Research On Change Detection Method Of Multi-source Remote Sensing Images

With the development of science and technology and the changes of the times,multisource remote sensing image data has been continuously expanded and enriched,and multisource remote sensing image change detection algorithms have also become diverse.The research on the change detection algorithm of space-borne SAR and ground-based SAR remote sensing image is regarded as the focus of this article.Although they all have the characteristics of all-time and all-weather observations,they are used in flood disasters,mining area monitoring,and landslide monitoring due to different search cycles and different data acquisition difficulties.However,the scattering characteristics of ground objects will change before and after the flood disaster and they are also affected by coherent speckle noise.Seriously incoherent areas cannot be monitored for a long time and are seriously affected by speckle noise.The above problems also lead to problems such as high error rate,poor detail retention,discontinuous edges,and low accuracy in both the space-borne SAR and ground-based SAR image change detection results.Aiming at the above problems,this thesis mainly studies the change detection algorithm of space-borne SAR and ground-based SAR remote sensing image from the aspects of obtaining difference map and extracting change information of difference map in unsupervised change detection algorithm.In this thesis,the unsupervised change detection algorithms for space-borne SAR remote sensing images and ground-based SAR remote sensing images in multi-source remote sensing images are mainly studied as follows:(1)Because the single difference map is disturbed by noise in the change detection of flood disaster and before and after the flood disaster occurs,the scattering characteristics of ground objects in some areas are changed.Therefore,directly performing cluster analysis on the difference map will lead to an increase in the error rate of the change detection results and a decrease in the retention of edges and details.This thesis proposes a space-borne SAR image change detection algorithm based on the fusion difference map.This method used wavelet transform to fuse the improved relative entropy difference map and mean ratio difference map in the frequency domain.Then the Pearson correlation coefficient is used to classify the clustering results of the FLICM algorithm twice,which can provide a better initial label distribution for the subsequent ICM-MRF algorithm.Finally,the validity and reliability of the method are verified by the measured space-borne SAR data,and the multiscene image change detection and analysis of the Poyang Lake flood disaster area in June2020 are carried out.It has estimated the affected area before and after the flood and the spreading trend of the flood during this period,and has a certain guiding role in the assessment of the disaster situation and the management plan after the disaster.(2)Aiming at the problem that ground-based radar images are affected by noise and the detection results of changes in the decoherence area remain poor in detail and the detection results are low in accuracy,an unsupervised change detection method for ground-based radar decoherence areas is proposed.The method first used an improved logarithmic ratio operation to enhance the noise suppression effect of the difference map.Secondly,the nonsubsampled contourlet transform is fused with the relative entropy difference map,and then the NLM algorithm is used to further smooth and suppress noise.Finally,the improved FLICM algorithm is used to obtain a more suitable initial label distribution,and the ICM model is extended by the similarity weight between pixels constrained by spatial distance under non-local conditions and combined with Markov random field to obtain the final change detection result.The qualitative and quantitative analysis of the measured data from a mining area in Inner Mongolia by LSA ground-based radar shows that the method in this thesis has a certain improvement compared with the traditional ICM algorithm in terms of detail preservation,noise suppression and detection accuracy.