Remote Sensing Information Extraction Of Zhangjiajie Geomorphology And Classification Based On Texture Analysis

Due to the uniqueness of Zhangjiajie Geomorphology,domestic and foreign scholars have done a lot of research on the geomorphology,genetic mechanism,geological structure and development status of Zhangjiajie Geomorphology,there are few researches on extracting geomorphology remote sensing feature information based on remote sensing technology.With the improvement of spatial resolution of remote sensing images,texture feature has been widely used as a basic component to increase the classification accuracy of multispectral remote sensing.This paper presents a case study implemented in Wulingyuan,a municipal district in zhangjiajie,hunan province.The GF-1 visible light data and the Sentinel-1 SAR microwave data were selected,and the gray level co-occurrence matrix(GLCM) was used to extract and count the five types(peak forest,construction land,water body,vegetation,farmland) of terrains in the two types of remote sensing images.Analyze the texture features of Zhangjiajie Geomorphology(Peak forest),and select the texture feature band and the corresponding multi-spectral(GF-1) band combination for classification research.The maximum likelihood method and obfuscation matrix are used to obtain the classification results and accuracy.Finally,analyze and discuss the classification accuracy.The results show:(1)Remote sensing image types suitable for analysis of Zhangjiajie geomorphology texture feature: By comparing and analyzing two remote sensing images(GF-1 and SAR),the SAR image contains better peak forest texture information,and among the 8 texture features,the Contrast difference is the most obvious,SAR(3.7573)> GF-1(1.1626) and the difference is 2.5947.The contrast feature is directly related to the sharpness of the image,reflecting the depth of the texture groove.This fully shows that the use of SAR images can better reveal the unique spatial texture structure information of the Zhangjiajie peak forest image,and also provide reference for date selection of remote sensing work.in Zhangjiajie geomorphology(2)The difference in texture between Zhangjiajie geomorphology and other land types: Zhangjiajie geomorphology highlights the shadow of the peak forest on the GF-1 image,which is similar to the water in characterization,and is easily interfered by water and vegetation.The texture features are relatively simple,darker tones and distinct from other land types.In the SAR image,the Zhangjiajie geomorphology shows a more prominent geomorphology structure,the texture features are more complete and less disturbed by vegetation.The texture features of the peak forest are consistent locally and overall,with strong texture directionality.(3)The combination of texture features applied to classification: by analyzing the difference of texture features of different land types,increasing the feature band to improve the classification accuracy,and selecting the texture feature values with greater degree of classification,the best classification of different land types in Wulingyuan District is obtained.The texture feature bands are combined into mean,variance,dissimilarity,entropy,and correlation..(4)The effect of texture on classification accuracy: The confusion matrix is used to evaluate the classification accuracy of the multispectral image(GF-1) before and after the texture is added.At 2 m resolution,the addition of texture features makes the overall accuracy of classification(OA) increased by 1.87%,Kappa increased by 0.0459,and 8 m resolution increased OA by 0.76% and kappa coefficient by 0.026.It can be seen from the accuracy of the classification results that the classification accuracy is better than the spectral classification based on the spectral and texture features,which indicates that the combination of texture features and spectral information can improve the classification accuracy of remote sensing images.(5)The influence of texture on the classification accuracy of different features: Based on the classification accuracy results,the changes of classification accuracy of different land types in the study area before and after the combination of texture features are analyzed.,Under the spatial resolution of 2 m,the classification accuracy of the five land types changed after adding the texture features.The peak forest,construction land,vegetation and farmland increased by 4.77%,2.66%,2.48%,2.90%,respectively,the classification accuracy of water is reduced by 0.6%.At the spatial resolution of 8 m,the five land classifications were increased by 4.38%,3.19%,0.58%,1.34% and 4.22% respectively.The results show that the addition of texture features is beneficial to improve the accuracy of land types,but its impact on different land types is different.