Ship abnormal behavior monitoring method and device, storage medium and computer equipment

Abstract

According to the ship abnormal behavior monitoring method and device, the storage medium and the computer equipment provided by the invention, when abnormal behavior monitoring is carried out on the target ship, the ship position data and the navigation radar data of the target ship at the current moment and in the first preset historical period before the current moment can be firstly acquired; then, the ship position data and the marine radar data can be subjected to weighted fusion by using a weighted track fusion algorithm of local information entropy to obtain fused track data of the target ship at each moment, and then the fused track data at each moment is input into a preset target ship track prediction model. Therefore, accurate predicted track data output by the target ship track prediction model can be obtained, and whether the target ship has abnormal behaviors can be accurately determined according to the predicted track data.

Description

Technical Field

[0001] The present application relates to the technical field of ship track prediction, and in particular to a method, device, storage medium and computer equipment for monitoring abnormal ship behavior. Background Art

[0002] Previous work on abnormal behavior analysis of ships has mainly focused on single-source data, such as AIS data, radar, vision, remote sensing satellite data, etc. A single data source cannot cover all the information that needs to be monitored and analyzed, resulting in incomplete data and affecting the accuracy of abnormal behavior detection. In addition, when using single-source data, the correlation between data from different sources is not analyzed, resulting in poor data quality. In addition, a single data source cannot cover all the information that needs to be monitored and analyzed, resulting in incomplete data, which affects the accuracy of abnormal behavior detection.

[0003] Based on this, today’s ship abnormal behavior analysis generally uses multi-source data for abnormal analysis, such as AIS data and radar data. However, due to the lack of unified standards for multi-source data, different data sources may use different data formats, parameters, and metrics, resulting in incompatibility between data, making it difficult to conduct effective comparison and analysis. In addition, affected by factors such as marine weather and ocean currents, the quality of AIS data and radar data may be affected to a certain extent, thereby affecting the accuracy of ship abnormal behavior analysis. In addition, maritime navigation involves interaction and coordination between multiple ships, and there are complex relationships and behavior patterns between ships. Algorithms require more complex models to make predictions and judgments, which makes it difficult for existing ship abnormal behavior recognition algorithms to cope with complex situations. Summary of the invention

[0004] The purpose of this application is to solve at least one of the above-mentioned technical defects, especially the technical defects that the ship abnormal behavior recognition algorithm in the prior art has low accuracy and is difficult to cope with complex situations.

[0005] The present application provides a method for monitoring abnormal behavior of a ship, the method comprising:

[0006] Acquire ship position data and navigation radar data of the target ship at the current moment and within a first preset historical period before the current moment;

[0007] Using a weighted track fusion algorithm of local information entropy to perform weighted fusion on the ship position data and the navigation radar data to obtain fused track data of the target ship at each moment;

[0008] Inputting the fused track data at each moment into a preset target ship track prediction model to obtain predicted track data output by the target ship track prediction model;

[0009] Determine whether the target ship has abnormal behavior according to the predicted track data.

[0010] Optionally, the weighted track fusion algorithm using local information entropy is used to perform weighted fusion on the ship position data and the navigation radar data to obtain fused track data of the target ship at each time, including:

[0011] The ship position data and the navigation radar data are spatially and temporally unified by a preset data time-space unification mathematical model;

[0012] Determine the associated track point matrix of the target ship at each time and the weight matrix corresponding to all track points in each associated track point matrix according to the unified ship position data and navigation radar data;

[0013] According to the weight matrices corresponding to all the track points in each associated track point matrix, the unified ship position data and the navigation radar data are weighted fused to obtain the fused track data of the target ship at each moment.

[0014] Optionally, the data time-space unified mathematical model includes a space unified model and a time unified model;

[0015] The spatial and temporal unification of the ship position data and the navigation radar data by using a preset data time-space unification mathematical model includes:

[0016] spatially unifying the ship position data and the navigation radar data through the spatial unification model;

[0017] The spatially unified ship position data and navigation radar data are temporally unified through the temporal unification model.

[0018] Optionally, determining the associated track point matrix of the target ship at each time according to the unified ship position data and navigation radar data includes:

[0019] Determine whether the track point corresponding to the unified ship position data and navigation radar data at each moment is within the wave gate of the target ship, and obtain the determination result of each track point at each moment;

[0020] According to the determination results of each track point at each time, the associated track point matrix of the target ship at each time is determined.

[0021] Optionally, determining the weight matrix corresponding to all track points in each associated track point matrix according to the unified ship position data and navigation radar data includes:

[0022] Determine the distance difference between each track point in each associated track point matrix according to the unified ship position data and navigation radar data, and construct the distance difference matrix of the target ship at each time according to the distance difference between each track point;

[0023] According to the distance difference matrix of the target ship at each time, the weight matrix corresponding to all the track points of the target ship at each time is calculated.

[0024] Optionally, the calculating, based on the distance difference matrix of the target ship at each time, a weight matrix corresponding to all track points of the target ship at each time includes:

[0025] According to the distance difference matrix of the target ship at each time, the distance information entropy corresponding to each track point of the target ship at each time and the system maximum entropy of the target ship at each time are calculated;

[0026] Calculate the residual degree function corresponding to each track point of the target ship at each time according to the distance information entropy corresponding to each track point of the target ship at each time and the system maximum entropy of the target ship at each time;

[0027] According to the surplus degree function corresponding to each track point at each time, the weight matrix corresponding to all track points of the target ship at each time is determined.

