“…During the past two decades, many methods [ 1 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 ] were proposed for the background initialization task. In general, these techniques can be classified into four main categories: pixel-based methods [ 1 , 7 , 8 , 9 , 10 , 11 ], iterative-based methods [ 12 , 13 , 14 , 15 ], low-rank/sparse data separation methods [ 16 , 17 , 18 , 19 , 20 , 21 , 22 ], and deep learning-based methods [ 23 , 24 , 25 , 26 , 27 ].…”