水下目标特征提取方法研究

Extraction of features of underwater target

有效的特征提取技术是水雷目标识别的基础。文章采用了两种前期研究中较为有效的水雷目标特征提取方法(频域离散小波变换法和常数Q滤波子带能量法),并引入了一种应用在水下目标识别领域中的特征提取方法(波形结构法)。应用此三种特征提取方法提取的特征来识别实雷目标以及假目标,分类器采用三层BP算法的前向神经网络,给出了具体的识别率,说明该特征提取算法是有效的,用波形结构法进行水雷目标的特征提取是可行的。

Effective feature extraction is fundamental in the underwater mine extraction techniques effective in the mine recognition are discussed, namely, recognition. Two feature frequency discrete wave transform （FDWT） and Q-filter. A feature extraction technique, wave structure （WS）, is introduced for underwater target recognition. The achieved recognition ratio with 3-layer BP network indicates the effectiveness of the applied techniques and feasibility of the WS technique used in feature extraction of mines.