Study On Surface Features Information Extraction From The Mining Area Using Object-oriented Classification Method

Along with the development of computer and space technologies, remote sensing hasbeen used to dynamically monitor the mining condition of surface coal mines and itssurrounding environment changes, with advantages of macro scale, periodicity, objectivity,and multilayer variation in space and time, which is important in rational exploitation andenvironment protection. Now remote sensing image data is more and more abundant, highresolution remote sensing images have more geometrical characteristics and textureinformation compared with the middle and low resolution remote sensing image. Thetraditional technology to extract information from remote sensing image is based on theanalysis of spectral information on each pixel, rarely use the shape of the terrain, texture andstructure and other space information, but object-oriented information extraction technologycan be a good solution to extract surface features information through the characteristics ofhigh spatial resolution images.This paper takes the Pingshuo Surface Coal Mine o in Shanxi Province as an example toextract information of the open mining area development situation using remote sensing.Remote sensing image of the CCD sensor and the high resolution remote sensing image ofHR sensor of China-Brazil Earth Resources Satellite (CBERS-02B) were used as data sources.The specific work include geometric correction, subset and other preprocessing of the remotesensing image. Several image fusion methods were applied on CCD image and HR imagethrough the IHS, Brovey, and Gram-Schmidt. A comparative analysis was conducted forabove-motioned fusion methods through the average and correlation coefficient, theGram-Schmidt fused image was chosen for further treatment. Then multi-scale segmentationand partitions level were created by using object-oriented classification software eCognition. Through comparison of different segmentation scales the segmentation scales for road,houses, and mines were determined as 45, 90, 190 respectively Spectral and specialcharacteristics of the specific surface features were then used to classify the image intovegetation, road construction, mining area etc six classes through the fuzzy classification ofthe membership function classification method. In the end, the accuracy assessment for theclassification results was conducted using Producer accuracy, User’s accuracy, Overallaccuracy, and Kappa coefficient, the results showed that the Overall accuracy is 88.03% withkappa coefficient of 0.88. The experimental results show that the methods adopted in thispaper is good to extract surface features information on mining areas. Research methods andremote sensing information extraction technology can well satisfy the survey demand of themining change.