Research of Adaboost integrated ELM based apple grading

Near infrared spectroscopy (NIRS) is widely used in apple grading research. The correlation between spectral information and soluble solids content (SSC) can be established, which can be used to build predictive models during the use of NIR spectroscopy. Extreme Learning Machine (ELM) methods can be used to establish the apple’s classification model. However, the established prediction model has poor applicability and low classification accuracy. For improving the classification accuracy of the predictive model, the multivariate scattering correction (MSC) will be used for baseline correction of the original spectral data, and the genetic algorithm (GA) will be used to screen the characteristic wavelength for data preprocessing. Moreover, Adaboost will be used to integrate the ELM model to establish a strong prediction model. The experimental results show that the classification accuracy of ELM model is 64.66%. The classification accuracy of daboost-ELM is 81.20%. Compared with the single ELM model, the Adaboost integrated model has higher accuracy and the accuracy, and the stability of apple grading are also improved.