On-line near-infrared sample size determination method

Abstract

An online near-infrared sample size determining method, comprising the following steps: step I, only acquiring spectrum information of all samples; step II, performing PCA projection on a matrix formed by the acquired spectrum information; step III, constructing statistics of infrared spectra; step IV, calculating a standard deviation of an overall sample space T value and a T value of the infrared spectrum of each sample; sampling all sample space T values according to different percentages; randomly taking samples for k times at each percentage and calculating the stability of the sample T values at each percentage; step V, determining the corresponding sampling rate when the difference between the standard deviation of the sample space T values at difference sampling percentages and the standard deviation of the overall sample T value is smaller than a preset value to be the required sampling rate of the online near-infrared sample size. The online near-infrared sample size determining method can accurately determine the required sample size for constructing a near-infrared model.

Description

technical field [0001] The invention relates to an online near-infrared sample size determination method based on spectral projection, belonging to the field of tobacco product production quality evaluation. Background technique [0002] Tobacco near-infrared spectrum contains a large amount of conventional chemical value information, physical information, and appearance information of tobacco leaves, so near-infrared plays a very important role in the field of routine chemical value detection, tobacco leaf stability evaluation, and tobacco leaf quality evaluation, but To establish a better near-infrared model, it often requires enough sample size and better basic data quality to build a near-infrared model detection and evaluation system; the size of the sample size in the model is more important, and often the model The sample size cannot be determined, which will either lead to too little sample size in the model, making the qualitative and quantitative model of the tobac...

AI Technical Summary

Problems solved by technology

The sample size of near-infrared is a problem worthy of research. In the process of constructing a complex qualitative and quantitative system of near-infrared, if the sample size is too small, the constructed infrared model often cannot contain the complex background system, and the robustness is not good, resulting in the model If the extrapolation and prediction ability of the near-infrared model system becomes poor, if the sample size is too large in the construction of the near-infrared model system, this will undoubtedly bring operational and computational complexity to the qualitative and quantitative construction of the model. At the same time, the advantages of the rapid analysis and detection of the near-infrared It cannot be reflected, because in the near-infrared field, the detection of basic data often has a certain degree of difficulty and cost, which will undoubtedly bring a substantial increase in costs for enterprises and increase the difficulty of maintaining near-infrared models;

Images