University Research Initiative Synthetic Dataset (HRR)

This page describes a collection of synthetic high resolution radar range-profiles known as the URISD. We have made extensive use of this dataset in the development and testing of algorithms for ATR using high resolution radar.

XPATCH

The HRR signature prediction tool known as XPATCH is a computer software package, developed jointly by Wright Laboratory and DEMACO, Inc., that is designed to simulate the radar return from a given target. The target is represented by a collection of triangular patches and outward normal vectors. Additionally, material parameters as well as edge and curvature information may be provided to further characterize the reflective properties of the target surface. The algorithm is based on a shooting and bouncing ray technique, in which rays are directed from the radar toward the target, and traced as they are reflected by the various surfaces on the target. Those rays which are eventually directed back toward the radar contribute to the return. This technique allows for incorporation of the effects of both shadowing, in which portions of the target remain unilluminated because they lie in the shadow of other structures, and multi-bounce, in which the impinging wavefront may reflect many times inside cavities on the target prior to reflection toward the radar. When simulating the range-profile for a particular target, the algorithm is invoked to simulate the complex scattered far-field at discrete frequencies that are uniformly spaced within the specified transmitter bandwidth. The range-profile is found from the frequency-domain data through the Inverse Discrete Fourier Transform.

The URISD

The University Research Initiative Synthetic Dataset (URISD) is a collection of simulated range-profiles produced by the HRR simulator XPATCH. The URISD includes range-profiles simulated for four ground vehicles over three frequency bands and at three different elevation angles. Each vehicle is represented by a CAD model consisting of triangular patches covering the surface of the vehicle and a simulated rough ground surface, included so that the range-profiles exhibit the effects of reflections from the terrain in the vicinity of the vehicle. The tank models are generic in nature and are not intended to represent any particular existing tank; they are assigned the arbitrary labels of "m1" and "t1". The 36 data sets comprising the URISD each contain range-profiles for target azimuth angles that are uniformly sampled around the circle at a density that depends on the frequency band. Additionally, the data sets contain range-profiles simulated for each of the four combinations of transmitted and received linear polarization states, {hh,hv,vh,vv}. The parameters for the various data sets are summarized in the following table. 
Frequency Band                   UHF            L              X 


Center Frequency                 600 MHz        1.5 GHz        10.0 GHz
Bandwidth                        739.5 MHz      739.5 MHz      1.4775 GHz
Frequency Sample Spacing         4.921 MHz      4.921 MHz      4.921 MHz
Number of Frequency Samples      151            151            301
Azimuth Angle Spacing (degrees)  0.3            0.15           0.02
Number of Azimuth Samples        1201           2401           18001

The Vehicles

The four vehicles for which the URISD contains range-profile data are depicted in the following images. Each image shows the CAD model for the vehicle, including nearby simulated terrain.
Fire Truck School Bus "M1" Tank "T1" Tank

The Data

The following eight images depict examples of URISD range-profiles. In each case, range-profiles were drawn from the dataset for each vehicle at an elevation angle of 10 degrees and for the "vv" polarization channel. Range-profiles from the L-band and X-band datasets are depicted. For the L-band data, range-profiles are shown for azimuth angles between 10 and 15 degrees. For the X-band data, range-profiles are shown for azimuth angles between 10 and 10.6 degrees. Each image shows the magnitude of the complex-valued range-profiles plotted against range bin and azimuth angle. Click on any of the thumbnail images to see a full size version.

L-Band Data

Fire Truck School Bus "M1" Tank "T1" Tank

X-Band Data

Fire Truck School Bus "M1" Tank "T1" Tank

Examination of these images reveals that there are significant differences between the groups of range profiles for each of the four vehicles. However, the range-profiles for a single vehicle exhibit extreme variability for small changes in azimuth. This demonstrates the primary challenge in developing ATR algorithms using range-profile data. The algorithm must recognize a target by capturing the variability between target types, yet be robust with respect to the variability over small changes in orientation.

To see the results of an estimation algorithm demonstrating joint tracking and recognition using these data, click here.

CIS (cis@cis.jhu. edu); page last updated on Wed Dec 13 10:10:54 2000.