[0028] Optionally, the training process of the target ship track prediction model includes:

[0029] Obtaining sample ship position data and sample navigation radar data of each ship within a second preset historical period;

[0030] The weighted track fusion algorithm of local information entropy is used to perform weighted fusion on the sample ship position data and sample navigation radar data of each ship to obtain the sample fusion track data of each ship at each time.

[0031] The sample fusion track data of each ship at each moment is input into the preset initial ship track prediction model, and the Adam algorithm is used to update the network parameters in the initial ship track prediction model until the preset training conditions are met, thereby obtaining the target ship track prediction model.

[0032] The present application also provides a ship abnormal behavior monitoring device, comprising:

[0033] A data acquisition module, used to acquire ship position data and navigation radar data of a target ship at a current moment and within a first preset historical period before the current moment;

[0034] A weighted fusion module, used for weighted fusion of the ship position data and the navigation radar data using a weighted track fusion algorithm of local information entropy to obtain fused track data of the target ship at each moment;

[0035] A track prediction module is used to input the fused track data at each moment into a preset target ship track prediction model to obtain the predicted track data output by the target ship track prediction model;

[0036] The abnormality monitoring module is used to determine whether the target ship has abnormal behavior according to the predicted track data.

[0037] The present application also provides a computer-readable storage medium, in which computer-readable instructions are stored. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the ship abnormal behavior monitoring method as described in any of the above embodiments.

[0038] The present application also provides a computer device, comprising: one or more processors, and a memory;

[0039] The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the ship abnormal behavior monitoring method as described in any one of the above embodiments are performed.

[0040] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0041] The ship abnormal behavior monitoring method, device, storage medium and computer equipment provided by the present application can first obtain the ship position data and navigation radar data of the target ship at the current moment and within the first preset historical period before the current moment when monitoring the abnormal behavior of the target ship; then, the present application can use the weighted track fusion algorithm of local information entropy to weightedly fuse the ship position data and the navigation radar data to obtain the fused track data of the target ship at each moment. The fused track data is obtained by unifying the ship position data and the navigation radar data in time and space, removing irrelevant track points through the weighted track fusion algorithm of local information entropy, and weighted fusion according to the weights of each relevant track point. Therefore, the fused track data at each moment is input into the preset target ship track prediction model to obtain the relatively accurate predicted track data output by the target ship track prediction model, and the predicted track data can be accurately determined whether the target ship has abnormal behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1 A schematic diagram of a process for monitoring abnormal behavior of a ship provided in an embodiment of the present application;

[0044] Figure 2 A data display diagram of the encapsulated AIS data provided in the embodiment of the present application;

[0045] Figure 3 A schematic diagram of a flow chart of a weighted track fusion algorithm based on local information entropy provided in an embodiment of the present application;

[0046] Figure 4 A schematic diagram of the training process of the initial ship track prediction model provided in an embodiment of the present application;

[0047] Figure 5 A schematic diagram of the structure of a ship abnormal behavior monitoring device provided in an embodiment of the present application;

[0048] Figure 6 A schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0050] In one embodiment, Figure 1 As shown, Figure 1 A schematic diagram of a process flow of a method for monitoring abnormal behavior of a ship provided in an embodiment of the present application; the present application provides a method for monitoring abnormal behavior of a ship, the method may include:

[0051] S110: Obtaining ship position data and navigation radar data of the target ship at the current moment and within a first preset historical period before the current moment.

[0052] In this step, when monitoring whether the target ship has abnormal behavior, the predicted track data of the target ship may be determined first, and then it may be determined whether the target ship has abnormal behavior based on the predicted track data.

[0053] Specifically, when determining the predicted track data of the target ship, the present application can first obtain the ship position data and navigation radar data of the target ship at the current moment and within the first preset historical period before the current moment. In this way, the predicted track data of the target ship can be jointly predicted based on the ship position data and the navigation radar data, thereby effectively improving the accuracy of the prediction results.

[0054] Among them, the specific duration of the first preset historical period here can be set according to the actual situation. For example, this application can determine the specific duration of the first preset historical period according to the accuracy of the prediction result and the time consumed by the calculation process, etc., and there is no limitation here. After the first preset historical period is determined by this application, the ship position data and navigation radar data of the target ship in the first preset historical period before the current moment can be obtained. The ship position data here can be AIS data, and the navigation radar data here can be data sent by the navigation radar.

[0055] Furthermore, after the present application obtains the AIS data, the AIS data may also be decoded to obtain specific propagation information. Figure 2 As shown, Figure 2 A data display diagram of the encapsulated AIS data provided in the embodiment of the present application; Figure 2 After the encapsulated AIS data is decoded, the decoding of each parameter is shown in Table 1:

[0056] Serial number meaning Serial number meaning 0 <![CDATA[ mmsi ]]> 7 heading 1 rot, turning rate, needs to be divided by 10# 8 speed, speed, needs to be divided by 10 2 class_Type,AIS data type# 9 status, navigation status 3 <![CDATA[ timeStamp , Vessel position report time]]> 10 <![CDATA[ shipType ,Ship Type]]> 4 <![CDATA[ lon , longitude, need to be divided by 600000]]> 11 breadth 5 lat, latitude, needs to be divided by 600000

[0057] Table 1 Decoding rules for each parameter

[0058] By using the decoding rules shown in Table 1 above, the AIS data obtained by the present application can be decoded and the ship position data can be obtained. In addition, for the marine radar data of the present application, since the acquisition process of the marine radar data is relatively convenient and the amount of marine radar data is higher than that of the AIS data, when performing subsequent calculations, linear interpolation and other operations can also be considered for the AIS data to keep the two consistent.

