A Feature Extraction Method of Intelligent Olfactory Atlas Based on Single Item Discriminant Method to Represent the Difference of Honey

Abstract

A feature extraction method based on the single-item discriminant method to characterize the difference of honey in the intelligent olfactory spectrum, which is characterized in that all feature points are pattern-recognized one by one, and the difference in discrimination accuracy is compared when each signal point is used as a single feature. The recognition method is embedded in the feature selection. By combining the discriminant method, the ability of each signal point to predict the sample can be obtained. This method is more dependent on the selected pattern recognition method, and the selection result will change with the discriminant method. Single vector selection The difference in the accuracy rate among different sensors is small, while the difference between the information points at different detection times in the same sensor is relatively large, but the change trend of the Bayesian discrimination accuracy rate at different time points in different sensors is consistent with the change trend of the variance ratio, that is, the detection The initial signal discrimination accuracy rate is relatively high, concentrated in the first 30s detection time of each sensor, but the effect in the later detection period is poor, and the signal points with a discrimination accuracy rate greater than 60% are selected as feature points, a total of 598 feature points, of which rapeseed honey 20 / 23, linden honey 13 / 17, acacia honey 34 / 39.

Description

technical field [0001] The present application relates to a feature extraction method of an intelligent olfactory spectrum for characterizing honey differences based on a single-item discriminant method. Background technique [0002] my country's honey production ranks first in the world. In recent years, the output has maintained a rapid growth trend, increasing from 252,000 tons in 2001 to 402,000 tons in 2009, accounting for more than 30% of the world's total output from nearly 20%. However, due to the drive of economic interests, the current honey market is seriously adulterated, resulting in adulterated honey accounting for 20% to 30% of the honey market. In some areas, adulterated and fake bee products account for about 50%, seriously damaging the interests of consumers and affecting The healthy development of the honey industry and the fight against foreign exchange earnings from export trade. [0003] Due to the lack of detection methods, it is difficult to crack do...

AI Technical Summary

Problems solved by technology

However, it is still weak in differential information mining, which is also a bottleneck restricting the development of electronic noses.

Images