Use of a Space-time adaptive processing algorithm to optimize the detection of slow moving targets

Jess Altman Mentored by Dr. Richard Czernik

A challenging problem faced by airborne ground moving target indicator (GMTI) radars is the ability to detect slow moving targets (i.e. dismounts) buried within a Doppler clutter ridge, which is an unwanted artifact of aircraft motion. Space-time adaptive processing (STAP) algorithms are advanced signal processing techniques that utilize a two-dimensional filtering technique and a phased-array antenna to mitigate platform-motion-induced-spread-Doppler clutter. This clutter interferes with the echo from slow moving ground targets, in addition to decreasing the severity of any jamming and interference that might also be present. Clutter, jamming, and interference reside in joint angle- Doppler space and together create one or more ridges within that space whose angle and behavior depends primarily upon the platform velocity, the Doppler frequency and the antenna orientation. The goal of this project is to use a Space Time Adaptive Processing (STAP) algorithm in a self-created simulated radar environment to optimize the rejection of clutter, jamming, and interference and thereby achieve a higher probability of successful location and identification.

Introduction

After learning the basics of radar detection theory, a simulation in MATLAB® was designed to model the hypothesized effect and the output of an airborne monostatic radar given realistic aircraft and radar inputs. Initially, the azimuth range, elevation, platform and target Doppler frequency and altitude, along with other quantities as a function of time had to be calculated in order to generate a target location in Angle-Doppler space for each simulation. The resulting signal to noise ratio (SNR) was then modeled, along with the ambient clutter to noise ratio (CNR) and jammer to noise ratio (JNR) and combined into a total interference covariance matrix. A model for the generation of hypothetical ambient Gaussian background clutter was also developed to serve as ‘training data’ for the simulated radar STAP algorithm. A complex valued data cube consisting of a series of space-time snapshots was then calculated using the appropriate spatial and temporal steering vectors based upon the total number of elements in the modeled antenna array, the number of radar pulses emitted during a coherent processing interval (CPI), and the total number of radar range resolution cells. Those key metrics are the Adapted Pattern (Angle-Doppler frequency response), the signal-to-interference plus noise power (SINR), the SINR Improvement Factor, the Minimum

Methods

Detectable Velocity (MDV), and the interference Eigenspectrum. These quantities allow one to determine the degree of successful clutter and jammer suppression achieved and thereby the overall effectiveness of radar performance. A graphical user interface was added to allow the user to test with multiple range rings in order to add additional training data to improve the ability to null the jammer and clutter, and also determine the effects of intrinsic clutter motion on overall interference suppression.

Methods (cont’d)

Figure 1a: Azimuthal view of a plane and radar system. Figure 1b: Normalized Doppler spectrum showing the effects of clutter and STAP. (Goldstein, Picciolo, Griesbach & Rustan, 2011)

I would like to thank my mentor, Dr Czernik for guiding me through this journey, and my faculty advisor, Mr. Davis, for providing any and all assistance whenever I needed it.

Acknowledgments

Goldstein, S., Picciolo, M., Griesbach, J., & Rustan, P. (2011, May). STAPii: Detection theory and advanced techniques. IEEE Radar 2011 conference briefing.

References

The purpose of this project was to successfully simulate a radar environment and optimize the probability of detection of slowly moving ground targets. As shown in Graph 1, both the clutter due to aircraft motion and the jammer were nulled, allowing slow moving targets to be detected and localized. The optimized data allowed for the most realistic test, meaning the next logical step would be to test this algorithm using ‘real world’ radar training data to verify the algorithm’s effectiveness. Such algorithms represent the state of the art in advanced airborne military radar systems.

Conclusion

MDVR spectra using training data from one range ring

Angle (degrees)

Optimized data was achieved in the simulation and is displayed below.

Results

Graph 1: Adapted pattern showing target location and jammer/clutter nulling.

Graph 2: Spectrogram showing the aliased clutter ridge that is being suppressed via STAP. Aliasing is due to the high aircraft velocity.

Adaptive pattern using training data from single target ring only

Dop

pler

freq

uenc

y (H

z)

2500 2000

500

1500 1000

-500 0

-1000 -1500

-2500 -2000

Angle (degrees) -80 -60 -40 -20 80 60 40 20 0

-140

-120

-100

-80

-60

-40

-20

0

20

-80 -60 -40 -20 80 60 40 20 0

2500 2000

500

1500 1000

-500 0

-1000 -1500

-2500 -2000

Dop

pler

freq

uenc

y (H

z)

-20 -15 -5 0 5 10

15 20

Results (cont’d) Graph 3: A well behaved Eigenspectrum showing the number of degrees of freedom available for interference suppression.

Graph 4: Plot of Adapted Pattern cross section showing near total suppression of the clutter ridge at 0 degrees and a null placed on the jammer at -40 degrees.

MDVR spectra using training data from one range ring

Angle (degrees)

(b) (a)

ADIANCETECHNOLOGIES

0 50 100 150 200 250 300 350 0

10

20

30

40

50

60

Eigenvalue Index

Mag

nitu

de(d

B)

Eigenspectra using training data from target ring only

-100 -80 -60 -40 -20 0 20 40 60 80 100 -100 -80 -60 -40 -20

0 20 40

dB

Azimuth (Degrees)

Adapted pattern principle cut for target Doppler (1 ring covariance matrix)