Integrating ultra weak luminescence properties and multi-scale permutation entropy algorithm to analyze freshness degree of wheat kernel

Abstract

With the improvement of agricultural productivity, the wheat yield in China is soaring, which extremely burdens the quality guarantee pressure both in circulation and storage. According to the incomplete statistics, we loss millions of tons wheat every year because of freshness degree, mould, insects intrusion and so on. Thus finding an effective method to detect wheat kernel’s freshness degree is of necessity for guiding scientific storage of wheat. Up to now, process of wheat kernel’s freshness degree analysis can be divided into sensory evaluation, biochemical detection and physical detection method. Due to the inevitable deficits existed among these traditional methods, a novel method to detect freshness degree of wheat kernel based on ultra weak luminescence(UWL) characteristics integrated with multi-scale permutation entropy(MPE) algorithm has been proposed. Through calculating the mean value, variance, standard deviation, and MPE of four year’s wheat kernel separately, then further resorting to back propagation neural network to make the classification. Compared with traditional analogy methods, the new method not only improves the correct rate from 82.5%–92.5%, but also saves running time nearly three times shorter than PE algorithm. Shown by the experiment results, the new method proposed above is an efficient and feasible way to analyze freshness degree of wheat kernel.