Research on nondestructive detection of pine nut quality based on terahertz imaging

Abstract

ObjectivePine nuts are of great nutritional and medicinal value, but they cannot avoid such defects as mildew and insect-eating during storage. Because of their hard shells, the internal quality detection of pine nuts is a major problem for industrial sorting. For this reason, it is of great significance to carry out rapid nondestructive detection of the internal quality of pine nuts. MethodsIn this paper, a rapid and nondestructive detection of pine nuts for mildew and plumpness based on terahertz transmission imaging technology was carried out. Firstly, the terahertz transmission images of pine nut samples were acquired and the terahertz spectral signals of four different regions of interest were extracted for analysis. Secondly, in order to reduce the interference of external environment on the acquisition of terahertz spectra, the terahertz spectra were pre-processedby several methods, such as ALS, AirPLS, BEADS and SNV + Detrending, and then three qualitative discriminant models, namely, Linear Discriminant Analysis (LDA), Support Vector Machine (SVM) and XGBoost integration learning, were established respectively to explore the optimal qualitative discriminant model for the detection of pine nut quality. Finally, the terahertz transmission image of pine nuts was subjected to image processing. The main feature gamut extraction combined with channel separation strategy were adopted. and then the automatic threshold segmentation algorithm was applied to perform binary threshold segmentation on the separated image, thus the plumpness of the pine nuts was calculated by calculating the ratio of the pixel points of the shell and the kernel. ResultsThe prediction set of BEADS + XGBoost model was established after data preprocessing with the optimal effect and the accuracy of 98.61%. The acquired terahertz images of pine nuts were extracted by the main feature gamut and the images of channel B were extracted by using channel separation. Finally, the automatic threshold segmentation of channel B was performed by using the maximum one-dimensional entropy, which can well realize the visual detection of the inner shell kernel of pine nuts. ConclusionTerahertz imaging technology can achieve rapid and nondestructive detection of pine mildew as well as pine nut plumpness. This study provides a new rapid and nondestructive effective method for pine nut quality detection, which can provide technical reference for other shelled nut quality detection and has significant practical value.