The objective of this project is to develop custom analog VLSI hardware for several tasks that occur early in the computational vision process. The work involves both hardware development to implement existing algorithms and algorithm modification to match hardware capabilities. A goal of this research is to build a single-chip motion sensor and analyzer that will be useful for robot navigation. The sensor array can be placed at the focal plane of lenses to acquire binocular image input directly from environment, and the outputs will consist of the intensity field, velocity field and depth map of the scene. The system can be used as a front-end processor for a variety of vision applications such as robot navigation and object recognition. The researchers in this team have impressive research achievements. The team has expertise in algorithms for vision, analog circuit design, system design and parallel computation. It is an excellent group with the potential to make significant discoveries along the way. The laboratory facilities and intellectual environment at the Massachusetts Institute of Technology are an excellent basis for the proposed work. The funding of this proposal is strongly recommended.
