Testowanie metod i algorytmów klasyfikacji wysokorozdzielczych zobrazowań satelitarnych Puszczy Niepołomickiej

• Autorzy: Wężyk P. , Bednarczyk P.

Warianty tytułu

EN

Testing of methods and classification algorithms of the vhr satellite images of the Niepolomice Primeval forest

• Języki publikacji: PL

Abstrakty

EN

Contemporary VHR satellite images very often offer resolution between 1 and 4 meters and very good dynamic range (11 bit). The new RS technology enables to extract new kind of information (especially thematic information) which can be used in the forest inventory process. This paper presents how two classification methods, i.e.: the.traditional. . supervised and the .newest. object-oriented and different algorithms were tested on the example of QuickBird-2 images (ground resolution 2,44 m MS; 0,62 PAN; 11 bit), for the stand mapping purpose. The Niepolomice Primeval Forest was selected for the study. The 10.000 hectares forest complex is located in southern Poland and represents a lowland forest in the Vistula valley. The dominant tree species in the Niepolomice Primeval Forest are: pine (Pinus sylvestris L.), oak (Quercus sp.) . 18% and black alder (Alnus glutinosa L.) . 9%. The paper describes options of the ER Mapper 6.4 (Earth Resource Mapping Inc.) in the supervised classification. Usefulness of the algorithms: Maximum Likelihood (Standard and Enhanced both with and without Neighbor filter), Mahalanobis, Parallelepiped, Minimum Distance and Minimum Distance (Standard deviation) . were tested. The Regions of Interest (ROI) were created using DGPS receivers. Forest stand mapping in the Niepolomice Primeval Forest based on the QuickBird image from September 15th 2003 required following ROI: Pine (So), Oak (Db), Alder (Ol), Larch (Md), Beech (Bk), Birch (Brz), shadow (cien), clearing (polana), built up areas (zabudowa) and agriculture (p_uprawne). The best result of supervised classification proved the Maximum Likelihood Standard algorithm with Neighbor filter (5x5 pixels and 6 passes). The classification accuracy of the Pine (So) was around 98% and of broad-leaves trees ca. 80%. However, the classification accuracy of species was good for pine (97%) and oak (73%). The result of the alder classification was only 63%. It was also proved that Parallelepiped algorithm was completely useless. In this work, the authors also tested usefulness of the .newest. object-oriented (O.O.) classification in automatic mapping of the areas with- or without tree cover, inside and outside the forest. In this case ROI (samples) were not used, but only empirical expressions, which describe main features. This solution allows transferring of class hierarchy on other satellite images. These techniques can save time and money in the future. It is useful in verification of the Forest Digital Maps (LMN) during forest inventory works and it retrieves some important parameters like forest type, crown density, forest gaps, etc. The research presented shows that modern RS techniques based on the VHR satellite images and image processing methods and algorithms are very useful for rapid semi- or fullautomatic forest mapping.

Słowa kluczowe

PL

QuickBird; klasyfikacja; segmentacja

EN

QuickBird; classification; segmentation

• Wydawca: Polskie Towarzystwo Informacji Przestrzennej
• Czasopismo: Roczniki Geomatyki
• Rocznik: 2005
• Tom: T. 3, z. 2
• Strony: 163--172
• Opis fizyczny: Bibliogr. 18 poz.

Twórcy

autor

Wężyk P.

• Laboratorium GIS i Teledetekcji, Katedra Ekologii Lasu, Wydział Leśny Akademia Rolnicza w Krakowie

autor

Bednarczyk P.

