The sensor measurement data supplied to an automatic tracking system is often provided by a peak detector or peak picker. Here it is assumed that digital processing is used for all steps from signal processing onwards. The peak detector follows the signal processing step and converts the received signals into radar measurements. It does this by interpolating the position about local peaks which are above a detection threshold. The peak detector output gives target and clutter measurements which are described by different probability density functions (PDFs). The probabilistic data association filter (PDAF) uses a priori models for the target and clutter distributions. The PDAF has since been extended to include a nonuniform clutter distribution. This paper outlines this extension and also describes another peak detector derived parameter, namely the curvature of a peak. It shows the difference between the peak curvature PDFs for targets and clutter from experimental over-the-horizon radar (OTHR) data. The extension of the PDAF to effectively use peak curvature is then outlined. The examples of performance show the impact of including these extra parameters from a peak detector. These results are based on data recorded from Australia's OTHR, Jindalee.
[1]
Samuel J. Davey,et al.
A comparison of track initiation with PDAF and PMHT
,
2002,
RADAR 2002.
[2]
Yaakov Bar-Shalom,et al.
Automated Tracking with Target Amplitude Information
,
1990,
1990 American Control Conference.
[3]
Samuel J. Davey,et al.
Advanced Jindalee Tracker: Enhanced Peak Detector
,
1999
.
[4]
S. B. Colegrove.
Advanced Jindalee Tracker: Probabilistic Data Association Multiple Model Initiation Filter
,
1999
.
[5]
Sb Colegrove.
Multi-target tracking in a cluttered environment
,
1987
.