Analysis of radar target tracking and classification

Abstract

This paper contains three general topics for analysis of radar target tracking and classification. Two topics are related to radar target classification and the other topic is the analysis of tracking performance for dual-mode homing guidance with target-orientation measurements. First, we propose a new target classification scheme exploiting the temporal dependency and spatial structure of radar cross section (RCS) measurements in low-frequency, passive or long-range surveillance radar systems. We employ target RCS modeling integrated into a hidden Markov model (HMM) with state-duration modeling. Assuming that the radar system ensures a sufficient sampling time, it is possible to utilize the spatial characteristic of airborne targets since there is consistency between successively sampled RCS measurements. Additionally, to exploit whole temporal characteristic of the sequence of RCS measurements that has been rarely considered in the literature, we adopt the HMM and target RCS modeling. To completely accomplish this task, we accurately develop target RCS models and establish the relationship between the target-sensor orientations, which are the hidden states of the HMM, and corresponding RCS measurements. The proposed target classification scheme which only uses the sequence of RCS measurements is demonstrated with simulation results and an analysis for various signal-to-noise ratios (SNRs) and target-fluctuation models. Second, we present three-dimensional (3-D) target tracking based on fused radar and infrared (IR) sensor data with the inclusion of the target orientation in the measurement vector. We provide noise statistic of IR-sensor measurements including target orientation measured from IR image. The track-to-track fusion with extended Kalman filter (EKF) is used to combine radar with IR sensor data. In conventional tracking approaches, there is a fundamental limitation in that it is difficult to accurately estimate the current acceleration of the target even with nearly perfect measurements of range and angle relative to the target. The correlation between target orientation and velocity can be used to overcome this limitation. In this paper, we evaluate tracking performance to show how much improvement is obtainable through the inclusion of the target orientation in the measurement data for a realistic 3-D scenario. Finally, we introduce a radar target classification technique based on the relevance vector machine (RVM) using high resolution range profiles (HRRPs). Although the radar target classification problem based on the support vector machines (SVMs) applied to the hyper-dimensional feature spaces has received much attention recently, RVM-based approaches have never been appeared in the open literature so far. An RVM typically utilizes significantly fewer basis functions than a comparable SVM and therefore can carry out classification with much faster learning time, while offering many additional advantages. Our simulation results confirm that the RVM is a valid and effective alternative to the SVM, and is more suitable for radar target classification.