Section: New Results

Visual tracking

3D model-based tracking

Participants : Antoine Petit, Eric Marchand.

Our 3D model-based tracking algorithm [3] was used in various contexts. First, it has been studied and tested on a mock-up of a telecommunication satellite using a 6-DOF robotic arm, with satisfactory results in terms of accuracy of the pose estimation and computational costs [41] , [42] . A potential application would be the final phase of space rendezvous mission using visual navigation. Then, it has been considered for designing a visual servoing scheme able to control the walking of a humanoid robot [29] .

Omnidirectional stereovision

Participants : Guillaume Caron, El Mustapha Mouaddib, Eric Marchand.

Omnidirectional cameras allow direct tracking and motion estimation of planar regions in images during a long period of time. However, using only one sensor leads to plane and trajectory reconstruction up to a scale factor. We proposed to develop dense plane tracking based on omnidirectional stereovision to answer this issue. The method estimates simultaneously the parameters of several 3D planes along with the camera motion using a spherical projection model formulation [20] .

Motion estimation using mutual information

Participant : Eric Marchand.

Our work with Amaury Dame related to template tracking using mutual information as registration criterion has been extended to motion estimation applications. It has been applied to mosaicing from an image sequence [28] . The main advantage is that this approach is robust to noise, lighting variations and does not require a statistically robust estimation process.

Augmented reality

Participants : Pierre Martin, Hideaki Uchiyama, Eric Marchand.

We developed an approach for detecting and tracking various types of planar objects with geometrical features[45] . We combine traditional keypoint detectors with Locally Likely Arrangement Hashing (LLAH) for keypoint matching. In order to produce robustness to scale changes, we build a non-uniform image pyramid according to keypoint distribution at each scale. It demonstrates that it is possible to detect and track different types of textures including colorful pictures, binary fiducial markers and handwritings. This approach was extended to consider non-rigidly deformable markers [46] .