An automatic identification method for tugboat operation state based on AIS data

Abstract

The present application relates to the technical field of ship traffic management, and particularly relates to a tug operation state automatic identification method based on AIS data, which comprises the following steps: collecting AIS data of a tug and performing pretreatment, and then adopting a sliding window method based on a speed threshold to divide a tug trajectory and preliminarily distinguish a tug state; based on the distinguishing result of the tug state, extracting statistical features and descriptive features from each navigation section of the tug, and constructing a tug state feature vector; according to the tug state feature vector, using a first state classification model to classify the navigation section of the tug into a cruising state and an auxiliary berthing and unberthing state; based on speed dynamic features of the tug, further dividing the auxiliary berthing and unberthing state into an auxiliary berthing state and an auxiliary unberthing state. The present application can accurately identify fine-grained operation behaviors of a tug, improve the efficiency and intelligent degree of port tug supervision, and further improve the port scheduling efficiency and safety management level.

Description

Technical Field

[0001] This invention relates to the field of ship traffic management technology, and in particular to an automatic identification method for tugboat operation status based on AIS data. Background Technology

[0002] Traditional manual monitoring methods are ill-suited to the short-distance, highly mobile operation characteristics of tugboats, and cannot achieve large-scale, real-time, and refined status monitoring and scheduling optimization. Therefore, developing data-driven automatic tugboat operation status identification technology is an urgent practical need to improve the transparency and intelligence of port operations.

[0003] Current research largely focuses on tugboat scheduling optimization, maneuverability analysis, and emission characteristic assessment, while research on automatic identification of operational status remains insufficient. Specifically, existing research suffers from the following shortcomings: First, some methods heavily rely on the accuracy of AIS status information, limiting their practicality; second, some methods identify the cooperative relationship between tugboats and large vessels through logistic regression, but fail to effectively distinguish the content of the cooperation (i.e., specific berthing and unberthing operations); most importantly, no research has systematically analyzed and quantified the differences in trajectory characteristics of tugboats during assisted berthing and unberthing processes, which is crucial for achieving high-precision identification. Furthermore, most studies have failed to fully utilize the multidimensional information in AIS data (such as speed, heading, and trajectory morphology) to construct a discriminative feature system, nor have they introduced advanced machine learning methods to address class imbalance and noisy data issues.

[0004] Although artificial intelligence technologies have been applied to AIS data mining, including anomaly detection based on clustering and random forest, trajectory prediction based on LSTM, and multimodal trajectory prediction frameworks, these methods have still failed to solve the core challenge in tugboat operation status identification: how to extract discriminative features from high-noise AIS data and achieve fine-grained distinction between berthing and unberthing operations. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an automatic identification method for tugboat operation status based on AIS data.

[0006] To achieve the above objectives, this invention provides an automatic identification method for tugboat operation status based on AIS data. The method includes the following steps: collecting and preprocessing AIS data of the tugboat; then using a sliding window method based on a speed threshold to divide the tugboat trajectory and initially distinguish its status; based on the distinction of the tugboat status, extracting statistical and descriptive features from each navigation segment of the tugboat to construct a tugboat status feature vector; according to the tugboat status feature vector, using a first state classification model to classify the tugboat's navigation segment into a cruising state and an auxiliary berthing / departure state; and further dividing the auxiliary berthing / departure state into an auxiliary berthing state and an auxiliary departure state based on the tugboat's dynamic speed characteristics. This invention can accurately identify fine-grained tugboat operation behaviors, improve the efficiency and intelligence of port tugboat supervision, and thus enhance port scheduling efficiency and safety management.

[0007] Optionally, the AIS data includes the tugboat's latitude, longitude, smooth speed, and heading at different points in time.

[0008] Optionally, the tugboat status includes a moored state and a sailing state.

[0009] Optionally, the process of collecting and preprocessing the AIS data of the tugboat, and then using a sliding window method based on a speed threshold to divide the tugboat trajectory and initially distinguish the tugboat's status, includes the following steps:

[0010] The AIS data of the tugboat is collected and preprocessed, and then the AIS data is sorted by time to obtain the tugboat trajectory data;

[0011] Set the sliding window step size, and start recording the AIS data when the smooth speed is greater than the speed threshold continuously, until the smooth speed is no greater than the speed threshold.

