COMPARISON AND EVALUATION OF SHIP TARGET DETECTION ALGORITHMS WITH SAR IMAGES

摘要

Nowadays many ship target detection algorithms have been developed based on SAR images. The typical algorithms include constant false alarm rate algorithm, wavelet transform algorithm and cross correlation algorithm. But there are no standard test methods to compare them. So in this paper, the three detection algorithms are used to extract ship signatures on a common set of ENVISAT ASAR images. Then those results are compared with in situ measurement data for analyzing the application scope and detection precisions of each ship target detection algorithm. The comparison show that choice of an appropriate threshold probably constitutes one of the major drawbacks of constant false alarm rate algorithm, and constant false alarm rate algorithm is not suitable in coarse sea state; On the contrary, the wavelet transform algorithm and cross correlation algorithm have great robustness and can extract several weak signatures which are easy to be disturbed by background. But they can miss some details information of vessel structure. Those details are very important for future vessel classification.