Joint Tracking and Classification Based on Recursive Joint Decision and Estimation Using Multi-Sensor Data

Joint target tracking and classification (JTC) is a joint decision and estimation (JDE) problem, in which decision and estimation affect each other and good solutions require solving both problems jointly. With the development of modern sensor technology, mixed data from heterogeneous sensors with different characteristics are available. In this paper, we solve a JTC problem using multisensor data in the JDE framework. A dynamic JTC problem based on kinematic and attribute measurements is formulated as a JDE problem, and the dynamic models and measurement models for both types of data are presented. We extend the original recursive JDE (RJDE) method to the multisensor scenario, and propose a multisensor data based RJDE method using the multiple model approach. To jointly evaluate the performance of multisensor data based JTC with unknown ground truth, we propose a joint performance metric (JPM) based on the idea of mock data. This metric unifies the distances in the continuous data space and the discrete data space. Simulation results demonstrate the effectiveness of the proposed approach and JPM. They show that the multisensor data based RJDE can outperform the traditional two-step strategies. Furthermore, the proposed approach can beat E&D (optimal decision and optimal estimation, respectively) in joint performance.