[0012] The tugboat trajectory corresponding to the recorded AIS data is recorded as a navigation segment, and the tugboat state of the navigation segment is the navigation state.

[0013] The tugboat status corresponding to the unrecorded AIS data is recorded as the mooring status.

[0014] Optionally, the sliding window step size is 1.

[0015] Optionally, the speed threshold is 0.3 knots.

[0016] Optionally, the statistical features include speed features, heading features, and spatial features. The speed features include average speed, maximum speed, maximum speed change, median speed change, sum of absolute speed changes, and mean speed change. The heading features include mean heading change, maximum heading change, median heading change, and heading change range. The spatial feature is the straight-line distance between the beginning and end points of the trajectory segment.

[0017] Optionally, the descriptive features include the speed change ratio, the head-to-tail distance ratio, the maximum distance ratio, and the overlap ratio.

[0018] Optionally, classifying the tugboat's navigation segment into a cruising state and an auxiliary berthing / unberthing state using a first state classification model based on the tugboat's state feature vector includes the following steps:

[0019] Collect historical AIS data and historical navigation status of tugboats during the navigation segment, and obtain historical statistical and descriptive characteristics of the tugboats;

[0020] The model training dataset is constructed using the historical navigation status, the historical statistical features, and the historical descriptive features;

[0021] The first state classification model is constructed using the model training dataset and a fully connected neural network;

[0022] The tugboat state feature vector is input into the first state classification model, and then the navigation state is output. The navigation state includes the cruise state and the assisted berthing and unberthing state.

[0023] Optionally, based on the dynamic speed characteristics of the tugboat, the assisted berthing and unberthing states are further divided into assisted berthing state and assisted unberthing state, including the following steps:

[0024] The navigation segment is divided into a first half and a second half at the midpoint of time, and the average speed of the first half and the second half is calculated respectively.

[0025] If the average speed of the first half is greater than the average speed of the second half, it is determined to be in assisted berthing state; otherwise, it is determined to be in assisted departure state.

[0026] The present invention has at least the following beneficial effects:

[0027] 1. A complete technical framework for tugboat operation status recognition was constructed, including trajectory division, feature engineering and classification model, which improved the efficiency and intelligence of tugboat supervision, thereby enhancing port scheduling efficiency and safety management level, and providing a reference method for similar ship behavior recognition research.

[0028] 2. A tugboat status identification scheme based on conventional AIS data was developed, which does not rely on additional hardware equipment and provides a low-cost solution for intelligent port supervision.

[0029] 3. The trajectory characteristics of tugboat operations were analyzed in depth, and a multi-dimensional feature system was designed, especially four descriptive features, which effectively captured the subtle differences in tugboat operations. Combined with the dynamic characteristics of tugboat speed, the system achieved accurate identification of fine-grained tugboat operation behavior. Attached Figure Description

[0030] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a flowchart illustrating an automatic identification method for tugboat operation status based on AIS data according to an embodiment of the present invention.

[0032] Figure 2 This is a schematic diagram illustrating the distribution differences of average speed under different conditions according to an embodiment of the present invention;

[0033] Figure 3 This is a schematic diagram illustrating the distribution differences of the maximum speed under different conditions according to an embodiment of the present invention;

[0034] Figure 4 This is a schematic diagram illustrating the distribution differences of the sum of absolute values ​​of speed changes under different states in an embodiment of the present invention.

[0035] Figure 5 This is a schematic diagram illustrating the distribution differences of the mean heading change under different states in an embodiment of the present invention;

[0036] Figure 6 This is a schematic diagram showing the distribution differences of the maximum heading change value under different states in an embodiment of the present invention.

[0037] Figure 7 This is a schematic diagram illustrating the distribution differences of the median heading change under different states in an embodiment of the present invention;

[0038] Figure 8 This is a schematic diagram illustrating the distribution differences of the heading change range under different states according to an embodiment of the present invention;

[0039] Figure 9 This is a schematic diagram illustrating the distribution differences of the mean change in speed under different conditions according to an embodiment of the present invention.

[0040] Figure 10This is a schematic diagram illustrating the distribution differences of the maximum speed change under different conditions according to an embodiment of the present invention;

[0041] Figure 11 This is a schematic diagram illustrating the distribution differences of the median change in air speed under different conditions according to an embodiment of the present invention.

