Abstract: It is one of the responsibilities of the navigation support department to ensure the correct layout position of the light buoy and provide as accurate position information as possible for ship navigation and positioning. If the position deviation of the light buoy is too large to be detected in time, sending wrong navigation assistance information to the ship will directly affect the navigation safety of the ship and increase the pressure on the management department. Therefore, mastering the offset characteristics of light buoy is of great significance for the maintenance of light buoy and improving the navigation aid efficiency of light buoy. Kernel density estimation can intuitively express the spatial and temporal distribution characteristics of buoy position, and indicates the intensive areas of buoy position in the channel. In this paper, in order to speed up deciding the optimal variable width of kernel density estimator, an improved adaptive variable width kernel density estimator is proposed, which reduces the risk of too smooth probability density estimation phenomenon and improves the estimation accuracy of probability density. A fractional recurrent neural network is designed to search the optimal bandwidth of kernel density estimator. It not only achieves faster training speed, but also improves the estimation accuracy of probability density.

Key words: kernel density estimation, buoy, bandwidth optimization, recurrent neural network, navigation aid efficiency, spatial information

摘要： 确保灯浮标位置正确，为船舶航行提供尽可能准确的位置信息，是航海保障部门的职责之一。如果灯浮标位置偏差过大，向船舶发送错误的助航信息，将影响船舶的航行安全，同时也会增大管理部门的压力。因此，掌握灯浮标的偏移特性对灯浮标的维护和提高灯浮标的助航效能具有重要意义。灯浮标位置的核密度估计能够直观地反应灯浮标位置的空间分布特征以及浮标位置的密集区域。为提高计算速度和核密度估计的精度，降低核密度估计过于平滑的风险，采用分数阶递归神经网络优化核密度估计带宽的方法，设计了一种自适应带宽核密度估计器。

关键词: 核密度估计，浮标，带宽优化，递归神经网络，助航效能，空间信息