A new software has been developed that allows for parameter-free generation of a digital terrain model from LiDAR data. This method uses an adaptive morphological filter, where the size of the structural element is defined by the distance of a point from it’s nearest edge. The software is capable of accurately estimating the terrain under the great majority of particularly hard circumstances exposed by the ISPRS benchmark dataset.
D.11 Other
COBISS.SI-ID: 16412438Light Detection and Ranging (LIDAR) has become an important technology for terrain data acquisition and mapping. High acquisition density of LIDAR provides high detail of the sampled surface but it also results in a vast amount of data. To efficiently store LIDAR data dedicated compression algorithms are applied. In the paper, FPGA based hardware compression architecture is presented. It consists of three modules: LIDAR data predictor, variable length encoder, and arithmetic coder.
B.03 Paper at an international scientific conference
COBISS.SI-ID: 26489639This dissertation considers two new methods for automatic generation of digital terrain models from LiDAR data. The first method iterates a thin plate-spline interpolated surface towards the ground, while points' residuals from the surface are inspected at each iteration, with a gradually decreasing structural element. Top-hat transformation is used to enhance discontinuities caused by surface objects. Finally, parameter-free ground point filtering is achieved by automatic thresholding, based on a standard deviation. The experiments show that this method correctly determines DTM even in those cases of difficult terrain features. The expected accuracy of ground point determination on those datasets commonly used in practice today is over 96%, while the average total error produced on the ISPRS benchmark dataset is under 6%. The second method uses an adaptive morphological filter, where the size of the structural element is defined by the distance of a point from it's nearest edge. The input data is arranged into a grid and compass edge detection based on the Sobel operator is applied for edge extraction. Morphological region-filling is used in order to segment grid-cells into foreground and background regions, while the distance transformation of the foreground regions defines the size of the structural element for each foreground grid-cell. Finally, LiDAR point-filtering is achieved using adaptive top-hat transformation, followed by a constant thresholding. As confirmed by experiments, the average CPU execution time decreases by more than 94% compared to the first method, while the accuracy improves by nearly 20% on low-density datasets, and by nearly 30% on high-density datasets.
D.09 Tutoring for postgraduate students
COBISS.SI-ID: 16270870