[0042] Figure 12 This is a schematic diagram illustrating the differences in the distribution of the straight-line distance between the first and last points of the trajectory segment under different states according to an embodiment of the present invention.

[0043] Figure 13 This is a schematic diagram of the characteristic distribution pattern of the speed change ratio in an embodiment of the present invention;

[0044] Figure 14 This is a schematic diagram of the characteristic distribution pattern of the distance ratio between the first and last points in an embodiment of the present invention;

[0045] Figure 15 This is a schematic diagram of the characteristic distribution pattern of the maximum distance ratio in an embodiment of the present invention;

[0046] Figure 16 This is a schematic diagram of the characteristic distribution pattern of overlap rate in an embodiment of the present invention;

[0047] Figure 17 This refers to the accuracy of the first state classification model in this embodiment of the invention on the training set and validation set during training.

[0048] Figure 18 This refers to the loss value on the training set and validation set during the training process of the first state classification model in this embodiment of the invention. Detailed Implementation

[0049] Specific embodiments of the present invention will now be described in detail. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been specifically described to avoid obscuring the invention.

[0050] Throughout this specification, references to "an embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in an embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale.

[0051] It should be noted in advance that, in one alternative embodiment, except for independent descriptions, the same symbols or letters appearing in all formulas have the same meaning and value.

[0052] In one optional embodiment, please refer to Figure 1 This invention provides an automatic identification method for tugboat operation status based on AIS data, the method comprising the following steps:

[0053] S1. Collect and preprocess the AIS data of the tugboat, and then use the sliding window method based on the speed threshold to divide the tugboat trajectory and initially distinguish the tugboat status.

[0054] The tugboat's status includes both moored and sailing states. Step S1 specifically includes the following steps:

[0055] S11. Collect AIS data of the tugboat and preprocess it. Then sort the AIS data by time to obtain the tugboat trajectory data.

[0056] Specifically, in this embodiment, AIS data of the tugboat is acquired in real time, and preprocessing operations such as outlier removal and missing value filling are performed on the acquired AIS data. Then, the preprocessed AIS data is sorted by time to obtain the tugboat trajectory data.

[0057] In this embodiment, the collected AIS data includes the tugboat's latitude, longitude, smoothed speed, and heading at different times. The tugboat trajectory data can be represented as follows:

[0058]

[0059] in, For the trajectory data of the k-th tugboat, For the kth tugboat at time point latitude, For the kth tugboat at time point longitude, For the kth tugboat at time point Smooth speed, For the kth tugboat at time point The course, , .

[0060] S12. Set the sliding window step size. When the smooth speed is greater than the speed threshold consecutively, start recording the AIS data until the smooth speed is no greater than the speed threshold.

[0061] Specifically, in this embodiment, the sliding window step size is set to 1, the speed threshold is set to 0.3 knots, the window is slid according to the set sliding window step size, and it is determined whether the smooth speed within the window is greater than the speed threshold. If from a certain point in time... Initially, if there are two or more consecutive instances where the smoothed speed exceeds the speed threshold, then starting from the time point... Start recording AIS data and stop recording when the smoothed airspeed does not exceed the airspeed threshold at a certain time point.

[0062] S13. The tugboat trajectory corresponding to the recorded AIS data is recorded as a navigation segment, and the tugboat state of the navigation segment is the navigation state.

[0063] Specifically, in this embodiment, the tugboat trajectory corresponding to the time period of the AIS data recorded in step S12 is recorded as a navigation segment, and the tugboat status of the navigation segment is the navigation status. It should be noted that there may be multiple navigation segments.

[0064] S14. Record the tugboat status corresponding to the unrecorded AIS data as the mooring status.

[0065] Specifically, in this embodiment, the tugboat status during the time period in which the unrecorded AIS data is located is the docked status, and it will not be involved in feature extraction and classification in the future.

[0066] S2. Based on the differentiation results of the tugboat's state, statistical features and descriptive features are extracted from each navigation segment of the tugboat to construct a tugboat state feature vector.

