Research On Freeze Thaw Landslide Detection Method Based On SAR Image Change

In the past 10 years,freeze thaw landslides of different scales have occurred in riparian mountain of Changbaishan vein in Jilin Province in late November,which has blocked roads repeatedly and seriously affected production and life of residents.With the rapid development of remote sensing technology,it is a common technique to real-time monitor the hazardous area of landslide on aspect of disaster investigation and evaluation with remote sensing images.However,the meteorological environment during the occurrence of freeze thaw landslide is complex,and the weather caused by cloud,fog,rain and snow can easily lead to the inability to obtain the effective optical remote sensing image of study area.Due to snow cover,the backscattering coefficients of various surface features in the SAR images before and after landslide have changed greatly.It is a challenging topic to accurately detect small-scale freeze thaw landslide on the SAR images with large-scale changes.This paper mainly studies an unsupervised change detection method for SAR image based on fusion of differential image feature,and successfully applies this method to the detection of freeze thaw landslide in the riparian mountain of Songhua River in HONGSHILAZI Town,Huadian city.The main research contents are as follows:(1)An unsupervised freeze thaw landslide detection method based on texture change of SAR image is proposed in this paper.The texture features of SAR image are extracted by gray level co-occurrence matrix,and the difference map is constructed by texture features.Combined with FLICM clustering algorithm,the landslide area is accurately extracted.The experimental results also show that the performance of FLICM clustering algorithm is better than FCM clustering method.(2)The author proposed a detection method for freeze thaw landslide based on Gabor feature fusion in SAR image.Firstly,the ENVI software is used to preprocess the sentinel1 data,and the decision tree classification is used to obtain the key research area.Then,the difference map is generated by the difference method,and the two-dimensional Gabor wavelet features of multi-scale and multi-direction are fused.Finally,the improved FLICM clustering algorithm is used to obtain the detection results.The experimental results show that it is a valid approach to accurately identify the area of freeze thaw landslide,and the improved FLICM clustering algorithm has a better effect in reducing the false alarm rate,improving the detection accuracy and kappa coefficient.