Algorithm and Decision Support for Vehicle Mounted Mine Detection Systems
Abstract
The work proposed herein is intended to address specific needs of the US Army RDECOM CERDEC Night Vision and Electronic Sensors Directorate (NVESD) related to the detection of landmines and other buried explosive hazards using vehicle-mounted sensors primarily employing ground penetrating radar sensors, and possibly including electromagnetic as well as other types of sensors. The needs of NVESD in support of design and development of such vehicle-mounted detection systems are quite varied. This document proposes to provide research support for efforts to build such systems by carrying out scientific research in support of the following areas: 1. Research in developing algorithms and methods to employ deep belief networks to leverage detection methods developed using existing hardware platforms employing specific sensor configurations to reduce the manpower and research effort to develop detection methods suited to hardware platforms employing modified sensor configurations. 2. Research to identify effective ways to analyze spaces of features associated with sensor data to identify manifold representations to effectively data dimensionality leading to more effective algorithm development. 3. Research into the theoretical and practical use of features based on approximations of Kolmogorov complexity for discriminating between buried explosive hazards and clutter. 4. Support for software systems used to evaluate detection and discrimination algorithms and assistance in transitioning such systems to government agencies. Our approach in each of these areas will be to apply relevant published techniques, methods, and theories to the problems at hand, to identify which are appropriate to the problem domain, where problems exist, where modifications and extensions are indicated, and where we can identify new methods and techniques that will aid in the location of landmines and other buried explosive hazards. As part of this work, we will develop prototype implementations appropriate for both demonstration of effectiveness and transitioning into deployable systems. The significance of this work is to improve the state of practice in detecting and discriminating target objects in a wide variety of data with an evolving set of sensors. Our efforts are expected to lead to improvements in both theoretical methods and their practical application to fielded detection systems as well improving the capability of the government to effectively evaluate the properties and behavior of proposed detection systems. This proposed work includes efforts of two faculty members working at an average of 15% FTE as well as one post-doc working at 100% FTE, three graduate students, and one IT Specialist working at 100% FTE.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 16, 2018
- Source ID
- W911NF1710103
Entities
People
- Joseph Wilson
Organizations
- Army Contracting Command
- United States Army
- University of Florida