Detection of Ship Wakes in SAR Images Using Grayscale Parallel Coordinate Transform

摘要

The detection of ship wakes is one of the most important targets in the area of SAR image applications, but the existing Radon Transform (RT) or Hough Transform (HT) based line detection methods can only be used to examine the collinearity of image points in the parametric space. Considering the characteristics of speckle noise and the spatial distribution of ship wakes along a line with certain width in SAR images, a new method using a nonparametric Grayscale Parallel Coordinate Transform (GPCT) is proposed. In this method, the original grayscale SAR image will be transformed into a parallel coordinate space by a new extended nonparametric GPCT, and the obvious points with high accumulation value will be detected by a modified iterative thresholding method. Then, the detected obvious points will be clustered according to its spatial distribution, and the center of each class will be determined to evaluate the lines of ship wakes. Experimental results using synthetic and real SAR images show that comparing with RT and HT based methods, our proposed method performs better in terms of the accuracy and correctness, better or equivalent in terms of the time efficiency.