You are here

New Radar Exploitation Methods for Combat Identification


OBJECTIVE: Develop automatic target recognition capabilities based on reduced feature sets (i.e. salient feature representation to reduce dimensionality and enhance performance) to support on-board combat ID and decision fusion for remotely piloted vehicles. DESCRIPTION: Emerging operational requirements using remotely piloted vehicles (RPV) for combat identification (ID) demand highly accurate, efficient multi-sensor automatic target recognition (ATR) systems. Conventional systems either provide on-board image-only information or ground station exploited-imagery information. Yet, reduced size, weight and power requirements of advanced computational systems combined with efficient methods and algorithms suggest opportunity to develop image and exploitation products directly on board the RPV. Rather than generate imagery and then exploit the image, novel approaches could perform feature extraction and exploitation jointly with imaging to provide enhanced ATR performance. This effort addresses the potential to enhance ATR performance through novel exploitation of high resolution radar (HRR) and synthetic aperture radar (SAR) data derived from common tactical radar systems that do not have full polarimetric capabilities. The primary applications are tracking and targeting of moving and stationary military and civilian ground vehicles using active, monostatic air to ground radars. Technical challenges to produce this type of multi-sensor ATR include real-time image formation and feature extraction/exploitation along with fast, efficient algorithms. Recent advances offer new approaches to performing joint image/feature extraction for ATR [1-2]. Integration of the image formation process with feature extraction to accomplish ATR could be advantageous in maintaining pedigrees needed to achieve high confidence and accuracy in the fused solution. One possible approach is the use of attributed imagery [3], in which each qualifying pixel is attributed during image formation as a canonical or primitive structure referred to as an attributed feature. Besides contextual information from the image, the attribution offers potential to convey additional information for increased confidence and accuracy in classification. This approach, applicable to both SAR and HRR data, could be used to form a joint ATR with compact feature space that operates on physics-modeling rather than stochastic modeling. This work will explore the feasibility for using compact feature target representation along with novel ATR and fusion algorithms to develop on-board combat ID capabilities. The effort should include the use of attributed features or other data constructs optimized to convey covariance, invariant feature data [4,5] or other information necessary to support the development of an efficient, high confidence ATR. The proposed concepts should function to the highest possible degree with data extracted from existing target data sets. ATR development must include metrics for measuring and validating performance (e.g. efficiency, speed, and accuracy) throughout the decision chain. Proposers must be familiar with target modeling, feature extraction and data fusion techniques used in ATR and have at least 3 years experience in the development of ATR algorithms for both HRR and SAR. A limited data set will be made available to successful Phase I proposers and no other government materials, equipment, data, or facilities will be provided under Phase I. PHASE I: Develop concepts for extracting and representing salient, compact feature sets from SAR and HRR data with potential to improve ATR performance. Develop ATR algorithm and data fusion concepts based on compact feature sets. Develop mathematical uncertainty models for feature extraction and propose metrics for tracing performance and uncertainty throughout the ATR classification process. PHASE II: Develop and demonstrate ATR object classification algorithms. Develop and validate uncertainty models and metrics for the ATR system from end to end (i.e. feature extraction to decision declaration). Develop concepts and requirements for on-board ATR capabilities supporting high confidence combat ID. Demonstrate and validate the ATR system performance using existing data sets. PHASE III: Military Application: Provide automated object discrimination and classification for targeting, target exploitation and situational awareness. Commercial Application: Provide automated object discrimination and classification for land and ocean rescue operations. REFERENCES: 1. Potter, L.E. and R.L. Moses,"Attributed Scattering Centers for SAR ATR,"IEEE Trans. Image Processing, vol. 6, no. 1, Jan. 1997. 2. Pasala, K.M. and J.A. Malas,"HRR radar signature database validation for ATR: An information theoretic approach,"IEEE Trans. Aerospace and Electronic Systems, vol. 47, no. 2, Apr. 2011. 3. Fuller, D.F."Phase history decomposition for efficient scatterer classification in SAR imagery,"PhD dissertation, Air Force Institute of Technology, AFIT-DEE-ENG-11-09, 2011. 4. V. Velten,"Geometric invariance for synthetic aperture radar (SAR) sensors,"Algorithms for Synthetic Aperture Radar Imagery V, E. Zelnio, ed., v. 3370, SPIE Proceedings, Orlando, FL, April 1998. 5. V. Velten,"Experiments with Synthetic Aperture Radar Geometric Invariants,"Seventh Automatic Target Recognizer Systems and Technology Conference, US Naval Postgraduate School, March 1-3, 1999.
US Flag An Official Website of the United States Government