Volume 2 · Issue 3 (2025)
An Optimal Study on Non-Uniform Sampling Target Motion Parameters Calculation Method Integrated with STC-BiLSTM
Ziheng Han, Huapeng Yu
Institute of National Defense Science and Technology Innovation, Academy of Military Sciences, Beijing 100071, China
Abstract: In recent years, with the complexity of Marine traffic and the popularization of asynchronous sensor acquisition, the prerequisite of traditional uniform sampling is broken, and the data time series presents the non-uniform characteristics of "disordered time interval and dense mutation features", which makes the Marine moving target calculation face more and more challenges of non-uniform sampling data. In order to cope with the shortcomings of traditional methods in abrupt change modeling and time structure adaptability, this paper systematically analyzes the key influencing factors based on a large number of simulation data, and designs a hybrid model architecture with the ability of short-term abrupt change response and long-term dependence modeling. In this paper, we propose a hybrid deep regression model (STC-BiLSTM-Attention) that integrates spatio-temporal convolution, bidirectional LSTM and attention mechanism, which can realize high-precision adaptive calculation method optimization. The model uses SCT module to extract local spatio-temporal features, BiLSTM to model long-range dependence, and self-attention mechanism to focus on key time points, so as to enhance the modeling ability of nonuniform sequence. The experimental results demonstrate that the proposed method achieves an average prediction accuracy of 89.75% across five independent datasets. In scenarios characterized by non-uniform sampling and high dynamic maneuverability, the position calculation error rate is reduced to 8.52%, which is substantially lower than that of the Kalman filter (21.36%) and the manual calculation approach (27.84%). Ablation studies further validate that the synergistic interaction between the STC module and the attention mechanism significantly enhances model accuracy. This highlights the method's superior capability in addressing tasks involving both short-term abrupt changes and long-term dependencies, thereby offering robust technical support for auxiliary decision-making under complex maritime conditions.
Keywords: Target Motion Parameters Calculation; STC-BiLSTM; Non-uniform sampling
References
[1]
National Space Science Center, Chinese Academy of Sciences. "A method and system for dead space prediction of sparse and non-uniform time series data." CN113221450B.2024-03-12.
[2]
Ye, Haibin, et al. "A daily reconstructed chlorophyll-a dataset in South China Sea from MODIS using OI-SwinUnet." Earth System Science Data Discussions 2024 (2024): 1-35.
[3]
Han M, et al. " Measurement and analysis of Marine ambient noise in underwater acoustic buoy over the South China Sea." Applied Acoustics 39.04(2020):536-542.
[4]
Zhaoxuan Zhu, Houjun Wang, Zhigang Wang."A Method for determining the optimal parameters for periodically nonuniformly sampled multi-band Signals." Acta Electronics 9(2010).
[5]
Wang Hongwei, Lian Jie." Control of Non-uniformly sampled Systems Based on switching Principle." Control and Decision 4(2017).
[6]
Erik F. Wilthil,Yaakov Bar-Shalom,Peter Willett,Edmund Brekke. "Estimation of Target Detectability for Maritime Target Tracking in the PDA Framework.".
[7]
Shi, Xingjian, et al. "Convolutional LSTM network: A machine learning approach for precipitation nowcasting." Advances in neural information processing systems 28 (2015).
[8]
Wu Jianyun, in double Ann. "based on ConvLSTM expressway traffic flow prediction simulation research." the laboratory research and exploration of 41.12 (2022) : 132-137.
[9]
Wang, Y., Zhang, K., & Li, X. SALA-LSTM: A Self-Adaptive Learning Approach for Maritime Radar Target Detection in Complex Clutter. IEEE Journal of Oceanic Engineering, 2023.
[10]
Yang, Jian, et al. "A ConvLSTM nearshore water level prediction model with integrated attention mechanism." Frontiers in Marine Science 11 (2024): 1470320.
[11]
Sani, Depanshu, and Saket Anand. "Graph-Based Multi-Modal Sensor Fusion for Autonomous Driving." arXiv preprint arXiv:2411.03702 (2024).
[12]
Chen, Jin, et al. "FRA-LSTM: A vessel trajectory prediction method based on fusion of the forward and reverse sub-network." arXiv preprint arXiv:2201.07606 (2022).
[13]
Zhou, Yi, et al. "Using LSTM with Trajectory Point Correlation and Temporal Pattern Attention for Ship Trajectory Prediction." Electronics 13.23 (2024): 4705.
[14]
Zhang, Zitong, et al. "Improved YOLOv7-Tiny for object detection based on UAV aerial images." Electronics 13.15 (2024): 2969.
[15]
Wang, Jun, et al. "Deep learning for multivariate time series imputation: A survey." arXiv preprint arXiv:2402.04059 (2024).
[16]
Zhang, Qin, et al. "Prediction of sea surface temperature using long short-term memory." IEEE geoscience and remote sensing letters 14.10 (2017): 1745-1749.
[17]
Bommidi, Bala Saibabu, et al. "Hybrid attention-based temporal convolutional bidirectional LSTM approach for wind speed interval prediction." Environmental Science and Pollution Research 30.14 (2023): 40018-40030.
Download PDF