Ship track prediction method based on spatial-temporal feature fusion

Abstract

The invention relates to the technical field of track prediction, and particularly discloses a ship track prediction method based on spatial-temporal feature fusion, which comprises the following steps of: processing AIS information of each time in a known ship track by adopting a neural network model based on deep learning to extract time dimension features and spatial dimension features of the ship track; hidden association dominant modeling analysis is carried out on time dimension features and space dimension features of a ship trajectory to excavate hidden key clues between the time dimension features and the space dimension features, and alignment interaction fusion of the time dimension features and the space dimension features of the ship trajectory is realized based on the hidden key clues. And predicting the future track of the ship based on the space-time fusion features of the ship track. Through the mode, the correlation among the ship navigation speed, the course change and the geographic position can be deeply excavated, so that more accurate feature alignment interaction fusion is realized, and the accuracy of ship track prediction is further improved.

Description

Technical Field

[0001] The present application relates to the technical field of track prediction, and more specifically, to a ship track prediction method based on spatiotemporal feature fusion. Background Art

[0002] In the context of the booming maritime transport industry, ship track prediction is crucial to ensuring maritime traffic safety, optimizing shipping resource allocation, and improving port operation efficiency. Accurate track prediction can avoid potential collision risks in advance and rationally plan ship routes, thereby effectively reducing congestion and improving overall shipping efficiency. In recent years, with the widespread application of the Automatic Identification System (AIS), a large amount of ship navigation data has been accumulated, providing rich data support for the development of ship track prediction technology. Track prediction methods based on the fusion of spatiotemporal features have become a research hotspot.

[0003] For example, the invention patent with publication number CN117933492A proposes a long-term prediction method for ship tracks based on spatiotemporal feature fusion, which converts the AIS information observation values ​​at each time in the known ship track into the corresponding high-dimensional embedding vector, and uses the ship track prediction model TCNformer to extract the time dimension features and space dimension features from the high-dimensional feature vector sequence, and further fuses the time dimension features with the space dimension features, and predicts the track information at the next time point according to the spatiotemporal fusion features until the prediction task is completed. In this way, the complexity of the ship motion pattern and the heterogeneity of the AIS data are effectively solved, and the long-term prediction of the ship track is realized.

[0004] However, in the prior art, the time dimension characteristics and space dimension characteristics of the ship trajectory are directly fused by splicing and fusion, which fails to fully explore the complex internal connection between time and space characteristics. Specifically, during the navigation of a ship, its speed and heading changes in the time dimension are closely related to its geographical location in the space dimension. The direct splicing and fusion method cannot deeply capture this internal connection, which may lead to inaccurate prediction results. In addition, due to the influence of factors such as the ship's navigation speed and route planning, the frequency and scale of changes in time and space characteristics are not consistent. Direct splicing and fusion will cause the fused features to have information redundancy or association loss problems, so that the fused features cannot fully and accurately reflect the actual situation of the ship's navigation, thereby affecting the accuracy of the track prediction.

[0005] Therefore, an optimized ship track prediction method based on spatiotemporal feature fusion is expected. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a ship track prediction method based on the fusion of time and space features, which uses a neural network model based on deep learning to process the AIS information at each time in the known ship trajectory to extract the time dimension characteristics and space dimension characteristics of the ship trajectory, and then, through the implicit correlation explicit modeling analysis of the time dimension characteristics and space dimension characteristics of the ship trajectory, the implicit key clues between the two are excavated, and based on this, the alignment and interactive fusion of the time dimension characteristics and space dimension characteristics of the ship trajectory are achieved, and then the future track of the ship is predicted based on the time and space fusion characteristics of the ship trajectory. In this way, the correlation between the ship's sailing speed, course change and geographical location can be deeply excavated, so as to achieve more accurate feature alignment and interactive fusion, thereby improving the accuracy of ship track prediction.

[0007] According to one aspect of the present application, a ship track prediction method based on spatiotemporal feature fusion is provided, which includes: Obtain known ship tracks; Based on the AIS information at each time in the known ship trajectory, extracting the time dimension features and space dimension features of the known ship trajectory; Performing a spatiotemporal feature interactive fusion based on implicit association explicit modeling on the time dimension feature and the space dimension feature to obtain a spatiotemporal fusion feature; Based on the spatiotemporal fusion features, the ship track information is predicted.

