Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.geogr.msu.ru/cafedra/karta/anniversary/docs/plushkavichute.pdf Дата изменения: Thu Nov 8 23:04:02 2012 Дата индексирования: Sun Feb 3 23:55:50 2013 Кодировка:

Y. A. Plyushkyavichyute yuratep@student.ethz.ch EVALUATION OF POSSIBLE CARTOGRAPHIC AND REMOTE SENSING METHODS FOR DETECTING TEMPORAL CHANGES IN THE URBAN AREA IN THE ALPINE REGION Swiss Federal Institute of Technology, Zurich ABSTRACT This project explores the potential of Cartographic and Remote Sensing approaches in the study of Alpine urban region in order to detect the temporal changes. Both approaches give an opportunity to find the most efficient type of representation of the spatial characteristics of urban phenomena. Furthermore this analysis of different methods provides the following main outputs: a dynamic framework to represent the evolution of the urban form; defines the change in planning data and information; gives the opportunity of systematical analysis of the wildland urban development processes. The result of the project allows to consider, that both of approaches are giving not just relevant information about temporal changes, but also provide the results with significant value of accuracy. A discussion of all the fundamental contribution of these methods is then presented. INTRODUCTION The motivation for this research has been identified due to the lacking functional information content of the temporal changes in Alpine environment. Therefore was decided to use two main approaches in order to detect the temporal changes: Cartographic and Remote Sensing. Both of them use different types of analysis. The research problem, therefore, deals with linking both results by using spatial configurations (georeferencing) in order to compare them and evaluate the best method. The case study area is mostly situated in the Gorduno municipality, on the south of Switzerland in the Canton Ticino. This region covers an area of 9.22kmІ, where the urban zone has diverse representation, 43% of the urban zone are build up areas (with houses and buildings) 57 % of the rest area are roads and other transportation infrastructure. In the project were used two types of input data: Cartographic data; Remote Sensing data. Cartographic data: Topographic maps were produced by swisstopo and used as an input for the generation the cartographic approach analysis. The scale of input maps was 1: 50 000. Raster maps from the years: 1995 and 2006 were used in this project. Additionally to the cartographic data some results from the first phase of the project were considered. The distances of 270/185/85 m from the closest building were identified as the areas including 90/70/60 % of the fire ignitions. These distances have been adopted for the buffer generation process in order to better analyze the changes of urban area. As a Remote Sensing data were used two main Landsat TM images , which were derived from USGS online archive , the following dates were selected: 09.07.1990 and 05.07.2003. All incoming data was pre-processed (geometric correction, radiometric calibration in ERDAS Imagine software). This step was very important, in order to establish more direct linkage between the data and biophysical phenomena, removal of data acquisition errors, image noise (Coppin et al., 2004). Additionally to the Landsat Data in current work the calculated NDVI (Normalized Difference Vegetation Index) images for the study area for the same dates as Landsat images were used: 09.07.1990 and 05.07.2003. METHODOLOGY


In order to receive the results was decided to use following methodology, by dividing it into to major parts: cartographic and remote sensing approaches (Fig.1). Cartographic approach: Within the cartographic approach in order to apply two types of raster and vector analyses were used three following steps: Georeferencing (made the topographic maps comparable between each other spatially); Color separation (separated the urban object on the map (houses, buildings, roads)); Raster Cleaning (created a raster layer only with houses and buildings object). Further temporal investigations within cartographic approach were done by using two different types of analysis: raster and vector. Raster analysis performed by using image differencing method, vector analysis was divided into two steps: first- vectorization of raster map and second buffering analyses in order to investigate the distance, where the new buildings were built. Remote Sensing approach: For Remote Sensing approach two types of change detection possibilities were used (supervised and unsupervised). As incoming data for unsupervised approach NDVI images and for Supervised- resulted classified images were used. Within the unsupervised approach were done: Qualitative analysis (FCC (False Color Composite) technique, which visualized the changed areas); Statistical variability (PCA (Principal Component Analysis), which increased the results of further step); Image math (subtractive difference, in order to get the difference map). For supervised approach two types of classifications were used: supervised and unsupervised. The last step in both Remote Sensing approaches was calculation of confusion matrix (error matrix) in order to verify the correctness of chosen methods. RESULT In order to evaluate the results form


