基于LSTM-KF的非共视情况下舰船目标跟踪

Ship Target Tracking under Non-common-view Based on LSTM-KF

为解决实际舰船航行环境中,因岛礁遮挡造成非共视情况导致时差定位无法正常运行,转换为测向定位时常出现轨迹跳变的问题,将卡尔曼滤波(KF)算法和长短时记忆(LSTM)网络相结合,通过分析共视情况下目标运动轨迹,对非共视情况下的目标位置预测进行了研究和仿真。通过对比分析测向定位结果及联合算法的测试数据,结果表明在非共视情况下,所提出的联合算法可以更加精准地对目标进行定位跟踪,有效减少了轨迹跳变的发生。

In order to solve the problem that the time difference positioning can’t be operated normally due to the non-common-view situation caused by the occlusion of islands and reefs in the actual ship navigation environment,and the trajectory jump often occurs when it is converted to direction-finding and positioning,the Kalman filtering(KF)algorithm is combined with long short term memory(LSTM)network.By analyzing the target trajectory in the common-view case,the position prediction of the target in the non-common-view case is studied and simulated.By comparing and analyzing the direction-finding and positioning results&the test data of joint algorithm,the results show that the joint algorithm proposed can locate and track the target more accurately in the case of non-common view,and effectively reduce the occurrence of trajectory jump.