Ship behavior feature representation method and feature vector

Abstract

This application provides a method for representing ship behavior features and a feature vector. The method for representing ship behavior features includes: obtaining the original dataset of the ship, the original dataset including... N The system generates L trajectory segments from the original dataset. Each trajectory segment contains multiple original data points, representing the same or different ship behavior states. A ship behavior feature vector is constructed based on each trajectory segment. The original data represents time-point data of ship behavior generated at specific times, and the ship behavior feature vector is used to represent the ship behavior state over a specific time period. This application achieves the technical effect of compressing a large amount of time-point data into ship behavior feature vectors representing time periods.

Description

Technical Field

[0001] This application relates to the field of marine vessel behavior data processing, and in particular to a method for representing vessel behavior features and a feature vector. Background Technology

[0002] In recent years, automated threshold detection and even large model detection have been widely used in various fields of industry. The monitoring of the behavior of marine vessels has also gradually moved away from the monitoring methods that rely on manual data analysis and introduced machine detection methods such as threshold detection or large model detection. Due to the large variety of ship types and the different maritime management policies of various countries, different countries or different companies need to select appropriate raw data for threshold judgment or large model training according to different needs.

[0003] In existing technologies, raw ship data typically originates from Automatic Identification System (AIS) data, BeiDou Navigation Satellite System, etc., resulting in inconsistent data formats. Data users need to repeatedly extract different data from the raw data to meet different needs, leading to a significant amount of repetitive work during threshold comparisons or large model training. Due to the massive volume of raw data, there is a large amount of redundant data during transmission and sharing, making data sharing difficult and causing problems for the management of marine vessel behavior data. Summary of the Invention

[0004] The purpose of this application is to provide a method for representing ship behavior characteristics and a feature vector to solve the technical problems in the prior art, such as the chaotic format of the original ship data, the need for data users to repeatedly extract data from the original data for different needs, resulting in a lot of repetitive work, and the difficulty of data sharing due to the large amount of original data and the large amount of redundant data in transmission and sharing.

[0005] For the purposes mentioned above, this application provides the following technical solution: Firstly, this application provides a method for representing ship behavior characteristics, including: Obtain the raw dataset of the ships, the raw dataset including N Each piece of raw data includes a timestamp, Maritime Mobile Service Identifier (MMSI), longitude, latitude, ground speed, ground heading, heading, turning rate, and ship type code, wherein the longitude and latitude values ​​together constitute the position value of the vessel. The original dataset is segmented to obtain L Each trajectory segment contains multiple raw data points, and different trajectory segments represent the same or different behavioral states of the ship. Construct ship behavior feature vectors using trajectory segments as units; The original data consists of time-point data on ship behavior generated at a specific time, and the ship behavior feature vector is used to characterize the ship behavior state over a specific time period. and .

[0006] Furthermore, after obtaining the original dataset of the ships, the process also includes: The original dataset N The original data is sorted according to timestamps to obtain an ordered original dataset, which includes... Group of adjacent original data; According to the order of the original dataset The timestamps, longitude values, and latitude values ​​of adjacent raw data sets are calculated to obtain the results. The average speed of the ships, The average speed of each ship is characterized by The average velocity of adjacent sets of raw data; Will The average speed of each vessel is respectively correlated with the maximum speed corresponding to the vessel's type code. By comparing, we can obtain The results of the speed comparison were based on... The speed comparison results determine the jump point raw data in the original dataset; From the original dataset in the stated order N The original data with skip points is removed from the original data to obtain the sequential original dataset after skip point removal. The sequential original dataset after skip point removal includes... M The original data, of which .

[0007] Furthermore, The average speed of each vessel is respectively correlated with the maximum speed corresponding to the vessel's type code. By comparing, we can obtain The results of the speed comparison were based on... The speed comparison results determine the original jump point data in the original dataset, including: In the K The speed comparison result is the first K The average speed of the ships is greater than In the case of determining the first K The first original data and the first The original data is the jump point original data, among which, .

[0008] Furthermore, the original dataset is segmented to obtain... L Each trajectory segment includes at least one of the following: The order of the original dataset after removing jump points M In the original data, in the first... H The timestamp and the first If the difference between the timestamps is greater than 30 minutes, then in the case of the first... After dividing the original data into segments, we obtain the original dataset. L A trajectory segment; The order of the original dataset after removing jump points M In the original data, in the first... H The ground-oriented heading and the first When the difference in heading relative to the ground is greater than 60 degrees, in the first case... After dividing the original data into segments, we obtain the original dataset. L A trajectory segment; The order of the original dataset after removing jump points M In the original data, in the first... H The position value and the first When each of the position values ​​belongs to a different region, in the case of the _th _ After dividing the original data into segments, we obtain the original dataset. L One trajectory segment; in, The area is a pre-defined area range, including at least one of a port area, an anchorage area, and a waterway area.

[0009] Furthermore, a ship behavior feature vector is constructed using trajectory segments as units, including: The first l In each trajectory segment s The original data is sorted by timestamp; according to s The original data is used to obtain a ship behavior feature vector, which includes the ship's trajectory motion features, trajectory morphology features, and static profile features; The trajectory motion characteristics include at least one of the following: average speed, speed standard deviation, maximum speed, speed coefficient of variation, average turning rate, heading change rate, dwell time percentage, and low speed time percentage. The trajectory morphology features include at least one of the following: straightness, rotation index, convex hull area, trajectory compactness, number of turns, and beginning-end closure. The static profile features include at least one of the following: ship tonnage, main engine power, and ship length.

