The analysis of ecological and economic factors of a forest has a long history: The methods of Forest Management Planning date back to the early 18th century. While the economic factors are obvious tools for companies to measure the yield of a forest patch, ecological properties became more and more important worldwide in recent years. In many forests, however, gathering information on timber is neglected because of the lack of qualified personnel, knowledge, or monetary or technological resources. Recently, an increasing number of countries started to look into a more controlled forest economy. Of particular interest is information regarding the diameter at breast height (DBH), stem shape, and tree positions to derive timber volume, quality, and distribution. In particular, the stem shape and existence/thickness of branches as indicators for the quality are missing, because it cannot be measured in an efficient way. Moreover, most of these parameters cannot be obtained accurately enough neither by conventional terrestrial methods nor by remote sensing assessments. Thus, research and engineering to drive technical development must be focused on improved accuracy and operationality at reasonable costs.
New Technology Required
The state-of-the-art for this type of forest inventorying is by means of satellite imagery or airborne lidar for wall-to-wall coverage based on simple calibrations of lidar height against conventional plot inventory. However, besides this method being of low efficiency it is missing any statement of timber quality. A single relation between plot basal area versus lidar height is not universal, and may be fundamentally different in different forest types which highlights the need for more direct observation of the factors in a forest. In this project, we therefore propose to investigate in novel cost and time efficient methods to automate the assessment of forest inventory parameters. We propose to use GPS independent, vision based navigation algorithms, path planning, and vision based 3D reconstruction methods, that will allow a small Unmanned Aerial Vehicle (sUAV) navigate autonomously in the GPS shadowed area under the canopy and reconstruct the tree stems by means of photogrammetry for automated forest parameter-extraction.
Challenging Research Questions
The addressed research questions are:
- GPS independent state estimation, robust against self-similar structure, abrupt light changes, and the highly cluttered environment
- Real-time local motion planners in full 3D, able to avoid obstacles at high speed
- Adaptive global planners for optimal coverage under uncertain initial (tree) distribution
- High-detail 3D reconstruction of the environment and accurate object segmentation for automated forest-parameter extraction
- Integration of above listed technologies into a prototype forest inventory system
We expect that the research in this project will lead to a research prototype aerial vehicle with algorithms that allow autonomous navigation through managed mature forest and that provide sufficiently dense 3D information to automatically extract ecological data like biomass index, position of trees, coverage of the bush layer, diameter as breast height, and stem shape.
The partners from the Alpen-Adria Universität Klagenfurt (project lead), Joanneum Research Forschungsgesellschaft mbH, Lakeside Labs GmbH, Umweltdata GmbH and the E.C.O. Institute for Ecology are working together to reach the goals of this project.
The Alpen-Adria Universität focuses on the GPS independent state estimation with the aim to enable the UAVs to navigate autonomously solely based on visual and inertial information. In addition, a real-time stereoscopic vision system should yield information for obstacle avoidance.
Lakeside Labs GmbH concentrates on the development of a path finding algorithm, which both avoids collision with obstacles based in the stereoscopic information and navigates the drone along a suitable route to retrieve as much usable forest data, as possible.
Joanneum Research Forschungsgesellschaft mbH will investigate on how to use this acquired data to reconstruct the forest with high detail. The goal is to segment and prepare this data such that it can be used in current forest inventorying tools by companies.
Umweltdata GmbH and E.C.O. Institute for Ecology as company partners of the project focus their effort on usage of the generated data and on user feedback such that the technology can be best adapted to the user requirements.
This research is receiving funding from the Austrian Ministry for Transport, Innovation and Technology (BMVIT) within the ICT of the Future Programme (4th call) of the Austrian Research Promotion Agency (FFG) under grant agreement n. 855468 (Forest-IMATE).