the two approaches of change detection methods it is necessary mention that in the project the goal was to evaluate different methodologies. As declared in the methodology phase (Fig.1), every approach needs to have a result, representing the temporal changes. As a consequence there are four main results, for the further discussion. Result 1: Cartographic approach, raster analysis. The result of this method is the raster image with changed (red) and non-changed (green) areas (Fig.2). This image qualitatively allowed us to distinguish the areas, which were changed. While analyzing this image it was possible do distinguish, that the symbol for the buildings on the map for the year 2006 has been changed by changing its thickness. Therefore there are some areas, which were detected as the new buildings, but in a reality it is still the same building. But the result still represents the buildings, which have been built between years 1995 and 2006. To conclude, this method requires a lot of manual work, which can not be done automatically. Therefore it strongly depends from the cartographer, his knowledge of the territory, while further improvement of the results might need manually to exclude not build-up objects. Possible further evaluation of the result can be the opportunity of calculating the changed cells in order to derive the value for the changed areas. Result 2: Cartographic approach, vector analysis. This type of analysis gave the opportunity to obtain two types of results. The first is the image with the buffering zones, where the resulting analysis shows, that the main distance, where the new buildings were built is 270 m (Fig.3). Second approach of buffering analysis allows the distinction of the zones, were new building have been built or also it is possible to obtain the zones, where buildings are not existing anymore (Fig.4). These types of results give a relevant representation of current and past situation. The only disadvantage of this work analysis is, that in order to get statistical information additional steps are needed to be done. However, the achieved results allow us to receive relevant information about the territory. Result 3: Remote Sensing Approach, NDVI analysis. For the NDVI it was decided to define the threshold for the urban zones. For each image the value of the threshold was found, but, the difference in the values between two years was not very high. However, it was possible to assume, that the general threshold for the build-up area is within this interval: 0.3-0.62 m. Afterwards the confusion matrix was done, where the value of total accuracy was calculated for both years, for the year 1990 the value of total accuracy was 78 %, which might be explained by the fact, that the year of publication of the map is 1995 and can therefore include recent information about newer objects


(Fig.5).

Result 4. Remote Sensing Approach. Classification Difference. After analyzing the confusion matrixes were decided to further consider only of the supervised classification, because the total accuracy was high (88% for the year 1990 and 87% for the year 2003) (Fig.6). The usage of classification difference method allowed us to distinguish the difference in buildup detected classes. One of the advantages of this method is that it is possible to get the statistical information about build-up area for each year, which allowed us afterwards to calculate the difference and distinguish the values of changes. In order to analyze the result it was decided to compare qualitatively the obtained classification. CONCLUSION This research work demonstrates the potential of using two approaches (Cartographic and Remote Sensing) in order to investigate temporal changes. Attempts were made to detect build-up area as it changes trough time as accurate as possible. As a result it can be concluded, that both selected approaches


can be used in order to estimate the changes. However, both analyses have their advantages and disadvantages. The Cartographic approach gives an opportunity of extracting the information about every single building with the opportunity of further analysis. As a disadvantage this type of method is time consuming, therefore it is hard implementing for the big areas of interest. Remote Sensing approach gives fast representation of temporal changes. One of the disadvantages is, that the accuracy of the results in this approach is strongly depends on the algorithms, which are used within this approach, while most of them are done automatically. ACKNOWLEDGEMENT I would like to thank Prof. Dr. Lorenz Hurni and Lorenzo Oleggini, for supervising this project, for guidance and support from the initial to the final level.