Evaluating the accuracy of hyperspectral and multispectral images in wetland cover classification using data mining models (Case study: Shadegan wetland)

Wetland cover classification is of special importance in order to identify the type of plant species inside the wetland and also to distinguish it from the wetland margin vegetation and to study their ecosystem changes.  Due to the spectral similarity between different plant species of wetlands and plants along the wetlands and agricultural lands, this is faced with problems using multispectral data and hyperspectral data can be very useful in this regard. in this study power of hyperspectral and multispectral sensors in identifying the characteristics of the wetland and the ability of ETM + (2011), Hyperion (2011) and ALI (2011) sensors to study the characteristics of Shadegan wetland during 1390 and different spectral indices with a suitable combination of The satellite imagery bands of these sensors were compared as input to a variety of classification methods including maximum likelihood, minimum distance, neural network and support vector machine. The results showed that the support vector machine and neural network methods with closer classification accuracy of 85% in all three images show closer results to reality. The classification accuracy for all three images was at its highest for the backup vector machine method, with a total accuracy of 95.73 for the Hyperion image, 88.03 for the ALI and 89.34 for the ETM +. Therefore, the characteristics considered for the wetland, in the three images obtained from the SVM algorithm showed that showing the differentiation of wetland vegetation use from irrigated agricultural land use is more ambiguous than other wetland features. Studies have shown that this part is less recognizable in ALI and ETM + images than Hyperion images, or in some areas these parts are not separable from aquaculture land at all, while Hyperion due to having 220 bands and having a higher level of Spectral details have the ability to distinguish between the two classes.