Multi-feature fusion based GMM for moving object and shadow detection

Shadow detection and removal plays an important role in segregating an object from background accurately. Traditional object and shadow detection algorithm based on single feature such as color is easily restricted by the scene and illumination changes. In this paper, a multi-feature fusion based Gaussian mixture background modeling method is proposed to lower the false detection rate using single feature by integrating color and texture. And then a double shadow judgment method is proposed to determine the suspected shadow and the true shadow. Firstly, the shadow is determined by the color angle of shadows, then, the shadow is detected according to the brightness between the shadow region and the background. Finally, the two results of shadow detection are combined, which offers a double guarantee for the accurate removal of shadow.