Scholarlink

<p>Many methods have been proposed and verified for sediment classification using single-beam acoustic sounder. This paper proposed a new MLP-KNN (Multi-Layer Perceptron–K-Nearest Neighbor) model for sediment classification in shallow water. In shallow water, the accuracy of sediment classification is reduced because echoes are overlapping. Hence, we first design an MLP neural network that takes six feature parameters from the overlapping echo as input layer and the area of the overlapping part as output layer, and estimate the area of the overlapping part. Then, based on the estimation results, we calculated <i>E</i><sub>1</sub>,<i>E</i><sub>2</sub> and classified sediment using the KNN model as input <i>E</i><sub>1</sub>,<i>E</i><sub>2</sub>. To evaluate the performance of the proposed method, we used data collected from single-beam acoustic sounder with a frequency of 200kHz and a beam angle of 10° in shallow water of Taedong River and true ground samples in the corresponding water area. And a comparative analysis was carried out with the sediment classification method by BP neural network, the previous method. The test results showed that the proposed method classified sediment with higher accuracy in shallow waters.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Acoustic sediment classification using MLP-KNN model on single-beam echosounder data from shallow water

  • 期刊论文

  • 发表日期：

    2025年5月21日

查看全文
  • Gum-Chol Jong,
  • Jong-Chol Choe,
  • Un-Ryong Rim,
  • Un-Song Ri

摘要

Many methods have been proposed and verified for sediment classification using single-beam acoustic sounder. This paper proposed a new MLP-KNN (Multi-Layer Perceptron–K-Nearest Neighbor) model for sediment classification in shallow water. In shallow water, the accuracy of sediment classification is reduced because echoes are overlapping. Hence, we first design an MLP neural network that takes six feature parameters from the overlapping echo as input layer and the area of the overlapping part as output layer, and estimate the area of the overlapping part. Then, based on the estimation results, we calculated E1,E2 and classified sediment using the KNN model as input E1,E2. To evaluate the performance of the proposed method, we used data collected from single-beam acoustic sounder with a frequency of 200kHz and a beam angle of 10° in shallow water of Taedong River and true ground samples in the corresponding water area. And a comparative analysis was carried out with the sediment classification method by BP neural network, the previous method. The test results showed that the proposed method classified sediment with higher accuracy in shallow waters.