Superpixel Segmentation and Texture Feature Learning forMulti-Aspect Underwater Scene Understanding
Abstract
Using high-resolution side-looking sensors such as synthetic aperture sonar (SAS), it is possible to image large areas of seabed with fine detail and use this imagery to map bottom sediments and seabed composition. Being able to perform automated seabed understanding that can characterize seabed and bottom sediment enables one to predict mine-hunting effectiveness in shallow water and very shallow water sea lanes and leverage this environmental information during mine-hunting (e.g., within an environmentally adaptive detection system). The overall goal of this research is to further develop an autonomous underwater scene understanding system that analyzes collections of multi-aspect side-look synthetic aperture sonar (SAS) imagery generated during a sonar survey. This goal will be achieved through investigation of the following research objectives:1. Develop an automated SAS image alignment algorithm2. Develop multi-scale, multi-image superpixel segmentation algorithms applicable to SAS imagery3. Develop automated texture feature extraction capabilities through the investigation and development of a histogram-based feature learning layer for deep learning networks
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 08, 2020
- Source ID
- N000142012386
Entities
People
- Alina Glenn
Organizations
- Office of Naval Research
- United States Navy
- University of Florida