Anomaly Detection in Multibeam Bathymetric Point Clouds Integrating Prior Constraints With Geostatistical Prediction

Abstract

Multibeam echo sounding sonar, as an active acoustic detection technology, is the mainstream equipment for current seabed topography surveying. The accuracy of the seabed topographic model is significantly influenced by anomalies in survey. Addressing the limitations in stability and comprehensiveness in anomaly processing of existing detection algorithms, this article proposes a novel anomaly detection algorithm based on prior constraints. This algorithm integrates the distribution characteristics of various types of anomalies and classifies them meticulously. A robust estimation of the detection model is achieved through geostatistical methods to enhance the accuracy of seabed topography representation. Experimental results demonstrate that our method excels in error control. Tests on one simulated dataset and two field-collected datasets showed type I errors between 1.6% and 3%, type II errors between 5.3% and 17%, and total error rates between 1.9% and 3.1%. Compared to existing classic algorithms, our method not only preserves the terrain features but also achieves comprehensive detection of different types of anomalies.

Keywords

• Anomaly detection
• geostatistical prediction
• multibeam bathymetry
• point cloud
• prior constraints