Research Of Mine Area Land Cover Information Extraction Based On Object-oriented Methord

With the rapid development of the mining industry, the problem of the mineecological environment are increasingly various. The application of remote sensingtechnique in mineral development order investigation and monitoring, which helpedthe management departments’ regulations of mining improved their administrationseffectively. But how to efficiently and accurately extract the thematic informationfrom the vast amounts of remote sensing data is remaining a difficult problem to dealwith. The traditional classification which base on pixel has many disadvantages suchas low classification precision, and serious "salt&pepper" phenomenon is difficult tomeet certain accuracy requirements. The extraction method of human-computerinteractive interpretation has higher classification precision, but it requires extensiveexperience and specialized knowledge, and there also has some weakness such as thecycle of the extraction method is too long, low efficiency for vast amounts of data. Ittakes the Li Wu mine in Jiu Long District as the object of study in this paper, and usesremote sensing data QuickBird trying to make based on object-oriented classificationmethod of mining area land cover information extraction research. The researchcontent and research results of this paper are as follows:(1)It systematically introduces the theory of the object-oriented classificationmethods. At first, overview of image segmentation technology is covered. Themulti-scale division technology is introduced emphatically. It expounds the concept ofmulti-scale segmentation and multi-scale segmentation parameter selection. And thenearest classification and fuzzy classification are introduced concise.(2)In the study area the experimental data is preprocessed. Dealing with the topographic map data is generated to DEM. The remote sensing image data is fused,shot corrected and cut out. Analysis of the image fusion method, it chooses theGram-Schmidt method for image fusion from the point of the qualitative.(3)The optimal partition scale is detailed discussion. And according to the ratioof the mean algorithm, an improved the optimal partition scale selection method isproposed. This method is applied in the study area data to validate by experiment.Through the experiment, three segmentation scales are choose to build objecthierarchy.(4)The object image information, geometry information and spectra textureinformation is made full use to construct fuzzy classification rules. And theadvantages and disadvantages of the fuzzy classifier and the nearest neighborclassifier taken together, two classifiers are combined to extract information of miningland cover.(5)It takes classification accuracy and kappa coefficient as evaluation index toevaluate the classification results which based on object-oriented classification andmaximum likelihood classification method. By comparisons, the object-orientedclassification method of mining area of land cover the accuracy of informationextraction is obviously better than the classification method based on pixel.Through the above findings, the paper considers that the results accuracy whichbased on the object-oriented classification is better than it which based on themaximum likelihood classification. The visual effect of object-orientedclassification’s result is also good.