In order to improve the weak magnetic detection ability under the background of Gaussian colored magnetic environment noise, a magnetic anomaly detection method based on feature fusion and isolation forest (IForest) algorithm is proposed in this paper. The method uses different feature algorithms to extract the statistical features, time-frequency features and fractal features of the signal, reduces the dimensionality of the features by principal component analysis (PCA) and generates feature fusion tensors. Finally the IForest algorithm is used to achieve target detection. The simulation and experimental results show that the method has a higher detection rate under different SNR of Gaussian color noise, which is approximately 5%-18% higher than that of the traditional feature detection algorithm. This method can train an effective detection model with only a small number of negative samples. Compared with the fully connected neural network (FCN) model trained with unbalanced samples, the detection rate increases by approximately 5%-12%, and it takes less time.

In the field of aviation anti-submarine warfare, magnetic detection systems often result in a large number of false alarms due to the inability to effectively distinguish between a local geomagnetic anomaly in the background field and a target magnetic anomaly. To address this problem, this paper proposes a matching detection method for the magnetic anomaly of underwater targets. This method is based on the measured data of the magnetic anomaly in the sea area, and the geomagnetic reference map of the area is generated by interpolation of the Kriging method; then, the background field signal extracted from the geomagnetic reference map and the real-time signal acquired by detection are matched by the similarity metric algorithm; finally, the target detection is realized by using the feature fusion-isolation forest algorithm. Compared with the conventional detection methods, such as time–frequency analysis, the detection rate is improved by nearly 15% and the false alarm rate is reduced by about 40%, indicating that matching detection can effectively filter out the magnetic anomaly in the background field, reduce the false alarm, and improve the detection rate in the real geomagnetic environment, which has certain application value.

Aiming at the problem of the limited application range and low accuracy of existing radar calibration methods, this paper studies the radar calibration method based on cooperative targets, and establishes the integrated radar measurement error model. Then, the improved sparrow search algorithm (ISSA) is used to estimate the systematic error, so as to avoid the loss of partial accuracy caused by the process of approximating the nonlinear equation to the linear equation, thus improving the radar calibration effect. The sparrow search algorithm (SSA) is improved through integrating various strategies, and the convergence speed and stability of the algorithm are also improved. The simulation results show that the ISSA can solve radar systematic errors more accurately than the generalized least square method, Kalman filter, and SSA. It takes less time the than SSA and has a certain stability and real-time performance. The radar measurement error after correction is obviously smaller than that before correction, indicating that the proposed method is feasible and effective.

The trajectory data mining and analysis of maritime targets are of great significance in furthering the construction of maritime traffic facilities, improving the ability of marine supervision and maintaining national marine security. However, due to factors such as detection means and environmental interference, a large number of trajectory data have problems such as large space-time span, uneven sampling, and poor continuity, which seriously restrict the effect of trajectory mining. Therefore, this paper proposes a method of trajectory reconstruction based on navigation state recognition and bidirectional kinematic interpolation. The method mainly includes three steps: (1) data preprocessing, (2) navigation state recognition, and (3) trajectory interpolation. The method can recognize the navigation state of the targets in different segments, and then adaptively select the interpolation method to reconstruct the trajectories, that is, linear interpolation in the straight segments and bidirectional kinematic interpolation in the turning segments. Among them, bidirectional kinematic interpolation uses the cubic Hermite function to nonlinearly fit the acceleration of the interpolation section, and then calculates the velocity and coordinates of the interpolation points by time weighting from the positive and negative directions. The proposed method is verified and analyzed on the contest dataset of “Intelligent classification and recognition of XX trajectories”. Compared with the existing methods, the reconstruction results of the proposed method are closer to the real trajectories, and it can effectively reconstruct the target trajectories with better accuracy and stability. At the same time, the effect of trajectories classification based on the Long Short-Term Memory (LSTM) model, which uses trajectories before and after reconstruction, is compared and analyzed. The results show that the model has a higher classification accuracy for reconstructed trajectory, which proves the necessity of trajectory reconstruction.

To improve the feature extraction method for ship trajectories and enhance trajectory classification performance, this paper proposes a ship trajectory classification model that combines a one-dimensional residual network (ResNet1D) and an attention-based Long short-term memory network (AttLSTM). The model aims to address the limitations of traditional methods in extracting feature patterns jointly represented by non-adjacent local regions in ship trajectories, optimized through the introduction of a self-attention mechanism. Specifically, the model first utilizes the ResNet1D module to progressively extract implicit motion pattern features from local to global levels, while the AttLSTM module captures temporal sequence features of ship trajectories. Finally, the fusion of these two types of features generates a more comprehensive and rich spatiotemporal motion feature representation, enabling accurate classification of five types of ship trajectories, including towing vessels, fishing vessels, sailing vessels, passenger ships, and tankers. Experimental results show that this model excels on extensive real-world trajectory datasets, achieving a classification accuracy of 89.7%, significantly outperforming models relying solely on single feature sets or lacking integrated attention mechanisms. This not only validates the model’s superior performance in ship trajectory classification tasks but also demonstrates its potential and effectiveness for practical applications.

Trajectory clustering is of great significance for identifying behavioral patterns and vessel types of non-cooperative ships. However, existing trajectory clustering methods suffer from limitations in extracting cross-spatiotemporal scale features and modeling the coupling relationship between positional and motion features, which restricts clustering performance. To address this, this study proposes a deep ship trajectory clustering method based on feature embedding representation learning (ERL-DTC). The method designs a Temporal Attention-based Multi-scale feature Aggregation Network (TA-MAN) to achieve dynamic fusion of trajectory features from micro to macro scales. A Dual-feature Self-attention Fusion Encoder (DualSFE) is employed to decouple and jointly represent the spatiotemporal position and motion features of trajectories. A two-stage optimization strategy of “pre-training and joint training” is adopted, combining contrastive loss and clustering loss to jointly constrain the embedding representation learning, ensuring it preserves trajectory similarity relationships while being adapted to the clustering task. Experiments on a public vessel trajectory dataset show that for a four-class task (K = 4), ERL-DTC improves ACC by approximately 14.1% compared to the current best deep clustering method, with NMI and ARI increasing by about 28.9% and 30.2%, respectively. It achieves the highest Silhouette Coefficient (SC) and the lowest Davies-Bouldin Index (DBI), indicating a tighter and more clearly separated cluster structure. Furthermore, its inference efficiency is improved by two orders of magnitude compared to traditional point-matching-based methods, without significantly increasing runtime due to model complexity. Ablation studies and parameter sensitivity analysis further validate the necessity of each module design and the rationality of hyperparameter settings. This research provides an efficient and robust solution for feature learning and clustering of vessel trajectories across spatiotemporal scales.