Object-Oriented Information Extraction Technology Of High Resolution Remote Sensing Image

Compared with the low or middle resolution image, the high resolution Remote Sensing image has the richer structure information and the texture information, and the traditional statistical classification technology based on pixels' spectrum can not obtain the ideal effect; it's produced thematic map contains lots of" pepper and salt " noises as a result of lose of the integrity, and the phenomena of "the foreign matter same spectrum" and "the same thing different spectrum" cannot be distinguished. Those are all the inevitable limitations of the traditional statistical classification algorithm based on pixels' spectrum.The idea of object-oriented is introduced into the information extraction technology from the high resolution image. This technology produces homogeneous image objects through multi-scale segmentation technology, and then carries out the information extraction by fuzzy classification using spectral features and shape features classification.Two typical regions of 470*400 sizes (urban and agricultural areas) are chose from Huairou County QuickBird images as study areas, and jobs be done as following:1. The multi-scale segmentation experiments are carried out in the city and the agricultural regions, different scales which are suitable for the different features are obtained. Take the city as the example: when the segmentation scale is 8, it is suitable to extract lawn, the crown and its shadow; When the segmentation scale is 20, it is suitable to extract the building information; when the segmentation scale is 75, to extract roads. To establishing a segmentation hierarchy after choosing idea segmentation scales. Establishing a segmentation hierarchy containing three levels in the city area, and the segmentation scales respectively is 8, 20, and 75; Establishing a segmentation hierarchy containing four levels in the agricultural region, and the segmentation scales respectively is 6, 20, 30, 50.2. The information extraction is carried out on the base of multi-scale segmentation. The object-oriented information extraction uses fuzzy classification method, which has two kinds of realization forms-the nearest neighbor and membership function. The nearest neighbor method is suitable for the extraction of information which has little diversity, such as garden, nursery, and so on in the agricultural region.Membership function is suitable for those whose features are different largely with other...