Skip to Main Content
Usually, moving object detection by exclusively background modelling is not enough to separate correctly background and foreground. On the other hand, strategies based on both background and foreground models are complex and computationally inefficient, and therefore need to be optimised for real-time performance. A novel and fast strategy for moving object detection by non-parametric background-foreground modelling is proposed. Whereas the background is modelled using only colour information, both colour and space are considered for the foreground modelling. Moreover, application of an efficient iterative multi-tracking algorithm allows the update of the spatial information, thus enabling an important reduction in computational requirements and improving the segmentation results.