[0059] S120: Using a weighted track fusion algorithm of local information entropy to perform weighted fusion on the ship position data and the navigation radar data to obtain fused track data of the target ship at each moment.

[0060] In this step, after obtaining the ship position data and navigation radar data of the target ship at the current moment and in the first preset historical period before the current moment through S110, the ship position data and the navigation radar data can be weightedly fused using the weighted track fusion algorithm of local information entropy to obtain the fused track data of the target ship at each moment.

[0061] It can be understood that when monitoring abnormal behaviors such as shutting down the AIS equipment of the target ship or tampering with the AIS report information, the present application can integrate the navigation radar data and the ship position data to realize the active analysis of the behavior of mobile targets at sea, improve the effective utilization rate of multi-source spatiotemporal data at sea, and improve the accuracy of identifying abnormal behavior of ships.

[0062] Furthermore, through analyzing the characteristics of the two data types of marine radar data and ship position data, the present application finds that the marine radar data is highly accurate, the ship position data has obvious lag, and the measurement errors of the marine radar data and the ship position data are relatively independent, which leads to the shortcomings of the existing track fusion algorithm, such as large computational scale, high complexity, failure to consider track fusion in the case of multiple ships, and failure to consider the interference of the errors of the sensing equipment itself.

[0063] To address the above problems, the present application constructs a weighted trajectory fusion algorithm based on local information entropy (WTF-LIE). The algorithm analyzes the navigation trajectory of the same target ship under different sensing devices based on the data matrix corresponding to the marine radar data and the ship position data, and determines the weight value of each sensing device when the track is fused according to the navigation trajectory, and then performs weighted fusion of the ship position data and the marine radar data according to the weight value, thereby obtaining the fused track data of the target ship at each moment.

[0064] S130: Inputting the fused track data at each moment into a preset target ship track prediction model to obtain predicted track data output by the target ship track prediction model.

[0065] In this step, the ship position data and the navigation radar data are weightedly fused using the weighted track fusion algorithm of local information entropy through S120 to obtain the fused track data of the target ship at each moment. The present application can also input the fused track data at each moment into a preset target ship track prediction model to obtain the predicted track data output by the target ship track prediction model.

[0066] Specifically, the present application adopts the weighted track fusion algorithm of local information entropy to fuse AIS data with navigation radar data to obtain fused track data including but not limited to the ship's Maritime Mobile Service Identity (MMSI), which is used as part of the ship's basic data and as a unique identifier to distinguish the navigation behavior, time (UpdateDateFormat), longitude (LON), latitude (LAT), COG course overground, SOG speed over ground and other information of different ships. The specific fused track data example is shown in Table 2:

[0067] MNSI <![CDATA[ UpdateDateFormat ]]> LON LAT COG SOG b339653b 2021 / 10 / 167：41 118.168625 24.43651333 264 1.7 b339653b 2021 / 10 / 167：43 118.1669643 24.43635715 258 1.7 b339653b 2021 / 10 / 167：45 118.1653462 24.43603839 258 1.8 b339653b 2021 / 10 / 167：47 118.1636711 24.43568635 259 1.8 b339653b 2021 / 10 / 16 7:49 118.1619546 24.4353584 261 1.9 b339653b 2021 / 10 / 16 7:51 118.1601527 24.43510413 265 2 b339653b 2021 / 10 / 167：53 118.1582649 24.4349719 267 1.7

[0068] Table 2 Partial fusion track data information table

[0069] As can be seen from Table 2, for each ship, the present application can determine the longitude (LON), latitude (LAT), COG course over ground, SOG speed over ground and other data according to its fused track data, and perform ship track prediction based on the data from various aspects to obtain more accurate predicted track data.

[0070] Furthermore, the target ship track prediction model of the present application is obtained by training with sample fused track data after fusion of sample ship position data and sample navigation radar data as training samples and real track data as sample labels. Therefore, when the present application inputs the fused track data obtained after fusion into the target ship track prediction model, the target ship track prediction model can be used to predict the predicted track data of the target ship.

[0071] S140: Determine whether the target ship has abnormal behavior based on the predicted track data.

[0072] In this step, the fused track data at each moment is input into the preset target ship track prediction model through S130. After obtaining the predicted track data output by the target ship track prediction model, the present application can also determine whether the target ship has abnormal behavior based on the predicted track data.

[0073] Specifically, after the present application uses the target ship track prediction model to predict the future track of the target ship and obtains the predicted track data, it can also continue to collect the actual track data of the target ship at the prediction time. Then, the present application can use a preset abnormal behavior identification method to compare the predicted track data and the actual track data, and determine whether the target ship has abnormal behavior based on the comparison results.

[0074] For example, the present application can determine whether the target ship has abnormal behavior by calculating the Euclidean distance between the predicted track data and the actual track data of each target ship. If the Euclidean distance exceeds the preset distance threshold, it is determined that the target ship has abnormal behavior. If the Euclidean distance does not exceed the preset distance threshold, it is determined that the target ship has not exhibited abnormal behavior, and the track of the target ship at the next moment is continued to be predicted and abnormality identification is performed.

