基于GA-KPCA的特征选择在水下目标识别中的应用

Application of Feature Selection Based on GA-KPCA in Underwater Target Recognition

水下辐射声场和水声信道的复杂性使得声呐接收的噪声信号相互耦合、调制畸变,具有很强的非线性。文中利用核函数将原始特征空间的非线性数据映射至高维特征空间,在高维特征空间进行主元分析(PCA)法提取特征,并采用遗传算法(GA)对核参数进行优化,形成了基于GA-核主元分析(KPCA)的水下目标特征选择方法。实际样本数据验证结果表明,该方法在一定程度上弥补了传统线性PCA方法处理非线性数据的不足,能够获得更高的识别正确率。

The complexity of underwater radiated acoustic field and underwater acoustic channel results in intercoupling of the noise signals received by sonar,modulation distortion,and strong nonlinearity.In this paper,the kernel function is used to map the nonlinear data of the original feature space to the high-dimensional feature space;the principal compo-nents analysis(PCA)method is used to extract the features from the high-dimensional feature space,and the genetic algorithm(GA)is used to optimize the kernel parameters,thus an underwater target feature selection method based on GA-kernel principal components analysis(KPCA)is established.Actual sample data validation shows that,to a certain extent,this method compensates the insufficiency of the traditional linear PCA method in dealing with nonlinear data,and it has higher recognition accuracy.