![Sample our Physical Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110819/TOC-Subject-Sample-2021-Physical-Sciences.jpg"})
Abstract
The quality of milk is largely determined by the protein content. The feasibility of predicting the protein content of milk by hyperspectral image has attracted more attentions from researchers for minor detection cost and high efficiency. In this paper, a prediction modeling method based on improved sparrow search algorithm (SSA) and optimized back propagation (BP) neural network is proposed, in which sine chaotic map is introduced to initialize the population position to improve the optimization performance of SSA. In the experiment, hyperspectral images of each kind of milk were collected by visible/near infrared hyperspectral imaging system to acquire hyperspectral data, then the spectral data were pretreated by Savitzky–Golay smoothing, and the competitive adaptive reweighted sampling combined with successive projections algorithm to select 13 characteristic bands. Subsequently, the spectral data corresponding to the characteristic bands are used as the input of back propagation neural network, optimized by the improved sparrow search algorithm for the initial weight and threshold of BP neural network, to establish three prediction models(BP model, the BP model based on SSA optimization and the BP model based on improved SSA optimization).Experimental results demonstrate that the BP model based on improved SSA optimization has better fitting ability and higher prediction accuracy for milk protein content. This research provides algorithm support and theoretical basis for the rapid nondestructive detection of milk protein content based on BP neural network.
Disclosure statement
No potential conflict of interest was reported by the author(s).