关键词: 目标检测, 背景模型, 阈值模型, 自组织视网膜映射图, 神经网络, 动态背景

Abstract: For reducing the impact on the performance of moving object detection by dynamic background,a moving object detection method is proposed.This method is fused with Self-organizing Retinotopic Maps(SORM) background model and Discrete Time Cellular Neural Network(DTCNN) thresholding model,builds background model according to SORM,ameliorates DTCNN in terms of requirement of moving object detection,constructs thresholding model to remove noise during the process of moving object detection,and updates the parameters of SORM and DTCNN models to adapt scene change.Experimental results show that this method has better performance of object detection,especially has stronger adaptation capacity for dynamic background,compared with traditional methods such as Gaussian mixture model,self-organizing background subtraction and growing self-organizing maps.

Key words: object detection, background model, threshold model, Self-organizing Retinotopic Maps(SORM), neural network, dynamic background