IMPACT OF SPATIALLY AND TEMPORALLY
NONSTATIONARY CLUTTER ON COVARIANCE
MATRIX ESTIMATION
Yan-ling Shi
College of Telecommunications & Information Engineering,Nanjing University of Posts and Telecommunications,
Nanjing 210003, China
E-mail: shiyanlingasd@yahoo.com
Telephone: 86-25-83492413
Abstract
The sample covariance matrix (SCM) and normalized sample
covariance matrix (NSCM) are the general estimators to
estimate the covariance matrix of high-resolution sea clutter.
Although two estimators have been proposed for a long time,
their impacts on the spatially and temporally nonstationary
sea clutter have not been analyzed until now. The adaptive
normalized matched filter (ANMF) and block-ANMF
detector are utilized to analyze the detection performance of
those two estimators in spatially and temporally nonstationary
sea clutter. It provides a theoretical guidance for the marine
radar to detect targets.
Keywords: High-resolution sea clutter, SCM, NSCM, Target
detection.
1 Introduction
High-resolution sea clutter exhibits the spatial-temporal
nonstationarity and is closely related to the dynamic
evolvement of sea surface in time. Greco has presented an
analysis of sea clutter nonstationarity with respect to clutter
covariance matrix estimation and its impact on the constant
false alarm rate (CFAR) property [1,2]. In the processing of
sea clutter, the clutter suppression is an average method at
present. The residual clutter, as the output of the clutter
suppression filter, remains to be spatially and temporally
nonstationary [3,4]. Target detection performance in residual
clutter is still an open research question. Due to the unknown
clutter covariance matrix, the two-step detection scheme is
presented [5-8]. Firstly, assume that clutter covariance matrix
is known and derive the canonical detector; secondly,
estimate clutter covariance matrix by the secondary samples.
There are many covariance matrix estimators, such as the
sample covariance matrix (SCM) and normalized sample
covariance matrix (NSCM) which have wide application in
the field of target detection, especially in multiple-input
multiple-output (MIMO) radar system [6,7] and space-time
adaptive processing (STAP) [8]. In the paper, the impact of
the spatially and temporally nonstationary residual sea clutter
on the SCM and NSCM estimators will be analyzed, and the
canonical detectors are the adaptive normalized matched filter
(ANMF) detector and block-ANMF detector.
The paper is organized as follows. In Section 2, we describe
the ANMF detector and block- ANMF detector in residual
clutter. SCM and NSCM estimators are given in Section 3.
Experimental results and performance evaluations are
presented in Section 4. Finally, we conclude our paper.
2 ANMF detector and block-ANMF detector in
residual clutter
In most methods available in sea clutter, moving target is
directly detected in sea clutter [5-8], as shown in Fig. 1. Sea
clutter is spatially and temporally nonstationary, which limits
the detection performance. If the clutter is suppressed, do the
nonstationary characters of sea clutter chance? How is the
detection performance? Shown in Fig. 2 is the idea of the
paper.
Fig. 1 Schematic diagram of traditional detectors
Fig. 2 Schematic diagram of the proposed detector
2.1 Residual clutter
Utilizing the clutter suppression filter in [4], we obtain the
residual clutter. The clutter suppression filter has the
following characters [4]. First, the clutter suppression filter
does not disarrange the phases of target echoes, which is
crucial to the subsequent target detection. Second, the
residual clutter can be modeled as spherically invariant
random vector (SIRV). Meanwhile, at all the Doppler
frequency bins, the residual clutter vectors have their power
spectra close to a constant. But, the residual clutter remains to
be spatially and temporally nonstationary.