[0075] In the above embodiment, when monitoring the abnormal behavior of the target ship, the ship position data and navigation radar data of the target ship at the current moment and within the first preset historical period before the current moment can be obtained first; then, the present application can use the weighted track fusion algorithm of local information entropy to weightedly fuse the ship position data and the navigation radar data to obtain the fused track data of the target ship at each moment. The fused track data is obtained by unifying the ship position data and the navigation radar data in time and space, removing irrelevant track points through the weighted track fusion algorithm of local information entropy, and weighted fusion according to the weights of each relevant track point. Therefore, the fused track data at each moment is input into the preset target ship track prediction model to obtain the more accurate predicted track data output by the target ship track prediction model, and the predicted track data can be accurately determined whether the target ship has abnormal behavior.

[0076] In one embodiment, in S120, weighted fusion of the ship position data and the navigation radar data using a weighted track fusion algorithm of local information entropy to obtain fused track data of the target ship at each time may include:

[0077] S121: Unifying the ship position data and the navigation radar data in space and time through a preset data time-space unification mathematical model.

[0078] S122: Determine the associated track point matrix of the target ship at each time and the weight matrix corresponding to all track points in each associated track point matrix according to the unified ship position data and navigation radar data.

[0079] S123: performing weighted fusion on the unified ship position data and navigation radar data according to the weight matrices corresponding to all track points in each associated track point matrix, so as to obtain fused track data of the target ship at each time.

[0080] In this embodiment, when weighted fusion is performed on the ship position data and the navigation radar data, since the ship position data and the navigation radar data of this application are obtained by collecting relevant data of the same target ship from multiple directions using multiple sensors, the types of measurement data obtained are also relatively large, and taking into account the non-ignorability between the relevant sensors, the density of the ship's trajectory under the condition of multi-target tracking, and the fact that different data sources may use different data formats, parameters and measurement indicators, etc., this application can use the weighted track fusion algorithm of local information entropy to first unify the ship position data and the navigation radar data in time and space, and then after obtaining the track points related to the target ship, determine the weight matrix corresponding to all the track points of the target ship at each time, and then perform weighted fusion on the unified ship position data and the navigation radar data according to the weight matrix.

[0081] Specifically, the present application can use a preset data time-space unified mathematical model to unify the space and time of the ship position data and the navigation radar data. Spatial unification can use the coordinate system conversion technology to convert the coordinate system reference of the ship position data and the navigation radar data to achieve spatial unification. Temporal unification can use the timestamp method to convert the sampling time in the ship position data and the navigation radar data into the time difference of the same time reference point to achieve temporal unification. In this way, the data can be easily processed and analyzed, and the utilization value of the data can be improved.

[0082] Next, the present application can determine the associated track point matrix of the target ship at each moment based on the unified ship position data and marine radar data, as well as the weight matrix corresponding to all track points in each associated track point matrix, and then perform weighted fusion on the unified ship position data and marine radar data based on the weight matrix corresponding to all track points in each associated track point matrix, thereby obtaining the fused track data of the target ship at each moment.

[0083] It can be understood that the unified ship position data and navigation radar data of the present application include multiple types of data. When the present application determines each track point of the target ship at each moment, the various types of ship position data and navigation radar data of all track points at each moment can be weighted and fused according to the weight of each track point, thereby obtaining the fused track data of the target ship of different types at each moment, such as the latitude, longitude, speed and heading of the target ship at the current moment and at each moment in the first preset historical period. In this way, the predicted track data of the target ship can be determined based on the fused track data of different types.

[0084] In one embodiment, the data spatiotemporal unified mathematical model may include a spatial unified model and a temporal unified model.

[0085] S121 unifies the ship position data and the navigation radar data in space and time by using a preset data time-space unification mathematical model, which may include:

[0086] S1211: spatially unifying the ship position data and the navigation radar data through the spatial unification model.

[0087] S1212: Temporally unifying the spatially unified ship position data and the navigation radar data through the temporal unification model.

[0088] In this embodiment, when weighted fusion is performed on the ship position data and the navigation radar data, a preset data spatiotemporal unification mathematical model may be used to unify the ship position data and the navigation radar data in space and time.

[0089] Specifically, the data spatiotemporal unified mathematical model of the present application may include a spatial unified model and a temporal unified model. Therefore, the present application can spatially unify the ship position data and the navigation radar data through the spatial unified model, and temporally unify the spatially unified ship position data and the navigation radar data through the temporal unified model.

[0090] For example, since the ship position data and the marine radar data have different spatial information references, the ship position data is the WGS 84 coordinate system and the marine radar data is the polar coordinate system centered on the radar base station. This application can use the WGS 84 coordinate system as a unified reference and use coordinate system conversion to achieve the unification of coordinate space references.

[0091] Assume that the geographic coordinates of a shore-based radar are (X lon0 ,X lat0 ), the distance between the radar and the target ship is set to d, the angle between the line connecting the radar to the target ship and the horizontal line is θ, and the position information of the target ship in the radar system is converted according to formula 4-1 and formula 4-2 to obtain the latitude and longitude coordinates of the target ship. The formula is as follows:

[0092] X lon =X lon0 +dsinθ(4-1)

[0093] X lat =X lat0 +dcosθ(4-2)

[0094] Furthermore, when achieving a unified time update standard for ship position data and navigation radar data, it is necessary to first select a time interval as a basis for calculation. Since the update time interval of radar target data is relatively stable, the average period of radar scanning time can be used as a calculation basis. In this application, the average period of radar scanning time is generally 2s, and the update time interval of radar target data is mostly 2s. The update time of ship position data is not fixed and depends on the current motion state of the ship. This application takes into account the relevant conditions of the ship during normal navigation and takes the update time of ship position data as a research basis.