[0010] Furthermore, after obtaining the original dataset of the ships, the process also includes: Obtain an objective dataset of the vessel, which includes the vessel's historical average trajectory segment duration, port location, anchorage location, fishing ban area location, and waterway location; Accordingly, the ship behavior feature vector also includes the ship's historical overall operational features and spatial correlation features; Among them, the overall characteristics of historical operations include the average trajectory segment duration; The spatial association features include at least one of the following: distance to the nearest port, distance to the nearest anchorage, distance to the nearest fishing ban area, time spent in the fishing ban area, and distance to the nearest channel centerline.

[0011] Furthermore, the distance calculation in the spatial association feature adopts the first... l In each trajectory segment s Using the centroid of the curve formed by the ship position values ​​in the original data as a reference point, calculate at least one of the following distances from the reference point: the nearest port, the nearest anchorage, the nearest fishing ban area, and the nearest channel centerline.

[0012] Furthermore, the objective dataset also includes detailed information on the ship's historical trajectory segments; Accordingly, the ship behavior feature vector also includes one-hot encoded features and detailed historical operation features; The unique hot coding features include at least one of the following: ship type features, ship registration features, historical main operation type features, historical main operation area features, and area type features; The detailed characteristics of the historical tasks include the percentage of historical task types and the percentage of months for each trajectory segment.

[0013] Secondly, this application provides a ship behavior feature vector, applied to the ship behavior feature representation method described in the first aspect, including: The trajectory motion features include average speed, speed standard deviation, maximum speed, speed coefficient of variation, average turning rate, heading change rate, dwell time percentage, and low speed time percentage. The trajectory motion features are 8-dimensional vectors. The trajectory morphology features include straightness, rotation index, convex hull area, trajectory compactness, number of turns, and beginning-end closure. The trajectory morphology features are 6-dimensional vectors. Static profile features include ship tonnage, main engine power, and ship length; these static profile features are 3-dimensional vectors. The overall characteristics of historical tasks include the average duration of historical trajectory segments, and the overall characteristics of historical tasks are a one-dimensional vector. Spatial correlation features include distance to the nearest port area, distance to the nearest anchorage, distance to the nearest fishing ban area, time spent in the fishing ban area, and distance to the nearest channel centerline. The spatial correlation feature region is a 5-dimensional vector. Among them, the trajectory motion characteristics, trajectory morphology characteristics, static image characteristics, overall characteristics of historical operations, and spatial correlation characteristics are all continuous.

[0014] Furthermore, the ship behavior feature vector also includes: The hot-unique coding features include ship type features, ship registration features, historical main operation type features, historical main operation area features, and area type features. The hot-unique coding features are 60-dimensional vectors. Detailed features of historical tasks include the percentage of historical task types and the percentage of trajectory segments by month. These detailed features of historical tasks are 15-dimensional vectors. The hot unique coding features and the detailed features of historical operations are both continuous.

[0015] This application has at least the following advantages or beneficial effects: This application designs a method for representing ship behavior features, which constructs ship behavior feature vectors based on trajectory segments. This solves the technical problems that data users need to repeatedly extract different data from the original data for different needs, and that there is a lot of repetitive work in threshold comparison or large model training. The original data volume is extremely large, and there is a lot of redundant data in transmission and sharing, making data sharing difficult. This method achieves the technical effect of compressing a large amount of time point data into ship behavior feature vectors that represent time periods. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 A flowchart of a method for representing ship behavior characteristics provided in Embodiment 1 of this application; Figure 2 A flowchart of a method for representing ship behavior characteristics provided in Embodiment 2 of this application; Figure 3 This is a schematic diagram of the structure of the original dataset for ships; Figure 4 An example graph of specific raw data; Figure 5 This is a schematic diagram of the structure of a ship's original AIS dataset; Figure 6 This is a schematic diagram of skip point removal in a sequential raw dataset; Figure 7 This is a schematic diagram of the structure of a ship behavior feature vector; Figure 8 A schematic diagram of another type of ship behavior feature vector; Figure 9 This is a schematic diagram of a structure representing a trajectory motion characteristic; Figure 10 A schematic diagram of a trajectory morphology feature; Figure 11 This is a structural diagram of a static portrait feature; Figure 12 This is a structural diagram of a spatial association feature; Figure 13 This is a schematic diagram of the structure of a one-hot encoded feature; Figure 14 This is a structural diagram illustrating the detailed features of a historical task. Detailed Implementation

[0018] The technical solutions of this application will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0020] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0021] Furthermore, terms such as "horizontal" and "vertical" do not imply that the component must be absolutely horizontal or suspended, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," not that the structure must be completely horizontal, but can be slightly tilted. Similarly, terms such as "front," "back," "left," and "right" do not imply that the component must be absolutely front, back, left, or right, but can be slightly tilted.

