Motion detection based on background model

Motion object detection is a very important part of Video Surveillance System. In this paper we provide a new modeling approach based on color, neighborhood feature to solve the shortage of the traditional Gaussian mixture model (GMM) on motion detection. Firstly, we build the initial GMM based on pixel color information and obtain initial parameters and initial detection result. Then we extract the gradient and modified local binary pattern (lbp) feature and build new GMM to achieve accurate detection result. Moreover, we use a unique voting strategy during the foreground pixel detection. We testify out approach in indoor and outdoor environment, the implement results show that our approach can remove the influence of illumination and camera shaking.