Kernel-Based Nonlinear Spectral Unmixing with Dictionary Pruning

Zeng, Li; Chen, Jie; Rahardja, Susanto

doi:10.3390/rs11050529

Cited by 9 publications

(2 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…11 Kernelbased unmixing methods consider interactions among endmembers, 12 and the nonlinear unmixing issue is transformed into a linear unmixing problem by mapping endmembers to a high-dimensional feature space. 13,14 For the issue of identifying sources of water pollution, the turbidity of surface water changes within a specific range due to its exposure to an open environment. In the previous study, we found that the turbidity of surface water influences the spectral mixing mechanism.…”

Section: ■ Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Identification of Multiple Surface Water Contamination Sources Based on UV–Vis Spectral Unmixing with Turbidity Adaptiveness

Li,

Shao,

Wei

et al. 2024

ACS EST Water

View full text Add to dashboard Cite

Surface water contamination incidents are typically caused by an enterprise's excessive or clandestine illegal discharge. Accurate and rapid identification of pollution sources is crucial for timely pollution control and the enforcement of penalties for the responsible enterprises. At present, the identification of water pollution sources based on UV−Vis spectroscopy mostly relies on pattern recognition. However, this method is suitable only for identifying a single pollution source and cannot differentiate mixtures of multiple pollution sources. To solve this problem, an enhanced spectral unmixing method is proposed in this paper to identify multiple sources of contamination in surface water. In this method, the spectra of multiple pollution sources are used as endmembers to unmix the spectrum of the polluted surface water. The proportions (abundances) of each pollution source are then calculated to identify the pollution sources and determine the types of pollution discharge enterprises. Generally, the linear unmixing model is suitable for low-turbidity surface water, while the nonlinear unmixing model is suitable for high-turbidity surface water. A single linear or nonlinear unmixing model is not adaptive to turbidity variations in polluted surface water. Therefore, an unsupervised threshold SK-Hype hybrid unmixing algorithm is proposed in this paper. A weighted nonlinear fluctuation of endmembers is introduced to the linear function to characterize the turbidity change of surface water, and its weight is adaptively determined by the mixing mechanism of the target mixture. In this paper, experiments were conducted to unmix turbid polluted water and polluted surface water. The experimental results demonstrate that the algorithm presented in this paper exhibits higher accuracy in identifying sources of surface water pollution with turbidity adaptiveness.

show abstract

Section: ■ Introductionmentioning

confidence: 99%

“…For the nonlinear mixing model, several spectral unmixing methods based on the kernel function are currently proposed . Kernel-based unmixing methods consider interactions among endmembers, and the nonlinear unmixing issue is transformed into a linear unmixing problem by mapping endmembers to a high-dimensional feature space. , …”

Section: Introductionmentioning

confidence: 99%