[0022] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0023] Currently, mandatory communication systems include Automatic Identification System (AIS), Long-Range Identification and Tracking of Ships (LRIT), Vessel Monitoring System (VMS), VHF Data Exchange System (VDES), and BeiDou Navigation Satellite System. The data transmission patterns of these mandatory communication systems are roughly as follows: ships broadcast AIS data to nearby ships or base stations every 2 seconds to 3 minutes; the IMO mandates that some internationally navigating vessels transmit LRIT data every 15 minutes to 24 hours; specific vessels transmit VMS data to their flag state or fisheries management authority every 15 minutes to 6 hours; VDES data does not have a fixed transmission frequency, with ships requesting or transmitting data to base stations as needed, making data transmission more flexible; and BeiDou satellite data is transmitted every 1 minute to 1 hour, depending on the ship type. However, regardless of whether it is AIS data, LRIT data, VMS data, VDES data, or BeiDou satellite data, the data that ships send out are all point-in-time data. The operational status of ships often requires comprehensive analysis of data from multiple points in time over a period of time to obtain reliable conclusions. Therefore, AIS data, LRIT data, VMS data, VDES data, and BeiDou satellite data, as raw data, cannot be directly applied to threshold detection or large model detection. In other words, there is no existing technology that can comprehensively represent point-in-time data of different types or sources in a way that can characterize the ship's behavioral status over a specific time period.

[0024] Example 1 This application discloses a method for representing ship behavior features. The method can be used to obtain ship behavior features for a specific time period based on time point data of ship behavior. The method is applicable to ship types including merchant ships and fishing vessels, and supports using raw data such as AIS data, LRIT data, VMS data, VDES data or Beidou satellite data as data sources.

[0025] Figure 1 A flowchart of a method for representing ship behavior characteristics provided in Embodiment 1 of this application is provided. The method for representing ship behavior characteristics includes: S10. Obtain the ship's original dataset, the original dataset including... N Each piece of raw data includes a timestamp, Maritime Mobile Service Identifier (MMSI), longitude, latitude, ground speed, ground heading, heading, rate of turn, and ship type code. The longitude and latitude values ​​together constitute the position value of the ship.

[0026] Figure 3 This is a schematic diagram of the structure of the original dataset for ships, such as... Figure 3 As shown, the original dataset includes N The original data are: original data 1, original data 2, original data 3, ..., original data... N -2. Raw Data N -1 and original data N For the raw data N The structure of -2 is expanded as shown in the figure, original data. N -2 includes a timestamp, Maritime Mobile Service Identity (MMSI), longitude, latitude, ground speed, ground heading, heading, rate of turn (ROT), and ship type code. The timestamp indicates the point in time when the raw data was collected; the MMSI is a unique 9-digit identifier for a vessel, allowing for global unique identification of a single vessel; the longitude and latitude values ​​together constitute the vessel's position value, indicating the vessel's location at that point in time; the ground speed (SOG) represents the vessel's actual speed relative to the ground or seabed; the ground heading (COG) represents the vessel's direction of travel relative to the ground or seabed; the heading (TH) represents the direction the vessel's bow is facing; the rate of turn (ROT) represents the speed at which the vessel changes its heading; and the ship type code indicates the vessel's category and the type of cargo it carries.

[0027] Figure 4 This is an example diagram of specific raw data, specifically AIS data. Of course, the raw data can also be LRIT data, VMS data, VDES data, or BeiDou satellite data, etc. Figure 4 As shown, the timestamp of this raw data is recorded in ISO 8601 format. 2026-01-22T02:32:08Z indicates that this AIS data was generated at 2:32:08 AM on January 22, 2026. The MMSI is 413XXXXXX, where the first three digits are the country code, with "413" representing China. The longitude value ranges from -180° to 180°, and is represented as a floating-point number, with positive values ​​for east longitude and negative values ​​for west longitude. Figure 4 The "122.5840" shown represents 122.5840° east longitude. Latitude values ​​range from -90° to 90°, and are expressed as floating-point numbers; north latitude is positive and south latitude is negative. Figure 4 The "29.9210" shown represents 29.9210° North latitude. The speed relative to ground ranges from 0 to 102.2 knots; 3.20 indicates the ship's instantaneous speed is 3.20 knots. The heading ranges from 0° to 360°, with true north at 0°. "087" indicates 87° east of north. The heading ranges from 0° to 359° or 511; 511 is invalid. Figure 4 The "089" shown indicates 89° east of north. The turning rate ranges from -127° / min to 127° / min, with positive values ​​for right turns and negative values ​​for left turns. "5" indicates that the ship's instantaneous turning rate is 5° / min for right turns. The ship type code ranges from 0 to 99. In AIS data, the first digit of the ship type code indicates the ship's major category, and the second digit is a sub-category, indicating whether the ship transports dangerous goods, etc. For example, "70" indicates a cargo ship, and "80" indicates a cruise ship, etc. Figure 4 The "30" shown indicates that the vessel is a fishing boat.

[0028] Figure 4 The raw data shown includes 9 fields, which is the minimum number of fields included in the raw data. The AIS data actually includes more fields such as IMO code, port of destination, and draft. Therefore, other fields can be added to the raw data as needed.

[0029] S20. Segment the original dataset to obtain... L Each trajectory segment contains multiple raw data points. Different trajectory segments represent the same or different behavioral states of the ship. .

[0030] The original data is time point data of ship behavior generated at a specific time. One of the inventive points of this application is to process discrete time point data into time period data that better reflects the characteristics of ship behavior. Therefore, the original dataset is segmented. Figure 5 The figure shows a schematic diagram of the structure of a raw AIS dataset for a ship. The dataset contains 487 raw data points, spanning 8 hours and 5 minutes. Figure 5 The original dataset shown is divided into L The system consists of four trajectory segments. The first segment spans 2 hours and 15 minutes and includes 132 raw data points; the second segment spans 4 hours and 13 minutes and includes 248 raw data points; the third segment spans 1 hour and 2 minutes and includes 62 raw data points; and the fourth segment spans 35 minutes and includes 45 raw data points. Each segment represents a different behavioral state of the vessel: for example, the first segment represents the vessel underway, the second under trawling, the third under anchor, and the fourth underway. Because... Typically, the number of trajectory segments L Much smaller than the amount of the original data N Therefore, the amount of data obtained from the trajectory segments will be much smaller than the amount of data in the original data.

