Report

Jonatan Gefen 28/11/2012 Outline Introduction to classification Whole Pixel Subpixel Classification Linear Unmixing Matched Filtering (partial unmixing) More Classification techniques Image Classification Spatial Classification Spectral Classification Image Classification Spatial Image classification: Based on the structures in the image (clear edges) Based on neighbor pixels Depends on the spatial resolution Can be done manually Image Classification Spectral Image classification: Increase of information per pixel Increase of dimensionality Can’t be done manually (but can be done Automatically) Based on spectral sig. Based on single pixel Spectral Classification Whole Pixel Sub-Pixel Others (advanced) Supervised / Unsupervised Based on known a priori through a combination of fieldwork, map analysis, and personal experience On-screen selection of polygonal training data (ROI), and/or On-screen seeding of training The seed program begins at a single x, y location Expands as long as it finds pixels with spectral similar to the original seed pixel This is a very effective way of collecting homogeneous training information From spectral library of field measurements Whole Pixel classification Assumes that each pixel contains single material and noise Tries to determine if a Target is in the pixel Whole Pixel classification Euclidean Distance SAM Spectral Feature Fitting Sub-pixel Tries to measure the abundance of the Target in the pixel Assumes that a pixel can represent more than one material Sub-pixel Linear Unmixing Filter Match Spectral classification Definitions: Target Endmember Infeasibility Linear Unmixing A model assumption that each pixel is a Linear-Combination of materials = = ∗ + – is the pixel value at band – spectral value of the endmember – the abundance factor of the endmember – noise Linear Unmixing Linear Unmixing is trying to solve linear equations to find possible endmembers and their fraction of the pixel. – the number of bands General Linear Unmixing Minimizing: , = − 2 Find Least min square. − − L1 Unmixing Assumes that all the elements are non negative. Minimizing: 1- called regulator Using NMF (Nonnegative matrix factorization) NMF(original form) , = − 2 Algorithm: = , = , = 1 = .∗ ./( + 10−9 ) = .∗ ./( + 10−9 ) (in our case already known) Match Filter(Partial Unmixing) This technique is used to find specific Targets in the image only user chosen targets are mapped. Matched Filtering “filters” the input image for good matches to the chosen target spectrum The technique is best used on rare Targets in the image. Match Filter(Partial Unmixing) Likelihood Ratio Using a threshold to decide if signal is present at the pixel. Match Filter(Partial Unmixing) The Matched Filter result calculation: The T(x) will hold the MF value of the endmember at pixel x if > 0 the endmember present. MNF (Minimum Noise Fraction) Λ is a diagonal matrix containing the eigenvalues corresponding to V MNF: is the covariance matrix of the signal (generally taken to be the covariance matrix of the image) is the covariance matrix of the noise (can be estimated using various procedures) Match Filter(Partial Unmixing) Mixture-Tuned Matched Filtering − matched filter vector - MNF Covariance matrix − the target vector in MNF space Match Filter(Partial Unmixing) − infeasibility value − the interpolated vector of eigenvalues − the target vector component - the MNF spectra for pixel After filter result More techniques Non-linear mixing Linear unmixing Non Linear unmixing Sub-Pixel Summery Can allow search of item that is a very small part of a given pixel Can give data about abundance of Targets Issues: Highly dependent on the contrast of the target to the background of the pixel One potential problem with Matched Filtering is that it is possible to end up with false positive results More techniques Spatial-spectral classification References N. Keshava - “A Survey of Spectral Unmixing Algorithms” P. Shippert, “Introduction to Hyperspectral Image Analysis” , Earth Science Applications Specialist Research Systems, Inc. Uttam Kumar, Norman Kerle , and Ramachandra T V – “Constrained Linear Spectral Unmixing Technique for Regional Land Cover Mapping Using MODIS Data” Yuliya Tarabalka, Jón Atli Benediktsson , Jocelyn Chanussot, James C. Tilton – “Hyperspectral Data Classiﬁcation Using Spectral-Spatial Approaches” Jacob T. Mundt, David R. Streutker, Nancy F. Glenn – “PARTIAL UNMIXING OF HYPERSPECTRAL IMAGERY: THEORY AND METHODS “ B. Ball, A. Brooks, A. Langville - Nonnegative matrix factorization Z. Guo, T. Wittman and S. Osher - L1 Unmixing and its Application to Hyperspectral Image Enhancement