Dr Fred Labrosse is interested in building representations of environments of mobile robots to allow navigation. The representations are topological maps only containing information about the appearance of the environments (and nothing about the geometry of the environments) and allow visual navigation along visual paths specified as successions of viewpoints. Representations are built on-line during exploration of previously unknown environments using an algorithm based on the idea of Artificial Immune Networks (link to mjn's bit). Find more at http://users.aber.ac.uk/ffl/frob/frob.php.
the main research interests lie in machine vision, especially 3D imaging. Once the 3D objects have been scanned, we can estimate the camera motion from the scanned images, reconstruct a full model of the 3D object, render the 3D model in different resolutions using different illumination models, and recognise the object from the 3D model (4RE: REgistration, REconstruction, REndering, and REcognition).
In the same line, instead of being given two overlapping 3D point clouds, we are given a single 3D point cloud and a 2D projective point cloud that are overlapping in 3D space. We can also do the similar work with regard to 4RE.
Key words: 3D imaging, pattern recognition, image processing, computer aided geometric design, active vision, 3D shape matching, 3D model reconstruction, automatic registration, overlapping, point cloud, free form surface.