Site modeling using centroid neural network

This paper presents a new method for the site modeling from aerial image data. Initially 3D lines are extracted by using elevation data obtained by area-based stereo. The grouping process is implemented by using centroid neural network algorithm to classify 3D lines into groups of lines. Using grouped 3D lines, hypothesis selection is carried out based on undirected graph, in which close cycles represent complete rooftops hypotheses. This approach makes hypothesis selection a simple graph search for close cycles. We test the proposed method with the synthetic images generated from Avenches dataset of Ascona aerial images. The experiment result shows that our method can be efficiently used for the task of site modeling from aerial images.