Abstract
The early symptoms of cork spot disorder in 'Akizuki' pear (Pyrus pyrifolia Nakai) are challenging to distinguish from those in healthy fruits, hindering early identification in production. In this study, samples of cork-browned 'Akizuki' pears, asymptomatic fruits and healthy fruits were examined to determine the content of relevant mineral elements. A micro near-infrared spectrometer collected spectral information, and various pretreatment methods were applied to the near-infrared spectral data. Support vector machine (SVM) modelling using the original data achieved the highest overall recognition accuracy of 84.65% and an F1 value of 84.06%. For identifying fruits without cork spot disease, Autokeras modelled data processed with the SG method, achieving the best accuracy of 90%. These findings establish a reliable basis for the early identification and diagnosis of cork spot disorder in 'Akizuki' pear, enhancing pear production management.