S变换在水下目标分类识别中的应用

目标特征提取是目标识别的重要部分。介绍了一种较新的时频分析方法——S变换,对莱蒙湖底四类沉积物的反射回波进行S变换,并提出了提取变换后以频谱图的时间能量谱和奇异值为特征的特征提取方法,分析了四类回波的时间能量谱和奇异值特征的差异,并进一步用距离可分性测度检验了所提取的特征性能。最后利用最近邻分类器分类,仿真结果显示,该特征提取方法是一种有效的、稳定的特征提取方法,将在水下目标识别领域有更多的应用。

Application of S-transform to feature extraction and classification of underwater target

Target feature extraction is an important part of target recognition. A comparatively new method of time-frequency analysis-S transform (ST) is proposed as a method of feature extraction in this article. For four types of echoes from the bottom sediments in Geneva Lake, the time-power spectrum and singular value feature based on S transform are firstly extracted as target features, and the differences of the features extracted are analyzed. Then they are estimated by the distance classifiability criterion and classified by the nearest distance classifier. The result demonstrates that these algorithms are stable and efficient methods for extracting feature, and the S transform would be more useful in recognition of the underwater targets.