利用BP神经网络对红外目标进行识别之前,若不对原始样本数据进行预处理与特征提取,一方面使识别结果准确性降低,另一方面使BP神经网络的结构复杂化,采用主成分分析法可解决这些问题.主成分分析法能较好地提取表征样本的少数几个主分量,由该方法的特点可知,这几个主分量彼此不相关,非常符合特征优化的要求.研究结果表明,用该方法处理后的结果数据输入BP神经网络,提高了识别正确率,减少了训练时间,同时也简化了网络结构.将两种常见的模式识别方法结合用于红外目标识别:先由主成分分析法对原始样本数据进行精简处理,然后再由BP神经网络法进行分类识别,与传统的单一识别方法相比,准确度得到提高,计算量大为减少.

Before the infrared target is recognized by BP neural network,the recognition precision will be low and the structure of BP neural network will become complex if samples＇ data is not preprocessed and features are not extracted. In the paper, the principal componenet analysis is used to solve these problems.This method can extract main factors that explain the targets＇ sample and these factors are not correlative each other which can well satisfy the features optimization. The study result indicates that while the processed data is put into the neural networks,the precision of recognition is improved, the training time is reduced,and the structure of neural networks becomes simple.The innovation of the paper is two common methods are combined to recognize the infrared target. Firstly the principal component analysis is used to process the sample data,then the BP neural network is used to recognize the target.Compared with the traditional simple method,it improves the precision, furthermore reduces the calculation.