[0067] Specifically, in this embodiment, in order to distinguish between cruising and assisted berthing / departure states, 15-dimensional features are extracted from each navigation segment, including 11 statistical features and 4 descriptive features, to characterize trajectory differences from three dimensions: speed, heading, and spatial morphology.

[0068] The statistical characteristics include speed characteristics, heading characteristics, and spatial characteristics. The speed characteristics include average speed, maximum speed, change in maximum speed, median change in speed, sum of absolute changes in speed, and mean change in speed. The heading characteristics include mean change in heading, maximum change in heading, median change in heading, and range of heading change. The spatial characteristic is the straight-line distance between the beginning and end points of the trajectory segment. Each statistical characteristic satisfies the following relationship:

[0069]

[0070]

[0071]

[0072]

[0073]

[0074]

[0075]

[0076]

[0077]

[0078]

[0079]

[0080] in, The average speed is given by n, and the number of AIS data entries within the flight segment is given by n. For the kth tugboat at time point Smooth speed, , For maximum speed, This represents the maximum change in speed. For the kth tugboat at time point Smooth speed, This represents the median change in speed. This indicates taking the median. It is the sum of the absolute values ​​of the changes in speed. This represents the average change in speed. This represents the average change in heading. For the kth tugboat at time point The course, For the kth tugboat at time point The course, This represents the maximum change in heading. This represents the median change in heading. For the range of course changes, The straight-line distance between the first and last points of the trajectory segment. For point With point The distance between them For the kth tugboat at time point longitude, For the kth tugboat at time point Latitude.

[0081] Descriptive features include the speed change ratio, the ratio of the distance between the bow and stern points, the maximum distance ratio, and the overlap ratio. Each descriptive feature satisfies the following relationship:

[0082]

[0083]

[0084]

[0085]

[0086] in, The ratio of the change in speed. The ratio of the distance between the first and last points. This represents the maximum straight-line distance within the navigation segment. The maximum distance ratio, The total distance traveled within the navigation segment. The overlap rate, Counting the number of times a vehicle enters an overlapping grid within a navigation segment. The first entry into the grid within the navigation segment is counted, and `const` is the amplification factor. The speed change ratio measures the degree of speed fluctuation, and this value is usually lower in assisted berthing and departure states; the head-to-tail distance ratio reflects the degree of trajectory reversal, and this value is smaller in assisted berthing and departure states; the maximum distance ratio describes the tortuosity of the trajectory, and this value is higher in navigation states; the overlap rate is calculated based on a 100 m × 100 m grid, and `const` is set to 100 to amplify the differences.

[0087] The 15 features obtained are Z-score standardized to obtain a 15-dimensional tugboat state feature vector.

[0088] S3. Based on the tugboat state feature vector, the first state classification model is used to classify the state of the tugboat's navigation segment into cruise state and assisted berthing / departure state.

[0089] Step S3 specifically includes the following steps:

[0090] S31. Collect historical AIS data and historical navigation status of tugboats in the navigation segment, and obtain historical statistical characteristics and historical descriptive characteristics of tugboats.

[0091] Specifically, in this embodiment, using AIS data of tugboats at a port in 2020 as the target, 572 tugboat operation trajectories within six months are selected as source data. After initially distinguishing the tugboat status in step S1, a total of 483 historical navigation segments and their historical navigation statuses are obtained. The historical AIS data and historical navigation status of one historical navigation segment constitute a raw data sample. To alleviate the class imbalance problem, the SMOTE oversampling technique is used to balance the number of raw data samples for each class to 242. Subsequently, for each raw data sample, the historical statistical characteristics and historical descriptive characteristics of the tugboats are obtained according to the description in step S2.

[0092] After obtaining historical statistical and descriptive features, this embodiment further conducted a detailed distribution analysis on 15 features.

[0093] Figures 2 to 12 The analysis results for 11 statistical characteristics are presented sequentially, including average speed, maximum speed, sum of absolute speed changes, mean course change, maximum course change, median course change, course change range, mean speed change, maximum speed change, median speed change, and straight-line distance between the beginning and end points of the trajectory segment. It should be noted that... Figures 2 to 12 In the diagram, because the bar charts for the two states (cruising state and assisted berthing / departure state) are displayed on the same numerical range, the colors are mixed to form an "overlapping area".

