Theoretical Optical Flow for Target Position Prediction on Fls Images

This study is related to the obstacle avoidance issue for an Autonomous Underwater Vehicle (AUV). In addition to its original mission, the vehicle must ensure its own survival and therefore understand the environment in safety. The use of a Forward Looking Sonar (FLS) on AUV is one of the most efficient solutions to detect unexpected and potentially dangerous changes of the environment, like the presence of obstacles or seabed slope. Like this, a FLS can prevent the vehicle from obstacles or terrain that may endanger the underwater vehicle. Besides navigation is performed knowing data from a Doppler Velocity Log (DVL) and a Motion Reference Unit (MRU). From these last data a process model has been derived in order to predict the motion of a ground target detected in a sonar image. The state vector is composed of the coordinates of the target in the image. Motion prediction leads to a theoretical optical flow that we can compare to an observed optical flow based on images correlation. This study is of high interest for the French organism GESMA which is involved in the development of decisional autonomy for AUV for several years. Results will be given on real data recorded in April 2006 during sea trials organized by GESMA and should be extended on data recorded during future campaigns involving the Rapid Environment Assessment (REA) AUV “Daurade”.