Research On Detection Technique Of Ship Targets In PolSAR Images

Synthetic Aperture Radar(SAR)is an active microwave imaging system,and its imaging is not limited by meteorological conditions such as light,climate,and clouds,and its imaging resolution is independent of distance.And it has all-day,all-weather data acquisition capabilities.Synthetic aperture radar has become one of the main highresolution observation devices for ocean areas,and it plays an important role in continuous and long-term monitoring tasks in ocean areas.With the on-orbit operation of various spaceborne SAR sensors and the use of various airborne SAR sensors,the acquisition of SAR images has become more and more convenient.Compared with single-polarization SAR data,full-polarization SAR(Polarimetric SAR,PolSAR)data contains not only the information of the backscattering intensity of the target,but also the information of the polarimetric scattering characteristics of the target.In this paper,the problem of ship target detection in PolSAR images is studied,and the detection problem of weak and small ship targets in full-polarization SAR images is mainly studied.In order to promote a practical and efficient full-polarization SAR image ship target detection system.The main research work of this paper includes the following four parts:1.For weak and small targets in single-channel SAR images,the signal-to-clutter ratio is low,and traditional pixel-level detection algorithms are sensitive to speckle noise.Using the physical scattering mechanism and structural information of the target,this paper proposes a PolSAR image ship detection algorithm constructed by using the reflection symmetry metric image and the keypoints extracted by the Scale-invariant feature transform(SIFT)algorithm.Due to the obvious difference in the degree of deviation of the ship target and sea clutter from the reflection symmetry,the signal-to-clutter ratio of the weak and small targets in the reflection symmetry metric image has a certain improvement compared with the original image.However,affected by speckle noise and strong sea clutter,there are still some strong scattering points on the sea area in the reflection symmetry metric image,which affects the performance of the pixel-level target detection algorithm.In order to solve the above problems,this paper uses the SIFT keypoint detector to extract the keypoints with blob structure on the reflection symmetry metric image,and uses the scale information of the keypoints to determine the range of the potential target regions,and completes target detection on the potential target regions.Since the ship target has obvious structural characteristics,and the SIFT keypoints are not sensitive to noise,the method proposed in this paper can effectively reduce false alarms and has a certain ability to detect weak and small targets.In order to verify the effectiveness of the proposed algorithm,this paper conducts experiments on two sets of measured data containing weak and small targets.The results show that the performance of the proposed method is significantly improved compared with the traditional pixel-level detection methods.2.The traditional polarimetric notch filter(PNF)suppresses sea clutter through orthogonal projection technology,and uses the energy of the target subspace to construct pixel-level detection statistics.However,pixel-level detection statistics are vulnerable to speckle noise and strong sea clutter.To solve the above problems,first,polarimetric whitening filtering is performed on the original data to reduce the influence of speckle noise on the energy of the target subspace.Then,a patch-level feature,patch variation indicator,is defined to describe the spatial distribution characteristics of ship targets and sea clutter in local regions,and the whitened subspace energy and patch-level features are used to construct the patch-level detection statistics.Finally,the experimental verification is carried out on a set of simulated data and two sets of measured data,and the proposed method is compared with the classical detection methods.The results show that the proposed method can suppress the strong sea clutter to some extent,and the detection performance is improved.3.In view of the fact that the blob structure of weak and small targets in the PolSAR image is not obvious,and the Gaussian filtering operation in the SIFT algorithm will cause the image to be blurred,the SIFT keypoint detector has insufficient detection performance for weak and small targets.Inspired by the SAR-SIFT keypoint detection algorithm,this paper proposes a keypoint detection algorithm,which is referred to as PolSAR-SIFT,to detect corner structures in PolSAR images,and designs a ship target detection algorithm based on the PolSAR-SIFT keypoint detector.First,a gradient for PolSAR image is defined using a distance metric between polarmetric covariance matrices and an exponentially weighted average operator to construct a PolSAR-SIFT keypoint detector.Then,the keypoints in the PolSAR image are extracted using the PolSAR-SIFT keypoint detector.According to the patch variation indicator values of the detected keypoints,the false alarm points are removed,and then the potential target area is extracted by the filtered keypoints.Finally,the target detection is completed by thresholding the detection statistics on the potential regions.Since the corner structure of weak and small targets in PolSAR images is relatively obvious,and the exponential weighting operator can reduce noise while preserving image details,the detection performance of PolSAR-SIFT keypoint detectors for weak and small targets is improved.Finally,the detection performance of the proposed algorithm for weak and small targets is verified on three sets of measured data.4.Aiming at the problem of low efficiency in the computation of PolSAR-SIFT keypoint detection and the problem of target detection in non-orthogonal subspaces,this paper proposes a ship detection algorithm based on non-circularity keypoint detection and oblique projection technology.First,inspired by the Harris corner detector,a keypoint detector is designed using a two-dimensional matrix characterizing polarimetric scattering characteristic and non-circularity to extract potential target regions.Then,the basis vector of the target subspace is set according to the ship detection task,and the oblique projection technology is used to suppress the clutter while ensuring that the energy of the target subspace is not projected to other subspaces,which improves the contrast between the target and the clutter.Finally,on a set of simulated data and two sets of measured data,the proposed method is compared with the PNF algorithm which is based on orthogonal projection technology and other classical target detection algorithms,and the detection performance improvement of the proposed algorithm is verified.