Co-registration of multi-sensor UAV imagery. Case study: Boreal forest areas

Abstract

ABSTRACT Monitoring the regeneration process of a forest is an important part of forestry management. Compared to traditional methods of counting tree species, UAVs have been a revolutionary means of saving time and costs due to the temporal and spatial flexibility of data collection. In turn, the integration of multispectral cameras allows the traditional vegetation indices that have been used with satellite imagery to be obtained. However, data from multispectral cameras must be combined with data from other types of sensors, such as RGB. It is therefore necessary to co-register all the information in order to obtain combined vegetation indices and carry out segmentation processes that allow the identification of the different tree species. In this study, the coordinate transformation methods available in QGIS software through the georeferencer plugin are evaluated. It also studies the influence of the number and distribution of control points on the accuracy of the transformation. It is concluded that of the transformation methods studied, TPS transformation has the highest accuracy with an MAE of 0.9 pixels and a deviation of 0.6 pixels, providing a minimum of 10 control points and a stratified or edge distribution.