Research On SAR Image Change Detection Algorithm Based On Graph Theory

Spaceborne remote sensing is a non-contact detection technology that receives the electromagnetic waves from the surface to indirectly obtain the space resources information.Synthetic Aperture Radar(SAR)is a microwave system that can perform remote sensing imaging processing at all times under any weather and light conditions.Therefore,SAR images have become one of the most important data sources of remote sensing image interpretation.SAR image change detection is a hot research topic in the field of remote sensing image interpretation.The purpose is to accurately and quickly extract change information between SAR images which obtain in the same geographic area but at different times.It has been widely used in the civil and military fields,and gets important research value and practical significance.This thesis aims at developing the SAR image change detection approach based on the graph theory.It improves the mode of change measurement and the performance of difference image analysis.The core research work of the thesis mainly includes the following four parts:(1)In order to solve the problem of the speckle noise in SAR images,a pointwise change detection approach is developed by employing a bilaterally weighted graph model and an irregular Markov random field(I-MRF).First,keypoints with local maximum intensity are extracted from one of the bi-temporal images to describe the textural information.Then,two bilaterally weighted graphs with the same topology are constructed for the bi-temporal images using the keypoints,respectively.They utilize both the spatial structural and intensity information to provide good performance for feature-based change detection.Next,a change measure function is designed to evaluate the similarity between the graphs,and then the nondense difference image(NDI)is generated.Finally,an I-MRF with a generalized neighborhood system is proposed to classify the discrete keypoints on the NDI.Experiments on real SAR images show that the proposed NDI improve separability between changed and unchanged areas,and I-MRF provides high accuracy and strong noise immunity for change detection tasks with noise-contaminated SAR images.On the whole,the proposed approach is a good candidate for SAR image change detection.(2)Aiming at the problem that it is difficult to use the spatiotemporal information in the change detection process.A novel pointwise approach is proposed for change detection in bi-temporal SAR images using stereo-graph model.Due to the fact thatSAR image suffers from the speckle noise,a pointwise approach based on a set of characteristic points only,not on the whole pixels,seems to be more efficient.Moreover,the correlations of neighbourhood points which have different locations in bi-temporal SAR images should be studied to repress the speckle in change detection.Therefore the stereo-graph model,which extends the graph model to three-dimensional space,is designed to connect the local maximum pixels on bi-temporal SAR images and can be used to capture the multiple-span neighbourhood information from the edges.Furthermore,a specialized change measure function is presented to quantify the neighbourhood information from stereo-graph model,and thus a novel nondense difference image is generated.Finally,a traditional classification method is used to analyse the NDI into changed class and unchanged class.Experiments on real SAR images show that the proposed NDI can improve separability between changed and unchanged areas,and the final results possess high accuracy and strong noise immunity for change detection tasks with noise-contaminated SAR images.(3)To alleviate the problem of information loss in difference image generation,a pointwise approach using stereo-graph cuts(SGC)with spatio-temporal information is proposed for SAR image change detection.The SGC is based on stereo-graph,which is designed to connect the local maximum pixels on two SAR images and can be used to capture the spatio-temporal information.With the support of stereo-graph,a novel SGC energy function is presented to quantify the spatio-temporal information and implicitly measure the difference information between the SAR images.Therefore,the change detection results can be obtained by minimizing the energy function with graph cuts technology.Experimental results on real SAR datasets confirm the validity of the proposed approach in which SGC and spatio-temporal information offer great contributions on improving the robustness and accuracy of detection.(4)In order to make full use of the high-order neighborhood information in SAR change detection,this thesis proposes an unsupervised framework for SAR image change detection in which each pixel is taken as a vertex and the collection of pixels is represented by hyperedges in a hypergraph.Thus,the task of SAR image change detection is formulated as the problem of hypergraph matching and hypergraph partition.First,instead of using the K nearest neighbour rule,we propose a coupling neighbourhood based on the spatial-intensity constraint to gather the neighbours for each vertex.Then,hyperedges are constructed on the pixels and their coupling neighbours.The weight of hyperedge is computed via the sum of the patch-basedpairwise affinities within the hyperedge.Through modelling the two hypergraphs on the bi-temporal SAR images,not only the change level of vertices is described,but also the changes of local grouping and consistency within hyperedge are excavated.Thus,the difference image with a good separability can be obtained by matching each vertex and hyperedge between the two hypergraphs.Finally,a generalized hypergraph partition technique is employed to classify changed and unchanged areas in the generated difference image.Experimental results on real SAR datasets confirm the validity of the proposed framework in improving the robustness and accuracy of change detection.