[0094] It is easy to see that, in terms of speed characteristics, the average speed during cruising is concentrated in the 6-8 knot range, and speed change indicators such as the median speed change and the sum of the absolute values ​​of speed changes are generally low, reflecting its stable navigation characteristics. Conversely, the average speed distribution during berthing and unberthing is more dispersed, mainly between 2-6 knots, and the sum of speed changes is significantly higher, confirming the frequent acceleration and deceleration characteristics of this type of operation.

[0095] The heading characteristics exhibit a similar pattern to the speed characteristics. During cruising, the average heading change is mostly below 5°, and the range is typically less than 30°, indicating stable heading. However, the heading change indicators are significantly higher during berthing and unberthing operations, with the average heading change often exceeding 10° and the range mostly greater than 90°. This is entirely consistent with the operational characteristics of tugboats assisting large vessels, requiring frequent adjustments to their pushing and pulling directions.

[0096] Among the spatial characteristics, the distribution of straight-line distances between the beginning and end points of the trajectory segment shows particularly significant differences. In cruise mode, this value mostly exceeds 1000 meters, reflecting its long-distance straight-line navigation characteristics; while in assisted berthing and unberthing mode, this value is mostly below 500 meters, confirming its limited operating range and frequent return navigation.

[0097] Figures 13 to 16 The analysis results of four descriptive features are presented in sequence: speed change ratio, head-to-tail distance ratio, maximum distance ratio, and overlap rate.

[0098] It is easy to see that the speed change ratio has a single-peak distribution (peak value of about 0.5) in cruise mode, which is consistent with the typical navigation pattern of "acceleration-uniform speed-deceleration"; while in assisted berthing and unberthing mode, the value is more dispersed and lower overall, reflecting its complex and variable speed pattern.

[0099] The distance ratio between the fore-and-aft points is highly concentrated around 1.0 during cruising, indicating near-straight-line navigation; while during assisted berthing and unberthing, this value is mainly distributed in the range of 0.2-0.6, showing obvious turning characteristics. The distribution of the maximum distance ratio further supports this conclusion, with the value mostly above 0.7 during cruising and mostly below 0.5 during assisted berthing and unberthing.

[0100] The overlap rate distribution is almost completely separated between the two states: in the cruise state, it is mostly below 10, while in the assisted berthing and unberthing state, it is mostly above 20, which strongly confirms the unique turnaround navigation mode of the latter.

[0101] Figures 2 to 16 The observed feature distribution pattern not only verifies the rationality of the feature design in this embodiment, but also provides an interpretable basis for the high classification performance of the first-state classification model in the subsequent model.

[0102] It should be noted that there are two categories described in this embodiment: cruising state and assisted berthing / departure state.

[0103] S32. Construct a model training dataset using the historical navigation status, the historical statistical features, and the historical descriptive features.

[0104] Specifically, in this embodiment, the historical statistical features, historical descriptive features, and historical navigation status of a unified historical navigation segment are used as a sample, and then multiple samples are used to construct a model training dataset.

[0105] S33. Construct the first state classification model using the model training dataset and a fully connected neural network.

[0106] Specifically, in this embodiment, a three-layer fully connected neural network FCNN is constructed, comprising an input layer, hidden layer 1, hidden layer 2, and an output layer. First, the input layer receives a 15-dimensional feature vector. Then, the data enters hidden layer 1, which has 64 neurons and uses the ReLU activation function. Dropout (with a rate of 0.4) and He initialization are applied to prevent overfitting and accelerate model convergence. Next, the data is passed to hidden layer 2, which contains 32 neurons and also uses the ReLU activation function to further extract deeper features from the data. Finally, in the output layer, the probability of a flight segment belonging to each category is output through two neurons and the Softmax activation function. To optimize the model training process, this embodiment uses the Adam optimizer with a learning rate of 0.0001. L2 regularization (weight decay set to 0.005) and an early stopping strategy (patience value set to 30) are also introduced to enhance the model's generalization ability and ensure excellent performance on unknown data.

[0107] Furthermore, the model employs a sparse classification cross-entropy loss function and introduces class weights to address the issue of imbalanced samples. Therefore, the model's loss function can be simplified as follows:

[0108]

[0109] in, Let r be the weighted loss value for the r-th sample. For category weights, Based on the true category of the r-th sample To determine, This is the loss value for the r-th sample calculated using the sparse classification cross-entropy loss function.