[0008] Compared with the prior art, the ship track prediction method based on spatiotemporal feature fusion provided by the present application uses a neural network model based on deep learning to process the AIS information at each time in the known ship track to extract the time dimension characteristics and space dimension characteristics of the ship track, and then, through the implicit correlation explicit modeling analysis of the time dimension characteristics and space dimension characteristics of the ship track, to dig out the implicit key clues between the two, and based on this, to achieve the alignment and interactive fusion of the time dimension characteristics and space dimension characteristics of the ship track, and then predict the future track of the ship based on the spatiotemporal fusion characteristics of the ship track. In this way, the correlation between the ship's sailing speed, course change and geographical location can be deeply excavated, so as to achieve more accurate feature alignment and interactive fusion, thereby improving the accuracy of ship track prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 The present invention is a flowchart of a ship track prediction method based on spatiotemporal feature fusion according to an embodiment of the present application.

[0011] Figure 2 Schematic diagram of data flow of a ship track prediction method based on spatiotemporal feature fusion according to an embodiment of the present application.

[0012] Figure 3 This is a flowchart of sub-step S3 of the ship track prediction method based on spatiotemporal feature fusion according to an embodiment of the present application.

[0013] Figure 4 This is a flowchart of sub-step S32 of the ship track prediction method based on spatiotemporal feature fusion according to an embodiment of the present application. DETAILED DESCRIPTION

[0014] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0015] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0016] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0017] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0018] It is worth noting that in this application, all actions to obtain data are carried out in compliance with the relevant data protection laws and policies of the country where the data is located, and with the authorization given by the owner of the corresponding device.

[0019] As mentioned in the background technology above, patent CN117933492A proposes a long-term prediction method for ship tracks based on spatiotemporal feature fusion, which converts the AIS information observation values ​​at each time in the known ship track into the corresponding high-dimensional embedding vector, and uses the ship track prediction model TCNformer to extract the time dimension features and space dimension features from the high-dimensional feature vector sequence, and further fuses the time dimension features with the space dimension features, and predicts the track information at the next time point according to the spatiotemporal fusion features until the prediction task is completed. In this way, the complexity of the ship motion pattern and the heterogeneity of the AIS data are effectively solved, and the long-term prediction of the ship track is realized.

[0020] However, in the prior art, the above method simply relies on splicing and fusion to integrate the temporal and spatial features of the ship's trajectory, which makes it difficult to deeply reveal the complex internal relationship between the temporal and spatial features. During ship navigation, the speed and heading changes in the time dimension are closely related to the geographical location in the spatial dimension, and the direct splicing method cannot effectively capture this internal connection, which may greatly reduce the accuracy of the prediction results. In addition, due to the constraints of factors such as ship navigation speed and route planning, there are differences in the frequency and scale of changes in temporal and spatial features. Direct splicing and fusion can easily lead to information redundancy or loss of association in the fused features, and cannot fully and accurately present the true situation of the ship's navigation, which ultimately has an adverse effect on the accuracy of track prediction. In response to the above technical problems, this application proposes an optimized ship track prediction method based on spatiotemporal feature fusion, which uses a neural network model based on deep learning to process the AIS information at each time in the known ship track to extract the time dimension characteristics and space dimension characteristics of the ship track, and then, through the implicit correlation explicit modeling analysis of the time dimension characteristics and space dimension characteristics of the ship track, to dig out the implicit key clues between the two, and based on this, realize the alignment and interactive fusion of the time dimension characteristics and space dimension characteristics of the ship track, and then predict the future track of the ship based on the spatiotemporal fusion characteristics of the ship track. In this way, the correlation between the ship's sailing speed, course change and geographical location can be deeply excavated, so as to achieve more accurate feature alignment and interactive fusion, thereby improving the accuracy of ship track prediction.

[0021] Figure 1 The present invention is a flowchart of a ship track prediction method based on spatiotemporal feature fusion according to an embodiment of the present application. Figure 2Schematic diagram of data flow of a ship track prediction method based on spatiotemporal feature fusion according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the ship track prediction method based on spatiotemporal feature fusion includes the following steps: S1, obtaining a known ship track; S2, extracting the time dimension features and space dimension features of the known ship track based on the AIS information at each time in the known ship track; S3, performing spatiotemporal feature interactive fusion of the time dimension features and the space dimension features based on implicit association explicit modeling to obtain spatiotemporal fusion features; S4, predicting the ship track information based on the spatiotemporal fusion features.

