Skip to Navigation Skip to Page Content

Getting to the bottom of it…

Photo: Alina Zare with Chao Chen looking at a computer screen

Electrical and computer engineering Assistant Professor Alina Zare, at right, reviews seabed images with doctoral student Chao Chen. Their imaging research is aimed at more clearly recognizing what lies at the bottom of the ocean. Photo by Hannah Sturtecky

Currently, mapping the ocean floor is a labor intensive, complex process and bathymetry — the measurement of oceanic topography — can be an inexact science, creating maps with non-existent rigid segmentation between seabed features.

Technology developed by a University of Missouri researcher, in collaboration with the Office of Naval Research (ONR), offers a better solution for mapping of the ocean floor and could also improve current underwater object detection performance.

Alina Zare, assistant professor in the MU Electrical and Computer Engineering Department, currently is in the first year of a three-year grant funded by ONR on the subject of multi-aspect underwater scene understanding. Her lab is developing algorithms that will allow researchers to get a clearer look at not only the bathymetry of the sea floor, but also more clearly delineate different materials that make up the floor as well as objects located at the bottom of the ocean.

“Where is there sea grass; where is there a sand ripple; where’s the hard-packed sand?” Zare said.

Image: Example results for the researchers’ autonomous Possibilistic Sea-bed Segmentation algorithm.

Above are example results for the researchers’ autonomous Possibilistic Sea-bed Segmentation algorithm. Regions that are likely shadow are highlighted in red and those that are not shadow are blue.

“We want to know where sand ripples are, what orientation does the sand ripple have, and how high are the ripple peaks and troughs and all these details. If there are any objects there, what are their characteristics? Where are their positions; how are they oriented?”

Chao Chen, a doctoral student working in Zare’s lab, said that eventually, they hope to have an algorithm that can detect these objects as well as differentiate between the different materials that make up the seafloor — a process called seabed segmentation that will separate and delineate the different materials that make up the seabed.

She also said the team is working on target detection.

“There might be some plant life or rocks on the seabed; this can target them. We can analyze the image and try and find and localize these targets,” Chao said.

Image: Example results for the researchers’ autonomous Possibilistic Sea-bed Segmentation algorithm.

Above are example results for the researchers’ autonomous Possibilistic Sea-bed Segmentation algorithm. Regions that are likely sand ripple are highlighted in dark red.

Currently, the Naval Surface Warfare Center [NSWC] in Panama City, Fla., provides Zare, her team and Tory Cobb, senior scientist in the Advanced Signal Processing and Automatic Target Recognition Branch at the NSWC, image data generated by its synthetic aperture sonar [SAS] system. The SAS data is collected worldwide and comprises various seafloor environments.

“What happens typically is sometimes the data is sensitive, so she’ll develop an algorithm on non-sensitive synthetic data to give to me, and we can test it here and see how it performs in a more real environment,” Cobb said.

Image: Examples of synthetic aperture sonar images.

Above are examples of synthetic aperture sonar images containing sea floor regions of sand ripple, sea grass and hard packed sand. Also, there are four man-made objects on the sea floor in this image.

The team can then apply additional algorithms to the map to go for greater description or, potentially, target detection. In May, Zare, Cobb and doctoral student Xiaoxiao Du presented a paper titled “Possibilistic context identification for SAS imagery” at the annual conference of SPIE — The International Society for Optical Engineering.

“If we know where the sonar system is, we know how high up off the ground it is, and we first detect shadows. Based on how big the shadows are, we can figure out how tall objects are on the sea floor and get a bathymetry map of the whole area,” Zare said.

“By collecting data at different angles and different ranges, we can get a good bathymetry map of the sea floor by putting all the images together. And then, if we apply all of our algorithms on the bathymetry map, we don’t have to deal with things looking different from far away or close in because the bathymetry map isn’t variant to range and the original imagery is.”

Cobb said the current underwater imaging process is very labor intensive in terms of personnel. A key component to this project is to eventually make the process fully automated, which would provide a clearer picture and a more cost-effective one.

Image: Examples of synthetic aperture sonar images.

Above are examples of synthetic aperture sonar images containing sea floor regions of sand ripple, sea grass and hard packed sand.

“If you give unmanned systems the ability to automatically map and characterize the undersea environment, it ultimately reduces the time spent by  researchers or Navy operators to do the same task and scales up your survey capability, ,” Cobb added.

There are a multitude of practical applications for having detailed bathymetry maps and object detection, particularly in terms of research. Zare said that one of the reasons her team is working on object detection is to enhance sea mine detection capabilities.

“If we know what sea floor types there are, we can use the right type of algorithms and the right type of sensors to do our best at finding any targets we’re looking for,” she said. “That’s the application we have in mind.”

Zare became involved with underwater scene understanding during the two summers — 2012 and 2013 — she spent working as a summer faculty fellow at the NSWC in Florida. It was there she began working with Cobb on estimating the characteristics of the NSWC’s sonar images, and that work eventually grew into her current project.

“Dr. Zare has great graduate students working on the projects, and they’ve made a lot of progress,” Cobb said. “We’ve published several joint papers, and the Office of Naval Research is very happy with the progress.”