Event Detail

Event Type: 
Department Colloquium
Tuesday, February 18, 2003 - 04:00

Speaker Info

Drexel University

We are all familiar with the advantages of curved sideview and security mirrors. In a similar fashion, one can design visual sensors that have a wider field of view by coupling a curved mirror with a conventional (digital or analog) camera, resulting in a catadioptric (pronounced "Cata-die-optrick") sensors. The advantage of such sensors is that it is easy to create wide-angle, panoramic images or even omnidirectional images from just one snapshot of a scene. Additionally, when dealing with digital images, one may perform software transformations to achieve different projections of the scene. I will give an overview of the recent work done on catadioptric sensor design and then describe some of my own work. For example, the mirror shape that gives rise to a uniform resolution sensor turns out to be any surface of revolution of constant Gaussian curvature. The main question that the research part of the talk will address is "Given a prescribed projection, how does one design a catadioptric sensor to realize it?" This problem may be approached using 2-distributions in R^3 and partial differential equations. Using these methods, we have designed several strange and interesting sensors. To find out more see http://www.mcs.drexel.edu/~ahicks/mirrors.html