XRF Spectral Feature Extraction and Classification Method Based on the EO-CNN Deep Learning Model

An XRF spectral feature extraction and classification method based on an EO-CNN model is proposed to address the complexities of soil spectral data and improve classification accuracy. The proposed model is applied to soil classification, achieving an overall accuracy of 95% for five distinct soil categories. The experimental results show robustness in handling imbalanced datasets, as evidenced by a 69% classification accuracy for the category representing only 1% of the dataset.