Abstract Synthetic Aperture Sonar (SAS) is a sensor that was designed for hydrographic survey of the seabed. It detects small targets, enabling high geometric resolution images. However, the images generated by SAS are susceptible to speckle noise, which makes digital processing difficult, since the noises are confused with the targets of interest. The goal of this study was to develop a semi-automatic routine for SAS image processing to verify the structural integrity of submarine pipelines. The method presented incorporates four stages: pre-processing to reduce noise and highlight the targets of interest; extraction of features aiming to recognize features related to the pipelines; post-processing to reduce the fragmentation generated during feature extraction; validation to quantify the results from the reference images to estimate the performance of the proposed methodology. The results showed that more than 80% of the submarine pipelines were mapped in the semi-automatic mode, which considerably reduced the time needed to manually identify a large number of pipelines operating on offshore oilfields.

中文翻译：

---

使用合成孔径声纳图像半自动识别海底管道

摘要 合成孔径声纳（SAS）是一种专为海底水文测量而设计的传感器。它检测小目标，实现高分辨率的几何图像。然而，SAS 生成的图像容易受到斑点噪声的影响，这使得数字处理变得困难，因为噪声与感兴趣的目标相混淆。本研究的目标是开发用于 SAS 图像处理的半自动程序，以验证海底管道的结构完整性。提出的方法包含四个阶段：预处理以减少噪声并突出感兴趣的目标；特征提取，旨在识别与管道相关的特征；后处理以减少特征提取过程中产生的碎片；验证以量化参考图像的结果，以估计所提出方法的性能。结果表明，80%以上的海底管道以半自动模式绘制，大大减少了人工识别海上油田大量管道所需的时间。