[0110] Furthermore, the model training dataset was randomly divided into training and test sets in a 7:3 ratio to complete the training and validation of FCNN. The model evaluation metrics included precision, recall, and F1 score. The training process is as follows: Figure 17 and Figure 18 As shown in Table 1, the evaluation metrics on the test set are as follows.

[0111] Table 1. Evaluation metrics of the model on the test set

[0112]

[0113] from Figure 17 and Figure 18As shown in the figure, the training process exhibits good convergence characteristics: the accuracy rises rapidly from the initial 0.47 and eventually stabilizes above 0.95; the loss value decreases significantly from 1.41 to around 0.32. The performance on the validation set remains highly consistent with that on the training set, with the final accuracy reaching 0.9412 and the loss value approximately 0.325, indicating that the model has strong generalization ability.

[0114] It is worth noting that the application of the SMOTE oversampling technique effectively alleviated the class imbalance problem. In the early stages of training, the model showed a preference for the majority class (cruising state), but with the application of oversampling and the weighted loss function, the model's learning ability for the minority class (assist berthing and unberthing states) was significantly improved. This improvement was directly reflected in the high recall rate of the "assist berthing and unberthing state" category on subsequent test sets.

[0115] As shown in Table 1, overall, the first-state classification model achieved an F1 score of 0.9, indicating that the model has good comprehensive classification ability. For the "cruising state" category, the model showed high precision (0.93) but relatively low recall (0.86), indicating that the model was cautious in judging this category and may have misclassified some boundary cases as auxiliary berthing / departure. For the "auxiliary berthing / departure state" category, the high recall (0.93) indicates that the model can effectively capture the key features of this category, but the relatively low precision (0.87) suggests that some abnormal cruising patterns may be misclassified as auxiliary operations. These phenomena are consistent with the actual complexity of tugboat operations. In the port environment, the cruising patterns of tugboats may exhibit diverse characteristics due to factors such as traffic control and temporary task scheduling, and the boundary with auxiliary operations may be blurred in some cases.

[0116] S34. Input the tugboat state feature vector into the first state classification model, and then output the navigation state, which includes the cruise state and the assisted berthing and unberthing state.

[0117] Specifically, in this embodiment, by inputting the real-time tugboat state feature vector into the first state classification model, it can be determined whether the tugboat is currently in a cruising state or an assisted berthing / unberthing state.

[0118] S4. Based on the dynamic speed characteristics of the tugboat, the assisted berthing and unberthing states are further divided into assisted berthing state and assisted unberthing state.

[0119] Based on the "assisted departure state" category output by the first-state classification model, the assisted berthing state and assisted departure state are further distinguished according to the dynamic characteristics of speed. Step S4 specifically includes the following steps:

[0120] S41. Divide the navigation segment into a first half and a second half according to the midpoint of time, and calculate the average speed of the first half and the second half respectively.

[0121] S42. If the average speed of the first half is greater than the average speed of the second half, it is determined to be in assisted berthing state; otherwise, it is determined to be in assisted departure state.

[0122] The rule in this embodiment that distinguishes between assisted berthing and assisted unberthing states is based on the typical speed change patterns of tugboats during berthing and unberthing operations, and has high interpretability and practicality.

[0123] It should be noted that in some cases, the actions described in the specification can be performed in different orders and still achieve the desired results. In this embodiment, the order of steps is given only to make the embodiment clearer and easier to explain, and not to limit it.

