Extraction of tree height from large viewing angle aerial images
Lists
Only you see this
Abstract
Due to the complexity and non-regularity of tree shapes, traditional digital photogrammetry using stereo matching method is difficult to obtain the accurate tree height, This fact therefore limits the application of the aerial digital photogrammetry technology in the power line survey. This paper presents a method of tree height extraction from large viewing aerial image using the knowledge of segmented tree crown. This method is based on a rough digital surface model (DSM) of tree crowns and the exterior orientation of the image. The basic steps of this method is that the DSM is first used to find the region of interest in the image based on the exterior orientation, and then the edges of the distinct trees or branches are extracted using image segmentation technology. An algorithm that uses both the rough DSM height information and exterior orientation data to calculate the accurate heights of the segmented trees or branches is presented. The algorithm assumes that most of the trees are upright, and the projection in the large viewing angle images of the crown and branches can therefore be used to calculate their heights relative to the averaged DSM height. Hence, the accurate height of the trees around the rough DSM can be refined. Some experimental results are given with the image captured from multi-angular imaging system mounted on a helicopter in which a Position and Orientation System (POS) is onboard to record the exterior element of the cameras. The experimental results demonstrated that this algorithm can largely improve the accuracy of tree height extraction. The application in power line monitoring system is promising.
BibTeX
@inproceedings{li2005extraction,
author = {Chaoyang Li and Guangjian Yan and Qiang Liu and Zhiqiang Xiao and Jingdi Wang},
title = {{Extraction of tree height from large viewing angle aerial images}},
booktitle = {Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE},
year = {2005},
doi = {10.1117/12.657847},
}Loading related content...