Abstract:
Solving the l_{1}-norm minimization problem or matrix basis pursuit is used as a sparsity-based spectral unmixing for hyperspectral imaging data. Kronecker least angle ...Show MoreMetadata
Abstract:
Solving the l_{1}-norm minimization problem or matrix basis pursuit is used as a sparsity-based spectral unmixing for hyperspectral imaging data. Kronecker least angle regression algorithm is used to solve this matrix basis pursuit using properties of Kronecker products. In this paper, we present the potential of using the Kronecker least angle regression algorithm dynamically to update the matrix basis pursuit problem solutions using the last available estimated signal. Instead of solving a new basis pursuit problem from scratch or starting from an initial estimate of zero at every iteration, we use the last available signal estimate as the starting point in a homotopy formulation. We use homotopy continuation for updating the l_{1}-norm unmixing problem, which requires a series of rank-one updates along the so-called homotopy path. We used the proposed algorithm to unmix synthetic and real AVIRIS hyperspectral data cubes. Compared to static KLARS, our novel dynamic KLARS algorithm reduced spectral unmixing times by 43% for synthetic hyperspectral imaging data and 74% for actual hyperspectral imaging data.
Date of Conference: 11-16 July 2021
Date Added to IEEE Xplore: 12 October 2021
ISBN Information: