Estimating Pollution Contents in an Urban Area using Airborne Hyperspectral Thermal Data

Abstract

The best remedy of pollution is its detection and control. The major pollutants in an urban area are asbestos and aerosol. This work introduces a technique to detect the pollutants in an area using hyperspectral data. Due to its enriched spectral information very minute contents are identified. The hyperspectral image is captured with the Telops' Hyper- Cam which contains thermal infrared bands which is used to create temperature map of that land surface thereby the different objects are classified according to different temperature ranges. The visible bands are also used to classify the image and the percentage of area under each class is calculated. To assess the accuracy of each classifier, confusion matrix is computed and identified that Support Vector Machine (SVM) classifier is the best other than Spectral Angle Mapper (SAM) and Spectral Information Divergence (SID) having accuracy of 94.89%. The amount of aerosol present in a locality is calculated with respect to a factor called $PM_{10}$ which gives the concentration of particles of dimension less than $10 \mu \mathrm {m}.$ Using the relation between $PM_{10}$ and atmospheric reflectance the value of $PM_{10}$ is obtained between $34 \mu \mathrm {g}/\mathrm {m}^{3}\mathrm {a}\mathrm {n}\mathrm {d}66\mu \mathrm {g}/\mathrm {m}^{3}.$ It's level is above $15 \mu \mathrm {g}/\mathrm {m}^{3}$which is the safe value according to Canadian jurisdiction so there is chance of hazardous health effects on human beings.