基于人工神经网络的多路光学相关器相关峰的实时识别

REAL TIME RECOGNITION OF THE CORRELATION PEAKS OF MULTI CHANNEL OPTICAL CORRELATOR USING NEURAL NETWORK

提出一种应用人工神经网络根据多路光学相关器相关面输出图像识别目标的方法．运用统计规律确定相关峰的位置，不但把自相关峰作为目标特征，还将互相关峰也引入了识别，并通过神经网络实现相关峰复杂特征的抽取．该方法能有效地排除目标背景干扰和器件性能波动的影响。

A method using neural network to recognize targets according to the output of the multichannel optical correlator was proposed. The position of the correlation peaks was found through statistical rules. Not only the self correlation peaks but also the cross correlation peaks were used to be the features of the targets and the complex feature extraction was realized through neural network. The proposed method can effectively exclude the effects of the target background interference and the fluctuation of the device performance so as to realize a real time target recognition.