[0031] S30. Construct ship behavior feature vectors using trajectory segments as units.

[0032] The ship behavior feature vector is used to characterize the ship's behavior state over a specific time period. Continuing with the previous example, let's take the third trajectory segment as an example for illustration. Figure 5 As shown, the third trajectory segment includes 248 original data points from original data 133 to original data 380. These 248 original data points are used to construct a feature vector.

[0033] Optionally, after obtaining the original dataset of the ship in step S10, steps S11, S12, S13 and S14 are also included. The purpose of steps S11 to S14 is to automatically remove data with obvious errors from the original dataset through an algorithm. Figure 6 This is a schematic diagram of skip point removal in a sequential raw dataset. For ease of illustration, Figure 6 The image only shows the timestamp, longitude, and latitude values ​​of the raw data.

[0034] S11, the original dataset... N The original data is sorted according to timestamps to obtain an ordered original dataset, which includes... Group adjacent raw data. Normally, raw data, such as AIS data, is arranged in chronological order. However, to prevent some raw data from becoming out of chronological order, the raw data is first sorted by timestamp before further processing. The resulting sorted data groups adjacent raw data within the original dataset. N Composition of raw data Groups of adjacent original data, such as Figure 6 As shown.

[0035] S12. According to the order of the original dataset The timestamps, longitude values, and latitude values ​​of adjacent raw data sets are calculated to obtain the results. The average speed of the ships, The average speed of each ship is characterized by The average velocity of adjacent original data sets.

[0036] Although the raw data includes a field for Earth velocity, S12 does not use this field to calculate the average velocity. Instead, it uses the timestamp, longitude value, and latitude value to calculate the average velocity. For example... Figure 5 As shown, when calculating the average velocity between original data 133 and original data 134, the distance between the corresponding locations of original data 133 and original data 134 is calculated using longitude and latitude values. The time interval between original data 133 and original data 134 is obtained based on the difference in timestamps. The average velocity of adjacent original data is then obtained by dividing the distance value by the time interval. Figure 6 As shown, for For adjacent original data sets, the calculation yields Average speed of ships.

[0037] S13, will The average speed of each vessel is respectively correlated with the maximum speed corresponding to the vessel's type code. By comparing, we can obtain The results of the speed comparison were based on... The speed comparison results determine the jump point raw data in the original dataset.

[0038] The original dataset of ships includes a field for ship type codes, such as... Figure 4 As shown, the ship type code is "30", indicating that the vessel is a fishing boat. Assuming the fishing boat's maximum speed... Then Figure 6 In The average speed of each ship is respectively with By comparison, if the third average speed is much greater than 20kn, it indicates that there is an error in the original data 3 or original data 4 used to calculate the third average speed, i.e., jump point data.

[0039] Optionally, S13 specifically includes: in the... K The speed comparison result is the first K The average speed of the ships is greater than In the case of determining the first K The first original data and the first The original data is the jump point original data, among which, .

[0040] This section presents a specific method for determining the original data of a jump point. For example... Figure 6 As shown, with For example, in the third speed comparison, the average speed of the third ship is greater than... In this case, the third and fourth raw data points are determined as jump point raw data. The third speed comparison result is that the third ship's average speed is less than or equal to... In this case, the original jump point data was not found in this comparison. This is because ships may accelerate in certain situations during navigation to cope with unexpected events, thus exceeding their speed. Therefore, This threshold is used to judge the original data at the jump point to prevent false judgments that could lead to the rejection of normal original data.

[0041] S14, from the original dataset in the stated order N The original data with skip points is removed from the original data to obtain the sequential original dataset after skip point removal. The sequential original dataset after skip point removal includes... M The original data, of which .

[0042] like Figure 6 As shown, assuming the third speed comparison result is that the average speed of the third ship is greater than... Therefore, the third and fourth original data points are the jump point original data points. Removing the third and fourth original data points from the sequential original dataset yields the jump point-removed sequential original dataset. .

[0043] Because the original data comes from various sources, such as AIS data, LRIT data, VMS data, VDES data, Beidou satellite data, and even a combination of multiple data sources, the quality of the data may be flawed. Under the influence of various factors such as radio interference and equipment failure, a common problem with time-point data of ship behavior is skipping points. Considering the spatial rationality of the data, by removing skipped data, the accuracy of the data in the original dataset can be guaranteed, thereby ensuring the accuracy of the ship behavior feature vector obtained by condensing a large amount of original data.

[0044] Optionally, the original dataset can be segmented to obtain... L Each trajectory segment includes at least one of the following three segmentation conditions, thereby segmenting the original dataset into the first segment. After dividing the original data into trajectory segments, among which Condition 1: The order of the original dataset after removing jump points. M In the original data, in the first... H The timestamp and the first If the difference between the timestamps is greater than 30 minutes, then in the case of the first... After dividing the original data into segments, we obtain the original dataset. L The trajectory segment is divided into segments. Taking AIS data as the original data as an example, considering the continuity of time, the base station can usually receive one AIS data every 2 seconds to 3 minutes. When the time difference between adjacent original data is greater than 30 minutes, the Automatic Identification System (AIS) may be shut down, resulting in a data gap. Since the behavior of the ship cannot be judged during the absence of AIS data, the trajectory segment is divided here.

