A camouflage target detection method based on local minimum difference constraints

To address the problems of missing inside and incomplete edge contours in camouflaged target detection results, we propose a moving algorithm based on local minimum difference constraints (LMDC). The first uses mean to optimize initial background model, removes stable region by global comparison, extracts point set potential so that each boundary (seed) grows along center target. Finally, define minor term, combine seed path space consistency, calculate attributes pixel area realize detection. this paper is verified public data sofa video test videos compared with five classic algorithms. experimental results show proposed yields good integrity, accuracy, number objective evaluation indexes, its overall performance better than