[0095] Therefore, in order to achieve the unified time update standard of ship position data and navigation radar data, the radar update speed is used as the benchmark, assuming that the ship moves at a uniform speed within the calculated time interval, and the interpolation and extrapolation method is used to calculate the unified time. The specific calculation formula is as follows:

[0096] X=X t +vΔt(4-3)

[0097] Where: X is the calculated position, X t is the position before the time interval Δt, v ​​is the target ship’s speed relative to the ground, in nmile / h.

[0098] This application constructs a unified mathematical model of data time and space to address issues such as different coordinate system references and sampling time intervals for ship position data and marine radar data, thereby achieving preprocessing of ship position data and marine radar data to meet the unification in time and space.

[0099] In one embodiment, Figure 3 As shown, Figure 3 A flow chart of a weighted track fusion algorithm based on local information entropy provided in an embodiment of the present application; determining the associated track point matrix of the target ship at each time according to the unified ship position data and navigation radar data in S122 may include:

[0100] S1221: Determine whether the track points corresponding to the unified ship position data and navigation radar data at each moment are within the gate of the target ship, and obtain the determination results of each track point at each moment.

[0101] S1222: Determine the associated track point matrix of the target ship at each moment according to the determination results of each track point at each moment.

[0102] In this embodiment, when determining the associated track point matrix of the target ship at each moment, it is possible to first determine whether the track points corresponding to the unified ship position data and navigation radar data at each moment are within the wave gate of the target ship, and construct the associated track point matrix based on the determination result.

[0103] Specifically, since the ship tracks are relatively dense under the condition of multi-target tracking, the present application can first determine the irrelevant track points, and then calculate the relevant track points. For example, the present application can establish a track relationship by determining whether the track point is in the gate intersection area, and use the associated track point matrix to describe this attribution relationship, which is mathematically expressed as follows:

[0104]

[0105] In the above formula, ω ij is a binary variable, ω i1 =1 means that measurement i is within the gate of target 1, ω i1 =0 means that measurement i is not within the gate of target 1. i0 =1 means that any measurement may be from a false alarm or clutter, so the value is 1. In this application, the track point related to the target ship j in all the measurement values ​​is the jth column of the confirmation matrix, and its data is expressed as follows:

[0106]

[0107] Based on this expression, the associated track point matrix of the target ship at each moment can be determined.

[0108] In one embodiment, determining the weight matrix corresponding to all track points in each associated track point matrix according to the unified ship position data and navigation radar data in S122 may include:

[0109] S1223: Determine the distance difference between each track point in each associated track point matrix according to the unified ship position data and navigation radar data, and construct the distance difference matrix of the target ship at each time according to the distance difference between each track point.

[0110] S1224: Calculate the weight matrix corresponding to all track points of the target ship at each time according to the distance difference matrix of the target ship at each time.

[0111] In this embodiment, Figure 3 As shown, when determining the weight matrix corresponding to all the track points in each associated track point matrix, the distance difference between each track point in each associated track point matrix can be determined based on the unified ship position data and navigation radar data, and the distance difference matrix of the target ship at each time can be constructed, and then the weight matrix corresponding to all the track points of the target ship at each time can be calculated.

[0112] In a specific implementation, the present application can regard the detection information from multiple sensors as feature information during the track fusion process. First, assuming that N sensors detect the same target ship from multiple directions, the number of feature information is N. Considering the non-ignorability between related sensors, the present application can calculate the cost function of the processed local target measurement of the target ship j at track point a and track point b: Then determine the weight matrix between related track points.

[0113] Furthermore, when constructing the distance difference matrix of the target ship at each time, the present application can first assume that there are M fusion targets j, and according to the confirmation matrix of the target ship j, the number of track points with correlation is N (i.e., ω ij = 1 is N), thereby establishing the distance difference between track point a and track point b at time k As shown below:

[0114]

[0115] Among them, (X a,lon (k),X a,lat (k)) represents the estimated position of track point a at time k; (X b,lon (k),X b,lat (k)) represents the estimated position of track point b at time k.

[0116] Next, the present application can calculate the distance difference between each track point of target ship j at time k: And establish the distance difference matrix D j (k), as shown below:

[0117]

[0118] After obtaining the distance difference matrix of the target ship at each moment through the above calculation process, the weight matrix corresponding to all the track points of the target ship at each moment can be calculated based on the distance difference matrix of the target ship at each moment, and then the fused track data of the target ship at each moment can be calculated through each weight matrix.

[0119] In one embodiment, in S1224, calculating the weight matrix corresponding to all track points of the target ship at each time according to the distance difference matrix of the target ship at each time may include:

[0120] S12241: Calculate the distance information entropy corresponding to each track point of the target ship at each time and the system maximum entropy of the target ship at each time based on the distance difference matrix of the target ship at each time.

[0121] S12242: Calculate the residual degree function corresponding to each track point of the target ship at each time according to the distance information entropy corresponding to each track point of the target ship at each time and the system maximum entropy of the target ship at each time.

[0122] S12243: Determine the weight matrix corresponding to all the track points of the target ship at each time according to the surplus degree function corresponding to each track point at each time.

