Skip to Main Content
Various implemented software programs and GIS software applications were employed to investigate an effective method of extracting features (primarily buildings) from a LIDAR dataset (Vosselmann, 1999). Boundaries or outlines define the features, with the result being an external output file containing vector data of these features. The method involved the use of various GIS and image processing functions. These functions include in sequence, classification and smoothing using majority and median filters. The goal of classification and smoothing is to condition the tile for edge detection. The fundamental problem with the LIDAR dataset is the classification of pixel heights, in order to provide clear-cut boundaries between topographical features, such as ground level and roof level of buildings. To overcome this, various sized filters were investigated (30, 5×5, and 70) as well the number of iterations of the filter to find the optimal classified and smoothed dataset. An edge detection function was performed by determining slope values for the tile. Thresholding was applied next to aid in the sequential processes, thinning and expanding, by removing spurious pixels. The resulting feature outlines were vectorised. Subsequent functions were applied to improve accuracy of feature outlines. Visibility analysis has been implemented, showing visible and obstructed pixels. Basic radiowave propagation techniques have also been implemented for use on the LIDAR data. These include LOS and single-knife edge diffraction. The vectorised features can now be utilised in developing further propagation techniques, such as reflection and scattering.