A method for establishing a qualitative analysis model of near-infrared spectra and spectral images of seeds

Abstract

The invention provides a method for establishing a qualitative analysis model of near-infrared spectra and spectral images of seeds. The method includes: collecting sample data of each category in different environments, and determining a model training data set and a test data set. The sample data is Near-infrared spectral data and / or spectral images; use different data preprocessing methods and / or feature extraction methods to process the sample data in the model training data set to obtain the separability and classification accuracy between different types of sample data and deviation degree; select the best data preprocessing method, the best feature extraction method and the optimal data dimension of the best feature extraction method according to the separability, classification accuracy and deviation degree, and extract the sample data in the model training data set The characteristics; the qualitative analysis model is established by using the improved bionic pattern recognition method. The invention improves the efficiency of establishing the qualitative analysis model and the robustness of the qualitative analysis model, and has broad application prospects.

Description

technical field [0001] The invention relates to the technical field of spectrum and spectral image detection, in particular to a method for establishing a qualitative analysis model of near-infrared spectra and spectral images of seeds. Background technique [0002] The conventional qualitative analysis model building process includes steps such as data set creation, data preprocessing, feature extraction, model building, system testing, and model optimization. It is difficult and complicated to establish a model that meets the requirements, and it is necessary to select the appropriate method and adjust the parameters in each step; and it requires analysts to have professional knowledge, and ordinary personnel cannot make good use of near-infrared spectroscopy (spectral image) technology. Qualitative analysis, the analytical efficiency of conventional qualitative analysis models is low. [0003] Conventional qualitative analysis model building methods only consider the cor...

AI Technical Summary

Problems solved by technology

It is difficult and complicated to establish a model that meets the requirements, and it is necessary to select the appropriate method and adjust the parameters in each step; and it requires analysts to have professional knowledge, and ordinary personnel cannot make good use of near-infrared spectroscopy (spectral image) technology. Qualitative analysis, the analysis efficiency of conventional qualitative analysis models is low

[0003] Conventional qualitative analysis model building methods only consider the correct rate of the model training set, but do not pay enough attention to the robustness of the model, resulting in poor adaptability of the established model and need to be continuously updated

Establishing a qualified seed analysis model that meets the requirements requires a large number of samples, and at the same time, it is necessary to consider the influence of factors such as genetic background, growth environment, seasonal changes, harvest time, storage methods, and instrument status changes on the spectrum. Therefore, establishing an ideal model often requires take a long time

[0004] Existing pattern recognition methods are not effective in identifying natural products such as seeds, and there are certain defects

The samples of different origins and years of the same category of seeds may have great differences, and the existing partition-based pattern recognition methods (expenditure vector machines, etc.) and class-by-class modeling methods (nearest neighbor methods) cannot classify these samples into one kind

Images