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

打开文本图片集

关键词：蜂窝网联无人机；轨迹识别；道路匹配；LSTM；隐马尔科夫；动态时间规整中图分类号：TN919-34；V279；TP183 文献标识码：A 文章编号：1004-373X（2026）07-0019-07

Cellular networked UAV trajectory recognition technique based on Hidden Markov modelandLSTM

HE Xiang，LEI Chaojun，LI Zeng，JIA Chunlei （Schoolof Police Equipment Technical， China People's Police University，Langfang O650oo，China）

Abstract：Inviewofthecomplexityandparticularityof theelectromagneticenvironmentinkeyareas，aswell asthe powerfulconcealmentofthecelularnetworkedUAV（unmannedaircraftvehicle）anditscharacteristicsthataredificulttobe detected bythetraditionaldetection means，anewUAVtrajectoryrecognitionmethodbasedonthehidden Markov model（HMM） andthelongshort-term memory（LSTM）neural network isproposed.Alargenumberofsignalingdata willbe generated when celularuserscommunicatewithnearbycellularbasestationsduringoperation.Positioninformationisextractedfromthese signalingdatafortrajectorykinematicsanalysis，soastoobtaincharacteristicparameters，includingspeed，andthedistancefrom startandtoend.Thecharacteristicparametersandroadnetworkdata withintheveryareaaresubjectedtoroadmatchingbasedon HMM.Thedynamictime warping（DTW）algorithmisusedtocalculatethesimilaritycharacteristicsof thetrajectoryandtheroad. Aftercomparison，thesecharacteristicsaretrainedastheinputdataofLSTM，soastoidentifysuspectedUAVusesfromthemany celularnetworkedusersinkeysecurityareas.Theexperimentalresultsshowthat，whenthepositioningeroris2Om，the clasificationrecognitionaccracyofUAVterminaltrajectoriesandothertraffcterminaltraectoriesareover9O%byselectingthe outputresultsoftwo-classficationandsix-classfcationwithorwithoutsilritycharacteristics.Thepropoedalgoriprovides a new idea for the development of UAV detection field.

0 引言

无人机（UnmannedAircraftVehicle，UAV）技术的快速发展和广泛应用给人们的生活带来了诸多便利。（剩余8666字）