[0045] Condition 2: The order of the original dataset after removing jump points M In the original data, in the first... H The ground-oriented heading and the first When the difference in heading relative to the ground is greater than 60 degrees, in the first case... After dividing the original data into segments, we obtain the original dataset. L The trajectory is divided into segments. A sudden change in course is often a hallmark of a sudden change in the ship's behavior, hence the segmentation of the trajectory.

[0046] Condition 3: The order of the original dataset after removing jump points M In the original data, in the first... H The position value and the first When each of the position values ​​belongs to a different region, in the case of the _th _ After dividing the original data into segments, we obtain the original dataset. L The trajectory segment is defined as follows: The area is a pre-defined region, including at least one of a port area, an anchorage area, and a channel area. Spatial crossings are often a hallmark of a sudden change in a ship's behavior; for example, when a ship enters an anchorage area, it typically begins preparing to anchor, hence the division of the trajectory segment.

[0047] In addition to the three conditions listed here, other conditions can be selected to segment the original dataset, such as dividing the trajectory segments based on velocity changes. H The ground velocity and the first If the difference in the ground velocity is greater than the threshold, then in the th case... The original data is then segmented. Several typical trajectory segmentation methods are given here, and you can choose and add them as needed in practical applications.

[0048] Optionally, a ship behavior feature vector is constructed on a segment-by-segment basis, specifically including steps S31 and S32. S31, ... l In each trajectory segment s The original data is sorted according to timestamps. Step S20 divides the original dataset into... L The trajectory segment, with the first one being... l Taking a trajectory segment as an example, this paper illustrates the method of constructing a ship behavior feature vector. Since the correct trajectory can only be obtained by arranging the position values ​​in order, the original data in the trajectory segment are arranged in chronological order to facilitate subsequent calculations to obtain the trajectory morphology features.

[0049] S32, according to s The original data is used to obtain the ship behavior feature vector, which includes the ship's trajectory motion features, trajectory morphology features and static profile features. Figure 7 The figure shows a schematic diagram of a ship behavior feature vector. The ship behavior feature vector includes trajectory motion features 51, trajectory morphology features 52, and static profile features 53. The trajectory motion features 51 include at least one of the following: average speed, speed standard deviation, maximum speed, speed coefficient of variation, average turning rate, heading change rate, percentage of dwell time, and percentage of low-speed time. The trajectory morphology features 52 include at least one of the following: straightness, turning index, convex hull area, trajectory compactness, number of turns, and bow-stern closure. The static profile features 53 include at least one of the following: ship tonnage, main engine power, and ship length.

[0050] Figure 9 The figure shows a schematic diagram of a trajectory motion feature. The trajectory motion feature 51 includes average speed, speed standard deviation, maximum speed, speed coefficient of variation, average turning rate, heading change rate, dwell time percentage, and low speed time percentage. The trajectory motion feature is an 8-dimensional vector. Figure 9 The vector below gives Figure 5 The structure diagram shown is a composite image of the trajectory motion features of the second trajectory segment in the original dataset, filled with specific numerical values. Figure 9As shown, during the second trajectory segment (4 hours and 13 minutes), the ship's average speed was 3.42 knots, the standard deviation of speed was 0.61 knots, the maximum speed was 4.85 knots, the coefficient of variation of speed was 0.178, the average turning rate was 8.4° / min, the rate of change of heading was 12.7° / min, the percentage of time spent in stillness was 0.04, and the percentage of time spent at low speed was 0.92. The data recorded in the trajectory motion characteristics fully characterize the common ship motion characteristics of the second trajectory segment, as shown in Table 1, which illustrates the calculation method for the trajectory motion characteristics.

[0051] Table 1. Calculation Methods for Trajectory Motion Characteristics

[0052] Figure 10 As shown in the figure, the trajectory morphology feature 52 includes straightness, rotation index, convex hull area, trajectory compactness, number of turns, and beginning and end closure. The trajectory morphology feature is a 6-dimensional vector. Figure 10 The vector below provides a structural diagram showing the specific numerical values ​​of the trajectory morphology features filled in for the second trajectory segment. As shown in the figure, the ship's straightness during the second trajectory segment (4 hours and 13 minutes) is 0.31, its turning index is 4.6, and its convex hull area is 12.4 km². 2 The trajectory compactness is 0.087, the number of turns is 5, and the initial-to-final closure is 0.42. The data recorded in the trajectory morphology features fully characterize the trajectory morphology of the second trajectory segment, as shown in Table 2, which illustrates the calculation method for the trajectory morphology features.

[0053] Table 2. Calculation Method of Trajectory Morphology Features

[0054] Figure 11 The figure shows a schematic diagram of a static image feature. The static image feature 53 includes the ship's tonnage, main engine power, and ship length. The static image feature is a 3-dimensional vector. Figure 11 The vector below provides a structural diagram of the specific values ​​for filling in the static profile features for the second trajectory segment. As shown in the figure, the ship's tonnage during the 4h13min period of the second trajectory segment is 95 tons, the main engine power is 360kW, and the ship's length is 32 meters. The data recorded in the static profile features fully characterize the static profile of the second trajectory segment. Table 3 shows the calculation method for the static profile features, which uses MMSI as an index and obtains the ship's tonnage, main engine power, and ship length through a public information platform.

