树(集合论)
霍夫变换
激光扫描
计算机科学
遥感
均方误差
点云
职位(财务)
人工智能
计算机视觉
数学
激光器
地理
统计
图像(数学)
光学
数学分析
物理
财务
经济
作者
Maja Michałowska,Jacek Rapiński,Joanna Janicka
标识
DOI:10.1016/j.rsase.2022.100863
摘要
Forest management and planning require information regarding the current state of the forest. Remote sensing techniques allow to obtain geospatial data, also for the forestry sector. As one of the remote-sensed technologies datasets, Terrestrial Laser Scanning data is widely used to derive detailed information about tree and forest stand parameters. This article presents the combination of circular Hough transform, denoising procedure, and robust least-square circle fitting method to extract stem positions from Terrestrial Laser Scanning data. In the proposed approach, initial tree stems position was detected with circular Hough transform. Then, obtained results were denoised to exclude most non-tree trunk points and analyze three-dimensional data from laser scanning to find exact circular tree stems with a robust least-square circle fitting method. The developed algorithm is effective in obtaining the trees' geodetic positions from laser scanning data. The results generated in this study can be used as basics for further automatic determination of tree characteristics, such as tree species, height, or crown range. In this study, 94.8% tree stems delineation was generated with a mean accuracy of 87.2%, 1.64 cm of root mean square error for stem position, and 1.15 cm for tree radius measured at ground level. The process conducted in this research can be used to detect other circle-shaped objects, such as lamps or power towers, for which obtaining dense Terrestrial Laser Scanning data is available. The detected positions of these objects can power the geographic information systems or thematic industry systems, where it is necessary to determine the geodetic object position results from legal regulations.
科研通智能强力驱动
Strongly Powered by AbleSci AI