ANN Based High Spatial Resolution Remote Sensing Wetland Classification

Abstract

RS (Remote Sensing) image classification based on ANN (Artificial Neural Network) is carried out with high spatial resolution images of the wetland, which is the most important ecological environment element within the land components. Wetland dynamic change monitoring is often built upon its classification result concerned here. The typical high spatial resolution image of the wetland in Nanjing is used as a study case by ANN method in comparison with MLC (Maximum Likelihood Classification). Furthermore, the optimal number of ANN hidden neurons are simulated for enhance the classification effectivity. Totally, the results show classification method of ANN with optimal hidden neurons can effectively distinguish ground objects and improve the classification accuracy. The overall accuracy of the ANN classification is up to 93% and the Kappa coefficient is over 0.89.