基于神经网络的修正当前统计模型 |
投稿时间:2021-10-08 修订日期:2021-11-04 点此下载全文 |
引用本文: |
摘要点击次数: 221 |
全文下载次数: 0 |
|
基金项目:国家自然科学基金资助项目(61701240); |
|
中文摘要:近年来,神经网络在目标检测以及目标跟踪领域得到了广泛的发展,可以突破传统算法自适应不足的限制。而相对于各类传统目标跟踪算法而言,神经网络不仅可以更好的拟合非线性函数,并且具有更好的自适应性。随着目标机动性的不断提高,传统的当前统计模型(current statistical model,CS)算法已经无法满足需求。因此针对当前统计模型最大加速度无法自适应调整的缺陷,本文提出了基于神经网络的自适应CS算法。该改进算法将神经网络作为CS模型的反馈模块,反馈模块能够实时监测目标新息的变化,并以前一时刻的目标状态估计以及当前目标量测为网络输入,实时更新CS模型的最大加速度。本文对目标阶跃机动情况进行了仿真,并与传统CS算法、基于CS的交互式多模型(interacting multiple-model,IMM)算法进行了比较。仿真结果表明,该改进算法在目标发生阶跃机动时能自适应调整CS算法的最大加速度,从而减小目标跟踪误差以及防止CS算法发散。 |
中文关键词:机动目标跟踪 当前统计模型 神经网络 自适应参数调整 |
|
Current Statistical Model Algorithm Based On Neural Network |
|
|
Abstract:In recent years, neural network has been widely developed in the field of target detection and target tracking, which can break through the limitation of traditional algorithm adaptive deficiency. Compared with all kinds of traditional target tracking algorithms, neural network can not only fit nonlinear function better, but also has better adaptability. With the continuous improvement of target mobility, the traditional current Statistical Model (CS) algorithm has been unable to meet the needs. Therefore, aiming at the defect that the maximum acceleration of the current statistical model cannot be adjusted adaptively, this paper proposes an adaptive CS algorithm based on neural network. In this improved algorithm, the neural network is used as the feedback module of CS model. The feedback module can monitor the change of target information in real time, and update the maximum acceleration of CS model in real time by using the previous target state estimation and the current target measurement as the network input. In this paper, the target's step maneuver is simulated and compared with the traditional CS algorithm and the INTERACTING multiple model (IMM) algorithm based on CS. The simulation results show that the improved algorithm can adaptively adjust the maximum acceleration of CS algorithm when the target takes a step maneuver, so as to reduce the tracking error and prevent the divergence of CS algorithm. |
keywords:Maneuvering target tracking Current statistical model Neural network Adaptive parameter adjustment |
查看全文 查看/发表评论 下载pdf阅读器 |