An LSTM-based Target Tracking Method for High-mobility UAV under Incomplete Information Conditions

In the case of an Unmanned Aerial Vehicle (UAV) with high mobility and incomplete observation information, existing tracking methods for the UAV target tracking problem show insufficient robustness and accuracy. In this paper, a long short-Term memory(LSTM)-based target tracking method is proposed, based on the addition of incomplete information and observation misalignment judgment method to the structure of the filter, using convolutional layers to extract the trajectory sequence features, and combining with the trajectory sequence features, the LSTM is used to output the results of the compensation of the motion model under the incomplete observation information. The experimental results show that the proposed method improves the robustness and accuracy of UAV target tracking compared to comparing models.