[0022] In the above-mentioned ship track prediction method based on spatiotemporal feature fusion, the step S1 is to obtain the known ship track. It should be understood that in practical applications, the ship automatic identification system (AIS) is an important source for obtaining ship track data. The AIS system can collect various types of information of the ship in real time, including the ship's location information (expressed by latitude and longitude), operating speed (SOG), navigation direction (COG), etc., and continuously update over time to form a known navigation track record of the ship. In addition, due to the uneven quality of the original AIS data, it contains various problems. For example, there may be information in the data whose speed value is obviously inconsistent with the actual situation (such as the ship speed exceeds the normal navigation speed range), data generated when the ship is in a moored or anchored state, and AIS observation information that is too close to the coastline (within 1 nautical mile). These data not only increase the burden of data processing, but also may interfere with subsequent trajectory analysis and prediction. Therefore, in order to improve the accuracy and efficiency of ship track prediction, it is necessary to clean and preprocess the collected AIS data. The specific operations include deleting AIS information with unrealistic speed values, removing moored or anchored ship information, and deleting AIS observation information within 1 nautical mile of the coastline. Then, the known AIS data of the ship to be predicted is extracted from the AIS data after data cleaning and preprocessing to form the known ship trajectory.

[0023] Specifically, on the one hand, it is necessary to select a suitable AIS data source. This not only includes the elimination of obviously erroneous or abnormal data points, but also involves checking the integrity of the data. For example, if the AIS signal of a ship is lost during a certain period of time, it is necessary to supplement the missing information through other means (such as satellite communication, radar, etc.) to ensure the continuity of the trajectory. On the other hand, in the process of acquiring data, it is also necessary to check the consistency of data from different sources. Due to the differences in the working principles and technical parameters of various sensors, direct merging and using them may lead to error accumulation. Therefore, before integrating multi-source data, necessary calibration work should be carried out to ensure that all data can be compared and analyzed in the same coordinate system. For example, using high-precision geographic information system (GIS) tools to convert location information from different sensors into a unified map projection system can effectively reduce the deviation caused by inconsistent coordinate systems.

[0024] In addition to technical considerations, data security issues cannot be ignored. With the increase in the amount of AIS data and the expansion of its application scope, privacy protection has become an important issue. Especially when it comes to sensitive areas or specific types of ships, in order to meet research needs while complying with relevant laws and regulations and avoiding infringement of personal privacy, anonymization measures can be taken at the beginning of data collection, retaining only necessary navigation information for academic research or commercial purposes, and removing any identifiers that may identify specific individuals.

[0025] Furthermore, considering the complexity and dynamic nature of the maritime environment, it is particularly important to establish a real-time updated database. Such a database can not only reflect the latest changes in the position and status of the ship in a timely manner, but also provide support for rapid response in emergencies. For example, when encountering severe weather conditions or emergency collision avoidance needs, the latest and most accurate data can be quickly called to assist in decision-making. To this end, it is necessary to develop an efficient data synchronization mechanism to ensure seamless information exchange between land-based receiving stations and mobile terminals, and to minimize delays throughout the process.

[0026] It is worth noting that obtaining known ship tracks is an ongoing process that requires regular data updates and adjustments to analysis strategies. As time goes by, new routes are constantly opened up, and navigation patterns on existing routes may also change. Therefore, it is particularly important to establish a dynamic data update mechanism, keep up with the latest AIS data, and adjust analysis methods and technical routes according to actual conditions. This not only helps to maintain the timeliness of research results, but also ensures that the decisions made based on this are always close to the actual situation, maximizing the actual value of ship track prediction.

[0027] In the above-mentioned ship track prediction method based on the fusion of time and space features, the step S2 extracts the time dimension features and space dimension features of the known ship track based on the AIS information at each time in the known ship track. It should be understood that during the navigation process, the position, speed, heading and other states of the ship change continuously over time. By analyzing the speed and heading information at different times, the time nodes and frequencies of the ship's acceleration, deceleration, turning and other behaviors can be understood, and the subsequent movement trend of the ship can be predicted. At the same time, by analyzing the position information, navigation direction, etc. of different points in the ship's navigation track, the navigation characteristics and preferences of the ship can be understood, thereby providing a spatial dimension information reference for track prediction.

