An Automatic Sparse Pruning Endmember Extraction Algorithm with a Combined Minimum Volume and Deviation Constraint

Abstract

Abstract:In this paper, an automatic sparse pruning endmember extraction algorithm with a combined minimum volume and deviation constraint (SPEEVD) is proposed. The proposed algorithm can adaptively determine the number of endmembers through a sparse pruning method and, at the same time, can weaken the noise interference by a minimum volume and deviation constraint. A non-negative matrix factorization solution based on the projection gradient is mathematically applied to solve the combined constrained optimization problem, which makes sure that the convergence is steady and robust. Experiments were carried out on both simulated data sets and real AVIRIS data sets. The experimental results indicate that the proposed method does not require a predetermined endmember number, but it still manifests an improvement in both the root-mean-square error (RMSE) and the endmember spectra, compared to the other state-of-the-art methods, most of which need an accurate pre-estimation of endmember number.

Sparse Unmixing for Hyperspectral Imagery via Comprehensive-Learning-Based Particle Swarm Optimization

“…With the combination of dimension division and two comprehensive learning strategies, the endmembers and abundances are updated alternately in two collaborative swarms. As the iterations progress, the endmember swarm updates its particles and determine the global best position gbest t E according to (5), ( 6), (8), and (9), and then gbest t E serves as a known variable for the calculation of the particles' fitness values of the abundance swarm. In turn, the particles of the abundance swarm are updated using (5), ( 6), (10), (12) and two comprehensive learning strategies.…”

“…However, during the data collection process, the limited spatial resolution of the imaging spectrometer and the complex surface conditions usually result in the instantaneous field of view containing more than one material. Therefore, a large number of mixed pixels exist in the HSIs and affect the accuracy of remote sensing applications severely at the pixel scale [8]. Spectral unmixing technology solves this problem by extracting characteristic spectra of each kind of typical material and their corresponding fractional proportions from the HSIs, which are called endmembers and abundances, respectively [9]- [11].…”

“…depends on the cross computation of all the particles' positions of two swarms according to(8) and(9). Then, 0 gbest E is regarded as the given endmember matrix to produce…”

Weighted Nonlocal Low-Rank Tensor Decomposition Method for Sparse Unmixing of Hyperspectral Images

“…It is important to extract the spectrum of each ground object from the mixed pixel and to obtain the corresponding abundance coefficients. That is, each mixed pixel is decomposed into products of different pure spectra (called endmembers) [6] and their corresponding proportions (called abundance coefficients) [7]. To solve the unmixing problem, two main basic models, the linear mixture model (LMM) [8]- [10] and nonlinear mixture model [11]- [14], are widely used.…”