[0124] In summary, this invention has at least the following beneficial effects: It constructs a complete technical framework for tugboat operation status recognition, including trajectory segmentation, feature engineering, and classification models, improving the efficiency and intelligence of tugboat supervision, thereby enhancing port scheduling efficiency and safety management, and providing a reference method for similar vessel behavior recognition research; it develops a tugboat status recognition scheme based on conventional AIS data, requiring no additional hardware equipment, providing a low-cost solution for intelligent port supervision; it deeply analyzes the trajectory characteristics of tugboat operations, designs a multi-dimensional feature system, especially four descriptive features, effectively capturing subtle differences in tugboat operations, and, combined with the dynamic characteristics of tugboat speed, achieves accurate recognition of fine-grained tugboat operation behaviors.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. An automatic identification method of tugboat operation state based on AIS data, characterized in that, The method comprises the following steps: AIS data of the tugboat is collected and preprocessed, and then a sliding window method based on a speed threshold is used to divide the trajectory of the tugboat and preliminarily distinguish the state of the tugboat; Based on the distinguishing result of the state of the tugboat, statistical features and descriptive features are extracted from each voyage section of the tugboat to construct a state feature vector of the tugboat; The statistical features include speed features, heading features and spatial features, the speed features include average speed, maximum speed, maximum speed change, median speed change, sum of absolute values of speed change and mean speed change, the heading features include mean heading change, maximum heading change, median heading change and heading change range, and the spatial feature is the straight-line distance between the start point and the end point of the trajectory section; Each statistical feature satisfies the following relationship in turn: wherein, is the average speed, n is the number of AIS data pieces in the sailing section, is the smooth speed of the kth tugboat at the time point , , is the maximum speed, is the maximum speed change amount, is the smooth speed of the kth tugboat at the time point , is the median of the speed change amount, indicates taking the median, is the sum of the absolute values of the speed change amount, is the mean of the speed change amount, is the mean of the heading change amount, is the heading of the kth tugboat at the time point , is the heading of the kth tugboat at the time point , is the maximum value of the heading change amount, is the median of the heading change amount, is the heading change range, is the straight-line distance between the start point and the end point of the trajectory section, is the distance between the point and the point , is the longitude of the kth tugboat at the time point , is the latitude of the kth tugboat at the time point ; The descriptive features include speed change ratio, start-end point distance ratio, maximum distance ratio and overlap rate; Each descriptive feature satisfies the following relationship in turn: wherein, is a speed change ratio, is a head-to-tail distance ratio, is a maximum straight-line distance within a voyage segment, is a maximum distance ratio, is a total distance within a voyage segment, is an overlap ratio, is a count of entering overlapping grids within a voyage segment, is a count of first entering a grid within a voyage segment, const is a magnification factor; Historical AIS data and historical navigation states of the tugboat in the voyage section are collected, and historical statistical features and historical descriptive features of the tugboat are obtained; A model training data set is constructed using the historical navigation states, the historical statistical features and the historical descriptive features; A first state classification model is constructed using the model training data set and a fully connected neural network; The state feature vector of the tugboat is input into the first state classification model, and then a navigation state is output, the navigation state including a cruising state and an auxiliary berthing and unberthing state; Based on the speed dynamic features of the tugboat, the auxiliary berthing and unberthing state is further divided into an auxiliary berthing state and an auxiliary unberthing state.

2. The method according to claim 1, wherein: The AIS data includes latitude, longitude, smooth speed and heading of the tugboat at different time points.

3. The method according to claim 2, wherein: The state of the tugboat includes a berthing state and a navigation state.

4. The method of claim 3, wherein, The method of collecting AIS data of the tugboat and preprocessing, and then using a sliding window method based on a speed threshold to divide the trajectory of the tugboat and preliminarily distinguish the state of the tugboat comprises the following steps: The AIS data of the tugboat is collected and preprocessed, and then the AIS data is sorted by time to obtain the trajectory data of the tugboat; A sliding window step is set, and when the smooth speed continuously appears to be greater than the speed threshold, the AIS data is recorded until the smooth speed is not greater than the speed threshold; The trajectory of the tugboat corresponding to the recorded AIS data is recorded as a voyage section, and the state of the tugboat in the voyage section is the navigation state; The state of the tugboat corresponding to the unrecorded AIS data is recorded as the berthing state.

5. The method according to claim 4, wherein: The sliding window step is 1.

6. The method according to claim 4, wherein: The speed threshold is 0.3 knots.

7. The method of claim 1, wherein, The dynamic characteristics of the speed based on the tug further divide the auxiliary berthing and unberthing state into auxiliary berthing state and auxiliary unberthing state, including the following steps: The navigation section is divided into the first half and the second half according to the time midpoint, and the average speed of the first half and the second half is calculated respectively. If the average speed of the first half is greater than the average speed of the second half, it is determined as auxiliary berthing state, otherwise it is determined as auxiliary unberthing state.

Images