Pear nondestructive testing method based on neural network and near infrared spectrum

Abstract

The invention discloses a pear nondestructive testing method based on a neural network and a near infrared spectrum. The method comprises the following steps: firstly obtaining the most representative near infrared spectrum for each sample in an actual operation process of a sorting line, and constructing an initial data set; carrying out pretreatment; then performing feature selection on the whole spectral wavelength interval by using an interval selection and wavelength selection combination mode, and constructing a feature data set according to a selection result; dividing for multiple times according to a K-fold inspection method to obtain multiple groups of training and testing data sets, wherein the input characteristics of the training and testing data sets are spectral data processed in the characteristic data set, and the labels area sugar degree and whether mould core is contained or not; and constructing two neural network models for predicating the sugar degree and predicating that whether a model with the mould core is used for subsequent practical detection. The sugar degree and the mould core are detected with high accuracy, a powerful basis is provided for quality screening of a sorting line, and great convenience is provided for selling of agricultural products.

Description

technical field [0001] The invention belongs to the technical field of intelligent sorting of agricultural products, and in particular relates to a nondestructive detection method for pears based on neural network and near-infrared spectrum. Background technique [0002] China is a large agricultural country, and the quality of agricultural products directly affects the taste of food, which in turn affects exports and domestic sales in China. Therefore, the quality inspection and intelligent sorting of agricultural products has always been an important research direction for agricultural practitioners and scientific researchers. For example, in the quality inspection of pears, the sugar content, hardness and water content of pears are all important chemical indicators to measure the quality. Traditional detection methods, such as sampling detection, have the problems of high detection cost, long detection time, inaccurate detection results, and the need to damage samples. ...

AI Technical Summary

Problems solved by technology

Due to the nonlinear influence of the sample state, concentration, angle and other factors on the spectrum, the conventional linear data analysis method cannot be used in the actual industrial assembly line

Images