DIGITALNA ARHIVA ŠUMARSKOG LISTA
prilagođeno pretraživanje po punom tekstu



ŠUMARSKI LIST 1-2/2017 str. 40     <-- 40 -->        PDF

Tarmizi, N. M., A. M. Samad, M. S. C. Mat, M. S. M. Yusop, 2014: Qualitative and Quantitative Assessment on Shoreline Data Extraction from Quickbird Satellite Images, International Journal Of Computer Science, 2 (9): 54-62.
Trimble: eCognition Essentials, (http://www.ecognition.com/), (datum pristupa: 10.11. 2015).
Tucker, C. J., D. M. Grant, J. D. Dykstra, 2004: NASA’s global orthorectified Landsat data set, Photogrammetric Engineering and Remote Sensing, 70: 313–322.
Tuno, N., 2014: Optimalne metode geometrijske obrade digitalnih geodetskih planova, Doktorska disertacija. Građevinski fakultet Univerziteta u Sarajevu.
Tutić, D., M. Lapaine, 2009: Kartografska generalizacija linija sa svojstvom čuvanja površina, Kartografija i geoinformacije, 11: 84-100.
Valožić, L., 2014: Klasifikacija zemljišnog pokrova urbanog i periurbanog prostora pomoću objektno orijentirane analize multispektralnih snimaka, Hrvatski geografski glasnik, 76/2: 27 – 38
Vojniković, S., J. Taletović, B. Balić, M. Ljuša, O. Žurovec, H. Čustović, 2013: Procjena površina šuma u Kantonu Sarajevu prema 4tom nivou nomenklature Corine zemljišnog pokrivača, Naše šume, 32-33: 12-22.
Summary
Knowledge about positional accuracy of forest geospatial information, obtained by interpretation of satellite imagery, is of great significance. The consequences of the decisions that are based on data with insufficient or unknown quality could be very negative. This paper investigates the accuracy of closed linear shapes that represented boundaries of forest cover. Forest areas are effectively extracted from Landsat image by implementing the process of multiresolution image segmentation (figure 4), using all bands. Multispectral classification of defined segments was performed by special rules. The results of object-oriented classification showed that an overall accuracy from 99 reference points was better than 90 % (table 1), which can be considered as a very good result. The number of forest polygons, obtained by satellite imagery classification, was reduced by 37 times by cartographic aggregation (figure 5). The Polynomial Approximation with Exponential Kernel (PAEK) method was used for cartographic smoothing of the forest polygons, which smoothes lines in relation to a softening tolerance (tolerances from 30 m to 180 m were used in this research) (figure 6). The positional accuracy assessment of the boundary of forest areas, based on procedure of comparing a tested lines to a reference lines, showed that the best results were obtained by PAEK smoothing with 150 m and 180 m tolerances (CMAS = 49 m, according to STANAG 2215) (tables 2 and 3, figure 8).
The findings of this empirical research showed that cartographic generalization contributes to improvement of the forest boundaries accuracy, as well as the appropriate processing of the medium spatial resolution remotely sensed data can result in satisfactory quality of vector data.
KEY WORDS: Landsat, object-oriented forest classification, cartographic generalization, positional accuracy