一种被动声纳目标识别数据融合方法的研究

文章首先对目标噪声信号采用五种不同的方法提取特征矢量,然后采用基于自适应遗传BP算法的神经网络分别对五种特征矢量并发地进行分类,再采用遗传算法对分类器组合过程中的多参数进行优化,最后由五种分类结果最优组合产生最终的分类结果。实验结果表明该系统具有很好的分类效果。

Research on Technique of Passive Sonar Target Recognition Based on data fusion

Feature extraction of targets radiated-noise and design of targets classifier are key techniques of passive sonar target recognition system. ln this paper, five feature extraction methods are applied to extract feature of targets radiated-noise at first. And then a novel method of a combinational classification system of multiple neural network target classifiers is proposed for passive sonar target recognition. Each neural network target classifier based on adaptive genetic back-propagation algorithm is used to classify for five kinds of feature vector separately. Then, the multiple parameters in the classification combination are optimized by using genetic-backpropagation algorithm.At last,the best combination of five classification results is the final result. The classification experiment for three different classes of targets is done, results of experiment show that the passive sonar target recognition system designed in the paper has higher correct classification rate.