基于多维度声学特征优选的多波束海底底质分类

Multi-beam benthic environment classification based on multi-dimensional acoustic feature optimization

中文关键词:  海底底质分类  多波束测深系统  特征优选  反向散射强度  海底地形

英文关键词:Seabed sediment classification  Multibeam echo-sounding system  Feature optimization  Backscatter intensity  Seabed terrain.

中文摘要:

      海底表层底质分布信息的准确获取在构建海洋基础地理数据库中发挥着重要作用。目前，多波束是实现大范围海底底质分类的有效手段之一，基于多波束测深和反向散射强度数据所派生的声学特征被广泛应用于底质分类建模。然而，随着特征维度的增加，特征空间中存在的无关和冗余特征严重影响底质分类精度。为了定量评估声学特征对底质类别的表征能力，并消除无效特征对分类结果的干扰，本文提出了基于多维度声学特征优选的海底底质分类方法。首先，结合实际底质样本的物理属性对多维特征进行排序和优选，排除冗余和无关特征。其次，分别应用支持向量机、随机森林和深度信念网络构建海底底质监督分类模型。通过利用爱尔兰海南部多波束调查数据和实地取样信息进行试验，结果表明提出方法对海底底质的总体分类精度和Kappa系数分别最高达到了86.20%和0.834，相较于主成分分析和熵指标特征选择方法有明显提高，突出了该方法在海底底质探测及制图的应用潜力。

英文摘要:

      The accurate acquisition of seabed surface sediment distribution information plays an important role in the construction of marine basic geographical database. At present, multi beam is one of the effective means to achieve a wide range of seabed sediment classification. Acoustic features derived from multi-beam bathymetry and backscatter intensity data are widely used in sediment classification modeling. However, as the feature dimension increases, the presence of irrelevant and redundant features in the feature space seriously affects the accuracy of sediment classification. In order to quantitatively evaluate the representation ability of acoustic features on sediment categories and eliminate the interference of invalid features on classification results, this paper proposes a seabed sediment classification method based on multi-dimensional acoustic feature optimization. Firstly, based on the physical properties of actual sediment samples, multidimensional features are sorted and optimized to eliminate redundant and irrelevant features. Secondly, support vector machine, random forest and depth belief network are respectively applied to construct the seabed sediment supervision classification model. Through experiments using multi beam survey data and field sampling information from the southern part of Ireland, the results showed that the proposed method achieved the highest overall classification accuracy and Kappa coefficient of 86.20% and 0.834% for seabed sediment, respectively. Compared with principal component analysis and entropy index feature selection methods, it has significantly improved, highlighting the potential application of this method in seabed sediment detection and mapping.