SAR Image Despeckling And Ship Wake Detection Methods

Detecting the ship wakes with SAR images,one can acquire information of the ship,the ship wake,the Doppler shift,etc.It's currently one of the most effective ways of monitoring,locating,detecting and recognizing ships.However,there are much speckle noises in the SAR image,while the ship wake signals are relatively weak,thus bringing certain difficulties in the recognizing process.In this article,first introduce the imaging principle of the SAR radar,and then the origin of the speckle noises in SAR images and the multiplicative noise model.According to the main focus of this research,respectively introduce the traditional spatial domain filtering algorithm,the evaluation standards for suppression of speckle noises,the feature of ship wakes,and the traditional detection algorithms of SAR images.Based on the above theories,propose a complete set of methods for detecting the ship wakes in SAR images.Our primary research is as follows:(1)Propose a denoising algorithm for SAR images based on bilateral filtering and wavelet multi-model threshold.A lot of denoising algorithms based on the wavelet threshold don't deal with the low-frequency sub bands,while there are much speckle noises in the low-frequency sub bands of SAR images.After two-layer wavelet decomposition of SAR image,the three high-frequency components of the wavelet sub band of the first layer contain mostly the noise component,thus apply the GCV threshold for soft threshold denoising.The three high-frequency components of the sub band of the second layer contain less noise component,thus apply the classical Bayes threshold for soft threshold denoising.Experiment results show that this algorithm have great denoising effect on the SAR images,at the same time more effectively preserving the ship wake information.(2)Use the morphological component separation based on analytical dictionaries to separate the components in SAR images,and obtain the structural component image containing the ship wake and the textural component image containing the speckle noises and sea clutter.For SAR images heavily polluted by speckle noises,solely applying the algorithm in(1)can't achieve satisfactory denoising effect.Apply the morphological component separation based on analytical dictionaries,constructing the dictionary of the ship wake structure using Contourlet transform and that of the speckle noise and sea clutter using biorthogonal wavelet transform,and solve iteratively,thus separating the ship wake component from the noise and clutter component.Experiment results show that compared to the original image,the ship wake structural component image obtained through the above method has a vastly improved ENL,at the same time preserving the structure information of the ship wake well,having prominent effect.(3)Propose a ship wake detection algorithm based on local Radon transform and peak clustering decision.The algorithm uses the ship wake structural component image obtained by(2)as the input,using local Radon transform to extract all the local peaks,and then clusters them,removing the duplicates.This is because real ship wakes in SAR image have a certain width,and a single ship wake may correspond to several local peaks in the Radon field,thus by clustering and removing the duplicates one can shorten the decision process.After that,evaluate the local peaks using a decision function,and if the result surpasses the threshold,then it is the real peak,and conduct the inverse Radon transform to obtain the real ship wake.If not,it is a false peak.On the choice of the decision function,choose the matching degree of the wave form of the Gaussian wavelet and that of the local peaks in the ? direction and the amplitude of the local peaks in the Radon field as the arguments.Experiment results show that this algorithm has good detection effect on SAR images heavily disturbed by speckle noises.To sum up,(1)-(3)constitute the ship wake detection system of the SAR image.Conduct simulations applying the whole process on the satellite images of ERS-2.Experiment shows that our algorithm achieves better results on aspects of ship wake detection rate and false alarm rate.