Label-free detection of maize kernels aging based on Raman hyperspcectral imaging techinique

The feasibility of Raman hperspectral imaging technique was explored to detect maize kernels aging during storage to eliminate its negative effects on maize sowing and storage. Both TTC test and germination test were employed to evaluate the viability of maize kernels and the anlysis based on pixel-level and object-level were conducted. Different variable selection algorithms were used for screening of key features related with viability and three modeling methods were performed to classify maize kernels viability. In addition, Whale Optimization Algorithm(WOA) optimization algorithm was brought in to improve the accuracy of viability classification. The results showed that object-level method was more suitable for the classification of maize kernels viability. The fusion SVM model optimized by WOA coupled with CARS and VCPA-IRIV algorithm achieved the satifactory performance. In general, Raman hperspectral imaging techinique could be used as a poweful alternative for the nondestructive detection of maize kernels aging.