The significant difference in economic and nutritional value between new and old cereals has led to a plethora of grain mix-ups on the market, and it is necessary to distinguish the old and new grains accurately and quickly. In this study, millet, which has high economic and nutritional value in cereals, was used as the research object. Terahertz time domain spectroscopy combined with chemometrics was used to detect its mixture. The experimental findings show that, compared to the partial least squares regression model, and the support vector regression models optimized by particle swarm and grid algorithms, the support vector regression model optimized by the coati optimization algorithm (COA) exhibits superior performance. In order to further improve the prediction accuracy of the support vector regression model, an improved COA was proposed. The quality and diversity of the initial population are improved by introducing chaotic mapping, tent mapping, and opposition based learning mechanisms, followed by the introduction of adaptive weights and Wright flight mechanisms, which enhance the flexibility and convergence speed of the algorithm, and finally, the introduction of the sine-cosine algorithm to improve the search range. The experiments demonstrate that a support vector regression model optimized by an enhanced coati’s algorithm provides superior quantitative analysis of millet mixtures as compared to other models tested. The optimized model achieves high prediction accuracy, with correlation coefficients reaching 0.9872 and 0.9742 for millet mixtures with gradient ratios of 10% and 2%, respectively. This research provides technical support and a reference for rapid, nondestructive quantitative detection of millet mixtures while opening up new possibilities for subsequent quantitative detection of other cereals.
Read full abstract