Improving discrimination accuracy of pest-infested crabapples using Vis/NIR spectral morphological features

Abstract

The visible/near-infrared (Vis/NIR) spectroscopy technique is effective for fruit quality detection. The distinct spectral features can reflect the internal composition of fruits, while variations in external orientation may induce interference. Considering both external and internal factors, we improved the discrimination accuracy of pest-infested crabapples by compensating for variations in orientation and amplifying differences in spectral morphological features (SMFs). Firstly, spectral intensity variations caused by orientations and morphological differences caused by pest infestation were analyzed. Based on these differences, the global model was established to mitigate the external orientation influence. Subsequently, SMFs, derived from spectral peaks and troughs, were employed to amplify spectral features. Finally, with the supplementation using 1^st deviation, SMFs improved the discrimination performance of the partial least square–linear discriminant analysis (PLS-LDA) model for pest infestation, yielding results of sensitivity, specificity, and accuracy as 95.14%, 96.32%, and 95.94%, respectively. Overall, compensating for external orientation variations and exploiting internal spectral features enhanced the detection accuracy of pest infestation, providing valuable insights for internal defect discrimination based on Vis/NIR spectroscopy.