Sunflower Origin Identification Based on Multi-Source Information Fusion Technique of Kernel Extreme Learning Machine DOI Creative Commons

Limin Suo,

Hailong Liu,

Ni Jin

и другие.

Agronomy, Год журнала: 2024, Номер 14(6), С. 1320 - 1320

Опубликована: Июнь 18, 2024

This study constructs a model for the rapid identification of origins edible sunflower (Helianthus) using Kernel Extreme Learning Machine (KELM) with multi-source information fusion technology. Near-infrared spectroscopy (NIRS) and nuclear magnetic resonance (NMRS) were utilized to analyze 180 samples from Xinjiang, Heilongjiang, Inner Mongolia regions. Initially, models origin sunflowers NIR NMR data compared between two algorithms: (ELM) KELM, combined various spectral preprocessing methods. The experiment found that preprocessed standard normal variate (SNV) KELM algorithm was most accurate, achieving accuracies 98.7% in training set 97.2% test set. spin-echo non-local means (NLMs) second best, 98.4% 96.4% To further improve accuracy models, innovative developed based on layer feature NIRS NMRS. In model, optimal, F1 score 98.2% 98.18%, respectively, an improvement 1.0% over best single source model. four types feature-layer information-fusion established extraction algorithms, Competitive Adaptive Reweighted Sampling (CARS) Variable Importance Projection (VIP), joint simple merging strategies. CARS-KELM method be 100% both sets, 2.8% Identifying NMRS is demonstrated as feasible by results. single-spectrum achieved SNV preprocessing. combining suitable handling task identification. significantly improves recognition fast accurate sunflowers. research results provide new origin.

Язык: Английский

Sunflower Origin Identification Based on Multi-Source Information Fusion Technique of Kernel Extreme Learning Machine DOI Creative Commons

Limin Suo,

Hailong Liu,

Ni Jin

и другие.

Agronomy, Год журнала: 2024, Номер 14(6), С. 1320 - 1320

Опубликована: Июнь 18, 2024

This study constructs a model for the rapid identification of origins edible sunflower (Helianthus) using Kernel Extreme Learning Machine (KELM) with multi-source information fusion technology. Near-infrared spectroscopy (NIRS) and nuclear magnetic resonance (NMRS) were utilized to analyze 180 samples from Xinjiang, Heilongjiang, Inner Mongolia regions. Initially, models origin sunflowers NIR NMR data compared between two algorithms: (ELM) KELM, combined various spectral preprocessing methods. The experiment found that preprocessed standard normal variate (SNV) KELM algorithm was most accurate, achieving accuracies 98.7% in training set 97.2% test set. spin-echo non-local means (NLMs) second best, 98.4% 96.4% To further improve accuracy models, innovative developed based on layer feature NIRS NMRS. In model, optimal, F1 score 98.2% 98.18%, respectively, an improvement 1.0% over best single source model. four types feature-layer information-fusion established extraction algorithms, Competitive Adaptive Reweighted Sampling (CARS) Variable Importance Projection (VIP), joint simple merging strategies. CARS-KELM method be 100% both sets, 2.8% Identifying NMRS is demonstrated as feasible by results. single-spectrum achieved SNV preprocessing. combining suitable handling task identification. significantly improves recognition fast accurate sunflowers. research results provide new origin.

Язык: Английский

Процитировано

1