[0028] Specifically, for the AIS information at each time in the known ship trajectory, firstly, various attributes (such as position, speed, heading, etc.) in the AIS information are converted into high-dimensional vector representations through embedded coding technology to capture the potential associations between these attributes and form a sequence of high-dimensional feature vectors of AIS information. Then, since the time dimension characteristics of the ship trajectory mainly focus on the dynamic changes of the ship at different time points. In this regard, the TCN (Temporal Convolutional Networks) network can be used to process the sequence of high-dimensional feature vectors of the AIS information. Through its causal convolution structure, the convolution operation is performed on the time dimension of the sequence of high-dimensional feature vectors of the AIS information, which can effectively extract the local dependencies in the sequence data while maintaining the time order, so that the time dimension characteristics reflecting the ship motion mode can be extracted from the sequence of high-dimensional feature vectors of the AIS information. At the same time, the spatial dimension characteristics of the ship trajectory focus on the distribution and movement trend of the ship in the geographic space. To this end, the powerful parallel computing capability of the Transformer network can be used to process the sequence of high-dimensional feature vectors of the AIS information using a self-attention mechanism to capture the long-term correlations in the known AIS observation data, and to mine the correlations and movement trends of ships in different geographical locations, thereby extracting spatial dimension features that reflect the spatial distribution characteristics of ships. It is worth mentioning that the method for extracting the time dimension features and spatial dimension features of the known ship trajectory involved in this application can adopt the method disclosed in Chinese patent CN117933492A, and of course other equivalent feature extraction methods can also be used, which is not limited to this application.

[0029] In the above-mentioned ship track prediction method based on spatiotemporal feature fusion, the step S3 performs the spatiotemporal feature interactive fusion of the time dimension feature and the space dimension feature based on implicit association explicit modeling to obtain the spatiotemporal fusion feature. It should be understood that the present application takes into account that the time dimension feature reflects the dynamic change of the ship's navigation status over time, which has obvious temporal sequence and periodicity; while the space dimension feature focuses on the position relationship and motion trajectory of the ship in the geographic space. Due to the influence of factors such as the ship's navigation speed and route planning, the change frequency and scale of its time dimension features and space dimension features are not consistent. For example, in different sea areas, the navigation speed of the ship may be different, which leads to differences in the distance the ship moves in the space dimension within the same time interval in the time dimension. However, the traditional simple splicing and fusion method often ignores the heterogeneity between such features, resulting in unsatisfactory feature fusion effect, thereby reducing the model's ability to predict the ship's track. In this regard, this application proposes a method for interactive fusion of spatiotemporal features based on implicit association explicit modeling, which establishes a deep connection between the time dimension features and the space dimension features by mining the implicit key clues between the time dimension features and the space dimension features, and then performs alignment interaction between the time dimension features and the space dimension features on this basis to obtain spatiotemporal fusion features. Figure 3 FIG. 4 is a flowchart of sub-step S3 of the ship track prediction method based on spatiotemporal feature fusion according to an embodiment of the present application. Figure 3 As shown, the step S3 includes the steps of: S31, performing feature principal component analysis on the time dimension feature to obtain a set of time dimension feature principal component components; S32, performing implicit association analysis on the spatial dimension feature and each time dimension feature principal component in the set of time dimension feature principal component components to construct a feature pair of {spatial dimension deep implicit feature, anchored time dimension principal component deep implicit feature}; S33, performing spatiotemporal feature interactive fusion on the feature pair of {spatial dimension deep implicit feature, anchored time dimension principal component deep implicit feature} to obtain the spatiotemporal fusion feature.

[0030] Specifically, the step S31 is expressed by the formula:

[0031] in, is the covariance matrix of the time dimension features, Express The matrix composed of the set of principal component components of the time dimension characteristics obtained by eigenvalue decomposition, represents transpose, Express The diagonal matrix consisting of the set of principal component eigenvalues ​​of the time dimension obtained by eigenvalue decomposition, and Respectively represent the first and the second principal component eigenvalues ​​in the time dimension. eigenvalues, , and Respectively The 1st, 2nd and The time dimension feature principal component, is the number of principal components of the time dimension feature, represents the principal component analysis network, Represents the time dimension feature.

