Combining multiple views and temporal associations for 3-D object recognition