Background Modeling Through Spatiotemporal Edge Feature and Color

In this paper, we propose a new spatiotemporal edge feature for background modeling that can extract spatial and temporal (motion) features by considering the background model and current information. Previous work on background modeling considers mainly the spatial domain, which misses key temporal information. In our proposal, we create spatiotemporal edge features by using current and past background information by identifying the amount of change from past to present. By finding these differences, we can accurately detect the movement of objects that is more robust to noise and illumination variations. Moreover, our proposed background-modeling technique adapts to background changes that occur over time through a dynamic model update strategy. Additionally, we are enhancing the spatiotemporal edge features with color to maintain the characteristics of each other. Finally, we evaluated our proposed method on the publicly available CDNET 2012 dataset and compared with state-of-the-art methods. Our extensive evaluation and analysis show that our method outperforms previous methods on this dataset.