Smart pointclouds

Point cloud data of a forest

Smart point clouds research group investigates novel point cloud technologies, and how they can benefit the society and different fields of industry, such as the forest, construction, game, energy, and transportation industries.

The group develops disruptive technologies to create solutions to current and future challenges such as climate change, urbanisation, natural disasters, energy distribution, safety, autonomous transport, ageing of population, and loss of biodiversity. Technological breakthroughs are expected to create novel fields of industry.

The concept of smart point clouds means collecting the point clouds in intelligent ways. It means exploring the best solutions for digitising the world accurately, efficiently and comprehensively by developing and testing intelligent platforms and novel sensors. New platforms such as UAVs or new sensors, such as single-photon lidar are some of our main focus areas.

The smart point clouds also means interpreting the point clouds using novel methods and algorithms, including the full spectrum of machine learning to provide new and smart solutions and services to public and private sector as well as common people.

Revolutionising practices by disruptive point cloud technologies

Applying modern platforms, such as UAVs and personal backpacks, to collect point clouds, will revolutionise current practices, for example, in forest management, mapping and energy industries, and in environmental monitoring and studies. Remarkable improvements in the efficiency and productivity are expected through the application of the novel technologies.

The focus of study of the Smart point clouds group is the analysis of autonomously collected  in-situ LIDAR and image data. The group has been pioneering in investigating the use of terrestrial image-based point clouds in the mapping of individual trees in a forest plot (See also). In cooperation with Mobile Mapping and Autonomous Driving research group, we were the first to apply personal (backpack) and mobile laser scanning (all-terrain vehicle) in the mapping of large forest plots> See also, >See also

In 2017, the group started to investigate the collection of forest in-situ measurements using a UAV platform.

UAV mapping a forest

all-terrain vehicle-based and personal laser scanning of forest
UAV-based, all-terrain-vehicle-based and personal laser scanning of forest. Pictures: UAV/ Timo Toivonen, Backpack/Antero Kukko, ATV/ Samuli Junttila.

Understanding point clouds

Through exhaustive studies, the group dedicates itself to gaining deeper knowledge of the point cloud technologies. This information helps practitioners in selecting proper techniques in their applications to maximize profit, and contributes to research by laying down milestones and by recognising new important directions for further studies.

ALS pointcloud of trees
Tree height measurement from dense airborne laser scanning (ALS) point cloud (red points) is more reliable than from conventional field observations.

Benchmarking and comparison studies

Smart point cloud group hosted the international benchmarking study on the performance of terrestrial laser scanning (TLS) in forest inventory between 2014 and 2018. 18 different algorithms were benchmarked using a common TLS data set, a highly detailed reference data, and a standardised evaluation procedure. The performance and applicability of the TLS technology was assessed in various forest conditions with respect to forest structure and management activity.

In 2015, the group compared various data sources, such as, laser, stereo optical, SAR, and InSAR point clouds from air- and space-borne platforms in the retrieval of forest inventory attributes (Yu et al.) The results confirmed the high potential of 3D remote sensing data from space-borne high-resolution images for forest inventory purposes, and also specified the needs for further algorithm development in forest data processing.

The group has also been a host of two international benchmarking studies on the performance of  aerial laser scanning (ALS) in individual tree detection (ITD) (Kaartinen et al., Wang et al.). The studies provided understanding of the performance of automatic individual tree detection by comparing more than 20 different algorithms.

In 2019, the group hosts an international benchmarking study which evaluates the performance of terrestrial image-based point clouds in plot-level forest inventory, funded by ISPRS Scientific Initiatives.

point cloud picture of a forest
Image-based, or Structure from Motion (SFM), point clouds on forest plots as a low-cost alternative to terrestrial laser scanning.

