Computational virtual measurement for trees

Abstract

National forest inventory (NFI) is a systematic sampling method to collect forest information, including tree parameters, site conditions, and auxiliary data. The sample plot measurement is the key work in NFI. However, compared to the techniques 100 years ago, measuring methods and data-processing (modeling) approaches for NFI sample plots have been improved to a minor extent. The limit was that the newly-developed methods introduced additional validation workflows and would increase the workload in NFI. That was due to that these methods were usually developed based on species-specific and site-specific strategies. In order to overcome these obstacles, the integration of the novel measuring instruments is in urgent need, e.g., light detection and ranging (LiDAR) and the corresponding data processing methods with NFI. Given these situations, this thesis proposed a novel computational virtual measurement (CVM) method for the determination of tree parameters without the need for validation. Primarily, CVM is a physical simulation method and works as a virtual measuring instrument. CVM measures raw data, e.g., LiDAR point clouds and tree models, by the simulation of the physical mechanism of measuring instruments and natural phenomena. Based on the theory of CVM, this thesis is a systematic description of how to develop virtual measuring instruments. The first work is to introduce the CVM theory. CVM is a conceptual and general methodology, which is different from a specific measurement of tree parameters. Then, the feasibility of CVM was tested using a conceptual implementation, i.e., virtual ruler. The development of virtual ruler demonstrated the two key differences between CVM and conventional modeling methods. Firstly, the research focus of CVM is to build an appropriate physical scenario instead of finding a mathematical relationship between modeling results and true values. Secondly, the CVM outputs can approach true values, whereas the modeling results could not. Consequently, in a virtual space, tree parameters are determined by a measuring process without mathematical predictions. Accordingly, the result is free of validation and can be regarded as true values, at least in virtual spaces. With the knowledge from the virtual ruler development, two exceptional implementations are further developed. They are the virtual water displacement (VWD) method and sunlight analysis method. Both of them employ the same CVM workflow, which is firstly measured in reality and secondly measured in virtual space. The VWD aims to virtually measure the point clouds using the simulation of water displacement methods in reality. There are two stages in this method. The first stage is to apply the simulation of water displacement using massive virtual water molecules (VWMs). Some empirical regressions have to be employed in this stage, due to the limitation of computer performance. In the second stage, a single (or few) VWM (or VWMs) is developed to remove those empirical processes in VWD. Finally, VWD can function as a fully automatic method to measure point clouds.The sunlight analysis method aims to virtually measure the tree models using the simulation of solar illumination during daylight. There are also two stages in this method. The first stage is to develop sunlight analysis for a single tree. The second stage is to analyze the interference from neighboring trees. The results include default tree attributes, which can be collected in the future NFI. The successful developments of CVM, along with implementations of VWD and sunlight analysis methods, prove the initial assumptions in this thesis. It is the conversion of mathematical processing of data into virtual measurements. Accordingly, this is a different philosophy, i.e., the role of data is extended to the digital representative of trees. It opens an avenue of data processing using a more natural approach and is expected to be employed in the near future as a standard measuring instrument, such as a diameter tape, in NFI.