Dynamic Measurement Clustering to Aid Real Time Tracking

Many parameter estimation problems admit divide and conquer or partitioning techniques in order to reduce a highdimensional task into several reduced-dimension problems. These techniques all employ a partition which is defined statically using human understanding of the task. We present a method for automatically determining an optimum partition of the parameter space using the properties of the available measurements (their Jacobians). Further, we show that by dynamically performing this clustering at each step of the estimation, substantial computational reductions can be achieved, even for problems which cannot be partitioned in the traditional sense. The technique is applied to the task of tracking camera motions in real-time and video sequences are used to compare the complete tracking system to previous methods.

• Type: Conference
• Year: 2005
• Where: ICCV
• Authors: Christopher Kemp, Tom Drummond
• Keywords: Camera Motions, Computer Vision, ICCV 2005, Optimum Partition, Parameter Estimation Problems, Parameter Space, Substantial Computational Reductions