Exploring disruptive technologies

The group develops and applies disruptive technologies to discover the solutions for environmental digitization and cognition. An extensive study on tree height estimation was performed in 2019 (Wang et al., 2019) using a dataset of over 1100 trees. Three measurement techniques were compared in the estimation of the heights of individual trees, ALS, TLS, and conventional field observations. It was shown that, when the point density is high (450 pts/m²), aerial point cloud technology outperformed the traditional field measurements.

Pointclouds of trees
Laser scanner in a UAV flying on top of the canopy records the upper parts of the trees thoroughly while for some species and forest conditions occlusions exist in the lower parts.


A comprehensive investigation of a plot-level forest inventory using a high-end UAV laser scanner was performed in 2019 (Liang et al., 2019). The performance of the technology was assessed in various forest conditions. Compared to the state-of-the-art terrestrial point cloud technologies, UAV exhibited comparable or higher performance in homogeneous forests and in the tree height estimation. The study also revealed challenges that the UAV platform is facing. The target visibility varies significantly. For example, the tree trunk may be totally or partly missed, or fully recorded depending on the forest structure, conditions and viewing geometry.


The Smart point clouds group applies the smart point cloud technologies to different environments, such as forest and build environment as well as different kinds of corridors, such as power line and road corridors, together with the Mobile Mapping and Autonomous Driving research group.

Accurate and efficient corridor mapping

The group has studied how point clouds could be applied in road corridor mapping, such as in the mapping of road infrastructure (Lehtomäki et al., 2010; Lehtomäki et al., 2011) and object recognition (Lehtomäki et al., 2016).

The point-cloud-based corridor mapping could aid in matching the future needs of efficiency and accuracy in, for example, inventory of road infrastructure and planning and maintenance of roads and streets, as well as, in the inspection of power line corridors. The accurate 3D point cloud data could also be used in applications such as 3D city modelling, virtual reality, and personal navigation.

Mobile laser scanning could be an efficient technique for power line corridor inspection—either alone or to support aerial inspection. In 2019, Lehtomäki et al. demonstrated how all-terrain mobile laser scanning can be used to map power lines automatically outside the road network.

Point cloud of a road
Pole-like objects (e.g., traffic signs and lamp posts) have been detected automatically from mobile laser scanning point cloud collected in a road environment. (Lehtomaki et al., 2010)

Smart point clouds in forest

The group applied mobile platforms in forest plots to collect in-situ measurements needed, for example, in national forest inventories (Liang et al., 2018a). Conventional field measurements are laborious and time-consuming. Mobile and personal laser scanning technology can boost the field reference data collection. The group is studying how point clouds could be collected intelligently and efficiently in order to meet the needs of future forest inventories.

Application of TLS point clouds in wood quality estimation has shown that wood quality could be linked linearly to the 3D-features retrieved from TLS point clouds (Pyörälä et al. 2018a, Pyörälä et al. 2019b). The group has also studied automated methods to obtain wood quality indicators from TLS point clouds. While the spatial resolution of TLS limited the range of visibility in upper canopies, the lower canopy metrics were accurately obtainable from the point clouds with automated methods (Pyörälä et al. 2018b, et al. Pyörälä 2019a). In a future outlook, forest inventories could provide detailed field references that could link, for example, wood quality data in sawmills or empirical models of wood properties.

Point clouds and 3D modeling of trees
High-density aerial point clouds enable delineating individual trees from canopy height models, and describing the tree crowns in 3D. In addition, terrestrial point clouds from sample plot enable modelling stem geometry and branching that can be linearly linked to as well to select wood quality variables in sawmill data as to the ALS crown features.


Publications of the group members:

Juha Hyyppä (Profile in ResearchGate); Xinlian Liang (Profile in ResearchGate); Markus Holopainen (Profile in ResearchGate); Matti Lehtomäki (Profile in ResearchGate); Jiri Pyörälä (Profile in ResearchGate); Yunsheng Wang (Profile in ResearchGate); Xiaowei Yu.


In co-operation with: Centre of Excellence in Laser Scanning Research ,  DroneFinland

Contact persons