Sonar Noise Classifier

The Sonar Noise Classifier is a trained convolutional neural network that identifies noise in 3D point clouds from acoustic sensors. It was trained on many manually classified datasets covering different geographies and sensors, and it integrates with HIPS and CARIS Onboard workflows.

Key features

• Automatic identification of noise in 3D point clouds from acoustic sensors
• Data is rasterised into a high-resolution 3D voxel grid with zero origin before inference, which anonymises it
• Data is encrypted, and no data stays in the cloud after processing
• Results are mapped back to the individual points

Applications

Reducing manual cleaning time in bathymetric processing for survey companies, hydrographic offices and ocean-mapping programmes. CARIS links this AI work to its partnership with the UK Hydrographic Office and its collaboration with the Nippon Foundation–GEBCO Seabed 2030 Project, where a jointly developed AI tool removes noise from sonar data and reduces processing time from days to hours.