Spatio-temporal multimodal mean

Existing background modeling techniques that model objects that have become stationary will incorrectly detect a new object if an existing stationary object is moved. A novel spatio-temporal reasoning mechanism is presented based on multi-layer background modeling and appearance model of the displaced object to conserve state of the objects in a scene. The mechanism models layers of the foreground objects that have become stationary, along with moving foreground and background. An object that changes its place, partially or fully, is recognized based on its appearance model. The end result is the correct modeling of objects even in the case of spatial displacements. This provides a richer mechanism of analyzing and visualizing spatio-temporal scene events than the traditional binary foreground/background segmentation.