The Study On Unmixing Of Mixed Pixels And Application Based On HJ1-A Data

Hyperspectral is one of the most important technological breakthrough on remote sensing system in the-end of the 20th century, with the deepening of quantitative study, hyperspectral remote sensing technology, because of its higher spectral resolution ,it's being widely attented at home and abroad, And in resources, environment, urban, ecological and other fields has been widely used.Remote sensing technology in various fields show the superiority and the importance of increasing, but also a prominent issue that is facing the mixed pixelproblem. Mixed pixel problem, if not well resolved, will be follow-up larger error to the after application.The research around the HJ-1A hyperspectral data, the hyperspectral data for selecting effective bands, epairing and removallingbad bands line, atmospheric correction, geometric correction. Select the atmosphere dark pixel correction method, geometric correction using the same data source with the projection system, human-computer interactive, the error is controlled below 0.1. In the fully study decomposition theory of the Pixel components , to grasp characteristics of hyperspectral data based on the use of MNF transform-PPI-n-D divergence visual analysis to select and adjust the end member, after determining the data ,inputing to the Linear Spectral Mixture Model for pixel components decomposed component diagram of components around the property, and based on the results of unmixing the study area by land cover maps, aerial photos and then using high-resolution analysis and the results do very the accuracy . Classification results with the supervised classification and maximum likelihood comparison of the ISODATA unsupervised classification compared to prove that the decomposition of mixed pixels has been greatly improved classification accuracy and verify the feasibility of mixed-pixel decomposition. The research is carried out under funding of scientific and technological cooperation and exchange projects between China and Germany (2007DFB70200) and Shandong Province Natural Science Foundation (Y2008E10).