Joint Tracking and Classification Based on Conditional Joint Decision and Estimation

In joint tracking and classification (JTC) problems, both decision and estimation are involved and they affect each other. Good solutions for JTC require solving the two problems jointly. A joint decision and estimation (JDE) framework based on a generalized Bayes risk was recently proposed for solving the problem of inter-dependent decision and estimation. In the JDE framework, a conditional JDE (CJDE) risk was proposed, and the corresponding optimal solution was obtained. Due to the development of modern sensor technology, multisensor data with different characteristics are available. In this paper, we solve a JTC problem using multisensor data by the CJDE method. First, a dynamic JTC problem based on kinematic and attribute measurements is formulated as a JDE problem. To solve this problem, we propose a multiple-model recursive CJDE (RCJDE) method, which is an extension of the original RCJDE to the multisensor scenario. For joint performance evaluation, we suggest two joint performance metrics (JPM) for the cases with known and unknown ground truth, respectively. Simulation results demonstrate the effectiveness of the proposed RCJDE method. They show that the multisensor data based RCJDE can outperform the traditional two-step strategies in JPM.