ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume II-2/W2
https://doi.org/10.5194/isprsannals-II-2-W2-175-2015
https://doi.org/10.5194/isprsannals-II-2-W2-175-2015
19 Oct 2015
 | 19 Oct 2015

IDENTIFICATION AND MAPPING OF TREE SPECIES IN URBAN AREAS USING WORLDVIEW-2 IMAGERY

Y. T. Mustafa, H. N. Habeeb, A. Stein, and F. Y. Sulaiman

Keywords: Urban tree species, Supervised classification, VHR imagery, Kurdistan Region of Iraq

Abstract. Monitoring and mapping of urban trees are essential to provide urban forestry authorities with timely and consistent information. Modern techniques increasingly facilitate these tasks, but require the development of semi-automatic tree detection and classification methods. In this article, we propose an approach to delineate and map the crown of 15 tree species in the city of Duhok, Kurdistan Region of Iraq using WorldView-2 (WV-2) imagery. A tree crown object is identified first and is subsequently delineated as an image object (IO) using vegetation indices and texture measurements. Next, three classification methods: Maximum Likelihood, Neural Network, and Support Vector Machine were used to classify IOs using selected IO features. The best results are obtained with Support Vector Machine classification that gives the best map of urban tree species in Duhok. The overall accuracy was between 60.93% to 88.92% and κ-coefficient was between 0.57 to 0.75. We conclude that fifteen tree species were identified and mapped at a satisfactory accuracy in urban areas of this study.