Spectral analysis using a near-infrared region (NIR) sensitive camera towards the identification of chemical pollutants

Abstract

The main objective of this project was to evaluate the potential of a Near Infrared Region (NIR) sensitive video surveillance camera available in the market to detect chemical pollutants in air. Preliminary studies focused on binary liquid systems to ascertain the potential of the NIR-capable video sensor. The camera, operated on a multispectral mode using an external filter wheel, was able to acceptably quantify water concentration in a water/ethanol mixture. The chemometric models build for this data were capable of predicting the water concentration with an adequate precision (R-squared (R2) in a relation of predicted values vs reference values was 0,993). The air pollutants studied were nitrogen oxides (NOx), water as vapor (H2O), nitrogen (N2), acetylene (C2H2) and Ammonia (NH3). We analysed only 9 wavelengths which were selected based on the absorption profile of the identified air pollutants. A typical experimental procedure consisted of capturing band-filtered images of a gas or vapor sample inside a gas cell. Then, the pictures were pre-processed by an algorithm developed during this project, and finally submitted to Principal Component Analysis (PCA), Partial Least Squares Discriminant Analysis (PLS-DA) and linear discriminant analysis (LDA). Validation of the PCA models was tested with independent samples using classification analysis SIMCA. These samples were also used to test the PLS-DA and LDA models. We also had the opportunity to use a commercial short-wave infrared (SWIR) camera, which has higher sensitivity to the NIR region, than the video surveillance camera. This sensor was also successful in the detection and quantification of water in water/acetonitrile mixtures. The chemometric models build for this data were capable of predicting the water concentration with an adequate precision (R-squared (R2) in a relation of predicted values vs reference values was 0,995). Preliminary studies revealed some potential in gas sample discrimination with both video sensors. The project is currently on-going, and, in the future, we expect to develop chemometric models capable of discriminating a set of specific air pollutants.