基于XGBoost算法的多波束声呐数据海缆自动识别

海上风电场海底电缆的全面探测，对于海上风电场后续的生产和维护至关重要。现阶段的海底电缆识别需要检测人员根据多波束数据的影像特征手动判断，效率低且主观性强。为此，本文基于极端梯度提升（eXtreme Gradient Boosting，XGBoost）算法，提出了一套适用于多波束点云数据的海缆自动识别方法。首先计算多波束点云数据的几何特征；之后建立XGBoost分类模型，将几何特征为参数输入模型进行训练；最后将所需探测工区的多波束声呐数据输入训练好的模型，从而实现海缆的自动识别。将本文方法应用于某海上风电场实际工区的海缆探测之中，结果表明：XGBoost 算法对于多波束数据裸露海缆识别具有较高的准确性，可以极大地节约人工检测成本。

英文摘要:

Comprehensive detection of underwater cables is crucial for the production and operation of offshore wind farms. Currently, manually extracting features from multibeam data is the most commonly used method, but it is inefficient and highly subjective. Therefore, this paper proposes an automatic identification method for cables in multibeam point cloud data based on the XGBoost algorithm. First, the geometric features of multibeam point cloud data are calculated. Then, XGBoost classification model is introduced, and the geometric features are used as input parameters for training. Finally, the multibeam sonar data from the target detection area are fed into the trained model to achieve the automatic identification of subsea cables. We apply this method to detect cables in a real offshore wind farm survey. The results show that the XGBoost algorithm achieves high accuracy in identifying exposed underwater cables and significantly reduces labor costs.

查看全文查看/发表评论下载PDF阅读器