[0055] Table 3. Calculation Method of Static Portrait Features

[0056] This embodiment designs a method for representing ship behavior features, constructing ship behavior feature vectors based on trajectory segments. It solves the technical problems of data users needing to repeatedly extract different data from the original data for different needs, the large amount of repetitive work in threshold comparison or large model training, the extremely large amount of original data, the large amount of redundant data in transmission and sharing, and the difficulty in data sharing. It achieves the technical effect of compressing a large amount of time point data into ship behavior feature vectors representing time periods.

[0057] Example 2 Figure 2 This is a flowchart of a method for representing ship behavior features provided in Embodiment 2 of this application. Compared to the method described in Embodiment 1, after obtaining the original dataset of the ship in S10, it further includes: S40. Obtain the vessel's objective dataset, which includes the vessel's historical average trajectory segment duration, port location, anchorage location, fishing ban area location, and waterway location. The objective dataset needs to be indexed by MMSI and obtained through a public information platform.

[0058] Accordingly, the ship behavior feature vector also includes the ship's historical overall operational features 54 and spatial correlation features 55. Figure 8 This is a schematic diagram of another type of ship behavior feature vector, relative to... Figure 7 The ship behavior feature vector shown also includes historical operation overall features 54, spatial correlation features 55, hot unique coding features 56, and historical operation detailed features 57.

[0059] Historical assignment general characteristics 54 include average trajectory segment duration, such as Figure 5 As shown, the original dataset is divided into L = There are 4 trajectory segments. The time span of the first trajectory segment is 2 hours and 15 minutes, the time span of the second trajectory segment is 4 hours and 13 minutes, the time span of the third trajectory segment is 1 hour and 2 minutes, and the time span of the fourth trajectory segment is 35 minutes. Therefore, the average trajectory segment duration of the ship is 2 hours and 1 minute.

[0060] The spatial association feature 55 includes at least one of the following: distance to the nearest port, distance to the nearest anchorage, distance to the nearest fishing ban area, time spent in the fishing ban area, and distance to the nearest channel centerline. Figure 12 The figure shows a schematic diagram of a spatial association feature. The spatial association feature 55 includes the distance to the nearest port, the distance to the nearest anchorage, the distance to the nearest fishing ban area, the time spent in the fishing ban area, and the distance to the nearest channel centerline. The spatial association feature is a 5-dimensional vector. Figure 12The vector below provides a structural diagram showing the spatial correlation features of the second trajectory segment, filling in the specific numerical values. As shown in the figure, during the 4 hours and 13 minutes of the second trajectory segment, the ship's distance to the nearest port was 84.3 km, the distance to the nearest anchorage was 71.2 km, and the distance to the nearest no-fishing zone was -2.4 km. The negative values ​​indicate that the ship has entered the no-fishing zone, spent 253 minutes within it, and is 18.6 m from the nearest channel centerline. This records the time spent within the no-fishing zone; in practical applications, the time spent in other areas, such as a specific port, can be selected as needed.

[0061] Optionally, the distance calculation in the spatial association feature 55 adopts the method of the first... l In each trajectory segment s The centroid of the curve formed by the ship position values ​​in the original data is used as a reference point. At least one of the following distances is calculated: the distance to the nearest port, the distance to the nearest anchorage, the distance to the nearest fishing ban area, and the distance to the nearest channel centerline. Table 4 shows the calculation method for spatial correlation features, taking the second trajectory segment as an example. Figure 5 The second trajectory segment shown has a time span of 4 hours and 13 minutes and includes 248 raw data points. This means the trajectory segment contains 248 location values. Therefore, the choice of which location value to use as the reference point for calculating the distance to each region is debatable. One of the 248 location values ​​could be chosen, such as the start, midpoint, or end point of the trajectory segment. This embodiment does not use a single location value from the 248 raw data points as the reference point, but instead uses the centroid of the trajectory segment. The centroid better characterizes the position of the trajectory segment, and the algorithm for calculating the centroid of a trajectory segment is relatively mature, allowing for convenient acquisition of the centroid coordinates from the trajectory points.

[0062] Table 4. Calculation Methods for Spatial Association Features

[0063] Optionally, the objective dataset also includes detailed information on the ship's historical trajectory segments; correspondingly, the ship behavior feature vector also includes one-hot encoded features 56 and detailed historical operation features 57. The one-hot encoded features 56 include at least one of ship type features, ship registration features, historical main operation type features, historical main operation area features, and area type features. Figure 13 The figure shows a schematic diagram of a one-hot encoded feature. The one-hot encoded feature 56 includes ship type features, ship registration features, historical main operation type features, historical main operation area features and area type features. The one-hot encoded feature is a 75-dimensional vector. Figure 13The vector below provides a structural diagram of the one-hot encoded feature filling values ​​for the second trajectory segment. As shown in the figure, the one-hot encoding of the ship type feature during the 4h13min period of the second trajectory segment is 10000000, with 8 bits representing 8 ship types. The first bit being "1" indicates that the ship type is a fishing vessel, corresponding to... Figure 4 The ship type code is "30"; the unique hot code for the ship registration feature is 00000100000000000000000000000000, with 32 bits representing 32 ship registrations, and the 6th bit being "1" indicating that the ship is registered in Zhoushan, Zhejiang; the unique hot code for the historical main operation type feature is 100000, with 6 bits representing 6 operation types, and the 1st bit being "1" indicating that the operation type is trawling; the unique hot code for the historical main operation area feature is 000000100000000000000000, with 24 bits representing 24 different operation areas; the unique hot code for the area type feature is 00001, with 5 bits representing 5 different areas, and the 5th bit being "1" indicating that the AIS data corresponding to this ship trajectory segment indicates that the ship is located in the "high seas" for the largest proportion of the time.

