Research On SAR Image Denoising Methods And Hidden Geological Properties Extracting

Synthetic Aperture Radar(SAR),as an active microwave remote sensing,can detect the target area in all-weather and all-weather conditions.And the formed SAR image have a high range-resolution and azimuth-resolution at the same time.Therefore,SAR has a more and more important role at the fields of military,geosciences and bionomics,etc.While,the subsequent SAR image interpretation and understanding is difficult because of the presentation of SAR speckle the imperfection of the imaging principle.So,SAR speckle denoising is an important research topic of SAR image pre-processing.The microwaves have the penetrability for it's long wavelength.Therefore,SAR images are widely used in the detection of resource environment.The geological properties extracting based on SAR images is a research hotspot in the application of SAR technology.This paper studies the coherent speckle noise and summarizes various speckle noise suppression techniques at the present stage.Based on the sparse representation theory,the paper proposes an overcomplete dictionary learning algorithm with high efficiency and good edge retention.Based on the theory of block matching,the paper proposes adaptive BM3 D algorithm with more excellent denoising effect;and research on the geological properties extracting based on the denoised SAR image;constructing a hidden geological body recognition model and verifying the feasibility of SAR image geological body recognition through experiments.The main research work and achievements of the paper are as follows:(1)A Review of SAR Image Denoising ResearchIn this paper,we review and summarize recent SAR image denoising methods,prominent trend and changing approaches.These methods are classfied into four categories according to the filtering principle: spatial domain filtering,transform domain filtering,anisotropic filtering,and compressive sensing.The characteristics and suitable applications of these filtering methods are analyzed.Both established and new trends for assessment of denoising are presented.Eventually,future development of SAR image speckle suppression is discussed.(2)The double iterative optimal dictionary learning-based SAR image filtering method has been proposed.The KSVD algorithm has been a focus of image filtering research ever since it was first developed.However,this algorithm has a slow convergence rate for sparse coding,and dictionary updating via singular value decomposition(SVD)is too complex.To address these issues,this paper proposes a double iterative optimal dictionary(DIOD)learning algorithm.First,during the sparse coding process,a single iteration is performed to select the optimal and second most optimal dictionary atoms for representing the residual error,and the selected atoms are then updated according to the accumulation of dictionary coefficients.Next,the normalized and weighted reconstruction error is used to update the dictionary.This approach effectively avoids the repetitive occurrence of dictionary atoms in the iteration process,accelerates the convergence rate,and increases the sparsity of the sparse coding coefficients while simplifying the dictionary updating method,enabling rapid filtering of large synthetic aperture radar images,and achieving an improved filtering effect.(3)The SAR image denoising via clustering-based BM3 D method has been proposed.Block-matching and 3D filtering(BM3D)can effectively suppress the noise in stationary signal.However,it is feasible for the speckle noise in synthetic aperture radar(SAR)image with random characteristics due to the single 3d transform threshold and the local neighborhood for searching similar blocks.To address this,this paper proposes a Clustering-Based BM3 D algorithm for SAR image denoising.First,the feature vector is calculated according to the mean,variance and range of each image block,and cluster analysis is undertaken based on K-means algorithm for the feature vectors.The 3d transform adaptive threshold will be determined through the estimated noise variance of each class of image blocks.Second,global similar image blocks of reference image block can be found in the corresponding class of image blocks because the similar blocks are assembled in the same classification.(4)Research on the geological properties extracting based on the filtered SAR images.The preliminary exploration of the application of SAR imagery in the identification of hidden geological properties was conducted.This paper summarizes and analyzes the theoretical knowledges about the hidden geological properties and constructs a hidden geological properties extracting model based on the filtered SAR images.Because the study area is mostly homogenous area,the clustered BM3 D algorithm which is more suitable for the smooth area is selected to filter the target area SAR image.Based on the hidden geological properties extracting model constructed in this paper,the SAR images can be used for the hidden geological properties extracting through the comparison of the distribution of the Quaternary rock minerals between the SAR images after polarization decomposition and the optical images,and the SAR image has a good boundary effect.Through polarization decomposition of SAR images filtered by different denoising algorithms,the results show that the texture retention characteristics of the filtering algorithm have an important influence on the geological body recognition.In addition,the influence of different band SAR images on the identification of buried geological properties was explored.The Jurassic rocks in the region were studied.By analyzing the results of the L-band and C-band SAR images in this region,it was found that L-band images have more better geological properties extracting boundary effects.