Tree species detection with UAV-based hyperspectral imagery

点击翻译按钮获取中文摘要

When trees are replanted after harvest, forest plots must be monitored and managed for the period of early**regrowth to ensure the forest is re-established with a species mix similar to that harvested, after which the plots**are considered free to grow until maturity. First Resource Management Group Inc. (FRMG) provides**sustainable forest management services to forest tenure holders, services that include bringing re-planted forest**plots to this free-to-grow state. To this end, FRMG has developed remote sensing-based technology to monitor**forest plots using a combination of satellite imagery, photos and lidar data acquired from fixed-wing platforms,**and photos taken from helicopter. This allows FRMG to map a) stem density and b) tree height with good**accuracy, and map c) tree species with reasonable accuracy. However, the present approach suffers from two**principal drawbacks. Most importantly, there is considerable risk to human life involved with the acquisition of**low-altitude imagery from helicopter, as several forestry-related helicopter crashes in Canada have resulted in**the deaths of pilots and crew in recent years. Secondly, the identification of tree species with the existing mix**of data sources is not sufficiently accurate, and it is therefore complemented by costly field data collection. The**present project aims to test and integrate two new data sources into the existing technology mix: a) UAV-based**hyperspectral imagery, and b) very high resolution satellite imagery, and to develop automated data processing**algorithms for species identification. We will test how these new data sources allow identification of tree**species on their own, in combination with the existing data sources, and in combination with the other data**sources but excluding the helicopter-based imagery. The project will use field observations from two existing**forest plots in Ontario for calibration and validation of data processing algorithms, and will allow FRMG to**optimize their remote sensing-based approach to forest plot monitoring. The result will also inform the forest**management sector more broadly about the potential of remote sensing technology in forest monitoring. While**the research is based in Ontario, it is application across Canada, and globally.