A Cognitive Tracking Method Based on Evaluation and Feedback

In complex tracking problems with multiple targets, the cognitive tracking method can adjust the algorithm with feedback by analysing tracking results, and obtain more accurate estimation results than conventional tracking methods. With the aid of historical results, we design an online evaluation variable to achieve feedback. We dynamically adjust the association threshold through the evaluation results to improve the association accuracy, and with the help of historical tracks and measurements, we implement the CNN-based clutter statistics and trajectory management method, which provides a more accurate discrimination of clutter and newborn targets. For individual targets, we analyse the environmental information through the time series of the LSTM network to obtain a more accurate estimation of noise covariance and achieve a high-precision adaptive filter. They constitute a complete cognitive tracking method based on the feedback. Simulations in single-target and multi-target tracking scenarios are carried out respectively to demonstrate the enhancement effect of feedback on the tracking method by comparing with methods such as IMM, and ablation experiments are conducted to illustrate the accuracy and effectiveness of the algorithm.