The Study On Unmixing Of Mixed Pixels Based On Hyperion Data

This paper focuses on application of Hyperion hyperspectral remote sensing data in mixed pixels unmixing technology, compares the traditional classification method and hyperspectral classification method of mixed pixels unmixing.It is a well-known fact that pixels in remote sensing image are typically mixed pixels due to both the limited spatial resolution of sensors and the heterogeneous surfaces of ground covers. Mixed pixels are a major source of inconvenience in conventional classification. There are two difficult points about mixed pixels unmixing: (1)Determine and pick-up the end members(2)solve the modle of mixed pixels unmixing.For the characteristics of Hyperion hyperspectral remote sensing data, at first the paper pretreated Hyperion image and provide the guarantee for further analytical and actual application image. With the premise of assuring pixel immixing model, wether good or bad it is, the choice of endmembers directly influence accuracy of pixel unmixing.For the choice problem of terminal component endmembers in the classification process of hyperspectral remote sensing, this paper study and analyze the source and choice method of terminal component endmembers spectrum. With two kinds original methods of withdraw terminal component endmember,this paper raise improvement PPI method of withdraw and make use of the combineness usefulness of the method of SAM to testify practicability and validity of PPI.For the unmixing of mixed pixels,the paper describes unrestricted linear mixing method and restricted linear mixing method of unmixing model,and give its unmixing methods available. With the hyperspectral remote sensing,the paper compares traditional classification with unmixing of mixed pixels ,compares unrestricted linear mixing method and restricted linear mixing method of unmixing model, and testify practicability and validity of unmixing of mixed pixels in hyperspectral classification.