[0064] The detailed features of the historical operations 57 include the percentage of historical operation types and the percentage of months for trajectory segments. Figure 14 The figure shows a schematic diagram of the structure of a detailed feature of a historical task. The detailed feature of the historical task 57 includes the proportion of historical task types and the proportion of trajectory segments by month. The detailed feature of the historical task is a 15-dimensional vector. Figure 14 The vector below provides a structural diagram showing the detailed features of the historical operations in the second trajectory segment, filled with specific numerical values. As shown in the figure, the proportion of historical operation types for the vessel during the 4h13min period of the second trajectory segment was 62% for trawl, 8% for purse seine, and 30% for other operations. The proportion of the month for the trajectory segment from January to December was 18%, 12%, 5%, 2%, 1%, 1%, 1%, 5%, 10%, 15%, 18%, and 12%, respectively, thus capturing the seasonal pattern of the vessel's operations.

[0065] The technical solution of this embodiment adds historical operation general features, spatial correlation features, one-hot encoded features and historical operation detailed features to the ship behavior feature vector, and uses the centroid of the trajectory segment as the reference point when calculating the spatial correlation features, so that the content of the ship behavior feature vector is richer and the distance value calculation is more accurate. By combining the original dataset and the objective dataset, all information can be included in one feature vector.

[0066] Example 3 This embodiment discloses a ship behavior feature vector, applied to the ship behavior feature representation method described in Embodiment 1 or Embodiment 2, including: The trajectory motion feature 51 includes average speed, speed standard deviation, maximum speed, speed coefficient of variation, average turning rate, heading change rate, dwell time percentage, and low speed time percentage. The trajectory motion feature is an 8-dimensional vector. The trajectory morphology feature 52 includes straightness, rotation index, convex hull area, trajectory compactness, number of turns and the beginning and end closure degree. The trajectory morphology feature is a 6-dimensional vector. Static image feature 53, ship tonnage, main engine power and ship length, the static image feature region is a 3D vector; The overall characteristics of historical tasks 54 include the historical average trajectory segment duration, and the overall characteristics of historical tasks are a 1-dimensional vector. Spatial correlation feature 55 includes the distance to the nearest port area, the distance to the nearest anchorage, the distance to the nearest fishing ban area, the time spent in the fishing ban area, and the distance to the nearest channel centerline. The spatial correlation feature is a 5-dimensional vector. Among them, trajectory motion feature 51, trajectory morphology feature 52, static image feature 53, historical operation overall feature 54, and spatial correlation feature 55 are all continuous.

[0067] The region containing any feature is continuous, such as Figure 9 and Figure 10 As shown, the 8-dimensional vector in trajectory motion feature 51 is continuous, and the 6-dimensional vector in trajectory morphology feature 52 is continuous, rather than having a single dimension vector from trajectory morphology feature 52 interspersed among the 8-dimensional vectors of trajectory motion feature 51. The advantage of this setting is that when using ship behavior feature vectors later, it is convenient to extract certain features as needed without causing confusion.

[0068] Optionally, the ship behavior feature vector further includes: Hot unique coding feature 56 includes ship type feature, ship registration feature, historical main operation type feature, historical main operation area feature and area type feature, and the hot unique coding feature is a 60-dimensional vector; Historical task detailed features 57 include the proportion of historical task types and the proportion of trajectory segments by month. The historical task detailed features are 15-dimensional vectors. Among them, the hot unique coding feature 56 and the historical job detailed feature 57 are both continuous.

[0069] This embodiment designs a ship behavior feature vector, which solves the technical problems of data users needing to repeatedly extract different data from the original data for different needs, the large amount of repetitive work in threshold comparison or large model training, the huge amount of original data, the large amount of redundant data in transmission and sharing, and the difficulty of data sharing. It achieves the technical effect of compressing a large amount of time point data into a ship behavior feature vector that represents a time period.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 this application.

Claims

1. A method of representing a behavior characteristic of a ship, characterized by, include: an original data set of a ship, the original data set comprising N a strip of original data, each strip of original data comprising a time stamp, a maritime mobile service identity, MMSI, a longitude value, a latitude value, a speed over ground, a heading over ground, a ship heading, a turning rate and a ship type code, wherein the longitude value and the latitude value together constitute a position value of the ship; Segmenting the original data set obtains L a plurality of track segments, each track segment including a plurality of original data, and different track segments representing same or different behavior states of the ship. Construct ship behavior feature vectors using trajectory segments as units; The original data is time point data of ship behavior generated at a specific time, the ship behavior feature vector is used to represent the ship behavior state in a specific time period, and the ship behavior state prediction model is used to predict the ship behavior state in a specific time period. And .

2. The method for representing ship behavior characteristics according to claim 1, characterized in that, After obtaining the raw dataset of the ship, the process also includes: The original dataset N The original data is sorted according to timestamps to obtain an ordered original dataset, which includes... Group of adjacent original data; According to the original dataset in order The timestamps, longitude values, and latitude values ​​of adjacent raw data sets are calculated to obtain The average speed of the ships, The average speed of each ship is characterized by The average velocity of adjacent sets of raw data; Will The average speed of each vessel is respectively correlated with the maximum speed corresponding to the vessel's type code. By comparing, we can obtain The results of the speed comparison were based on... The speed comparison results determine the jump point raw data in the original dataset; From the original dataset in the stated order N The original data with skip points is removed from the original data to obtain the sequential original dataset after skip point removal. The sequential original dataset after skip point removal includes... M The original data, of which .

