Motion estimation is critical for the localization of autonomous underwater vehicles. Current SONAR-based techniques exclusively utilize either Doppler or spatial measurements. However, these measurement domains are complementary to each other; Doppler measurements directly measure radial motion, whereas spatial measurements uniquely observe angular motion. Therefore, this article presents SONARODO (SONAR Odometry), a novel real-time motion estimation algorithm for 2-D forward-looking SONARs. It depends on a largely decoupled motion estimation process that better utilizes each measurement domain for their respective strengths. Specifically, it estimates translational motion from Doppler-azimuth images and rotational motion from range-azimuth images. While this method does require a SONAR that can provide both image types, it was designed to ensure robustness to featureless seafloor environments and low-resolution images. This article's validation with high-fidelity simulation data demonstrated that SONARODO offers accuracy and computational cost advantages over related motion estimation techniques.

中文翻译：

---

基于SONAR图像中多普勒和空间数据的运动估计

运动估计对于自主水下航行器的定位至关重要。当前基于SONAR的技术仅利用多普勒或空间测量。但是，这些测量域彼此互补。多普勒测量直接测量径向运动，而空间测量唯一地观察角运动。因此，本文介绍了SONARODO（SONAR里程表），这是一种用于二维前瞻性SONAR的新型实时运动估计算法。这取决于很大程度上分离的运动估计过程，该过程可以更好地利用每个测量域的各自优势。具体而言，它从多普勒方位角图像估计平移运动，并从距离方位角图像估计旋转运动。尽管此方法确实需要可以提供两种图像类型的SONAR，它旨在确保对无特征的海底环境和低分辨率图像的鲁棒性。本文对高逼真度仿真数据的验证表明，SONARODO与相关的运动估计技术相比，具有准确性和计算成本优势。