基于无监督学习的目标轨迹预测

摘要

针对情报与侦察监视领域中目标轨迹预测问题,提出了一种基于无监督学习的预测方法.首先,根据历史信息分析目标历史活动规律;其次,构建隐马尔科夫模型,通过无监督学习自动实现预测目标在栅格网中的运动方向;最后,根据学习得到的运动方向和目标运动速度信息,计算未来短期内的目标轨迹.数值仿真验证了该方法能够有效地预测目标在未来短时刻内(通常为5 min)的运动轨迹.%Aiming at the problem of target trajectory prediction in the field of intelligence and reconnais-sance surveillance,a prediction method based on unsupervised learning is proposed.Firstly,a target histori-cal activity pattern is analyzed according to historical information.Secondly,the hidden Markov model is constructed and the motion direction of the prediction target in the grid is automatically realized by unsu-pervised learning.Finally,The target trajectory in the short term is calculated according to the learning di-rection of motion and the speed of target motion.Numerical simulations show that the proposed method can effectively predict the trajectory of the target in the short time (usually 5 minutes).