Underwater SAS Imagery Analysis

Examples of possibilistic context identification maps. The first row shows the original underwater SAS imagery and the second row highlights our identification (in red) for hard-packed sand, sand ripple, and shadow contexts.

Synthetic Aperture Sonar (SAS) imaging systems produce high-resolution seabed imagery useful for underwater target detection and scene understanding. One of the challenges for underwater context identification is that the boundaries between seabed contexts (e.g., sand ripple, sea grass, hard-packed sand) are usually gradual with wide regions of transition. We developed a Multiple Instance Learning (MIL) based approach that can produce possibilistic seabed context identification maps.

In addition, the characteristics of target objects in SAS imagery vary across different seabed contexts. We developed an end-to-end environmentally-adaptive target recognition system for SAS imagery that performs target recognition while accounting for environmental contexts.

Illustration of our anomaly detection window.

Example of our detected anomalies on the sea floor (marked by red).

Associated Publications

Environmentally-Adaptive Target Recognition for SAS Imagery, SPIE DCS Conference, 2017.
Multiple-instance Learning-based Sonar Image Classification, SPIE DCS Conference, 2017.
Possibilistic context identification for SAS imagery, SPIE DCS Conference, 2015.