[0123] In this embodiment, Figure 3 As shown, when calculating the weight matrix corresponding to all the track points of the target ship at each moment, the present application can first calculate the distance information entropy corresponding to each track point of the target ship at each moment and the maximum entropy of the system of the target ship at each moment based on the distance difference matrix of the target ship at each moment. In this way, the residual degree function corresponding to each track point of the target ship at each moment can be calculated according to the distance information entropy corresponding to each track point of the target ship at each moment and the maximum entropy of the system of the target ship at each moment, and the weight matrix corresponding to all the track points of the target ship at each moment can be constructed according to each residual degree function.

[0124] Specifically, information entropy is one of the metrics for measuring the uncertainty of random variables, and its calculation formula is as follows:

[0125] H(X)=-Σp(x)log 2 p(x) (4-8)

[0126] Among them, H(X) represents the information entropy of random variable X, p(x) represents the probability of a certain value of X, log 2 Represents the logarithm operation with base 2. This formula can be understood as, for each value x, calculating its probability p(x), and then adding p(x) and log 2 Multiply and accumulate p(x), and finally add a negative sign to get the information entropy of the random variable X.

[0127] It can be understood that information entropy is a numerical indicator used to measure the uncertainty or amount of information of a random variable. If the random variable has more possible values ​​and the probability of each value occurring is more balanced, its information entropy will be greater; conversely, if the range of values ​​of the random variable is smaller and the probability of each value occurring is more uneven, its information entropy will be smaller. Information entropy usually uses negative numbers as units, such as bits or nats, which represent logarithmic operations with a base of 2 or the natural logarithm e, respectively.

[0128] Furthermore, the present application can establish the probability p of the track point in the characteristics of the target ship j based on the distance difference matrix and formula 4-8: a,j(k), the formula is as follows:

[0129]

[0130] Calculate the distance information E corresponding to the track point a a,j (k):

[0131] E a,j (k) = -p a,j (k)lnp a,j (k) (4-10)

[0132] Then the maximum entropy of the system is as follows:

[0133] E max,j (k) = ln H (4-11)

[0134] Wherein, H represents the attribute length of the feature in fusion. The attributes used in this application are mainly the number of track points, so the value of H is N.

[0135] In addition, the remainder function is a function used to describe the range of values ​​of a random variable, which can be used to study the distribution properties and probability density function of random variables. Specifically, for a continuous random variable X, its remainder function can be expressed as:

[0136] R(x)=P(X>x) (4-12)

[0137] Where x is a real number representing the range of values ​​of the random variable X, and P(X>x) represents the probability that the random variable X is greater than . In other words, the remainder function describes the probability of the "remaining part" of the random variable X outside the range of values ​​x. For example, for the remainder function of the standard normal distribution, a numerical table or computer software can be used to find or calculate it. For a discrete random variable X, its remainder function can be expressed as:

[0138] R(x)=P(X>x)=1-P(X≤x)(4-13)

[0139] Where x is an integer representing the range of values ​​of the random variable X, and P(X>x) represents the probability that the random variable X is greater than x, which can be calculated by accumulating probability distribution functions. Because discrete random variables can only take a finite or countable number of values, their remainder function is a step-like curve composed of discrete points, such as the remainder function of the binomial distribution. The remainder function has a wide range of applications in statistics, signal processing, communication engineering, machine learning and other fields. For example, it can be used to evaluate the performance of classifiers, study the distribution characteristics of random variables, design signal detectors, etc.

[0140] On this basis, the present application can calculate the residual degree function of track point a according to the distance information entropy of track point a, formula 4-12 and formula 4-13, as shown in the following formula:

[0141]

[0142] If the correlation between track point a and the motion of target ship j is higher, it is closer to 1, indicating that track point a has more weight in the weighted fusion process of target ship.

[0143] In this application, the weight of the track point a can be obtained by solving the residual function, and the weight formula is as follows:

[0144]

[0145] For fusion target j, at the fusion center, information fusion is performed on each track point according to the weight calculation value. The track update formula of system target j at the fusion center is shown below.

[0146]

[0147] In one embodiment, the training process of the target ship track prediction model may include:

[0148] S310: Obtain sample ship position data and sample navigation radar data of each ship within a second preset historical period.

[0149] S320: Use the weighted track fusion algorithm of local information entropy to perform weighted fusion on the sample ship position data and sample navigation radar data of each ship to obtain the sample fused track data of each ship at each moment.

[0150] S330: Input the sample fusion track data of each ship at each time into the preset initial ship track prediction model, and use the Adam algorithm to update the network parameters in the initial ship track prediction model until the preset training conditions are met, so as to obtain the target ship track prediction model.

[0151] In this embodiment, when determining the target ship track prediction model, an initial ship track prediction model may be obtained and the initial ship track prediction model may be trained to obtain a final target ship track prediction model.

[0152] Specifically, Figure 4 As shown, Figure 4Schematic diagram of the training process of the initial ship track prediction model provided for the embodiment of the present application; the present application can first obtain the sample ship position data and sample navigation radar data of each ship in the second preset historical period. For example, the present application can obtain data every 10 minutes through a self-compiled program, and use the ship flow in the waters of Xiamen Port from September 1 to December 31, 2022 (a total of 122 days) as the input source of the sample ship position data, and the sample navigation radar data is collected from 0:00 on September 1, 2022 to December 31, 2022. The data is sent by the navigation radar for a total of 122 days, and the data is obtained every 2 minutes, and 145 ship flow statistics are generated in one day. Then, the present application can use the weighted track fusion algorithm using local information entropy to perform weighted fusion on the sample ship position data and sample navigation radar data of each ship, and obtain the sample fused track data of each ship at each time. The specific process can be found above, and it will not be repeated here.

