Efficient implementation of rank-only OS-CFAR with dedicated noise estimation

verfasst von

Daniel Köhler, Frank Meinl, Holger Blume

Abstract

Differentiating targets from background noise is an essential task of radar signal processing. Typically, constant false alarm rate (CFAR) detectors, which estimate local noise characteristics to determine an adaptive threshold, are employed for this purpose. A commonly used variant for automotive radar applications is the ordered-statistic CFAR (OS-CFAR) due to its good performance in multi-target scenarios and near clutter edges. However, obtaining the order statistics is associated with computationally intensive sorting of the CFAR training data. With the rank-only implementation, an efficient OS-CFAR algorithm is given, which does not require to calculate the order statistics explicitly and thus removes the necessity of sorting. This has the drawback, that the local noise estimates are not calculated, which may be required in some applications, e.g. to compute the signal-to-noise ratio. In this work we propose a dedicated noise estimation stage as an extension to the rank-only OS-CFAR to compensate for this disadvantage. We show that by including the detection information, noise estimates of comparable or even better quality in the case of spectral regions containing targets can be obtained with minimal computational effort. Furthermore, an efficient FPGA-based implementation of this two-stage approach is presented and compared against other implementations of OS-CFAR.

Organisationseinheit(en)

Institut für Mikroelektronische Systeme

Externe Organisation(en)

Robert Bosch GmbH

Typ

Aufsatz in Konferenzband

Band

2022

Seiten

312-317

Anzahl der Seiten

6

Publikationsdatum

2022

Publikationsstatus

Veröffentlicht

Peer-reviewed

Ja

ASJC Scopus Sachgebiete

Ingenieurwesen (insg.)

Elektronische Version(en)

https://doi.org/10.1049/icp.2022.2336 (Zugang: Geschlossen)