An Apple Grading Method Based on Single-Model Evidential Ordinal ELM Model

China is a large apple producing country, and apples are loved by the nation because of their high nutritional value. In order to meet the people’s demand for the high quality of fruits, it is of great importance to study the rapid non-destructive testing of the internal quality of apples. In this paper, the evidential ordinal extreme learning machine is applied to predict the grades of apples based on features extracted from the near-infrared spectra. As the evidential ordinal extreme learning machine model combines the extreme learning machine with DS evidence theory, the epistemic uncertainty in grade labels transferred from soluble solid content can be modelled by mass function. This paper takes Yantai Red Fuji apples as the research object, with features extracted from pre-processed near-infrared spectrum as inputs, and coding bits generated with mass function as outputs. Experiments show that the application of the evidential ordinal extreme learning machine can achieve a prediction accuracy of 88.64%, which is 20% higher than the prediction accuracy of the traditional extreme learning machine, and the model stability has also been improved substantially.