[0153] Furthermore, the present application can use the LSTM circular neural network as the initial ship track prediction model, take the sample fusion track data as input, and extract the future track data, and establish a mapping relationship between the ship track data and the future track data by comparing with the real track data. Specifically, for a ship, its track characteristics at point t can be expressed as:

[0154] Y(t)={lat,lon,sog,cog}

[0155] Among them, lat, lon, sog, and cog are the four characteristics of the ship at time t: latitude, longitude, speed, and heading.

[0156] When predicting the ship's track, the present application can use the track characteristic data at multiple time points and associate them with it, taking Y(t-n+1),...,Y(t-1),Y(t) track characteristic data as input, where n is the step size of the input layer, and taking the track characteristic data at time t+1 as output. Thus, an initial ship track prediction model is established:

[0157] Y(t+1)=f({Y(t-n+1),...,Y(t-1),Y(t)})

[0158] By updating the neural network parameters in the initial ship track prediction model and solving it, it is made to approach or approximate the optimal solution, thereby improving the prediction accuracy. On this basis, this application proposes an Adam algorithm that can adaptively adjust the learning rate, and adopts a method based on historical momentum to improve the convergence speed of LSTM. This method requires less storage space and has the characteristics of fast convergence. The following is the update rule of the Adam algorithm:

[0159] E t (g) = β t E t-1 (g)+(1-β 1 )g(ω t )

[0160] E t (g 2 )=β 2 E t-1 (g 2 )+(1-β 2 )g(ω t )⊙g(ω t )

[0161]

[0162] Updated weights:

[0163]

[0164] Among them, β 1 , β 2 is a variable parameter between [0, 1], and the first-order matrix of the gradient vector g is E t (g), the second-order matrix is ​​E t (g 2 ).

[0165] In a specific implementation, the present application can first define a Sequential model, then add an LSTM layer, with input of (train_seq.shape[1], train_seq.shape[2]), output of 108 neurons, return_sequences=False (i.e., the last layer does not return the entire sequence, but only returns the output value), and then add a Dense layer with output of train_label.shape[1]. Then, add an activation function, use relu, and define an optimizer adam with a learning rate of 0.01. This application can use the compile method to compile the model, the loss function is mse, the optimizer is adam, the evaluation index is acc, and the callback functions are defined, one is CSVLogger for log storage, and the other is ReduceLROnPlateau for reducing the learning rate. Then use the fit method to train the model, the input parameters are train_seq, train_label, batch_size is 32, the number of iterations is 20, the validation set ratio is 0.1, the callback functions are log and reduce, and the evaluate method is used to calculate the loss and acc on the test set and print them out. Finally, use the save method to save the model to obtain the target ship track prediction model.

[0166] In the above embodiments, intelligent algorithms are used to process and analyze data, which can quickly and automatically identify abnormal ships, classify and evaluate them, making supervision more intelligent and efficient. By integrating AIS and radar data and combining ship abnormal behavior analysis technology, ship behavior can be comprehensively analyzed to reduce the misjudgment rate and improve the precision and accuracy of supervision.

[0167] The ship abnormal behavior monitoring device provided in an embodiment of the present application is described below. The ship abnormal behavior monitoring device described below and the ship abnormal behavior monitoring method described above can be referenced to each other.

[0168] In one embodiment, Figure 5 As shown, Figure 5 A schematic diagram of the structure of a ship abnormal behavior monitoring device provided in an embodiment of the present application; the present application also provides a ship abnormal behavior monitoring device, which may include a data acquisition module 210, a weighted fusion module 220, a track prediction module 230, and an abnormality monitoring module 240, specifically including the following:

[0169] The data acquisition module 210 is used to acquire the ship position data and navigation radar data of the target ship at the current moment and within a first preset historical period before the current moment.

[0170] The weighted fusion module 220 is used to perform weighted fusion on the ship position data and the navigation radar data using a weighted track fusion algorithm of local information entropy to obtain fused track data of the target ship at each moment.

[0171] The track prediction module 230 is used to input the fused track data at each moment into a preset target ship track prediction model to obtain the predicted track data output by the target ship track prediction model.

[0172] The abnormality monitoring module 240 is used to determine whether the target ship has abnormal behavior according to the predicted track data.

[0173] In the above embodiment, when monitoring the abnormal behavior of the target ship, the ship position data and navigation radar data of the target ship at the current moment and within the first preset historical period before the current moment can be obtained first; then, the present application can use the weighted track fusion algorithm of local information entropy to weightedly fuse the ship position data and the navigation radar data to obtain the fused track data of the target ship at each moment. The fused track data is obtained by unifying the ship position data and the navigation radar data in time and space, removing irrelevant track points through the weighted track fusion algorithm of local information entropy, and weighted fusion according to the weights of each relevant track point. Therefore, the fused track data at each moment is input into the preset target ship track prediction model to obtain the more accurate predicted track data output by the target ship track prediction model, and the predicted track data can be accurately determined whether the target ship has abnormal behavior.