3. The method for representing ship behavior characteristics according to claim 2, characterized in that, Will The average speed of each vessel is respectively correlated with the maximum speed corresponding to the vessel's type code. By comparing, we can obtain The results of the speed comparison were based on... The speed comparison results determine the original jump point data in the original dataset, including: In the K The speed comparison result is the first K The average speed of the ships is greater than In the case of determining the first K The first original data and the first The original data is the jump point original data, among which, .

4. The method for representing ship behavior characteristics according to claim 3, characterized in that, The original dataset is segmented to obtain L Each trajectory segment includes at least one of the following: The order of the original dataset after removing jump points M In the original data, in the first... H The timestamp and the first If the difference between the timestamps is greater than 30 minutes, then in the case of the first... After dividing the original data into segments, we obtain the original dataset. L One trajectory segment; The order of the original dataset after removing jump points M In the original data, in the first... H The ground-oriented heading and the first When the difference in heading relative to the ground is greater than 60 degrees, in the first case... After dividing the original data into segments, we obtain the original dataset. L One trajectory segment; The order of the original dataset after removing jump points M In the original data, in the first... H The position value and the first When each of the position values ​​belongs to a different region, in the case of the _th _ After dividing the original data into segments, we obtain the original dataset. L One trajectory segment; in, The area is a pre-defined area range, including at least one of a port area, an anchorage area, and a waterway area.

5. The method for representing ship behavior characteristics according to claim 4, characterized in that, Construct ship behavior feature vectors on a segment-by-segment basis, including: The first l In each trajectory segment s The original data is sorted by timestamp; according to s The original data is used to obtain the ship behavior feature vector, which includes the ship's trajectory motion features (51), trajectory morphology features (52), and static image features (53). The trajectory motion features (51) include at least one of the following: average speed, speed standard deviation, maximum speed, speed coefficient of variation, average turning rate, heading change rate, dwell time percentage, and low speed time percentage. The trajectory morphology features (52) include at least one of the following: straightness, rotation index, convex hull area, trajectory compactness, number of turns, and beginning and end closure. The static image features (53) include at least one of the following: ship tonnage, main engine power and ship length.

6. The method for representing ship behavior characteristics according to claim 5, characterized in that, After obtaining the raw dataset of the ship, the process also includes: Obtain an objective dataset of the vessel, which includes the vessel's historical average trajectory segment duration, port location, anchorage location, fishing ban area location, and waterway location; Accordingly, the ship behavior feature vector also includes the ship's historical overall operational features (54) and spatial correlation features (55). Among them, the overall characteristics of historical operations (54) include the average trajectory segment duration; The spatial association feature (55) includes at least one of the following: distance to the nearest port, distance to the nearest anchorage, distance to the nearest fishing ban area, time spent in the fishing ban area, and distance to the nearest channel centerline.

7. The method for representing ship behavior characteristics according to claim 6, characterized in that, The distance calculation in the spatial association feature (55) adopts the method of the first l In each trajectory segment s Using the centroid of the curve formed by the ship position values ​​in the original data as a reference point, calculate at least one of the following distances from the reference point: the nearest port, the nearest anchorage, the nearest fishing ban area, and the nearest channel centerline.

8. The method for representing ship behavior characteristics according to claim 7, characterized in that, The objective dataset also includes detailed information on the ship's historical trajectory segments; Accordingly, the ship behavior feature vector also includes one-hot encoded features (56) and detailed historical operation features (57). Among them, the unique hot coding feature (56) includes at least one of ship type feature, ship registration feature, historical main operation type feature, historical main operation area feature and area type feature; The detailed characteristics of the historical operations (57) include the percentage of historical operation types and the percentage of months for trajectory segments.

9. A ship behavior feature vector, applied to the ship behavior feature representation method according to any one of claims 5-7, characterized in that, include: The trajectory motion features (51) include average speed, speed standard deviation, maximum speed, speed coefficient of variation, average turning rate, heading change rate, dwell time percentage and low speed time percentage, and the trajectory motion features are 8-dimensional vectors; Trajectory morphology features (52) include straightness, rotation index, convex hull area, trajectory compactness, number of turns and end closure, and the trajectory morphology features are 6-dimensional vectors; Static profile features include ship tonnage, main engine power, and ship length; these static profile features are 3-dimensional vectors. The overall characteristics of historical operations (54) include the historical average trajectory segment duration, and the overall characteristics of historical operations are a 1-dimensional vector; Spatial association features (55) include the distance to the nearest port area, the distance to the nearest anchorage, the distance to the nearest fishing ban area, the time spent in the fishing ban area, and the distance to the nearest channel centerline. The spatial association feature area is a 5-dimensional vector. Among them, the trajectory motion features (51), trajectory morphology features (52), static image features (53), historical operation overall features (54), and spatial association features (55) are all continuous.

10. The ship behavior feature vector according to claim 9, characterized in that, Also includes: The hot-unique coding feature (56) includes ship type feature, ship registration feature, historical main operation type feature, historical main operation area feature and area type feature, and the hot-unique coding feature is a 60-dimensional vector; Detailed features of historical tasks (57) include the proportion of historical task types and the proportion of trajectory segments by month. The detailed features of historical tasks are 15-dimensional vectors. Among them, the hot unique coding feature (56) and the historical operation detailed feature (57) are both continuous.

Images