RESULTS AVAILABLE FOR INTERNATIONAL BENCHMARKING OF TERRESTRIAL LASER SCANNING METHODS
The potential of terrestrial laser scanning (TLS) in forest applications, both in industry and in national forest inventories (NFIs), has gained increasing awareness in the last two decades. Boosting this development, there has been tremendous research efforts and progress on the topic.
In 2014, a TLS benchmarking project was launched by the European Spatial Data Research Organization (EuroSDR) and hosted by the Finnish Geospatial Research Institute (FGI). 18 groups from Asia, Europe and North America have successfully processed the data and submitted their results for evaluation. Latest results of the work are to be published (link to publication coming soon) in ISPRS Journal of Photogrammetry and Remote Sensing along with open data on the sample plots.
Three complexity categories of the sample plots in the TLS benchmarking: “Easy” (left) represents sparser stem densities and little understory vegetation, “Medium” (middle) represents moderate stem densities and sparse understory vegetation, and “Difficult” (right) represents high stem densities with dense understory vegetation. © Centre of Excellence in Laser Scanning Research, Xinlian Liang.
Achievements and remaining barriers to TLS-based forest investigations
Terrestrial laser scanning is an effective technique for measuring forest plots. A terrestrial laser scanner automatically documents its surrounding environment in three-dimensional (3D) space with millions to billions of 3D points. The technology is anticipated to be practically used in forest in-situ observations, e.g. national forest inventories.
According to the benchmarking results, TLS is great at extracting forest attributes of trees that are recorded in the point cloud data. These attributes, i.e. diameter at breast height and stem curve estimates, can be measured at 1-2 cm accuracy level which is close to what is required in practical applications, e.g. national forest inventories. Remaining barriers are mainly from the occlusion effects that prevail and hard to eliminate in forests; therefore, attributes of trees such as tree heights are difficult to estimate.
Five criteria extracted from TLS data at the plot- and tree-level. © Centre of Excellence in Laser Scannning Research, Xinlian Liang.
Open data on sample plot pointclouds
The single- and multi-TLS data, as well as the reference data, of six sample plots of this benchmarking project can be downloaded from here (link coming soon).
The data is opened to anyone who is interested in TLS based forest investigation for non-commercial use. Since the reference information is also opened, users can evaluate their own feature extraction approaches and results by themselves, and compare their own results with the other results reported in the project publication. When making a publication based on the open data set, please reference the date source of the ISPRS journal paper (link to publication).
Getting ready for operational use
The results of this international benchmarking suggest that TLS as well as sophisticated automated algorithms are ready to be used in practice under easy forest conditions, e.g., forests with good visibility or single trees. The operational use of TLS in forest mensuration is a complex process. The current bottleneck is the lack of practical software and it requires time to develop such software. The emerging mobile and personal laser scanning (MLS/PLS) have the potential to solve the occlusion problems of TLS, but have yet to achieve similar accuracy . It is worth noting that TLS currently provides the best quality terrestrial point clouds in comparison with all other technologies, meaning that all the benchmarks labeled in this our publication (link) can also serve as a reference for other terrestrial point clouds sources.
More information: Research manager Xinlian Liang, firstname.lastname(a)nls.fi
