The Intelligent Systems Laboratory (ISL) at RPI includes theoretical developments in computer vision, in probabilistic reasoning using graphical models, and in applications of these theories to different fields. Specifically, in computer vision, our research focuses on theoretical developments in object motion analysis and tracking, image segmentation, and object recognition. In probabilistic reasoning, our research focuses on three aspects: active inference, efficient inference, and model learning. Specifically, in active inference, we focus on developing algorithms and techniques that can identify the most informative evidences to use in order to perform effective inference in an efficient and timely manner. We are particularly interested in active inference by managing and controlling the sensing algorithms so that visual interpretation can be performed in a timely and efficient manner. For efficient inference, our research studies the issue of how to perform efficient belief propagation of the effects of the observed evidences. For model learning, our current research focuses on learning the graphical models by combining quantitative and qualitative data. We are also developing a unified probabilistic framework based on combining the directed and undirected graphs through the factor graph model. Finally, in applications, we have applied the theories we developed to human computer interaction (specifically on human state monitoring), information fusion for situation awareness and decision making, transportation, medicine, biology, military planning, and biometrics.
From systems perspective, we concentrate on two aspects of an intelligent system: sensing (perception) and understanding. For sensing, we develop computer vision algorithms to compute various visual cues (e.g. motion, shape, pose, position, and identity) typically characterizing the state of the objects. Given these visual observations, we then develop graphical models to model the relationships between the sensory observations and the high level situation that produces the observations as well as to model the related contextual information. Finally, high level visual understanding and interpretation are performed through a probabilistic inference using the graphical model and the available sensory observations.