Appearance Based SLAM (ASLAM) for Indoor/Outdoor Urban Terrain

The objective of this project is the research of multi-sensor Appearance Based SLAM (ASLAM) system for day/night operations in indoor and outdoor environments. These algorithms would perform place recognition based on multiple sensor data (imagery, laser, and radar) gathered from a UGV (Unmanned Ground Vehicle) as it travels through the environment. When the vehicle returns to a previously visited scene, the ASLAM algorithm will recognize the scene, update its internal representation, report this to the UGV.

Practical Methods for man-made feature extraction from remote sensing imagery and applications to national topographic map updating

The objective of this project is to participate in the development of automated methods for the extraction of cartographic features from satellite images with high spatial resolutions. Emphasis will be placed on man-made features of interest, including roads, settlement extent and large infrastructures, etc. This will involve review and performance assessment of existing methods, data preprocessing, software implementation and evaluation of new image processing methods suitable for high resolution images.

3D building/structure model reconstruction using multiple aerial images

The objective of this research is to propose and implement a set of new methodologies to create a 3D hazard map that includes models of structures and obstacles in the airport area. The project uses multiple oblique and nadir views high resolution aerial imageries (pictometery images) to reconstruct such models. Various views of an obstacle are combined using image processing and computer vision algorithms to create an accurate 3D wireframe model of the obstacle. The building blocks of the project include two sub‐systems.