Publications

Proceedings of the Institute of Geodesy and Cartography - No 110, Vol. LII

price: 40 zł

Zbigniew Bochenek
Analysis of classification methods of urban areas registered on high-resolution satellite images
Results of the research works aimed at preparation of the optimal method for classifying urban areas on the basis of digital analysis of high-resolution satellite images were presented in the article. Data from three different satellite systems: Landsat ETM+, ASTER and QuickBird were used these works. The following methods of digital analysis and classification of satellite images were investigated in the course of the works:

Spectral mixture analysis,
Classification of urban areas with inclusion of texture measures,
Object-oriented analysis with the use of multi-resolution segmentation, fuzzy logic and membership functions,
Neural networks.

As a results of the conducted works it was found, that hybrid method, combining elements of object-oriented analysis and neural networks, is optimal for classifying urban areas with adequate accuracy. Application of multi-resolution segmentation enables to distinguish homogeneous objects within urban areas and to determine their several spectral and non-spectral parameters. Neural Networks analysis allows to find optimal solution concerning type of network for classification and importance of various parameters used in the classification process. The mean accuracy of classification with the use of this method, with division of urban areas into 3 levels of built-up land density, reaches 85 %. The crucial for obtaining high accuracy, while classifying high-resolution satellite image, is to select properly parameters of segmentation and a set of features characterizing land cover classes, used in neural network analysis.

Andrzej Ciołkosz, Zenon F. Poławski
Methodology of integration and analysis of multisource cartographic data for evaluation of land use changes in Poland
Land use maps can be a basis for studies of changes in land use forms in Poland. Among variety of maps two cartographic materials are worthy of particular attention.
First of these two is land use map prepared in 50’s of 20th century by F. Uhorczak. Thematic contents of this map was based on the contents of topographic maps at a scale of 1:100 000, prepared before Second World War. Hence, the map presents state of land use in Poland, which was in 30’s of last century.
Database prepared within CORINE Land Cover 2000 (CLC-2000), containing information on land cover and land use in Poland in 2000, is the second type of map data. High-resolution satellite images collected by Landsat satellite were the source data used for preparation of this database.
Different methods of data acquisition and processing, various forms of their recording and visualization, as well as differences in projection and non-uniform rules of generalization create serious problems while comparing these cartographic materials directly. So, in order to make the comparison, feasible data ought to be standardized.
This process comprised at first conversion of analogue data to digital ones and next geometrical matching of both sets of data. It implied transformation of scanned small-scale maps to the reference system, which was applied for preparation of CORINE Land Cover 2000 database. Next step was aimed at obtaining both ranges of thematic contents comparable; it led to preparation of rules of thematic aggregation of CLC 2000 database.
In order to compare both sets of data preparation of methods of analysis was indispensable.
As a result of interpretation the database was prepared, which contains information on changes of land use forms in Poland between 1930 and 2000. This database enabled to determine transfigurations of land use forms and to analyze their multiyear spatial changes in Poland.

Janusz Ostrowski, Albina Kinga Mościcka, Anna Wrochna
Impact of map scale and method of cartographic presentation on mapping soil cover structure
Authors present in the article analysis of the way of reflecting variability and differentiation of soil cover, dependent on map scale and method of cartographic presentation. Experiment was done for soil maps of Poland at scales 1:250 000, 1:500 000 and 1:2 500 000, covering area of swietokrzyskie voivodship. Information stored in two databases: regional and national database has been used in this study. Results of the works and fragments of the prepared maps were presented in the article.
The presented analysis and interpretation of the results of the study leads to conclusion, that generalization of soil map contents at a regional scale to the level adequate for national scale allows for presentation of soil properties on the regional map with greater spatial precision. Visual evaluation of soil cover structure obtained through generalization and transformation of raster map a scale of 1:250 000 to its vector form enables to draw conclusion, that change of presentation from raster to vector form facilitates perception of map contents. However, this process implies area generalization, which makes structure of soil cover more coarse, still keeping its regional specific features.

Artur Karol Karwel, Ireneusz Ewiak
Assessment of usefulness of SRTM data for generation of DEM of the territory of Poland
Accuracy assessment of relief model acquired from the Shuttle Radar Topographic Mission for the territory of Poland is the subject of the article. Estimation of accuracy of this model was based on precise reference data – terrain profiles measured with GPS technique. Accuracy investigations covered test fields representing various forms of relief, located on open areas within 14 voivodships. Due to character of interferometric height measurements urban and forest areas were not included into these studies. Mean errors determined on the basis of height differences between points of terrain profiles and corresponding interpolated points of SRTM model were the measure of accuracy assessment. Comparative analyses were carried out within Intergraph MGE Terrain Analyst software and with the use of 3 DEM Terrain Visualization software tools created by Richard Horne. It was found, that absolute accuracy of SRTM model on the territory of Poland is mH = 2.9 m for flat terrains and mH = 5.4 m for undulating/hilly areas. It was also found, that accuracy of interferometric measurement of heights of node points of SRTM model grid is mainly degraded in the process of interpolation of heights of points inside grid, and the degree of this degradation depends on height differences and on size of grid mesh. Estimation of accuracy of SRTM model conducted with the use of big statistical representation revealed, that interferometric measurements are influenced by line component of systematic error. It was found, that while removing this component real absolute accuracy of SRTM model on the territory of Poland reaches mH = 1.0 m for flat terrains and mH = 2.7 m for undulating/hilly areas.

Jan Ziobro
Pre–adjustment detection of gross errors in GPS measurement for aerial triangulation
Author presents in the article a method for pre–adjustment detection of gross errors in the measured projection centers. The method is based on analyzing differences of the results of two independent measurements: one obtained from adjustment of aerial triangulation without determining projection center and the second, which considers measurement of projection centers during photogrammetric mission.
The technique of measuring projection centers for aerial triangulation with the use of GPS method exists since 1993 and is still improved, as far as precision and reliability is concerned. In standard work real verification of quality of measurement is done only at the stage of adjustment of aerial triangulation. The main aim of adjustment is to obtain the result with the highest probability, and it depends on removing gross errors from calculations. As it can be seen from practice, this condition is difficult to fulfill; the procedure is time–consuming and not fully efficient.
Detection and location of gross errors is difficult due to improper division of GPS measurements into profiles, multiple gross errors, mistakes in GPS measurement, or insufficient reliability level of network.
In the proposed method distances between neighboring points of profile are compared, obtained from two independent determinations. In addition, increments of coordinates between neighboring projection centers are also compared. These differences, which prove to be higher than triple mean error, are considered as gross errors.
The method has been tested on 26 blocks, which were prepared during last years in Poland. The aim of testing was to verify magnitude and number of gross errors of projection centers, which remain in the network after applying the method. Analysis of non–detected gross errors was done using W.Baarda data snooping method, i.e. the method of standardized residuals. In the test blocks at a scale of 1:13 000 level of detectability of gross errors in the measured projections centers was ca. 6 times mean error of coordinate of projection center, while for 1:26 000 photographs it was 12 times mean error, respectively.
The method enables to detect in one calculation step all mistakes and most of gross errors, which results in decreasing number of adjustment cycles in cases, when many data errors exist.