Abstract
Multi-layer decision fusion (MLDF) is an important method in smart surveillance system for a robust decision-making about objects behavior. Grey System Theory has recently attracted substantial interest in wide research areas. Grey Mixed Center Point Triangular Whitening Functions} (GMCPTWF) approach for decision fusion is presented in this paper to perform multi-layer decision fusion. After features are extracted from trajectory of the object of interest, three reference behavior vectors are constructed Three behavior categories under consideration in this paper are: abnormal, unknown, and normal. These vectors are utilized to train the system, and are used as an input for the decision-making subsystem. Finally, multi-layer decision fusion is performed GMCPTWF of three classes are tested by eighteen trajectories. The experiment on eighteen trajectories which are classified into three categories proves that this method can effectively improve the decision making performance.