Visible and near-infrared hyperspectral imaging as an intelligent tool for parasite detection in sashimi
Sai Xu, Huazhong Lu, Changxiang Fan, Guangjun Qiu, Christopher Ference, Xin Liang, Jian Peng
Abstract
There are parasites found on sashimi which can cause a series of health problems to those who consume them. Because the parasites are too small to be visible to the naked eye, the labor and time-consuming use of a microscope is required for detection. This study proposes a visible and near-infrared (VIS/NIR) hyperspectral imaging method to quickly and intelligently detect parasites in sashimi. The research results show that the VIS/NIR spectrums for fish meat and parasite images were different at certain wavelengths. The ability of a probabilistic neural network (PNN) combined with multiple detection models was better than that of partial least squares regression (PLSR) combined with a single detection model for the true positive detection of parasites on sashimi. A synthesis between PNN and a combination of detection models, including Savitzky-Golay, standard normal variate, and first derivative pre-processing, is able to optimally detect parasites in sashimi. Using this strategy, the detection accuracy of the validation set for Anisakis nematodes on the top and bottom of a sliced piece of sashimi were 91.67% and 82.14%, respectively. Thus, VIS/NIR hyperspectral imaging allows for intelligent, accurate, efficient, and rapid detection of Anisakis nematodes on sashimi.