Atmospheric effects on land classification using satellites and their correction.

Haze occurs almost every year in Malaysia and is caused by smoke which originates
from forest fire in Indonesia. It causes visibility to drop, therefore affecting the data
acquired for this area using optical sensor such as that on board Landsat - the remote
sensing satellite that have provided the longest continuous record of Earth's surface.
The work presented in this thesis is meant to develop a better understanding of
atmospheric effects on land classification using satellite data and method of removing
them. To do so, the two main atmospheric effects dealt with here are cloud and haze.
Detection of cloud and its shadow are carried out using MODIS algorithms due to
allowing optimal use of its rich bands. The analysis is applied to Landsat data, in
which shows a high agreement with other methods. The thesis then concerns on
determining the most suitable classification scheme to be used. Maximum Likelihood
(ML) is found to be a preferable classification scheme due to its simplicity, objectivity
and ability to classify land covers with acceptable accuracy. The effects of haze are
subsequently modelled and simulated as a summation of a weighted signal component
and a weighted pure haze component. By doing so, the spectral and statistical
properties of the land classes can be systematically investigated, in which showing
that haze modifies the class spectral signatures, consequently causing the
classification accuracy to decline. Based on the haze model, a method of removing
haze from satellite data was developed and tested using both simulated and real
datasets. The results show that the removal method is able clean up haze and improve
classification accuracy, yet a highly non-uniform haze may hamper its performance.