A simplified hyperspectral identification system based on mathematical Transformation: An example of Cordyceps sinensis geographical origins

Hyperspectral measurement system is an effective technique to acquire the information for identifying food quality and medical material authenticity. However, the high economic cost of hyperspectral equipment and the identification accuracy affect the application. In the study, we chose identifying the Naqu Cordyceps sinensis geographical origins as the application example. First, we experimented three typical mathematical transformation on hyperspectral reflectance curves, that can help us to find the sensitive spectral range. Meanwhile, we combined hyperspectral characteristics and several popular machine learning models to identify the authenticity of Cordyceps sinensis. Secondly, we analyzed the influence of different spectral resolution on the identification accuracy. Finally, our experimental results allow us to optimize and simplify the hyperspectral system specifically for the SWIR1 spectral range (1000–1800 nm) with a sampling interval of 1 nm. This simplification is to reduce costs while maintaining a high accuracy of identification. Through rigorous data analysis, it is possible to effectively determine the geographical origins of Cordyceps sinensis Naqu with an accuracy of 100 % using independent samples.