Scene Analysis And Feature Extraction Of High-Resolution Remote Sensing Images In Typical Landslide Area

Taiyuan Xishan Mining Area is an important large-scale mining area in Shanxi Province.The landslide hazard in the area will affect the normal production activities of the mining area.After the landslide hazard occurs,the identification of the landslide area is of great significance to the emergency rescue and damage assessment after the disaster.In recent years,high-resolution ground observation technology has been rapidly developed,and high-resolution remote sensing images have also been widely used in the fields of resource investigation,emergency rescue and disaster assessment.Because landslide hazard areas often have complex material composition and diverse surface coverage,higher resolution remote sensing images may not have higher accuracy in landslide information extraction.Therefore,it is necessary to analyze the landslide characteristics and map them to high-resolution remote sensing images to realize the extraction of single landslide features from high-resolution remote sensing images.The main research contents of this article are as follows:Analyze the typical landslide area from three levels of local landslide environment,landslide body and landslide sub-area.The local regional geoenvironmental layer pays attention to the topography and the damage of the landslide to other features.The landslide body layer pays attention to the type of landslide body,the geometric characteristics of the movement and the differences between different areas.The landslide sub-region layer pays attention to the material composition and structural characteristics of specific areas inside the landslide.The differences in spectral and spatial characteristics of the landslides in lowand medium-resolution remote sensing images and high-resolution remote sensing images are compared and analyzed.Based on the special spectral,color,morphological and texture features of landslide in high-resolution remote sensing image,the characteristics of different levels of field landslides are mapped to high-resolution remote sensing images.Taking the Pianqiaogou landslide in Xishan,Taiyuan as an example,the topography and surface coverage of the area where the landslide is located are analyzed in the local geographical environment layer,expounds the feature differences of the landslide and its surface features within a certain range,and quantitatively compares the spectral characteristics of the landslide and the surrounding forest land;in the basic feature layer of the landslide body describes the morphology and movement characteristics of the landslide,focusing on analyzing the texture,color and spatial location characteristics of the landslide body on the image;in the landslide sub-region feature layer,the landslide is divided by the comparison of field photos and UAV images,and the spectral,color,texture,geometry and context characteristics of different sub-regions of the landslide are analyzed.The 0.03m drone image was used to extract and analyze the landslide feature on the image from two aspects: texture feature and saliency detection.The subregions of the landslide body are divided by image segmentation based on texture features,and the stones in the landslide accumulation area are detected using the salient features,and the program design experiment is carried out in MATLAB.The texture features of the landslide are extracted through Gabor filtering analysis,and the landslide is segmented by K-means clustering based on color and spatial position features.The results show that Gabor texture features can,to a certain extent,divide the homogeneous regions of the landslide,and realize the extraction of landslide information;the brightness and distribution characteristics of landslide accumulation block stones are analyzed,and different distribution of block stones is extracted through visual significance detection combined with shape features.Experiments have proved that the brightness characteristics of landslide accumulation block stones have a good effect on the detection of block stones.