Tree Analysis Using Lidar

Tomáš Kafka
February 1, 2022

Tree protection zones detection from Airborne Laser Scanning Data

Introduction

   For efficient and comprehensive planning of utility networks, it is necessary to take into account not only the factors affecting the total cost of network construction (e.g. by minimizing the length of excavations, optimizing routes with respect to the cost of surface fix, efficient use of the capacity of the infrastructure to be laid, etc.), but also the factors related to the approval process (taking into account land ownership, considering the presence of other networks) and also the envisaged technology used for the construction based on local conditions (manual or machine excavation, drilling). A typical complication of the construction works, especially in the deep interconnectedness of towns, is the presence of mature greenery, where a protective buffer zone applies following national standards and laws. The protective zone is derived from the type of tree, the circumference of its trunk and the area of its crown.

   Passing through a tree protection zone usually requires the application of appropriate protective measures and excavation can mostly only be done by hand to avoid damaging the tree and its root system. If the designer has enough high quality information about the greenery in the area, then trenching through the protective zone around the trees can be avoided in the initial design of the routes, thus facilitating, and speeding up the subsequent work associated with the detailed survey and subsequent redesign.

Automatic tree detection from aerial data collection

   If a database of tree features is not available for a given site, raster data from aerial (orthophoto) imagery (RGB, CIR) are often used as input for this activity. However, the automatic processing of image data for tree detection often encounters problems such as insufficient resolution (see Fig. 2), significant shadows in the image (Fig. 3), oblique representation of tall trees or even the fact that the image was taken during the dormant season and the deciduous trees are leafless (Fig. 4). In such a case, the analysis of poor-quality image data for the purpose of accurate mapping of the buffer zones of mature trees does not provide accurate data or it could even provide and information that is misleading from reality.

Fig.2 - Low resolution raster data [Bing maps]
Fig.3 – Dark shadows cast by the tree canopy [Google maps]
Fig. 4 - Aerial image taken during the dormant season [Ortofoto DOP NRW]

   A very accurate technological procedure for tree detection is offered by the aerial lidar data. Lidar data is collected in dense point clouds where each point carries additional added information besides the relative height, it is also reflection index, scan angle, interpreted classification, etc. Different techniques, described in the literature, e.g. Roussel J.R.; Auty D.; Coops N.C.; Tompalski P.; Goodbody T.R.H.; Sánchez Meador A.; Bourdon J.-F., de Boissieu F., Achim A., 2020 [1], can be used to derive a Canopy Height Model (CHM), which can be interpreted as the difference between a digital terrain model and a digital surface model. Further algorithms applied to the CHM can then be used to progressively arrive at a detailed three-dimensional tree model. This procedure is also described by You, H.; Li, S.; Xu, Y.; He, Z.; Wang, D., 2021 [2]: "Firstly, the points in flat areas are distinguished from those in undulating areas, based on the integration of the flat distance and elevation difference, which filters out ground and roof points. Then, incorporating morphological characteristics of tree canopies and other interfering objects, a tree point extraction on the basis of point count is established in the undulating areas. Finally, a further tree point refinement step is deployed, using a qualified Euclidean cluster extraction (ECE) algorithm."

Fig. 5 - Extracted trees. You, H.; Li, S.; Xu, Y.; He, Z. 2021 [2].

   For utility planning, 3D tree models are unnecessarily too detailed information that is algorithmically challenging to derive. A polygon representing the area of the tree buffer zone would be just fully sufficient information. In this pilot study, the following simplified yet efficient procedure was used to derive it.

Demonstration of the use of publicly available lidar data to detect buffer zones around trees

Fig.6 –ALS data coverage in NRW
[https://www.bezreg-koeln.nrw.de/brk_internet/geobasis/hoehenmodelle/3d-messdaten/index.html]

   The German Regional Office in Cologne and the Regional Office for Information and Technology North Rhine-Westphalia publish lidar data on opengeodata.nrw.de portal. The dataset is freely available and covers the entire state of North Rhine-Westphalia (NRW). The measurements were taken between 2015 and 2021 (Figure 6) and are scheduled to be regularly updated on an ongoing basis until 2027. The data are acquired by airborne laser scanning (ALS), with a point density of 4 to 10 points per square meter. The point classification distinguishes between different types of surface points (see Table 1).

Source and more info at: https://www.bezreg-koeln.nrw.de/brk_internet/geobasis/hoehenmodelle/3d-messdaten/index.html

Table 1 -Classification of ALS data
[https://www.bezreg-koeln.nrw.de/brk_internet/geobasis/hoehenmodelle/nutzerinformationen.pdf]

The classified classes of ALS points according to the above table are:

  • 2 - terrain, ground points,
  • 24 - basement points (drains, shaft under the natural terrain),
  • 17 - bridges (elevation points assigned to the roadway),
  • 26 - derived land points (under bridges),
  • 21 - derived land points (under buildings),
  • 9 - derived water surfaces,
  • 20 - last focused points not lying on the ground,
  • 1 - unclassified elevation points (e.g. middle returns from vegetation)
  • 18 - High ground points considered as noise (e.g. flying bird).

