You are here

Mitigation of GMTI Radar False Alarms Due to Wind-Blown Foliage with Machine Learning Techniques


TECHNOLOGY AREA(S): Electronics, 

OBJECTIVE: The objective of this task is to develop a machine learning technique for Ground Moving Target Indicator (GMTI) radars that will mitigate false alarms caused by windblown foliage and other nonstationary clutter while maintaining the ability to detect slowly-moving ground targets. GMTI radars will typically adapt a Constant False Alarm Rate (CFAR) threshold and/or form a conventional clutter map system. This clutter map contains the Radar Cross Section (RCS), Doppler bandwidth of the clutter return, the extent of region in which the persistent false alarms occur, and the probability density functions of the clutter data. Radar operators will commonly desensitize particular areas of coverage due to these high false alarms. While this approach will minimize the false alarms caused by dynamic clutter, large portions of the radar coverage area can be desensitized. This desensitization can allow an area to be penetrated by hostile forces. The objective of this topic is to determine whether machine learning techniques are able to differentiate between areas of dynamic clutter that are unoccupied from those that contain ground moving targets. This effort will compare the performance of the machine learning approach versus that of a CFAR / conventional clutter map system as a function of the clutter characteristics (e.g., RCS, Doppler bandwidth, and temporal variability), the target speed, heading, RCS, time required to make an initial detection, and the tracking accuracy. 

DESCRIPTION: Airborne and ground-based MTI radars are designed to detect, locate, and track slowly moving targets such as walking dismounts. A critical issue for these radars is persistent false alarms caused by nonstationary clutter such as windblown foliage, moving water, and rotating objects. This effort will design a machine learning technique that will improve radar detection, location, and tracking performance by first using simulation, and then demonstrating the technique on collected radar MTI data. A thorough understanding of both machine learning techniques and the operation and performance of GMTI radar must be demonstrated to successfully perform this effort. 

PHASE I: Demonstrate through simulation a viable and robust machine learning technique to mitigation of GMTI radar false alarms. The simulation is expected to use statistical clutter based on user-specified RCS and Doppler bandwidth. Targets are expected to be simulated as constant speed with a user-specified RCS. The developer will compare these simulations against a conventional clutter map approach. 

PHASE II: Further develop the machine learning technique to process various dynamic clutter / ground moving target data. Data sets will be provided by the Government or obtained by the performer from any available sources including a customer-owned GMTI radar. The developer will compare these simulations against a conventional clutter map approach. 

PHASE III: Work with Army and industry partners to incorporate the machine learning technique within an existing or developmental radar system. The offeror will demonstrate the capability and applicability of the machine learning technique for both government and commercial applications. The offeror will provide a comprehensive commercialization program plan to ensure transition of this technology from military to commercial applications such as perimeter security, border security, or other military or civilian application. 


1: Skolnik, Merrill. Introduction to Radar Systems. McGraw-Hill Inc, New York, 1992, Pp. 392-395.

KEYWORDS: GMTI Radar, Clutter Suppression, Machine Learning 

US Flag An Official Website of the United States Government