Occlusion detection in visual scene using histogram of oriented gradients

Object detection is an important step in any video analysis. In this paper, we propose a novel framework for blob based occluded object detection. It detects and tracks the occluded objects in video sequences captured by a fixed camera in crowded scene with occlusion. Moreover the occlusion of an abandoned object is a critical aspect in the video surveillance. We present the system used to identify the abandoned object highlighting how the system can recognize a problem of occlusion and detect the object when it is visible again. Initially Pedestrians are detected using the pedestrian detector by computing the Histogram of Oriented Gradients descriptors (HOG), using a linear Support Vector Machine (SVM) as the classifier. In our system, the background subtraction is modeled by a Mixture of Gaussians technique (MOG). Several experiments were conducted to demonstrate the proposed method using huge video dataset show the robustness and effectiveness.