Application of the combination method based on RF and LE in near infrared spectral modeling

Cotton-Net: efficient and accurate rapid detection of impurity content in machine-picked seed cotton using near-infrared spectroscopy

Widespread adoption of machine-picked cotton in China, the impurity content of seed cotton has increased significantly. This impurity content holds direct implications for the valuation of seed cotton and exerts a consequential influence on the ensuing quality of processed lint and textiles. Presently, the primary approach for assessing impurity content in seed cotton primarily depends on semi-automated testing instruments, exhibiting suboptimal detection efficiency and not well-suited for the impurity detection requirements during the purchase of seed cotton. To address this challenge, this study introduces a seed cotton near-infrared spectral (NIRS) data acquisition system, facilitating the rapid collection of seed cotton spectral data. Three pretreatment algorithms, namely SG (Savitzky-Golay convolutional smoothing), SNV (Standard Normal Variate Transformation), and Normalization, were applied to preprocess the seed cotton spectral data. Cotton-Net, a one-dimensional convolutional neural network aligned with the distinctive characteristics of the seed cotton spectral data, was developed in order to improve the prediction accuracy of seed cotton impurity content. Ablation experiments were performed, utilizing SELU, ReLU, and Sigmoid functions as activation functions. The experimental outcomes revealed that after normalization, employing SELU as the activation function led to the optimal performance of Cotton-Net, displaying a correlation coefficient of 0.9063 and an RMSE (Root Mean Square Error) of 0.0546. In the context of machine learning modeling, the LSSVM model, developed after Normalization and Random Frog algorithm processing, demonstrated superior performance, achieving a correlation coefficient of 0.8662 and an RMSE of 0.0622. In comparison, the correlation coefficient of Cotton-Net increased by 4.01%. This approach holds significant potential to underpin the subsequent development of rapid detection instruments targeting seed cotton impurities.

Wavelength selection method for near-infrared spectroscopy based on Max-Relevance Min-Redundancy

Near-infrared spectroscopy (NIRS) is a rapid, nondestructive analytical technique utilized in various fields. However, the NIR data, which consists of hundreds of dimensions, may exhibit considerable duplication in the spectrum information. This redundancy might impair modeling effectiveness. As a result, feature selection on the spectral data becomes critical. The Max-Relevance Min-Redundancy (mRMR) method stands out among the different feature selection techniques for dimensional reduction. The approach depends on mutual information (MI) between random variables as the basis for feature selection and is unaffected by modeling methods. However, it is necessary to clarify the benefits of the maximum correlation minimal redundancy algorithm in the context of near-infrared spectral feature selection, as well as its adaptability to various modeling methods. This research focuses on the NIR spectral dataset of maize germination rate, and the mRMR method is utilized to select spectral features. Based on the preceding foundation, we create models for Support Vector Regression, Gaussian Process Regression, Random Forest, and Neural Networks. The experimental findings demonstrate that, among the feature selection methods employed in this paper, the Max-Relevance Min-Redundancy algorithm outperforms others regarding the corn germination rate dataset.

Rapid discrimination of urine specific gravity using spectroscopy and a modified combination method based on SPA and spectral index

To achieve high-accuracy urine specific gravity discrimination and guide the design of four-waveband multispectral sensors. A modified combination strategy was attempted to be proposed based on the successive projections algorithm (SPA) and the spectral index (SI) in the present study. First, the SPA was used to select four spectral variables in the full spectra. Second, the four spectral variables were mathematically transformed by SI to obtain SI values. Then, SPA gradually fusions the SI values and establishes models to identify USG. The results showed that the SPA can screen out the four characteristic wavelengths related to the measured sample attributes. SIs can be used to improve the performance of constructed prediction models. The best model only involves four spectral variables and 1 SI value, with high accuracy (91.62%), sensitivity (0.9051), and specificity (0.9667). The results reveal that m-SPA-SI can effectively distinguish USG and provide design guidance for 4-wavelength multispectral sensors.

Improvement of the prediction of a visual apple ripeness index under seasonal variation by NIR spectral model correction

Apple ripeness assessment is essential to ensure its post-harvest commercial value, and the visible/near-infrared(NIR) spectral models that are effective in achieving this goal are prone to failure due to seasonal or instrumental factors. This study has proposed a visual ripeness index (VRPI) determined by parameters such as soluble solids, titratable acids, etc., which vary during the ripening period of the apple. The R and RMSE of the index prediction model based on the 2019 sample were 0.871 to 0.913 and 0.184 to 0.213 respectively. The model failed to predict the next two years of the sample, which was effectively addressed by model fusion and correction. For the 2020 and 2021 samples, the revised model improves R by 6.8% and 10.6% and reduces RMSE by 52.2% and 32.2% respectively. The results showed that the global model is adapted to the correction of the VRPI spectral prediction model under seasonal variation.