SOZO database and its cartographic model – a sozological map is a thematic data source that represents the state of natural environment, as well as the causes and effects of both positive and negative transformations occurring in the environment. There is no doubt that the SOZO database should be a tool to investigate and evaluate the state of natural environment in a quantitative manner. However, the analysis of chosen features in the database reveals that they seem to be insufficient for conducting spatial analysis. The crucial problem is related to classes of forest damage represented by point features in a database without the extent of forest degradation. This type of presentation has limited use for geospatial analysis of environment therefore the Authors propose to use remotely sensed data to enhance SOZO database with updated data together with the spatial occurrence of the studied phenomenon. This research presents the potentials of determining class of forest damage based on the vegetation indices (NDVI and NDII) that successfully replaces the point-feature class of forest damage with polygon-based classes, thereby introducing the new quality of the data into the environmental database and make SOZO database a useful product in spatial analysis.
AN ALGORITHM OF 2D BUILDING MODELING IN AIRBORNE LASER SCANNING POINT CLOUD OF THE ISOK PROJECT
Krzysztof Sochiera, Andrzej Borkowski
airborne laser scanning, alpha shape, total least squares, ISOK, 2D-modelling
Airborne laser scanning data (ALS) are acquired mostly for the purpose of digital elevation models generation. In Poland, ALS data have been obtained for the whole country within the ISOK project, established for natural hazards risk mitigation. These data were used in this study to model the outlines of buildings. For this purpose an algorithm is proposed, that is a combination of α-shape algorithm and iterative total least squares adjustment. α-shape is used to detect points representing building outlines while the total least squares method is performed to receive regularized 2D building vector models. Identification of points representing outlines of buildings was performed on the point cloud thresholded at the given height with rejection of points above that height. Identification of a building as a gap (internal hull) in ALS data set is a better approximation of real building shape. For the algorithm verification a point cloud with a density of 4 points /m2 is utilized. This point cloud represents a city urban area, covering 21 large buildings. The results of 2D modeling of buildings have been compared with their representation in the cadaster data base. The linear deviation between corresponding corners of modeled and represented in cadaster data base buildings have been measured. The received mean value of the deviation equal 0.56 m is consistent with the nominal planar accuracy of ISOK ALS data. RMSE of building outline modelling calculated on the basis of linear deviations was equal 0,64 m. The results of modeling meet the requirements of Topographic Database Objects 1: 10000 (BDOT10k) and can be used for verification and updating of this data base.
AN ALGORITHM FOR FULL-WAVEFORM LASER SCANNING SIGNAL DECOMPOSITION AND MODELING
Agata Walicka, Andrzej Borkowski
airborne laser scanning, full waveform, approximation, signal decomposition
Airborne laser scanning is one of the most powerful techniques for acquiring information about Earth’s surface and land cover. Dynamic development of technology enabled the broader use of full-waveform’s type systems, which register the entire reflected waveform. In order to provide some additional information about the structure of the illuminated surface, discrete values should be approximated by parametric functions. Research is focused on algorithm development that would allow to carry out a rapid decomposition of the wave while detecting and approximating weak and overlapping echoes. Most of existing methods for full-waveform signal modeling requires knowledge of the number of peaks and approximate parameter values. In this paper new algorithm for signal decomposition has been investigated. It allows to carry out the decomposition effectively without preprocessing. This algorithm can be considered as a progressive algorithm modification. The method involves an iterative curve fitting using weighted Levenberg-Marquardt algorithm. Two-step validation of decomposition method has also been carried out on test data. Firstly, the quantity and distribution of approximation error have been investigated. Furthermore the results have been compared to standard procedure. Basing on algorithm validation it can be stated that the method allows proper detection of all components and their correct approximation.