   Another important parameter of the measured lidar points is the return number. When a lidar beam passes through vegetation, it passes through a specific type of surface that partially reflects it back and partially transmits it further. As shown in Figure 7, this results in several times reflection. The lidar data carries information whether the point represent the first, last or middle reflection.

Fig. 7 - Type of data by response
[https://www.bezreg-koeln.nrw.de/brk_internet/geobasis/hoehenmodelle/nutzerinformationen.pdf]

   An illustrative explanation of data classification and response type is given in Figure 8 along with Table 2.

Fig. 8 + Table 2 -Explanation of ALS data
[https://www.bezreg-koeln.nrw.de/brk_internet/geobasis/hoehenmodelle/nutzerinformationen.pdf]

   From the publicly available airborne lidar data (published at opengeodata.nrw.de) with point classification and return number, an algorithm for automatic tree detection was developed. A test area near the city of Minden (Fig. 9) was selected as the area of interest, where a telecommunication network design project was simultaneously processed.

Fig. 9 – Minden [Opentopomap.org]

A quick preview of the downloaded data (without the need to install specific lidar data processing software) is provided by the online application https://plas.io/, where the data can be for a clearer visualisation colour-classified - for example by height (Fig. 10) or by class (Fig. 11).

Fig. 10 - Viewing lidar data in plas.io - colour representation of height
Fig. 11 - Viewing lidar data in plas.io - colour representation of class

   The software "LAS tools" [http://lastools.org/] can be used to analyse, extract, or convert the data from LAZ format to GIS environment. According to the description and explanation of the data properties from above, only points that do not represent a surface (not the first or last return, the signal has passed through the object more than once) and points that do not have a specific classification assigned (the point has not been identified as part of a building object or surface type) are relevant for the purposes of tree detection (see Fig. 12).

Fig. 12 - Filtered points for tree detection purposes

   The extracted point clouds are further processed in the GIS environment. First, the data is cleaned using cluster analysis (DB SCAN), which will separate only significant spatially large and numerous data clusters from rather solitary points or small spatially insignificant groups. The result of the cluster analysis is converted into a polygon representation with a certain buffer zone (Fig. 12). These polygons then represent the protection areas around mature trees that can be taken account in possible trenching plan, or the cost of the expected hand excavation works can be included in the calculation of construction work costs.

Figure 13 - Protection zones of detected trees

Summary

   The use of lidar aerial photography (ALS) data is an accurate method with a clearly defined mathematical basis using the physical property of laser beam reflections when the beam passes through vegetation. The method described above for tree detection is a simplified process that uses an already established data classification and the return number attribute. For deriving the protective area around a tree, this method provides a reliable fully automated procedure that can be processed by publicly available software (QGIS, LAStools). However, for wider application in planning services, the (affordable) availability of lidar data is the limitation. The approach of NRW Regional Office for Information and Technology with opening the lidar data to the public is to this date extra ordinary service that offers a great playground for lidar data processing testing. In a real project, depending on the size of the territory and the total volume of the lidar data, some challenge for the IT infrastructure (disk space, processor, operating memory) can be faced. The outputs of the lidar analysis for tree detection will facilitate the planner to plan the routing more efficiently and to make more accurate estimate of the costs.

   If you are planning in NRW region and you are interested in tree polygon data, feel free to contact us through Yungo website. If you are interested in more details or future cooperation, don’t hesitate to do as well.

[1] Roussel J.R.; Auty D.; Coops N.C.; Tompalski P.; Goodbody T.R.H.; Sánchez Meador A.; Bourdon J.-F., de Boissieu F., Achim A.: lidR: An R package for analysis of Airborne Laser Scanning (ALS) data. Remote Sensing of Environment, Volume 251, 2020, 112061, ISSN 0034-4257. https://doi.org/10.1016/j.rse.2020.112061. https://www.sciencedirect.com/science/article/pii/S0034425720304314

[2] You, H.; Li, S.; Xu, Y.; He, Z.: Wang, D. Tree Extraction from Airborne Laser Scanning Data in Urban Areas. Remote Sens. 2021, 13,3428. https://doi.org/10.3390/rs13173428 https://www.mdpi.com/2072-4292/13/17/3428

Editor:
Dmytro Kovtoniuk
Translated:

You may also like

September 5, 2022

YUNGO turns 6 years old!

Today we are celebrating YUNGO's 6th birthday. It's been a great journey so far

Read more
September 1, 2022

Your Poop Next to Fibre Loop

How a unique solution for the installation of optical cables in sewers was created in Czechia

Read more
August 10, 2022

YUNGO - A Digital Riser

YUNGO grows by taking on challenges. We push ourselves to tackle new tasks and push the boundaries of our work!

Read more