Extraction of Field Boundary Information from Classified Satellite Images

Remotely sensed images and the thematic maps derived from such images are invaluable sources of information for GIS databases, in terms of providing spatial and temporal information about the nature of Earth surface materials and objects. One of the techniques that has emerged recently, and which has made a great impact on scientific community, is that of artificial neural networks (ANNs). ANNs have been found to be more robust than conventional statistical methods. ANNs have the advantage of being employed in almost all the stages of a GIS system, such as the data preparation, analysis and modelling stages. This study describes a method to accurately extract field boundaries from thematic maps produced from ANN classification results. A feed-forward network structure that learns the characteristics of the training data through the backpropagation learning algorithm is employed to classify six land cover features present within the scene. This study also illustrates the role of ANNs in classifying land cover objects. A number of factors affecting classification accuracy, including the determination of the optimum network structure are discussed. It is observed that classification accuracy of up to 90% is achievable for thematic maps produced by ANNs.