Sensors (Sep 2024)

Nondestructive Detection of Corky Disease in Symptomless ‘Akizuki’ Pears via Raman Spectroscopy

  • Yue Yang,
  • Weizhi Yang,
  • Hanhan Zhang,
  • Jing Xu,
  • Xiu Jin,
  • Xiaodan Zhang,
  • Zhengfeng Ye,
  • Xiaomei Tang,
  • Lun Liu,
  • Wei Heng,
  • Bing Jia,
  • Li Liu

DOI
https://doi.org/10.3390/s24196324
Journal volume & issue
Vol. 24, no. 19
p. 6324

Abstract

Read online

‘Akizuki’ pear (Pyrus pyrifolia Nakai) corky disease is a physiological disease that strongly affects the fruit quality of ‘Akizuki’ pear and its economic value. In this study, Raman spectroscopy was employed to develop an early diagnosis model by integrating support vector machine (SVM), random forest (RF), gradient boosting decision tree (GBDT), extreme gradient boosting (XGBoost), and convolutional neural network (CNN) modeling techniques. The effects of various pretreatment methods and combinations of methods on modeling results were studied. The relative optimal index formula was utilized to identify the SG and SG+WT as the most effective preprocessing methods. Following the optimal preprocessing method, the performance of the majority of the models was markedly enhanced through the process of model reconditioning, among which XGBoost achieved 80% accuracy under SG+WT pretreatment, and F1 and kappa both performed best. The results show that RF, GBDT, and XGBoost are more sensitive to the pretreatment method, whereas SVM and CNN are more dependent on internal parameter tuning. The results of this study indicate that the early detection of Raman spectroscopy represents a novel approach for the nondestructive identification of asymptomatic ‘Akizuki’ pear corky disease, which is of paramount importance for the realization of large-scale detection across orchards.

Keywords