Tracking multiple objects in urban traffic environments using dense stereo and optical flow

Stereo vision based sensors provide large amounts of data, a fact which is advantageous when trying to extract semantic information about the imaged scene. However, this data is corrupted by errors, caused especially by the uncertainties in the stereo reconstruction process. Temporal information can be used in order to minimize these errors. This paper presents an advanced object model, a novel association mechanism and the design of a Kalman filter based tracking algorithm, for tracking multiple objects, in complex, urban traffic scenarios.