SUMMARY
The main aim of the Bachelor thesis Optimisation of digital terrain models for the needs of orthogonalization of aerial images was to create filters and filter strategies that remove required points from point cloud. Second aim of the thesis was to compare filter strategies to reference data.
The software DTMaster was used to create filters and filter strategies. Filter strategies testing was made with data in LAS format in two different areas. Software DTMaster uses three kind of filters – Gross Error Filter, Building Filter and Vegetation Filter. These filters can be combined into filter strategies.
There were two kind of filter strategies evaluation, statistical and visual. Statistical evaluation was made in CloudCompare software. The comparison was made between ground points from automatic filtration and reference data from manual filtration. Characteristics counted in every compared point cloud were point mean distance, its standard deviation and maximum distance. According to these statistical characteristics the best filter strategies were chosen.
Visual evaluation was made in orthomosaic created from data marked as ground. Five objects were chosen to be compared to reference data. Reference orthomosaic was created from manual filtration points. In the end the best filter strategies were chosen according to both kinds of evaluation.
The outputs of Bachelor thesis are set of filter strategies for two case studies together with the written text of their suitability and use. The aim of automatic filter strategies is to lessen the amount of time to create required results.