Practice session on hyperspectral imaging algorithms

Practice session on hyperspectral imaging algorithms “Endmember extraction, spectral unmixing and target detection using ENVI and Matlab” Technical School of Cáceres, University of Extremadura Room: Laboratorio C1 (Pabellón Servicios Comunes) Monday October 29 th , 2007. (Session 1: 15:45 – 16:30) (Session 2: 16:45 – 17:30) (Session 3: 17:45 – 18:30) Instructors: David Valencia, & Maciel Zortea, davaleco@unex.es mzortea@unige.it UNEX UNEX 1 The problem: linear spectral unmixing...................................................................................................................................2 2 Objectives of this section.........................................................................................................................................................2 3 Background................................................................................................................................................................................3 3.1 PPI 3 3.2 N-FINDR ..........................................................................................................................................................................3 3.3 ATGP .................................................................................................................................................................................3 4 Proposed activities ....................................................................................................................................................................4 4.1 Hyperspectral data

description........................................................................................................................................4 4.1.1 Simulated data: ..........................................................................................................................................................4 4.1.2 Real data:....................................................................................................................................................................4 4.2 ACTIVITY # 0: analysis of simulated data using ENVI ............................................................................................5 4.3 ACTIVITY # 1: analysis of the simulated data using MATLAB...............................................................................6 4.4 ACTIVITY # 2: analysis of the real data using MATLAB .........................................................................................8 5 References..................................................................................................................................................................................9 First HYPER-I-NET School on Hyperspectral Imaging Complejo Cultural San Francisco, Ronda San Francisco s/n, 10002, Cáceres, Spain October 29-31, 2007 Practice session on hyperspectral imaging algorithms Page 2 of 9 1 The problem: linear spectral unmixing Linear spectral unmixing is a commonly accepted approach to mixed-pixel classification in hyperspectral imagery. This approach involves two steps. First, to find spectrally unique signatures of pure ground components, usually known as endmembers, and, second, to express mixed pixels as linear combinations of endmember materials. The linear mixture model can be defined as follows. Let h(x,y) be the hyperspectral signature collected by the sensor at the pixel with spatial coordinates (x,y). This signature can be considered an N-dimensional vector, where N is the number of spectral bands; it can also be modelled as a linear combination of endmember vectors e i , using the following expression: 1 (, ) (, ) E ii i hxy xy e = =Φ ⋅ ∑ (1) where Фi(x,y) is a scalar value representing the fractional coverage of endmember vector e i at pixel h(x,y). Two constraints are usually imposed in the previous equation [2]. These are the abundance nonnegativity constraint (ANC) and abundance sum-to-one constraint (ASC), respectively defined as 1 ( , ) 0, for all 1 (, ) 1. i E i i x yiE xy = Φ≥ ≤≤ Φ= ∑ (2) The above, and more details, can be found in Plaza et al., (2004). 2 Objectives of this section • Provide key ideas on endmember extraction • Review of classical approaches • Practice on computer simulated and real hyperspectral data First HYPER-I-NET School on Hyperspectral Imaging Complejo Cultural San Francisco, Ronda San Francisco s/n, 10002, Cáceres, Spain October 29-31, 2007 Practice session on hyperspectral imaging algorithms Page 3 of 9 3 Background Let us briefly review some of the classical endmember extraction algorithms proposed in literature. This survey has been previously reported in Plaza et al., (2004, 2006). The three algorithms that will be presented have in common the fact that they search for a set of target pixels of interest in the input image. 1 3.1 PPI The Pixel Purity Index...

Website: hyperinet.multimediacampus.it | Filesize: -
No of Page(s): 9
Download Practice session on hyperspectral imaging algorithms ... - Hyper-I-Net.pdf