Design of Connected Component Analysis Based Clustering of CFAR Image in Pulse Doppler Radars

Narasimhan, R S and Vengadarajan, A and Ramakrishnan, K R (2017) Design of Connected Component Analysis Based Clustering of CFAR Image in Pulse Doppler Radars. In: IEEE Aerospace Conference, MAR 04-11, 2017, Big Sky, MT.

PDF IEEE_Aer_Con_2017.pdf - Published Version Restricted to Registered users only Download (310kB) | Request a copy

Abstract

In this paper we discuss the design of clustering scheme of CFAR image map to identify distinct targets present in the coherent processing interval (CPI). Detections out of a CFAR are conventionally grouped together using a fixed size window. This window scans over the set of detections and groups all those that lie in proximity to the peak detections. This clustering process is carried out till all detections lie in any one cluster, and each cluster is used for further processing. The major limitation of the approach comes from the use of fixed correlation window for clustering CFAR image. The process does not ensure the possibility that an extended return or two targets with overlapping returns are clustered and processed together. In this paper we propose the application of connected component labeling for clustering CFAR image to form final distinct detections. Connected component labeling is a classical algorithm used in image processing for segmentation of binary images. The algorithm groups pixels of a binary image based on 4 or 8 connectivity. Thus connected component analysis is most suitable algorithm for segmentation of CFAR image to identify distinct clusters or targets. The algorithm does not require the prior knowledge of number of clusters/targets present and the correlation window to form clusters. The modifications to the classical CCA algorithm is carried out to fold the Doppler edge to take care of periodic spectrum behavior in Doppler direction. The algorithm allows the clustering of range extended returns or overlapped target returns in to single cluster if the returns are connected. The valley detection algorithm is devised to identify the cause for extended range clusters. If the extended range cluster is formed the valley detection algorithm works on the return strength from the detected range cells and the trough in between indicates the returns are from two nearby range targets. This information is used to improve range resolution performance of the radar. The resulting scheme does not require a fixed window. The scheme provides for a robust method for clustering extended or arbitrary target returns on the CFAR map. Simulation studies have been carried out to evaluate the improvement achieved in resolving targets in range axis. The simulation studies show that the target return peaks separated by a range cell is resolvable as against conventional approach where in the peaks shall be separated by multiple range cells.

Item Type:	Conference Paper
Series.:	IEEE Aerospace Conference Proceedings
Additional Information:	Copy right for this article belongs to the IEEE, 345 E 47TH ST, NEW YORK, NY 10017 USA
Department/Centre:	Division of Electrical Sciences > Electrical Engineering
Date Deposited:	13 Oct 2017 04:51
Last Modified:	13 Oct 2017 04:51
URI:	http://eprints.iisc.ac.in/id/eprint/58034

Actions (login required)

View Item