Monitoring Technology Of Land Cover Change In Guilin Based On Remote Sensing Image

With the increasing maturity of remote sensing earth observation technology,the research on land coverage change monitoring has gradually become a research highlight in recent years.It can not only provide important reference for the rational control of natural resources,but also study important data on global environmental changes.In recent years,China has developed rapidly,and the land and resources monitoring and management situation is grim.On the basis of the increasing quality of remote sensing image data,the research of long-term land coverage change is supported by sufficient data.Therefore,this thesis is mainly based on the remote sensing images of Gongcheng Yao Autonomous County of Guilin City as a research area in 2006 ～ in 2018.Guilin has become a world-class tourist city because of its unique karst landform,which its research.First of all,this thesis studies the remote sensing image dehazing algorithm.Because Guilin is located in the Lingnan region,the climate is mild and humid throughout the whole year,which affects the quality of image acquisition to a certain extent.The quality of remote sensing images in Guilin will be severely affected during the rainy season.Although a large amount of image is obtained,the actual available images are few.After a lot of experimental analysis,this thesis proposes a remote sensing image dehazing algorithm combined with curvature filtering optimization,using Gaussian curvature filtering method to optimize the image,and obtaining the dehazing image by estimating the global atmospheric light value and transmittance.After comparing with the dark channel dehazing algorithm,neural network dehazing algorithm and other image reconstruction physical model algorithms,it is concluded that the proposed algorithm has better dehazing ability.Secondly,this thesis conducts an experimental analysis of the spatial domain edge detection,as prewwit operators by weighted average has better noise suppression performance compared to other edge detection operators.The watershed segmentation algorithm is improved,and a vector-based morphology method is proposed to calculate the gradient of multispectral images.The morphological gradient vectorization of the traditional watershed segmentation algorithm is quantized,and the filtering enhancement operation of the image is increased,then the image will help the differentiation of various objects.Later,the ground is marked by threshold marking method,arrange the gradient values in reverse order,and the local minimum are marked on the image.Objectively analyze the accuracy of the watershed segmentation algorithm and the improved watershed algorithm,the improved watershed segmentation algorithm is more accurate.Then,the Fuzzy C-means clustering algorithm is improved and a multi-feature fusion classification algorithm is proposed.The method mainly analyzes from the three perspectives of color feature,texture feature and shape feature,and conducts experimental analysis on remote sensing images in different scenes.In the remote sensing image land cover mapping section,a classification method based on random forest was used to achieve large-area land cover classification from 2006 to 2018,and the classification results were evaluated using kappa coefficient,overall accuracy and other indicators,and the classification accuracy met the requirements.Finally,this thesis analyzes the land cover classification map and uses Markov model to analyze the characteristics of land cover spatial pattern,which can quantitatively indicate the conversion between land cover categories.Using the land cover change dynamic model analysis method to analyze the land cover change situation of the selected study area in Guilin,then obtain the land cover change trend of the study area.