The following information was submitted:
Transactions: INTERNATIONAL JOURNAL of MATHEMATICAL MODELS AND METHODS IN APPLIED SCIENCES
Transactions ID Number: 19-672
Full Name: Mitsuhiro Tomosada
Position: Researcher
Age: ON
Sex: Male
Address: 2-11-1, Iwado KIta, Komae-shi, Tokyo
Country: JAPAN
Tel: +81-(0)3-3480-2111
Tel prefix:
Fax: +81-(0)3-5497-0318
E-mail address: tomosada@y7.dion.ne.jp
Other E-mails: tomosada@criepi.denken.or.jp
Title of the Paper: novel mixture model for mixed pixel classification of multispectral image data
Authors as they appear in the Paper: Mitsuhiro Tomosada, Hiroe Tsubaki
Email addresses of all the authors: tomosada@y7.dion.ne.jp,tsubaki@ism.ac.jp
Number of paper pages: 9
Abstract: We propose a novel mixture model for use in the mixed pixel classification (MPC) of a multispectral image such as remotely sensed multispectral image data and Magnetic Resonance Image (MRI). Although the MPC method utilizes a generic statistical model of mixture such as a linear mixture model or a finite mixture model, the proposed mixture model of a single pixel in this paper is established on the basis of the process of mixed pixel generation in a real multispectral image. The variance-covariance structure for a pixel vector is considerably different from the variance-covariance structure derived from existing mixture models. Furthermore, we present an MPC method using the generalized method of moments (GMM), which satisfies the proposed mixture model and estimates the mixing ratio for each component in a single pixel, and the expected value and variance-covariance matrix of the pixel vector for each component. First, the MPC method is applied to a simulated ima!
ge data. Estimated parameters close to actual values are obtained, and the simulated image data is found to be in agreement with the constructed mixture model since the evaluation function is close to the actual value. The proposed mixture model and MPC method are applied to real multispectral image data acquired by Enhanced Thematic Mapper Plus (ETM+) onboard Landsat-7satellite as one example of a multispectral image. As a result, it was found that pixels in an ETM+ image, for which the mixing ratio of one component is high, are consistent with pixels in an image which are assumed to have the same component by visual inspection.
Keywords: Generalized method of moment, Image processing, Mixed pixel classification, Mixture model, Multispectral image
EXTENSION of the file: .pdf
Special (Invited) Session: novel mixture model for mixed pixel classification of remotely sensed multispectral image
Organizer of the Session: 635-377
How Did you learn about congress:
IP ADDRESS: 111.103.8.15
