Scene Reconstruction Using Accumulated Line - of - Sight

Standard scene reconstruction methods attempt to determine the positions which visible elements of the environment occupy in space. Instead of determining where things are, we choose to model where things are not. We model the open space in a scene. This is accomplished by tracking a human participant moving in the environment. Based on the position of the tracked object and where it is visible to the cameras, a 3-D vacancy grid is modified which represents the empty space in the environment. This method can produce useful spatial models itself or can be used to complement other reconstruction methods. This thesis begins with a survey of scene reconstruction followed by a general description of this method and its advantages and disadvantages. Its application to a single camera system is illustrated by an example. The numerous considerations for extending to multiple camera systems are discussed. Possible applications and extensions are outlined and conclusions are drawn. Thesis Supervisor: W. Eric L. Grimson Title: Professor of Electrical Engineering and Computer Science