The Research Of Rice Variety Identification Method Based On Artificial Olfactory System

With the continuous development of planting industry, the number of ricevarieties is increasing, in addition to promoting the development of production, to alarge extent, there are the phenomenons of doping, and identical to each other andeven infringement in varieties. This not only poses serious obstacles for graincirculation area, but also damages fundamental economic interests of the country andthe people. However, the traditional physical and chemical identification methods ofrice quality are time-consuming,cost and there are other disadvantages in it, so it isimperative to explore a fast and efficient method for identification of rice varieties.Currently, the e-nose technology is developing rapidly, it’s widely used in the food,agricultural products quality inspection, not only does it get general information ofvolatile odors from samples which is fast, convenient, safe and so on, but also providesan effective technique for the detection and identification of varieties of rice method.Electronic nose is used for odor detection of several different varieties of paddyrice and rice, classification and identification on different varieties. Prove thefeasibility of electronic nose used for detection and identification of rice varieties,which will not only provide identification of rice varieties with a new detectionmethods, but also for the electronic nose used for food quality inspection and qualitycontrol in the production process provides a theoretical base and reference value.1、Five samples information representation methods currently used are used.Study results indicate that the five sample information representation methods areeffective, to a certain extent, they all reflect the differences in information amongdifferent rice varieties. The results of integral value method is better.2、Compare and analyze four kinds of sensor array optimization methods.Analysis shows: Wilkes minimum statistic array optimization is easier to achieve,identification result after optimizing is better, and to a large extent, it effectivelyeliminates redundant information, facilitates the subsequent pattern recognition. 3、Four kinds of model identification including principal components analysis,Fischer differential analysis,BP neural network and RBF Neural network are used forclassification and identification of different rice varieties. Results indicate that thedifferent rice varieties can be distinguished, especially RBF Neural network effectsbest.4、Preliminaryly discuss the identification of rice varieties based on nucleartechnology, methods which is combined of nuclear technology, and Fisherdiscriminant analysis and principal component analysis are used for the identificationand analysis of paddy rice and rice samples. Results show that pattern recognition offeature extraction by the kernel principal component analysis and nuclear Fischeranalysis effect much better than the others, and to a large extent, improve the accuracyof the identification results.