather than using LBP for feature extraction, we use CS-LBP because LBP produces long histogram and are not too robust on flat image areas. The CS-LBP feature captures both gradient information and texture information and works well on flat image areas. So we combine HOG and CS-LBP method for better human detection with occlusion. Experiments on the INRIA pedestrian dataset show that the combination of CS-LBP feature and HOG feature with linear support vector machines (SVMs) gives better result for human detection under occlusion.
Keywords: Center Symmetric - Local Binary Pattern (CS-LBP), Histogram of Oriented Gradients (HOG), INRIA pedestrian dataset, Feature Vector, Occlusion, Support Vector Machine (SVM).
Title: HOG and CS-LBP Methods for Human Detection with Occlusion
Author: S.K UMA, SRUJANA B J, Dr. B R RAMACHANDRA
International Journal of Computer Science and Information Technology Research
ISSN 2348-1196 (print), ISSN 2348-120X (online)
Research Publish Journals
