Detection of abandoned objects in real time

In this paper we describe an intuitive model for accurate and efficient detection of abandoned objects. The system is built on the backbone of the Gaussian Mixture Model for background subtraction. We apply a simple and robust method for shadow detection. Next, detection of stable blobs is carried out using Mathew et al's method, and an important modification is suggested that is resistant to temporary occlusions, and removes unnecessary parameters from the model. Changes to the background itself are identified via a `ghost' removal procedure that can distinguish between true and removed objects. Results of testing the model are presented with conclusions.