ITL team gets their hands dirty with soil classification effort

U.S. Army Engineer Research and Development Center
Published June 28, 2022
Global maps are shown colored by soil "labels" found using product quantization.

Global maps are shown colored by soil "labels" found using product quantization.

Global maps are shown colored by soil "labels" found using product quantization.

Global maps are shown colored by soil "labels" found using product quantization.

Soil exhibits immense diversity across the Earth’s surface, naturally developing under varying climate regimes, geological materials, landscape portfolios, time intervals and more. A U.S. Army Engineer Research and Development Center (ERDC) team is working to help remotely identify soil in an effort to enable the Department of Defense to confidently and accurately predict its potential impacts on various operations, particularly in foreign countries and access-denied areas.

While numerous approaches have been developed to describe different soil types, most classification is based on criteria delineating from the morphological development of the complete soil profile — an approach forming the basis of most modern soil taxonomic systems.  However, while employed around the world, those systems contain conflicting diagnostic criteria, inconsistent naming conventions and irregular data requirements and availability.

“The ERDC Information Technology Laboratory (ITL) contribution to this effort was the classification and prediction of soil surface analogs using machine learning algorithms, along with a novel regional locality-based sensitive hashing technology that determines regions of similar soil on a global scale,” said Dr. Andrew Strelzoff, a computer scientist in ITL. “Our team became involved in the work being done by the ERDC Environmental Laboratory’s Dr. Mark Chappell, lead for this project, and quickly grew interested in some of the current problems in soil classification that machine learning could potentially help solve.”

The project, known as Intelligent Environmental Battlefield Awareness — Classification Engineering Using Regional Locality-Based Sensitivity Hashing, will allow the DoD to utilize accessible sites, where raw material can be collected, for laboratory experiments and to model predictions for soil behavior in restricted areas around the world. The knowledge gives leaders access to critical data that will allow them to make more informed decisions and, ultimately, better enable the regiment through improved mobility and equipment.  

“The technology has far-reaching implications for both military and civil applications,” said Dr. Haley Dozier, an ITL computer scientist. “For example, the ability to locate accessible areas that allow recreation of environmental conditions of inaccessible areas is critical to better enabling the regiment through improved mobility and equipment. We also envision expansion of the application portfolio to include new areas such as sound and voice, classifying zones by climate change susceptibility, microwave and energy weapons and sources, face recognition, pose and posture for individual similarity and more.”

“The innovative technology developed in this effort led to the filing of a provisional patent,” added Ashley Abraham, also a computer scientist in ITL. “Particular features of the developed methodology and code base are widely applicable to analog data searches across many types of ERDC and U.S. Army Corps of Engineers data due to scalability, flexibility of application-specific reconfiguration for achieving different levels of target accuracy and versatility for signal processing transformation to arbitrarily complex data and sensor signals.”

This success of this project is the direct result of collaboration across multiple ERDC laboratories, a unique capability that allows the organization to serve as a “one-stop shop” for some the nation’s toughest problem. The team was recognized with a 2022 ERDC Research and Development Achievement Award for this effort and plans to expand in the future. 

“I think our ability to collaborate across labs is one of ERDC’s biggest strengths — where else can you get access to all of this in one place?” said Althea Henslee, a computer scientist in ITL. “We are now working on a follow up effort for the same program, which will extend our previous work in this area and make new and impactful contributions to the fields of soil and computer science.”