• Laboratorium GIS i Teledetekcji, Katedra Ekologii Lasu, Wydział Leśny Akademia Rolnicza w Krakowie

Bibliografia

• 1. Baatz, M., Schäpe, A., 2000: Multiresolution Segmentation: an optimization approach for high quality multiscale image segmentation. In: Strobl, J. and Blaschke, T. (Eds.): Angewandte Geogr. Informationsverarbeitung XII, Wichmann, Heidelberg, ss. 12-23.
• 2. Blaschke, T., Hay, G. J., 2001: Object-oriented image analysis and scale-space: theory and methods for modeling and evaluating multiscale landscape structure. Int. Archives Photogrammetry and Remote Sens., 34, part 4/W5, ss. 22-29.
• 3. Burnett C., Blaschke T., 2003: A multi-scale segmentation/object relationship modeling methodology for landscape analysis Ecological Modelling, Volume 168, Issue 3, 15 October 2003, ss. 233-249
• 4. De Kok R., 2005: eCognition . object oriented classification training. Training Manual. ProGea ConsultinglandConsult. Krakow 2005.
• 5. Dickson E., Franklin S.E., Moskal L.M., 1999: Monitoring of forest biodiversity using remote sensing: Forest stand (high spatial resolution) protocol and examples. Alberta Biodiv. Monitoring Prog. http:// www.abmp.arc.ab.ca/chapter09.pdf, ss. 26.
• 6. Instrukcja Urz¹dzania Lasu, 2003: Ministerstwo Ochrony .środowiska, Zasobów Naturalnych i Leśnictwa. Dyrekcja Generalna Lasów Państwowych. Warszawa.
• 7. Jaworski A., 1994: Charakterystyka hodowlana drzew leśnych. Gutenberg, Kraków.
• 8. Kayitakire F., Farcy C., Defourny P., Culvenor D.C., 2002: IKONOS-2 imagery potential for forest stands mapping. http://www.enge.ucl.ac.be/staff/curr/kayitaki/forestsat.pdf., ss. 11.
• 9. Kristóf D., Csató E., Ritter D., 2002: Application of high-resolution satellite images in forestry and habitat mapping . evaluation of IKONOS images through a Hungarian case study; http://www.isprs.org/commission4/ proceedings/pdfpapers/349.pdf
• 10. Kurczyński Z., Wolniewicz W., 2002: Co oznacza piksel poniżej metra? [W:] Geodeta . Magazyn Geoinformacyjny 8(87).
• 11. Sasakawa H., Tsuyuki S. 2003: Development of forest type classification technique for the mixed forest with coniferous and broad-leaved species using the high resolution satellite data; http://definiens-imaging.com/ documents/reference2003.htm
• 12. Toutin T., Cheng P., 2002: QuickBird . A Milestone for High Resolution Mapping, Earth Observation Magazine, vol. 11, no. 4, ss. 14-18.
• 13. Wężyk P., 1998: Kompleksowe wykorzystanie informacji ze zdjęć lotniczych. Część III. Szczegółowe aplikacje zdjęć lotniczych w różnych dziedzinach gospodarki. Skrypt przygotowany na zamówienie Głównego Geodety Kraju w ramach projektu PHARE PL.9206-02-04/II. A3-10 . A3-37. Kraków-Sieradz.
• 14. Wężyk P., 2004: GPS w leśnictwie i ochronie przyrody. Mity i fakty Roczniki Geomatyki , Tom II, Zeszyt 4, ss. 19- 32.
• 15. Wężyk P., de Kok R., Zajaczkowski G., 2004: The role of statistical and structural texture analysis in VHR image analysis for forest applications. A case study on QuickBird data in the Niepolomice Forest. [In:] Strobl et al. Hrsg. Angewandte Geoinformatik 2004; ss. 770-775. Wichman. Heidelberg.
• 16. Wężyk P., de Kok R., 2005: Automatic mapping of the dynamics of forest succession on abandoned parcels in south Poland. [In:] Strobl et al. Hrsg. Angewandte Geoinformatik 2005; ss. 774-779. Wichman.Heidelberg .
• 17. Zająlczkowski G., Wężyk P., 2004: Techniki teledetekcyjne w inwentaryzacji urządzeniowej lasu. Roczniki Geomatyki, Tom II, Zeszyt 4, ss. 41-50.
• 18. Zawila-Niedzwiedzki T., 1994: Ocena stanu lasu w ekosystemach zagrożonych z wykorzystaniem zdjęć satelitarnych i systemu informacji przestrzennej. Prace Instytutu Geodezji i Kartografii, Tom XLI, Zeszyt 90.