[0032] That is, the present application takes into account that under normal circumstances, the spatial dimension features mainly reflect the positional relationship of the ship, and the correlation within the features is relatively clear, that is, the features are relatively stable and depend on the geographic spatial structure. The time dimension features mainly reflect the dynamic changes in the navigation status of the ship, and the correlation within the features is relatively complex and may contain more redundant information. Therefore, the present application first performs principal component analysis on the time dimension features to convert the time dimension features into a set of more representative principal component components of the time dimension features, so as to extract the core trends and periodic changes in the time dimension features, separate redundancy and noise, and thus reduce the complexity and computational complexity of subsequent interactive analysis.

[0033] Figure 4 FIG. 4 is a flowchart of sub-step S32 of the ship track prediction method based on spatiotemporal feature fusion according to an embodiment of the present application. Figure 4 As shown, the step S32 includes the steps of: S321, respectively extracting the deep implicit features of each time dimension feature principal component in the set of the spatial dimension feature and the time dimension feature principal component to obtain a set of the spatial dimension deep implicit features and the time dimension principal component deep implicit features; S322, performing implicit key clue anchor encoding on the set of the spatial dimension deep implicit features and the time dimension principal component deep implicit features to obtain the feature pair of {spatial dimension deep implicit features, anchored time dimension principal component deep implicit features}.

[0034] More specifically, in a specific example of the present application, the step S321 includes: performing implicit feature extraction based on point convolution coding on each time dimension feature principal component in the set of the spatial dimension feature and the time dimension feature principal component to obtain the set of the spatial dimension deep implicit feature and the time dimension principal component deep implicit feature, which is expressed by the formula:

[0035]

[0036] in, represents the spatial dimension feature, represents the point convolution operation, express Activation function, Represents the deep implicit features of the spatial dimension, Represents the first in the set of principal component components of the time dimension feature The time dimension feature principal component, Represents the set of deep implicit features of the principal components in the time dimension, , , and They represent the first, second, and third deep implicit features of the principal component in the time dimension. and The deep implicit features of the principal components in the time dimension.

[0037] That is, the present application further extracts the deep implicit features of the spatial dimension features and the principal component components of each time dimension feature through point convolution coding processing. In particular, point convolution does not focus on the locality of the original feature space of the feature, but directly performs information fusion in the global range of the feature through cross-dimensional weight mapping, and further enhances the feature expression ability through nonlinear activation functions, so that deep implicit features with strong representation ability can be efficiently extracted, and a collection of deep implicit features of the spatial dimension and deep implicit features of the principal component of the time dimension can be obtained.

[0038] More specifically, in a specific example of the present application, the step S322 includes: calculating the feature correlation factor between each time dimension principal component deep implicit feature in the set of the spatial dimension deep implicit feature and the time dimension principal component deep implicit feature, and selecting the time dimension principal component deep implicit feature corresponding to the largest feature correlation factor as the time dimension principal component deep implicit feature anchored by the spatial dimension deep implicit feature, so as to construct the {spatial dimension deep implicit feature, anchored time dimension principal component deep implicit feature} feature pair, which is expressed by the formula:

[0039]

[0040] in, represents the calculation of the inner product, represents the smoothing coefficient, represents the calculation norm, Indicates the index corresponding to the maximum value. Indicates the index of the deep implicit feature of the principal component in the time dimension corresponding to the largest feature correlation factor, Represents the deep implicit features of the anchored time dimension principal component, Represents the feature pair of {deep latent features in spatial dimension, anchored deep latent features in principal components in time dimension}.

[0041] That is, in order to filter out the noise and redundant components in the set of deep implicit features of the time dimension principal component, the present application further performs implicit correlation analysis on the set of deep implicit features of the spatial dimension and the deep implicit features of the time dimension principal component to capture the implicit key clues between the spatial dimension features and the time dimension features, so as to achieve semantic alignment between the features. In a specific implementation, by calculating the correlation between the deep implicit features of the spatial dimension and the deep implicit features of the principal components of each time dimension, the deep implicit features of the principal components of the time dimension that are most relevant to the deep implicit features of the spatial dimension are extracted as key clues, and used as anchor points to construct a feature pair of {deep implicit features of the spatial dimension, anchored deep implicit features of the principal components of the time dimension}, thereby giving the feature interaction a clear structural prior through the alignment operation and eliminating the ambiguity between the features.

