A Fast Two-Parameter CFAR Algorithm Based on FFT for Ship Detection in Large-Scale SAR Images

Abstract

Ship target detection in synthetic aperture radar (SAR) images via the classical constant false alarm rate (CFAR) algorithm is an important technique. However, each pixel in local sliding window detection is involved in parameter estimation multiple times, which leads to computational inefficiency, especially for large-scale SAR images. To solve the above issue, this paper proposes a fast two-parameter CFAR algorithm based on fast Fourier transform (FFT), named F2T-CFAR. First, FFT is employed to replace the sliding window detection and statistical properties are obtained by global computation. Second, we extract potential target points from the global detection results by threshold decision and subsequently cluster these potential points using the density-based spatial clustering of applications with noise (DBSCAN) algorithm to obtain the final ship target centers. Our F2T-CFAR significantly improves the computational speed compared to the two-threshold CFAR method. Experiments on the measured large-scale SAR images demonstrate the effectiveness of our F2T-CFAR algorithm.