Nondestructive measurement of internal quality attributes of apple fruit by using NIR spectroscopy

Abstract

In this paper, a hybrid approach, which combines back propagation neural network (BPNN), generalized regression neural network (GRNN) and particle swarm optimization (PSO), is proposed to determine internal qualities in apples by using NIR diffuse reflectance spectra in the wavelength range of 400-1022 nm. The essence of the hybrid approach incorporates six phases. Firstly, the original spectral data should be submitted to Savitzky-Golay smoothing method to reduce noise. Secondly, using multiplicative scatter correction (MSC) on de-noised spectral data to modify additive and multiplicative effects. Thirdly, principal component analysis (PCA) is used to extract main features from the pretreated spectral data. Fourthly, obtaining forecasting results by using BPNN. Fifthly, obtaining forecasting results by using GRNN. Finally, these respective results are combined into the final forecasting results by using the principle of PSO. The hybrid model is examined by determining soluble solid content (SSC) and total acid content (TAC) of Green apples. Experimental results illustrate that the hybrid model shows great potential for internal quality control of apple fruits based on NIR spectroscopy.