[0042] Specifically, in a specific example of the present application, the step S33 includes: inputting the {spatial dimension deep implicit feature, anchored time dimension principal component deep implicit feature} feature pair into a feature fine-grained interaction network to obtain the spatiotemporal fusion feature. More specifically, the step S33 includes: using the spatial dimension deep implicit feature as a query vector, the spatial dimension feature as a value vector, and the anchored time dimension principal component deep implicit feature as a key vector, inputting the query vector, the key vector, and the value vector into a feature attention interaction module based on the Transformer architecture to obtain the spatial dimension-time dimension attention interaction feature; performing linear weighted fusion on the spatial dimension-time dimension attention interaction feature and the time dimension feature principal component corresponding to the anchored time dimension principal component deep implicit feature to obtain the spatiotemporal fusion feature, which can be expressed as follows:

[0043] in, represents the normalized exponential function, and Represent different weight parameters respectively. represents the feature scale scaling parameter, represents vector multiplication operation, Indicates the The corresponding time dimension feature principal component, represents the spatiotemporal fusion feature.

[0044] Here, the present application uses a feature fine-grained interaction network to process the anchored feature pairs, so as to mine more complex nonlinear feature associations between spatial dimension features and temporal dimension features based on the existing semantic alignment. In the specific implementation, the feature fine-grained interaction network is based on the self-attention mechanism. By calculating the attention weights between the anchored feature pairs, it dynamically adjusts the association strength between different features to capture more sophisticated feature interaction patterns between spatial dimension features and temporal dimension principal component features, thereby improving the accuracy and efficiency of feature interaction analysis. In addition, the feature fine-grained interaction network further enhances the integrity of feature representation by introducing the original spatial dimension features and the anchored temporal dimension feature principal component components, and obtains spatiotemporal fusion features, which helps to more comprehensively understand and predict the navigation behavior of ships.

[0045] In particular, in a preferred example of the present application, the step S33 includes: first, performing feature power law distribution alignment constraints on the spatial dimension deep implicit features and the anchored time dimension principal component deep implicit features to obtain {optimized spatial dimension deep implicit features, optimized anchored time dimension principal component deep implicit features} feature pairs, which are expressed as:

[0046]

[0047] in, represents an exponential function with a natural constant as base, and They are vectors and Middle The eigenvalues ​​at the positions, Represents the optimized spatial dimension deep implicit features, Represents the optimized deep latent features of the principal components anchored in the time dimension.

[0048] Next, the {optimized deep latent features of spatial dimension, optimized deep latent features of principal components of anchored time dimension} feature pair is input into a feature fine-grained interaction network to obtain the spatiotemporal fusion features.

[0049] Here, the present application introduces a weakening and virtualization mechanism to perform weak virtualization power-rate expansion in order to address the alignment ambiguity between spatial and temporal dimension features caused by source uncertainty. That is, the 3 / 8 exponent is used as a precursor a priori extension to perform power-prior responsive fuzzification convergence on the feature parameter boundary alignment conditions, and the 1 / 4 exponent is used as the main alignment extension to strictly constrain the feature power-rate distribution alignment attenuation relaxation. In this way, when there is ambiguity in the definition of the alignment boundary condition constraints within the associated effective range, the value relevance mechanism is used to avoid the a priori ambiguity of the system behavior under a single mechanism, so as to correct the semantic distribution consistency fuzzification within the alignment interval, thereby improving the intuitiveness of mining the implicit fine-grained interactive correlation features between the deep implicit features of the spatial dimension and the anchored deep implicit features of the time dimension principal components.

[0050] In the above-mentioned ship track prediction method based on spatiotemporal feature fusion, the step S4 predicts the ship track information based on the spatiotemporal fusion feature. It should be understood that predicting the ship track information based on the spatiotemporal fusion feature is a prior art. For example, it can adopt the principle disclosed in Chinese patent CN117933492A to predict the ship track information. Of course, it can also adopt other principles to predict the ship track information, which is not limited by this application. For example, a prediction model can be constructed using LSTM (Long Short-Term Memory Network) to utilize the excellent long-term memory ability of the LSTM model to capture the long-term dependency of the ship's navigation behavior in the spatiotemporal fusion feature, thereby generating a prediction result for the next time point based on the known spatiotemporal state of the ship's track. Repeat until the predetermined prediction length is reached. It is worth noting that after each prediction is completed, the newly generated track information can be fed back to the model as input for the next stage of prediction, thereby realizing rolling prediction.

