Application Of Object-oriented Combined SVM In Information Extraction Of Open-pit Mine

In recent years,with the continuous development of remote sensing technology,high resolution remote sensing image with its abundant information,and the characteristics of high resolution advantages,has become the main means of mine environment monitoring.Traditional image classification methods have limitations,it is difficult to apply to high-resolution images.Therefore,the object-oriented classification method for high-resolution images has arisen.This method can make full use of space and texture information of the image.It can improve the classification accuracy,and can quickly and effectively extract the information of land use change in mining area.For open-pit mine development and environmental governance provides effective theoretical basis and technical support.It has important theoretical and practical significance.we based on multi-scale segmentation object oriented combine with support vector machine(SVM)classification method,used the GF-1data,took the fifth Minefield of Jiangcang diggings of Muli coalfield in Qilian Mountain,Qinghai province as an example.Then extracted the information of mining area and got a classification map.The classification results are compared with the traditional pixel-based supervised classification results.(1)Make data preprocessing for the study area of high resolution remote sensing image.Through the methods of image fusion,radiation enhanced to improve the quality of the image and interpretation.Provides the high quality for the experimental research on the basis of the data;(2)The fractal net evolution approach(FNEA)algorithm is applied to multiscale segmentation.Establish a network layer division level.According to different object objects to select different object characteristics,establish classification rules;(3)The method of object-oriented combined with support vector machine(SVM)was used to extract and classify the object information in the experimental area.Then,adopt the method of confusion matrix combined with the field survey to evaluate precision.The traditional pixel-based maximum-likelihood classification and object-oriented classification results are compared and analyzed;(4)Based on the research results,the statistical analysis of land use and utilization information in the second phase of 2009 and 2013 in Jiangcang No.5 opencast mine is carried out.The study concluded: Object-oriented combined with support vector machine(SVM)classification accuracy of 88.45%,Kappa coefficient of 0.8643.Object-oriented classification results precision is higher than the maximum likelihood classification accuracy.Two data contrast display: Development land area accounted for 16.60%,the affected area accounted for 34.11%,serious environmental damage.On the whole,classification results can meet the actual production needs,the classification results can provide technical support for environmental management and ecological restoration in the mining area.Achieve the research purpose.