| The mining industry economy has already become a major pillar industry in our country, but there are various problems in it. The remote sensing technique which has successfully applied in investigation and monitoring of mining development order, which helped the management departments'regulations of mining improved their administrations effectively.But how to efficiently and accurately extracting the thematic information from the multi-resolution, vast amounts of remote sensing data is remaining a difficult problem to deal with. The traditional classification which base on pixel has many disadvantages such as low classification precision, and serious "pepper" phenomenon is difficult to meet certain accuracy requirements. The extraction method of human-computer interactive interpretation has higher classification precision, but it requires extensive experience and specialized knowledge, and there also has some weakness such as the cycle of the extraction method is too long , low efficiency for vast amounts of data.It takes part of open-pit mining in Beijing Miyun District as the object of study in this paper, and tries to carry on the research on the information extraction method of the mining development area for GeoEye-1, RapdiEye, Landsat-5 three kinds of satellite image. It based mainly on object-oriented classification and maximum likelihood classification, and compare with the results of two kinds of classification result precision, achieved the following results:1)It systematically summarized the principles of the traditional classification and the object-oriented classification methods. It elaborates in detail the technical flow of the object-oriented classification, and research on image segmentation, the information extraction and other key technologies.2)It compares different fusion methods for high-resolution images-GeoEye-1, and evaluates the results of different fusion methods on 4 evaluation indexes: Information entropy, luminance information, correlation coefficient, relative deviation and so on, and considers that the result of GeoEye-1 data of study area which based on IHS fusion method is better.3)It takes the medium-resolution image—RapidEye data as the example to study the optimal image segmentation technology. Through comparing with multi-scale segmented image objects, the paper ultimately confirms three segmented scales:20,30,40 to construct image object hierarchy for extracting the information of the mining development area in the classification system.4)It constructs classification rules of the mining development area's types on the image object hierarchy. The object-oriented classification is full use of the image's spectrum, space, texture, context and other information, covering various features into the same type of the information of mining development area, and through combining the two classifiers -standard nearest neighbor and fuzzy membership function to extract information.5)It takes classification accuracy and kappa coefficient as evaluation index to evaluate the classification results which based on object-oriented classification and maximum likelihood classification method. It takes the results of visual interpretation which has through field validation as a reference to verify the object-oriented classification results's location and its area.Through the above findings, the paper considers that the results accuracy which based on the object-oriented classification is better than it which based on the maximum likelihood classification.The visual effect of object-oriented classification's result is also good. The method of object-oriented classification is able to play an assistant role to the way of man-machine interactive classification in the investigation and monitoring of mining development order, improving the indoor- interpretation accuracy while reducing the field validation pressure in the field, helpful to investigating the information of mining development area timely and effectively.
|
PDF Full Text Request