Moving object detection using vector image model

Abstract Motion detection is a challenging problem in video sequel with complex background. Different background subtraction techniques are introduced to detect moving objects from video sequel, but none of these has been adequate to address the complex nature of the dynamic scene in real-surveillance task. So a simple and efficient vector-based method is proposed to address real-surveillance challenges. In this paper, the concept of linear dependence of vectors is used to build background models corresponding to each pixel. Consecutively, linear independence is used to detect moving object from incoming video sequel. The proposed method is spatio-temporal based background subtraction method, as it uses spatial and temporal information for constructing background model. The background models are updated using an adaptive update rate with region diffusion to produce efficient results in the dynamic background. The benchmark datasets taken from complex scenario are experimented using the proposed algorithm and the generated results are compared with recent background subtraction methods. The presented method attains better performance in terms of qualitative and quantitative results.