Data fusion system performance evaluation: part I - correlation index

Performance evaluation (PE) has been regarded as a focus in multi-sensor data fusion area for a long time. As the first part of a series reports, the aim of this paper is to create the correlation index, which is the basis in the evaluation of multi-sensor track association. The main contributions of the current paper are as follows: first, two kinds of mutually inverse mappings are introduced to describe the sensor assignment and data fusion, and correlation and similar matrixes are applied to quantify the association process; secondly, the probability of specious correlation, which is provided to deal with the cross association of correct and error correlations of multi-sensor tracks, along with the probabilities of the correct, error and leaky correlation are combined together to construct the correlation index; and, finally, several examples are given to demonstrate the validity of the indexes.