Heterogeneous data source fusion for impedance indicators
Type de document
CHAPITRE D'OUVRAGE (CO)
Résumé / Abstract
The emergence of new, highly accurate detectors, as well as the abundance of works setting out several theories on data processing have increased the need for multi-source data merging techniques. The key-point of those techniques is how to combine different data from various sources, which have different scales, different accuracies, resulting from different experimental designs in order to improve the estimation or decision of a certain variable under consideration. This problem is known as the data fusion process. The main advantage justifying the fusion of multi-source data is the benefit that can be gained from the complementary nature of these different sources which can contribute to achieve a more accurate decision. The aim of this article is to put forward various statistical solutions for data fusion. The application to impedance indicators (essentially travel time) motivates the analysis and shows the ability of the proposed solutions to be turned in different ways for different specific applications. Performance of the proposed data fusion scheme have been evaluated using measured data. Preliminary results indicate a significant improvement over the models using data exclusively from one source or those selecting the single best estimator.