|RESEARCH - Online Mosaicing|
The main goal of the project is to provide scientists with a robotic system for large scale underwater mosaic image acquisition in a spatial and temporal context. The system should be easy to use even for robotic non-experts which implies that the robot requires elaborate autonomous skills. In particular, these include fully automatic mission and trajectory planning as to enable the autonomous scanning of an interesting area without gaps, and also repeated monitoring of the same area for temporal change analysis.
We will use online mosaic approaches so that the robot autonomously plans its trajectories and ensure full visual coverage of selected areas. For allowing the robot to autonomously plan its trajectories and to ensure full- coverage of selected areas online mosaicing approaches will be applied. This approach includes the generation of a rough mosaic image of the scanned area incrementally built during a mission which keeps track of the robot's trajectory. For this the robot will acquire visual images as well as georeference data with its different sensors during an exploration tour. The georefence data (e.g., DGPS1 or USBL2 data) provides an estimate of the robot's position at a given time, which can be directly associated with the visual images acquired. Visual images thus indexed with georefence information an initial mosaic images be constructed can quite easily and fast as the georefence data provides a suitable base for placing each image at its actual position within a global coordinate frame. The mosaic traces the robot's position within this global frame and, by this, yields the starting point for automatically identifying mapped areas and gaps in the survey, and automatically adapting its track to optimize the coverage.
To achieve the main goal of this second phase, which is to provide the robot with online mosaicing facilities, on the one hand techniques for acquiring and processing the georeference and visual image data have to be developed. On the other hand algorithms for data association are needed that allow to place the visual images at the correct position within the online mosaic according to the georefence data. Finally, gap detection algorithms will be investigated enabling the robot to automatically locate and close any uncovered sections of the scene by itself overcoming the need for tedious and time-consuming human intervention.