基于隐马尔科夫模型和LSTM的蜂窝网联无人机轨迹识别技术研究OA

In view of the complexity and particularity of the electromagnetic environment in key areas,as well as the powerful concealment of the cellular networked UAV(unmanned aircraft vehicle)and its characteristics that are difficult to be detected by the traditional detection means,a new UAV trajectory recognition method based on the hidden Markov model(HMM)and the long short-term memory(LSTM)neural network is proposed.A large number of signaling data will be generated when cellular users communicate with nearby cellular base stations during operation.Position information is extracted from these signaling data for trajectory kinematics analysis,so as to obtain characteristic parameters,including speed,and the distance from start and to end.The characteristic parameters and road network data within the very area are subjected to road matching based on HMM.The dynamic time warping(DTW)algorithm is used to calculate the similarity characteristics of the trajectory and the road.After comparison,these characteristics are trained as the input data of LSTM,so as to identify suspected UAV users from the many cellular networked users in key security areas.The experimental results show that,when the positioning error is 20 m,the classification recognition accuracy of UAV terminal trajectories and other traffic terminal trajectories are over 90%by selecting the output results of two-classification and six-classification with or without similarity characteristics.The proposed algorithm provides a new idea for the development of UAV detection field.