Characterization of colorimetric sensor arrays by a multi-spectral technique

Abstract

A new method based on a multi-spectral technique was proposed to characterize the signal of colorimetric sensor arrays for gas detection. Firstly, the characteristic wavelengths, which are most relevant to the detected substance, were extracted from the hyperspectral information of the colorimetric sensor arrays. Then, narrowband filters with the corresponding central wavelengths were selected to isolate the effective signal of the sensor arrays. In this study, ammonia (NH₃) was taken as an example to test the performance of the proposed multi-spectral method. Prediction of NH₃ concentration based on the hyperspectral method and normal tri-color (R/G/B) method was also performed for comparison. Compared with the tri-color method, the correlation coefficient for the testing set (R_t) based on the multi-spectral method increased from 0.902 to 0.976, root mean squared error of prediction (RMSEP) decreased from 1.213 to 0.548, and residual predictive deviation for the testing set (RPD_t) increased from 2.903 to 6.151, which means that the results were notably improved both in accuracy and stability. Furthermore, the multi-spectral method possesses the advantages of low cost, easy operation and greatly reduced data size. The proposed multi-spectral method could be used to characterize the signal of colorimetric sensor arrays for gas detection.