[0174] In one embodiment, the present application also provides a computer-readable storage medium, in which computer-readable instructions are stored. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the ship abnormal behavior monitoring method as described in any of the above embodiments.

[0175] In one embodiment, the present application also provides a computer device, including: one or more processors, and a memory.

[0176] The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the ship abnormal behavior monitoring method as described in any one of the above embodiments are performed.

[0177] Indicatively, Figure 6 As shown, Figure 6 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. The computer device 300 may be provided as a server. Figure 6 The computer device 300 includes a processing component 302, which further includes one or more processors, and a memory resource represented by a memory 301, for storing instructions executable by the processing component 302, such as an application. The application stored in the memory 301 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 302 is configured to execute instructions to perform the ship abnormal behavior monitoring method of any of the above embodiments.

[0178] The computer device 300 may further include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate based on an operating system stored in the memory 301, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM, or the like.

[0179] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0180] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0181] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

[0182] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring abnormal behavior of a ship, characterized in that: The method comprises: Acquire ship position data and navigation radar data of the target ship at the current moment and within a first preset historical period before the current moment; Using a weighted track fusion algorithm of local information entropy to perform weighted fusion on the ship position data and the navigation radar data to obtain fused track data of the target ship at each moment; Inputting the fused track data at each moment into a preset target ship track prediction model to obtain predicted track data output by the target ship track prediction model; Determine whether the target ship has abnormal behavior according to the predicted track data.

2. The method for monitoring abnormal ship behavior according to claim 1, characterized in that: The weighted track fusion algorithm using local information entropy performs weighted fusion on the ship position data and the navigation radar data to obtain fused track data of the target ship at each time, including: The ship position data and the navigation radar data are spatially and temporally unified by a preset data time-space unification mathematical model; Determine the associated track point matrix of the target ship at each time and the weight matrix corresponding to all track points in each associated track point matrix according to the unified ship position data and navigation radar data; According to the weight matrices corresponding to all the track points in each associated track point matrix, the unified ship position data and the navigation radar data are weighted fused to obtain the fused track data of the target ship at each moment.

3. The method for monitoring abnormal ship behavior according to claim 2, characterized in that: The data time-space unified mathematical model includes a space unified model and a time unified model; The spatial and temporal unification of the ship position data and the navigation radar data by using a preset data time-space unification mathematical model includes: spatially unifying the ship position data and the navigation radar data through the spatial unification model; The spatially unified ship position data and navigation radar data are temporally unified through the temporal unification model.

4. The method for monitoring abnormal ship behavior according to claim 2, characterized in that: The step of determining the associated track point matrix of the target ship at each time according to the unified ship position data and navigation radar data includes: Determine whether the track point corresponding to the unified ship position data and navigation radar data at each moment is within the wave gate of the target ship, and obtain the determination result of each track point at each moment; According to the determination results of each track point at each time, the associated track point matrix of the target ship at each time is determined.

5. The ship abnormal behavior monitoring method according to claim 2 or 4, characterized in that: The method of determining the weight matrix corresponding to all track points in each associated track point matrix according to the unified ship position data and navigation radar data includes: Determine the distance difference between each track point in each associated track point matrix according to the unified ship position data and navigation radar data, and construct the distance difference matrix of the target ship at each time according to the distance difference between each track point; According to the distance difference matrix of the target ship at each time, the weight matrix corresponding to all the track points of the target ship at each time is calculated.

6. The method for monitoring abnormal ship behavior according to claim 5, characterized in that: The step of calculating the weight matrix corresponding to all track points of the target ship at each time according to the distance difference matrix of the target ship at each time includes: According to the distance difference matrix of the target ship at each time, the distance information entropy corresponding to each track point of the target ship at each time and the system maximum entropy of the target ship at each time are calculated; Calculate the residual degree function corresponding to each track point of the target ship at each time according to the distance information entropy corresponding to each track point of the target ship at each time and the system maximum entropy of the target ship at each time; According to the surplus degree function corresponding to each track point at each time, the weight matrix corresponding to all track points of the target ship at each time is determined.

7. The method for monitoring abnormal ship behavior according to claim 1, characterized in that: The training process of the target ship track prediction model includes: Obtaining sample ship position data and sample navigation radar data of each ship within a second preset historical period; The weighted track fusion algorithm of local information entropy is used to perform weighted fusion on the sample ship position data and sample navigation radar data of each ship to obtain the sample fusion track data of each ship at each time. The sample fusion track data of each ship at each moment is input into the preset initial ship track prediction model, and the Adam algorithm is used to update the network parameters in the initial ship track prediction model until the preset training conditions are met, thereby obtaining the target ship track prediction model.

8. A device for monitoring abnormal behavior of a ship, characterized in that: include: A data acquisition module, used to acquire ship position data and navigation radar data of a target ship at a current moment and within a first preset historical period before the current moment; A weighted fusion module, used for weighted fusion of the ship position data and the navigation radar data using a weighted track fusion algorithm of local information entropy to obtain fused track data of the target ship at each moment; A track prediction module is used to input the fused track data at each moment into a preset target ship track prediction model to obtain the predicted track data output by the target ship track prediction model; The abnormality monitoring module is used to determine whether the target ship has abnormal behavior according to the predicted track data.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the ship abnormal behavior monitoring method as described in any one of claims 1 to 7.

10. A computer device, characterized in that: include: one or more processors, and memory; The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the ship abnormal behavior monitoring method as described in any one of claims 1 to 7 are performed.

Images