|
|
Surface Reconstruction from Point Clouds
Jarůšek, Tomáš ; Zemčík, Pavel (referee) ; Španěl, Michal (advisor)
Surface reconstruction from point clouds represents very modern and widespreadly used method for digitalization of real surfaces. The main focus of this work is on a comparison of several algorithms (Poisson surface reconstruction, Ball Pivoting Algorithm, Fourier surface reconstruction, Wavelet surface reconstruction and Multi-level Partition of Unity implicits) that reconstruct original surface in a form of triangle mesh from an oriented point cloud. Every algorithm was used in a series of tests on both real and synthetic datasets which were also modified to suit more complicated experiments. The following evaluation is done in various categories. The results are evaluated both visualy and with calculated measures (such as Hausdorff distance). Together these experiments depict a detailed analysis of the state of surface reconstruction algorithms. Our experimets show that the selection of optimal algorithm depends on a concrete task. Overall Possion surface reconstruction and Fourier surface reconstruction provide the best results.
|
|
|
Large Graph Data Visualisation on the Web
Jarůšek, Tomáš ; Bartík, Vladimír (referee) ; Burget, Radek (advisor)
Graph databases provide a form of data storage that is fundamentally different from a relational model. The goal of this thesis is to visualize the data and determine the maximum volume that current web browsers are able to process at once. For this purpose, an interactive web application was implemented. Data are stored using the RDF (Resource Description Framework) model, which represents them as triples with a form of subject - predicate - object. Communication between this database, which runs on server and client is realized via REST API. The client itself is then implemented in JavaScript. Visualization is performed by using the HTML element canvas and can be done in different ways by applying three specially designed methods: greedy, greedy-swap and force-directed. The resulting boundaries were determined primarily by measuring time complexities of different parts and were heavily influenced by user's goals. If it is necessary to visualize as much data as possible, then 150000 triples were set to be the limiting volume. On the other hand, if the goal is maximum quality and application smoothness, then the limit doesn't exceed a few thousand.
|
|
|
Large Graph Data Visualisation on the Web
Jarůšek, Tomáš ; Bartík, Vladimír (referee) ; Burget, Radek (advisor)
Graph databases provide a form of data storage that is fundamentally different from a relational model. The goal of this thesis is to visualize the data and determine the maximum volume that current web browsers are able to process at once. For this purpose, an interactive web application was implemented. Data are stored using the RDF (Resource Description Framework) model, which represents them as triples with a form of subject - predicate - object. Communication between this database, which runs on server and client is realized via REST API. The client itself is then implemented in JavaScript. Visualization is performed by using the HTML element canvas and can be done in different ways by applying three specially designed methods: greedy, greedy-swap and force-directed. The resulting boundaries were determined primarily by measuring time complexities of different parts and were heavily influenced by user's goals. If it is necessary to visualize as much data as possible, then 150000 triples were set to be the limiting volume. On the other hand, if the goal is maximum quality and application smoothness, then the limit doesn't exceed a few thousand.
|
|
|
Surface Reconstruction from Point Clouds
Jarůšek, Tomáš ; Zemčík, Pavel (referee) ; Španěl, Michal (advisor)
Surface reconstruction from point clouds represents very modern and widespreadly used method for digitalization of real surfaces. The main focus of this work is on a comparison of several algorithms (Poisson surface reconstruction, Ball Pivoting Algorithm, Fourier surface reconstruction, Wavelet surface reconstruction and Multi-level Partition of Unity implicits) that reconstruct original surface in a form of triangle mesh from an oriented point cloud. Every algorithm was used in a series of tests on both real and synthetic datasets which were also modified to suit more complicated experiments. The following evaluation is done in various categories. The results are evaluated both visualy and with calculated measures (such as Hausdorff distance). Together these experiments depict a detailed analysis of the state of surface reconstruction algorithms. Our experimets show that the selection of optimal algorithm depends on a concrete task. Overall Possion surface reconstruction and Fourier surface reconstruction provide the best results.
|
guest ::
