Optimized wavelength selection for eggplant seed vitality classification using information acquisition techniques.

Area of Science:

• Agricultural Science
• Spectroscopy
• Machine Learning

Background:

• Eggplant seed vigor is critical for germination and seedling quality.
• Current assessment methods can be inefficient or destructive.
• There is a need for advanced, non-invasive techniques.

Purpose of the Study:

• To evaluate eggplant seed viability using hyperspectral imaging.
• To develop and optimize feature selection and classification models.
• To establish an efficient and nondestructive method for seed quality assessment.

Main Methods:

• Collected hyperspectral data (395.24-1008.20 nm) from treated eggplant seeds.
• Applied data preprocessing techniques: MSC, SG, and SNV.
• Proposed an Enhanced Information Acquisition Optimization (EIAO) algorithm for feature selection, identifying 23 key wavelengths.
• Developed classification models using Extreme Learning Machine (ELM), Random Forest (RF), and Support Vector Machine (SVM).

Main Results:

• The MSC-EIAO-RF model achieved the highest classification accuracy of 91.45%.
• This performance was significantly better than the MSC-IAO model (82.41%).
• EIAO algorithm demonstrated superiority over conventional methods on UCI datasets, confirming robustness and generalizability.

Conclusions:

• Hyperspectral imaging combined with EIAO is a robust and effective method for nondestructive seed viability detection.
• This technology provides an intelligent and efficient solution for agricultural seed quality assessment.
• The findings support the advancement of precision agriculture techniques.