摘要
该文针对毫米波/红外传感器融合目标识别问题,提出一种新的用于决策层目标识别的神经网络融合算法.该网络结构新颖,网络训练时修改的是门限而不是连接权值.融合后的识别率可比毫米波和红外子源提高9.7％到11.3％,因此,该算法是有效可行的.
In the light of the target recognition based on MMW/IR fusion, a new neural network algorithm for the decision fusion is presented in this paper. The architecture of this network is novel. It is the thresholds, not the conjunction weights, that are modified, when the network is being trained. The mean correct recognition rate after fusion is higher than that of MMW and IR subsources by 9.7% and 11.3% respectively, which indicates this algorithm is effective and feasible.
关键词
神经网络
决策层融合
目标识别
Neural network, Decision fusion, Target recognition
