Developing full waveform, Bayesian analysis for Multi-Spectral Canopy LiDAR (MSCL) images

点击翻译按钮获取中文摘要

We aim to develop image and signal processing algorithms for a new type of air or space borne, remote sensing, 3D imaging LiDAR system designed to measure forest photosynthetic activity in three dimensions. We want to perform full waveform, multi-spectral signal analysis to conduct detailed structural and physiological measurements on forest ecosystems. The requirement is to better interpret data describing the geometry (forest canopy height, height profile, and fractional cover) and physiological signature (photosynthesis, transpiration, somatal response) of trees and vegetation above the earth's surface. By providing better understanding of the data collected from aerial and satellite imaging of forest ecosystems, the proposed research will allow us to better monitor landscape dynamics and the carbon cycle, which is a key factor in the prediction of climate change.Our project is a discipline 'hop' and has 5 phasesPhase 1: Establishing an Inter-Disciplinary Programme: This is a period of familiarisation as the applicant works with researchers at the Edinburgh Earth Observatory (EEO) to better understand the instruments and use of existing software for measurement and interpretation. Phase 2: Developing Bayesian techniques for processing Multi-spectral Canoy LiDAR (MCSL) data: we shall then extend and apply existing processing techniques to the remotely sensed LiDAR imagery to see whether we can gain significant improvement in structural imagery. Allied to this, we should investigate the use of better structural models for the forest canopy scenario, and so develop the algorithms. Task 3: Encoding the algorithms for use within EEO instruments: We need to incorporate the mutual information inherent in several wavelengths in reconstructing better MCSL imagery. The reconstruction of spatial structure and the reflectance analysis (classification) becomes one of drawing posterior inferences from data. Given the mathematical model that we propose, a significant activity will be the development of well structured and documented code to process the LiDAR data. As the project proceeds, we need to encode and document the original software developed in this project so it can be readily used by other researchers.Task 4: Evaluation and trials: As we develop the methodology, we need to assess its effectiveness on data provided by EEO. Structurally, we need to assess whether we can create more accurate 3D forest canopy and ground structure in the presence of significant visual 'clutter' and other confusing factors. Spectrally, we must go beyond current practice in extracting useful data from a series of spectral profiles. Throughout the proposed programme, EOl will be carrying out laboratory and field investigations with MCSL instruments, both existing and new. Measurements will be carried out over an extensive wavelength range in the range 0.4-2.5um using contrasting vegetation types at different growth stages, to be examined over time with different hydrological conditions to observe hyperspectral backscatter.Task 5: Pump-priming and collaboration: A key task will be to bring together the signal processing and geoscience communities to develop further cross-disciplinary activities. At HWU and within the ERPem pooling inititiative (www.erp.ac.uk) we have many staff studying the theory of signal processing in a single and several dimensions, the representation and modelling of sensors and scenes, innovative image and signal processing technologies, image and signal controlled autonomous systems, and systems that model the human-technology collaboration. We would organise pump-priming workshops on key problems with in-house and invited speakers, followed by break-out sessions to develop research and technology transfer proposals. The applicant would assume primary responsibility for their organisation, in consultation with academic staff at the home and host institutions.