An Image Expansion Method Based On Shadow Removal Imagery In Mountainous Area

Landsat satellite series has the characteristics of long time in orbit,stable operation,and abundant remote sensing image data.It has broad research prospects,deep application level,and great potential.At present,for this kind of mid-resolution remote sensing image,the expert system classification method has a strong affinity.Through the application,review,and analysis of existing land use information extraction,tree species information extraction,Landsat imagery in-depth analysis,expert system introduction and improvement,etc.at home and abroad,the current interpretation of Landsat images by the traditional expert system is discovered.There are some shortages in the classification and classification.At present,the traditional expert system method has fragmented results of fragmentation,prominent salinization,and problemssuch as inefficient extraction and classification of tree types and other information.The image expansion method is based on the improvement of the traditional expert system.This method analyzes the insufficiency of the Landsat image classification by the traditional expert system method and processes the completely preprocessed Landsat image.The steps include:optimal band extraction and knowledge base Establish rules,expand images,superimpose sequentially,and eliminate leaks.The method can not only deeply explore the richness of information in each band,but also strengthen the spatial relationship between various features,and can make a certain degree of improvement to the inadequacies of traditional expert system methods.The verification and application of this method takes Changting County of Fujian Province as an example.After the basic pre-processed Changting County Landsat8 OLIimages,the brightness of the mountain shadows is recovered and the shadows of the mountains are restored,according to the established image inflation method flow.Classification and extraction of four types of forest lands,impervious layers,water bodies,and other land types in Changting County,as well as pine,bamboo,fir,broad-leaved trees,combined with visual interpretation and collection verification The plot coordinates were verified and compared with the results obtained by the traditional expert system method in ERDAS.The main conclusions of the study are as follows:1.The mountain shading extracted by the proposed shading detection model has an extraction accuracy of 99.06%for mountain shading and a Kappa coefficient of 0.98.By introducing the distance coefficient to evaluate the restoration of mountain shading brightness,it is found that the mountain shading brightness is restored.The average distance coefficient of the model is reduced by 96%.Compared with other methods,the final effect is effectively improved.2.The image expansion method effectively suppresses the fragmentation and salt and pepper phenomenon in the traditional expert classification.By exploring the spectral information of the Multispectral Landsat8 OLI,the accuracy of the final classification is effectively improved,and the accuracy of the four types of ground feature extraction by the image expansion method is improved.For 95.8,the Kappa coefficient was 0.94,which was 15.7%and 0.24 higher than that of the traditional expert classification.The extraction accuracy of the fourtree types was 85.83 and the Kappa coefficient was 0.82,which was 9.13%higher than that of the traditional expert classification.0.11.