Abstract: Image Understanding research in pose estimation, object recognition, various forms of learning, predictive techniques, image-based view synthesis is continuing. The recent VSAM project will tie together several of these techniques but requires new work.
The objective of the most recent IU research at Rochester is to enhance battlefield awareness through the augmentation of surveillance and monitoring capabilities by the automatic recognition of objects and activities, and then by presenting graphic presentations of them in live video. Several research questions arise: the familiar object- and activity-recognition problems; the problem of real-time tracking of natural features and the determination of the state (pose and velocity) of remote objects; the representation of objects for computer processing; the registration of graphic objects with live video; the role of prediction in robust recognition schemes. We evaluate our work by comparision with other competing schemes, by quantifying how each modification or technique affects system performance, and by quantifying the reliability, speed, effectiveness, and versatility of demonstration integrated systems.