Für den Inhalt der Angaben zeichnet die Projektleitung verantwortlich.
Dieses von der Gebert Rüf Stiftung geförderte Projekt wird von folgenden weiteren Projektpartnern mitgetragen: Hochschule für Agrar-, Forst- und Lebensmittelwissenschaften HAFL; senseFly SA (drone manufacturer); Pix4D SA (processing software company), ecorisQ (developer of the single tree detection software).
Förderbeitrag: CHF 30'000
Dauer: 03.2018 - 03.2019
Handlungsfeld: Pilotprojekte, 1998 - 2018
Hochschule für Agrar-, Forst- und Lebensmittelwissenschaften HAFL
3052 Zollikofen (Schweiz)
- mark.guenter@bfh. ch
Building with wood is booming, but paradoxically Swiss forest owners and sawmills benefit only to a limited extent. 70% of the processed wood comes from abroad and Swiss forest owners, especially private individuals, are not currently doing much to ensure that more Swiss wood comes onto the market - the current timber price is too low for them. The abolition of the minimum euro exchange rate in January 2015 has aggravated the situation even further, because Swiss timber processors had to reduce their prices by an additional 15%. Political measures against the misery of the wood processors are currently being hotly debated. Parliament is discussing a revised forest law in which Swiss timber is to be given preference. However, only those who are efficient, innovative and rely on automation can survive on the market in the long term. With their prompt, cost-effective and yet very precise terrain and inventory data, drones offer the possibility of digitally recording the dynamic processes in the forest and thus enable the precise planning of timber harvesting, rejuvenation or other interventions. Single tree detection software uses drone data as the basis for estimating important tree parameters (tree position, height and diameter). This research project investigates how the workflow with drone data can be optimized and how single tree detection software can be used to increase the efficiency of forest management.
Was ist das Besondere an diesem Projekt?
Drone images are commonly used today as optical support in the forestry sector. However, the potential of drone data and parameters that can be generated from single tree detection software is far from exhausted. The innovative and creative aspect of this project is to create a digital twin of the forest. This twin provides all important tree parameters for the researchers to model the forest, make estimations of interventions, plan and make predictions.
Milestone 1 (March-June 2018): Analysis of the accuracy and speed (efficiency) of common terrestrial measuring methods. Measurement of tree parameters (position, breast height diameter, height) in hardwood, mixed wood and softwood forest stands.
Assessing terrestrial measurement methods: In this extensive analysis in a forest stand in Zollikofen, 103 trees were precisely measured using the most accurate methods in forestry today. Tree positions were measured with a Leica theodolite, tree diameters with a measuring tape and tree heights with a Vertex ultrasound device. These measurements were used as benchmarks to assess the accuracy and speed (efficiency) of popular measuring methods. Recommendation for breast height diameter (BHD): Caliper – fastest method and only slightly less accurate than the measuring band. Recommendation for tree height: TruPulse laser – fastest method and marginally less accurate than the Suunto clinometer, Leica Disto laser or MOTI GPS. Recommendation for determining tree positions: TruPulse laser – slower than GPS, but significantly more precise.
Terrestrial measurement of tree parameters in the hardwood, mixed wood and softwood forest stands: The recommended methods were applied to measure BHD, tree height and tree position in the three types of forest stands. 847 trees were measured in the hardwood forest stand in Rona (GR), 720 trees in the mixed wood forest stand in Zollikofen (BE) and 947 trees in the softwood forest stand in Renan (BE). The results from the measurements were subsequently used to assess the accuracy and efficiency of the proposed digital method with drone data and single tree parameter software.
Milestone 2 (July-September 2018): Optimal flight planning and execution.
Flight planning and execution: Ground sampling distance (GSD) and flight overlaps that directly affect the quality of data were the two parameters studied in this task. GSD is a function of altitude. The higher the altitude, the lower the resolution of the data, so a decrease in the quality of data can be expected. Inversely, flying at higher altitudes can save money since the same area can be covered in less time. Flights were conducted at different altitudes (100m, 140m and 180m) to assess a potential difference in the quality of the data. The tests revealed no significant difference in all three datasets. Furthermore, flights were conducted with varying degrees of lateral and longitudinal overlaps, as well as perpendicular vs. parallel flight trajectories. Results: perpendicular flights are not necessary; minimal lateral overlap = 85% and minimal longitudinal overlap = 80%.
Milestone 3 (October-December 2018): Improvement of the generation of the digital surface model (DSM) and optimal preparation of the drone data in combination with LiDAR data.
The processing of the drone images was done with Pix4D software. Different processing options were tested to find out which settings produced the best results. These “optimal” settings were then saved and applied as a template for the subsequent processing of RGB and NIR images. The digital surface model (DSM) from the drone images was used in combination with the swisstopo LiDAR digital terrain model (DTM). The normalized digital surface model (nDSM) is calculated by subtracting the DTM from the DSM. The nDSM is used as the main input for the single tree detection software FINT. Research was conducted to find out which resolution of the nDSM provided the best results for the different types of forest stands. Tests were also done to see if it is necessary to filter the nDSM before using it in FINT. The results showed that the 50cm nDSM is generally best for softwood stands, and the 1m nDSM for hardwood stands. It is recommended to use the 3x3 low pass filter to delete excessive “noise” in the data. For the mixed-wood stand, the most accurate results were obtained by “calibrating” the nDSM to a certain resolution with the results from sample plots, either taken terrestrially, or optically digitized in GIS. In general, it is good practice to calibrate the nDSM in any case, regardless of the type of trees.
Am Projekt beteiligte Personen
, Projektleiter, wissenschaftlicher Mitarbeiter GIS, HAFL Prof. Dr. Luuk Dorren
, Qualitätsmanagement, Dozent für Naturgefahren und Geoinformationssysteme, HAFL
Peter Zürcher, BSc student in forest sciences, term paper “efficiency comparison of terrestrial measuring methods for recording single tree parameters”
HAFL assistants and students, support in the terrestrial measurement of the forest stands
Letzte Aktualisierung dieser Projektdarstellung 14.01.2019