Abstract: An apparatus and method for modeling, interacting with and testing market behavior has a system defining a virtual market that may be used to study and test algorithmic trading and market behavior at the microstructure level. The system may use real data and time sequences and features a trading mechanism implemented by a database server, an information center, client computers and a matching engine through which a live stream of orders is matched against a static historical stream of orders. In one embodiment, the system uses real servers on a real network with inherent latency.

Abstract: A method and apparatus is disclosed for tracking an arbitrarily moving object in a sequence of images where the background may be changing. The tracking is based on visual features, such as color or texture, where regions of images (such as those which represent the object being tracked or the background) can be characterized by statistical distributions of feature values. The method improves on the prior art by incorporating a means whereby characterizations of the background can be rapidly re-learned for each successive image frame. This makes the method robust against the scene changes that occur when the image capturing device moves. It also provides robustness in difficult tracking situations, such as when the tracked object passes in front of backgrounds with which it shares similar colors or other features. Furthermore, a method is disclosed for automatically detecting and correcting certain kinds of errors which may occur when employing this or other tracking methods.

Abstract: A method and apparatus is disclosed for tracking an arbitrarily moving object in a sequence of images where the background may be changing. The tracking is based on visual features, such as color or texture, where regions of images (such as those which represent the object being tracked or the background) can be characterized by statistical distributions of feature values. The method improves on the prior art by incorporating a means whereby characterizations of the background can be rapidly re-learned for each successive image frame. This makes the method robust against the scene changes that occur when the image capturing device moves. It also provides robustness in difficult tracking situations, such as when the tracked object passes in front of backgrounds with which it shares similar colors or other features. Furthermore, a method is disclosed for automatically detecting and correcting certain kinds of errors which may occur when employing this or other tracking methods.