Identification Of Lakeshore Features Based On Data Mining Technology And Analysis Of Landscape Pattern

With the development of society and the increase of population,lakeshore resources need to be exploited urgently,but at the same time,the maintenance of lake life and health and the construction of aquatic ecological civilization can not be ignored,and the contradiction between them is constantly highlighted.Identification of different types of land features and analysis of landscape pattern in lakeshore zone can provide certain basis for river and lake management,alleviate the problems of imperfect lake management laws and regulations and frequent illegal reclamation of lakes.In this paper,GF-1 image and Landsat image are used as data sources.The feature optimization algorithm is added to the data mining algorithm to establish a new data mining classification method suitable for high-resolution images of domestic satellites.The applicability of the algorithm based on Landsat image is used.Based on the research,an index system suitable for analyzing the landscape pattern change around the lakeshore zone is established for the Taihu Lake shoreline.Through the analysis of the landscape pattern change,suggestions for the optimization of the landscape pattern of the lakeshore zone are proposed.The paper mainly achieved the following results:(1)Research on Extraction and feature optimization algorithm of terrain feature analysisIn this paper,watershed segmentation algorithm is used to segment the GF-1 image.On this basis,spectral,texture,geometric and exponential features of seven types of land features,such as lake water and vegetation,are extracted.A total of 76 features are obtained.The characteristics of four bands of the GF-1 image in seven types of land features are described respectively.On this basis,ReliefF algorithm is improved according to the characteristics of remote sensing image classification,and a feature selection algorithm combining ReliefF algorithm and J-M distance algorithm is proposed.(2)Method for identifying ground object types in the shore of Taihu Lake based on data miningIn this paper,the data mining of remote sensing images is accomplished by using the feature selection algorithm and the C4.5 decision tree algorithm.The classification accuracy of lakeshore features reaches 90.85%.The applicability of this classification method in Landsat image is analyzed.The experimental results show that the classification accuracy reaches 85.68%,which indicates that this method has better applicability in Landsat image.(3)Analysis of the change of landscape pattern in the shore zone of Taihu LakeIn this paper,the feature selection algorithm is used to effectively remove irrelevant and redundant features,and to ensure the accuracy and efficiency of the subsequent mining of land feature classification rules.The data mining of remote sensing images is completed by combining C4.5 decision tree algorithm.The classification accuracy of land features in lakeshore zone reaches 90.85%.The applicability of this classification method in Landsat image is analyzed.The experimental results show that the classification accuracy reaches 85.68%,which indicates that this method has better applicability in Landsat image.