Ahmed El-harby

Automatic Extraction of Vector Representations of Line Features: Classifying from Remotely Sensed Images Ahmed El-harby

This book describes the development and evaluation of a system that can classify line features from remotely sensed images in raster format using neural networks and transform the classified features into vector representations automatically using a new Square Scan Algorithm (SSA). The SSA was designed to deal with branching and crossing lines in order to transfer the line features in raster images into vector representations automatically. This algorithm was tested and it was found that the algorithm could successfully remove most noise pixels and detect branching, crossing, and isolated lines. In addition, it connected disconnected lines that have a small gap between them. A new method was proposed to collect the training data automatically from new images that depended on the neural network results. The above approach was applied for continuous classification from new images over time by selecting the training data positions automatically. This book helps students to apply neural networks for classifying features and to understand the automatic extraction process of vector representations of line features from remotely sensed images.