Using prior information to enhance tracking

摘要

The accuracy of tracking can be enhanced by the incorporation of prior information in the form of a local road map. For instance, the performance of ground target tracking algorithms can be improved by incorporating hard constraints (describing roads and junctions) into the tracking model. We describe an approach to enhancing tracking algorithms that uses 'probability fields' to represent the local road map information. The resulting 'Bayes-filter' equations are solved using particle niters and compared with a basic particle filter with no additional map information.