This paper describes complex networks in regards to their fractal characteristic. Firstly, a short background concerning fractal dimensionality based on box-counting is considered, while in the continuation its application on spatially embedded networks is proposed using rasterization at multiple scales. Finally, the obtained results are presented.
B.03 Paper at an international scientific conference
COBISS.SI-ID: 18157078Modern remote sensing technologies such as airborne Light Detection And Ranging (LiDAR) laser scanning enables the acquisition of large unstructured point cloud datasets representing highly detailed surfaces. Various methods have been developed to fuse LiDAR data with optical imagery in order to enrich the points with color information. However, optical imagery is not always available for some datasets. This paper proposes a new method for approximate reconstruction of LiDAR points’ colors by using supervised machine learning. Support vector machine (SVM) classifier is considered together with differential evolution (DE) in order to estimate the best parameters. The method successfully reconstructed LiDAR point cloud’s colors, as shown in the results.
B.03 Paper at an international scientific conference
COBISS.SI-ID: 18368790This work describes the visualization methods for airborne LiDAR terrain data. Common techniques used in commercial software are presented together with their advantages and drawbacks. Point-based rendering with splats is a recent alternative that improves these methods, representing all terrain data appropriately at various levels of details. The algorithms of point-based rendering are presented in details, with computational demands highlighted as the main disadvantage. To overcome this, the main chapters describe a procedure, which adds the option for selective hybrid rendering with triangle meshes. We can display parts of data as triangle meshes without unwanted graphical artefacts, at an equivalent level of quality to that in classical point-based rendering but with lower rendering times. These areas are identified by pre-processing the original data and can be converted from a point-based to a mesh representation at run-time without interfering with user interaction. The possibility of generating textures for the meshes enables additional simplification and even greater speeds without significant loss in quality. As the textures and triangulations are not stored on disk, the additional storage requirements are low.
D.09 Tutoring for postgraduate students
COBISS.SI-ID: 17982230