[0051] Specifically, when processing ship track data, the fusion of time dimension features and space dimension features makes it possible to understand complex navigation patterns. The LSTM model, through its unique gating mechanism, can effectively learn and remember important information in long time series while ignoring irrelevant details, making it an ideal choice for dealing with such problems.

[0052] The core of the LSTM model lies in its cell state, which is an information flow throughout the chain structure, allowing information to flow in a controlled manner. At each time step, LSTM decides which new information will be added to this state and which old information should be forgotten. This mechanism makes LSTM particularly good at processing data with long-term correlation, such as the position, speed and direction of a ship at sea that changes over time.

[0053] When designing the architecture of the LSTM model, you can choose LSTM layers with different numbers of layers and units according to the specific task requirements and available computing resources. Generally speaking, increasing the number of LSTM layers can improve the model's ability to capture complex patterns, but it will also increase training time and the risk of overfitting. Therefore, it is necessary to weigh the pros and cons according to the actual situation. In addition, you can also consider adding a fully connected layer (Dense Layer) after the LSTM layer to further extract high-level features and output the final prediction results. In this process, the choice of activation function is also very important. For example, using ReLU (Rectified Linear Unit) as the activation function of the hidden layer can speed up training, while Sigmoid or Softmax are often used in the output layer.

[0054] It is worth noting that although the LSTM model has a strong long-term memory capability, in practical applications, it is still necessary to prevent the problem of gradient vanishing or exploding. These problems may cause the model to be difficult to converge or fall into a local optimal solution. To this end, some technical means can be taken, such as setting reasonable initial weights and using gradient clipping to alleviate these problems.

[0055] Finally, when generating the prediction result for the next time point based on the known spatiotemporal state of the ship's track, the LSTM model not only relies on the input data at the current moment, but also refers to the memory information of multiple previous time steps. This means that for each ship, the LSTM model can infer the possible future navigation path based on the ship's navigation history over a period of time, combined with the spatial position and motion parameters obtained in real time. This feature makes LSTM particularly powerful in handling prediction tasks in dynamically changing environments, helping to discover potential risks in advance and optimize route planning, thereby improving overall shipping efficiency and safety.

[0056] In summary, the ship track prediction method based on the fusion of spatiotemporal features based on the embodiment of the present application is explained, which uses a neural network model based on deep learning to process the AIS information at each time in the known ship track to extract the time dimension characteristics and space dimension characteristics of the ship track, and then, through the implicit correlation explicit modeling analysis of the time dimension characteristics and space dimension characteristics of the ship track, the implicit key clues between the two are excavated, and based on this, the alignment and interactive fusion of the time dimension characteristics and space dimension characteristics of the ship track is realized, and then the future track of the ship is predicted based on the spatiotemporal fusion characteristics of the ship track. In this way, the correlation between the ship's sailing speed, course change and geographical location can be deeply excavated, so as to achieve more accurate feature alignment and interactive fusion, thereby improving the accuracy of ship track prediction.

[0057] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0058] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0059] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0060] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A ship track prediction method based on spatiotemporal feature fusion, characterized in that: include: Obtain known ship tracks; Based on the AIS information at each time in the known ship trajectory, extracting the time dimension features and space dimension features of the known ship trajectory; Performing a spatiotemporal feature interactive fusion based on implicit association explicit modeling on the time dimension feature and the space dimension feature to obtain a spatiotemporal fusion feature; Based on the spatiotemporal fusion features, the ship track information is predicted.

2. The ship track prediction method based on spatiotemporal feature fusion according to claim 1 is characterized in that: The time dimension feature and the space dimension feature are interactively fused based on implicit association explicit modeling to obtain a time-space fusion feature, including: Performing a principal component analysis on the time dimension feature to obtain a set of principal component components of the time dimension feature; Performing implicit correlation analysis on each time dimension feature principal component component in the set of the space dimension feature and the time dimension feature principal component component to construct a feature pair of {space dimension deep implicit feature, anchored time dimension principal component deep implicit feature}; The spatial-temporal feature interactive fusion is performed on the feature pair of {spatial dimension deep implicit feature, anchored time dimension principal component deep implicit feature} to obtain the spatial-temporal fusion feature.

