Automated tree crown delineation and detection in UltraCam-D Digital image

Document Type : Scientific article

Authors

1 Assistant Professor, Faculty of Agriculture, University of Yasooj

2 MSc. Student of photogrammetry, Faculty of Civil Engineering, University of Tafresh

3 Assistant Prof., Faculty of Environment and Energy, Tehran Science and Research Branch, Islamic Azad University

4 Ph.D. Student of forestry, Faculty of Agriculture, University of Khoramabad

Abstract

This study aimed to evaluate the results of applying Local maximum filtering, Template matching and Watershed segmentation algorithms on aerial image of UltraCam-D to delineate and detect automatically the single tree crowns of Persian Oak (Quercus brantii Lind) in comparison with the results of visual interpretation techniques and the filed measurement method of crown covers. After preprocessing of image, in a terrain with an area of 10 ha inside Yasooj forest park, 100 trees of Persian oak were selected randomly. The crown area of the selected trees was determined and calculated using visual interpretation and was accepted as the control data. Using the field inventory, the areas of the tree crowns were measured and the numbers of sprouts in each coppice form were counted. Using the field inventory, the areas of crowns of trees were measured and the numbers of sprouts in each coppice form were counted and recorded. Moreover, to delineate and recognize the crown of trees automatically, using programming, the Local maximum filtering, Template matching and Watershed segmentation algorithms, were applied on the mentioned image. The results showed that the accuracy of Watershed segmentation algorithm is better than the method of field measurement and it was 2.41 percent of the control method. Total accuracy and kappa coefficient obtained by error matrix algorithms for each algorithms of Local maximum filtering, Watershed segmentation and Template matching showed that Template matching algorithm was more accurate to detection crowns, or in other words, to detect single base or coppice trees than other algorithms used in this study.

Keywords

Forest Research and Development
Volume 2, Issue 3
December 2016
Pages 241-256

Files

Share

How to cite

Statistics