Terrestrial laser scanning to characterise three-dimensional foliage and woody material distributions in trees

Sasse, F 2019, Terrestrial laser scanning to characterise three-dimensional foliage and woody material distributions in trees , PhD thesis, University of Salford.

Download (18MB) | Preview


Three-dimensional characterisation of foliage and wood distribution within forests is essential for understanding, managing and monitoring forest ecosystems. The recent advances in terrestrial laser scanning (TLS) technologies have provided new opportunities to measure the 3D structure of forest canopies, which in turn can be correlated to tree attributes. In addition to estimation of variables such as stem density and the diameter at breast height and tree height, dual- and multi-wavelength systems are now being tested to distinguish foliage and wood based on their reflectance properties. Previous studies have suggested that using spectral information to distinguish foliage from wood materials is unlikely to provide an accurate classification on its own. In this thesis, a spectral approach was designed based on the frequency distribution of the reflectance and spectral ratios to distinguish between the foliage and woody materials. Additionally, a spatial classifier (CANUPO) approach was applied to describe the geometric relationships between the points of the TLS point clouds and characterise the local dimensionality at a given location and scale. TLS point cloud data of small broadleaf and needle-leaf trees in the laboratory, three single isolated oak trees with different structure and appearance and a full forest stand plot were used for foliage/wood classification in this research. The spectral and spatial classifications were compared to investigate the compatibility between them for all data sets. The results showed a clear separation of foliage and wood using 1063 nm and NDI data for the broadleaf tree and 1545 nm data for the needle-leaf tree. In contrast, the 1545 nm for the broadleaf and 1063 nm and NDI of the needle-leaf tree produced classification errors. A large number of foliage points were classified as wood for both trees using the spatial approach, with comparative errors of 67.35% and 73.18% for the broadleaf and needle-leaf tree respectively. For the three single trees, the 1545 nm data provide a clear separation for all trees while there was a variation in the classification using 1063 and NDI data for every tree. In general, the spatial classifier showed a clear separation for all of the trees with a few apparent errors in the canopy and on the stems with different results according to their structure and appearance. It was unlikely to be possible to separate foliage and wood using spectral data and ratios for the full forest data at ranges greater than 17m from the scanner. CANUPO classified 15% of the points as foliage and 85% as wood at a range of less than 15m. The classification showed a compatibility of 55.63% for the full stand data. Overall, the results highlight the potential of a dual-wavelength laser scanners for providing a wide range of data for forest ecology.

Item Type: Thesis (PhD)
Schools: Schools > School of the Built Environment
Depositing User: Fadal Sasse
Date Deposited: 12 Nov 2019 09:13
Last Modified: 27 Aug 2021 21:29
URI: http://usir.salford.ac.uk/id/eprint/52623

Actions (login required)

Edit record (repository staff only) Edit record (repository staff only)


Downloads per month over past year