Use of Ground Imagery to Study Wood Raft and Ice Dynamics in Fluvial Systems: Potential and Challenges.
Thursday, 18 December 2014
Automatic cameras allow acquisition of large amounts of information at high resolution in both temporal and spatial dimensions, with a roughly close range. Recently, ground cameras have been used to study the morphological evolution of fluvial environments (e.g. bank erosion, bar mobility, braided pattern changes) or to quantify components of fluvial dynamics (e.g. flow velocity, wood transport or river ice development). As the amount of information increases, automation of the data processing becomes essential, but many challenges arise to improve features detection, taking into account light contrasts, shadow and reflection, or to calculate surfaces and volumes from image orthorectification. This study illustrates the high potential of ground cameras to observe and quantify rapid, stochastic or complex events in fluvial systems and the numerous challenges we have to face. In order to automatically monitor such key fluvial processes, two ground cameras were installed. The first one was placed on the Genissiat dam (Rhône River, France) focusing on the reservoir where pieces of wood are trapped, creating a large raft. The objective is to survey wood raft area over time as a surrogate of the basin wood production. The second camera was installed along the St Jean River (Gaspesia, Québec) focusing on a pool section. The objective here is to characterize the evolution of ice cover, in terms of growing rate and ice types. The snowy environment is particularly challenging because of brightness or fairly homogeneous radiometric conditions amongst ice types. In both cases, remote sensing technics, especially feature based classification are used. Radiometric and texture indexes are used to discriminate both wood and water and ice types.