New Pose Estimation Methodology for Target Tracking and Identification

Abstract

Ground Moving Target Indicator (GMTI) and High Resolution Radar (HRR) can track position and velocity of ground moving target. Pose, angle between position and velocity, can be derived from kinematics estimates of position and velocity and it is often used to reduce the search space of a target identification (ID) and Automatic Target Recognition (ATR) algorithms. Due to low resolution in some radar systems, the GMTI estimated pose may exhibit large errors contributing to a faulty identification of potential targets. Our goal is to define new methodology to improve pose estimate. Besides applications in target tracking, there are numerous commercial applications in machine learning, augmented reality and body tracking.

Target Signatures and Pose Estimation

“…In a typical airborne application (USA JSTAR is good example) [15][16][17][18], a ground target is detected from a distance, identified and classified using stored target data base, and then tracked after that. In this paper we expand our work reported in [1,2], where we analyzed certain number of targets in spatial domain in particular. Section 2 in this paper reviews several targets signatures from USA public Moving and Stationary Target Acquisition and Recognition (MSTAR) program database [15][16][17][18].…”

“…Basic target and sensor platform geometry is given in Figure 1. In [1] we used three signatures for analysis, and they are repeated here as a reference in Figures 2 (raw continuous data) and Figure 3 (16 digitized signature values). In [2] we expanded our spatial analysis to many more targets, digitization was at the level of 32 and 64 data points, and we also defined a synthetic target as a reference in statistical testing.…”

“…The MSAR data base [15][16][17][18], has more targets and our methodology can be applied to any of them, or to similar target data base. In order to analyze target signatures, let us first define two families of stochastic processes X(i,j)=X(Ti,θi) and Y(i,j)=Y(Ti,θi) as follows: X(Ti,θi)= Radar X signature, i th target, j th pose Y(Ti,θi)= Radar Y signature, i th target, j th pose (1) where I = 1,2,3,4 (1-3 are real targets while 4 is synthetic), j = 1, 2… 274 (4 quadrants, 360 o ). Table 1 summarizes all target and pose combinations [5].…”

“…Hence we missed by one pose position, error of 1 in 274 or 0.365%, which is an excellent estimate between (i) geometrical and (ii) correlation comparisons. See our papers [1,2] for more information on correlation coefficients. In general, this still does not mean we found the correct pose, it is an initial estimate only.…”

“…Table 7 has corresponding statistics summary. We can start with X(1,1) and correlate it with X (1,2) , then with X (1,3) , etc., and finally with X (1,274) forming envelope of correlations. This results in a 4x274 = 1096 correlations.…”