3. The ship track prediction method based on spatiotemporal feature fusion according to claim 2 is characterized in that: Performing implicit correlation analysis on each time dimension feature principal component component in the set of the spatial dimension feature and the time dimension feature principal component component to construct a feature pair of {spatial dimension deep implicit feature, anchored time dimension principal component deep implicit feature}, including: Respectively extracting the deep implicit features of each time dimension feature principal component in the set of the spatial dimension feature and the time dimension feature principal component to obtain a set of spatial dimension deep implicit features and time dimension principal component deep implicit features; The set of the spatial dimension deep implicit features and the temporal dimension principal component deep implicit features is subjected to implicit key clue anchor encoding to obtain the feature pair of {spatial dimension deep implicit features, anchored temporal dimension principal component deep implicit features}.

4. The ship track prediction method based on spatiotemporal feature fusion according to claim 3 is characterized in that: The deep implicit features of each time dimension feature principal component in the set of the space dimension feature and the time dimension feature principal component are extracted respectively to obtain a set of space dimension deep implicit features and time dimension principal component deep implicit features, including: The implicit feature extraction based on point convolution coding is performed on each time dimension feature principal component in the set of the space dimension feature and the time dimension feature principal component to obtain the set of the space dimension deep implicit feature and the time dimension principal component deep implicit feature.

5. The ship track prediction method based on spatiotemporal feature fusion according to claim 4 is characterized in that: The set of the spatial dimension deep implicit features and the time dimension principal component deep implicit features is subjected to implicit key clue anchor encoding to obtain the {spatial dimension deep implicit features, anchored time dimension principal component deep implicit features} feature pair, including: The feature correlation factor between each time dimension principal component deep implicit feature in the set of the spatial dimension deep implicit feature and the time dimension principal component deep implicit feature is calculated, and the time dimension principal component deep implicit feature corresponding to the largest feature correlation factor is selected as the time dimension principal component deep implicit feature anchored by the spatial dimension deep implicit feature, so as to construct the {spatial dimension deep implicit feature, anchored time dimension principal component deep implicit feature} feature pair.

6. The ship track prediction method based on spatiotemporal feature fusion according to claim 5 is characterized in that: The spatial-temporal feature interactive fusion is performed on the feature pair of {spatial dimension deep implicit feature, anchored time dimension principal component deep implicit feature} to obtain the spatial-temporal fusion feature, including: The feature pair of {spatial dimension deep implicit features, anchored time dimension principal component deep implicit features} is input into the feature fine-grained interaction network to obtain the spatiotemporal fusion features.

7. The ship track prediction method based on spatiotemporal feature fusion according to claim 6 is characterized in that: The {spatial dimension deep implicit feature, anchored time dimension principal component deep implicit feature} feature pair is input into the feature fine-grained interaction network to obtain the spatiotemporal fusion feature, including: Taking the spatial dimension deep implicit feature as a query vector, taking the spatial dimension feature as a value vector, taking the anchored time dimension principal component deep implicit feature as a key vector, inputting the query vector, the key vector and the value vector into a feature attention interaction module based on the Transformer architecture to obtain a spatial dimension-time dimension attention interaction feature; The spatial dimension-temporal dimension attention interaction feature and the time dimension feature principal component corresponding to the anchored time dimension principal component deep implicit feature are linearly weighted fused to obtain the spatiotemporal fusion feature.

8. The ship track prediction method based on spatiotemporal feature fusion according to claim 5 is characterized in that: The spatial-temporal feature interactive fusion is performed on the feature pair of {spatial dimension deep implicit feature, anchored time dimension principal component deep implicit feature} to obtain the spatial-temporal fusion feature, including: Performing feature power-law distribution alignment constraints on the spatial dimension deep implicit features and the anchored time dimension principal component deep implicit features to obtain a feature pair of {optimized spatial dimension deep implicit features, optimized anchored time dimension principal component deep implicit features}; The {optimized spatial dimension deep latent feature, optimized anchored time dimension principal component deep latent feature} feature pair is input into the feature fine-grained interaction network to obtain the spatiotemporal fusion feature.

Images