A regional warning method based on AIS civilian ship behavior prediction

Abstract

The present invention belongs to the field of ship management technology, and specifically relates to a civilian ship behavior prediction and regional alarm method based on AIS. The method includes constructing a database and selecting an AIS target, using the target's historical route track points through DBScan clustering to obtain the target's attribution abnormality level, and obtaining the target's position abnormality value, motion trend abnormality value, current motion abnormality value, and current behavior abnormality value in turn according to the current state of the target, and obtaining each level according to the size of each abnormality value, and then searching for the selected AIS target, and obtaining the historical credit level of the selected AIS target according to the historical credit record searched; and then combining the target credit record to comprehensively generate the threat level of the target, and judging whether to perform the alarm task according to the threat level of the target. The method effectively distinguishes different ships through a comprehensive combination of multiple factors, and uses the ship threat level as the basis for task allocation, providing decision support for the coast guard and warships to take necessary preventive measures.

Description

Technical Field

[0001] The invention belongs to the technical field of marine ship management, and in particular relates to a regional warning method based on AIS civilian ship behavior prediction. Background Art

[0002] During maritime navigation, ships are the main actors in the navigation field. Abnormal ship behavior refers to abnormal deviations from the channel and heading, sudden acceleration and deceleration, and appearance in areas that should not be entered. The prediction of abnormal ship behavior is to analyze the speed, heading, and track of the ship based on the input ship motion-related data to determine whether the ship's movement conforms to the normal law of navigation activities, and then identify the safety of the ship itself, or whether there is abnormal behavior, and evaluate and warn of possible dangers. At present, the relevant technology mainly monitors key areas in real time, displays ships that have abnormal behavior to the operator, and then the operator determines whether a warning is needed to drive them away. This is inefficient and requires the operator to always pay attention to the situation of ships at sea. At present, there is a lack of a method that can autonomously judge the abnormal level of abnormal behavior ships and guide operators to warn and drive away abnormal ships. Summary of the invention

[0003] The purpose of the present invention is to provide a method for predicting the behavior of civilian ships and regional warning based on AIS in view of the lack of a method for autonomously judging the abnormal level of abnormal behavior ships in the prior art. The method includes building a database and selecting an AIS target, obtaining the target's attribution abnormal level by using the target's historical route track points through DBScan clustering, obtaining the target's position abnormal value, motion trend abnormal value, current motion abnormal value, and current behavior abnormal value in sequence according to the current state of the target, obtaining each level according to the size of each abnormal value, then searching for the selected AIS target, and obtaining the historical credit level of the selected AIS target according to the historical credit record searched; then comprehensively generating the target's threat level in combination with the target's credit record, and judging whether to execute the alarm task according to the target's threat level.

[0004] In order to achieve the above object, the present invention adopts the following technical scheme.

[0005] The AIS-based civilian ship behavior prediction and regional warning method comprises the following steps:

[0006] S1 builds a database, compiles and processes historical binary AIS track information and stores it in the database to obtain a track library table, and enters historical credit record documents into the database to obtain a credit library table;

[0007] S2 selects a batch of current AIS targets, retrieves the historical route track points of the targets in the database track library table according to the target's maritime mobile identification code MMSI, uses the DBScan clustering algorithm to calculate dense polygon areas, and calculates three types of area sets, namely core area set, important area set, and general area set, by setting different scanning radii and minimum number of included objects. Then, the location of the current track point is determined and the track point attribution anomaly level is calculated.

[0008] S3 calculates the shortest distance between the selected AIS target and the navigation area, and obtains the distance level according to the distance;

[0009] Determine whether the heading of the selected AIS target is towards the navigation area, and obtain the heading level according to the heading;

[0010] According to the speed of the selected AIS target, the speed level is obtained;

[0011] S4 uses the matrix method to obtain the target's position anomaly value, motion trend anomaly value, current motion anomaly value and current behavior anomaly value according to the selected AIS target's belonging anomaly level, distance level, direction level and speed level, and then obtains the position anomaly level, motion anomaly level, current motion anomaly level and current behavior anomaly level according to the size of each anomaly value.

[0012] S5 searches for the selected AIS target in the credit library table, and obtains the historical credit rating of the selected AIS target based on the historical credit records in the credit library table;

[0013] S6 obtains the target threat value by using the matrix method according to the current behavior anomaly level and historical credit level of the selected AIS target, obtains the threat level of the selected AIS target according to the threat value, and executes the alarm task according to the threat level.

[0014] Furthermore, the compilation and processing of the historically recorded binary AIS information in S1 is to interpret the message using the AIS standard protocol, extract the data of each field, and store it in the database track library table, that is, to convert each frame of the AIS message into a database record.

[0015] Furthermore, the historical credit record documents include credit documents recorded in previous tasks and credit documents recorded in daily AIS monitoring.

[0016] Furthermore, a batch of current AIS targets are selected in S3, and the historical route track points are retrieved from the track library table of the database according to the MMSI of the target, and the DBScan clustering algorithm is used to calculate the dense polygon area. The specific method is as follows:

[0017] Set the basic scanning radius r and the minimum number of included objects n, and use DBScan to calculate the density:

[0018] When the scanning radius is r and the minimum number of included objects is n, the general area set Area is obtained by scanning. com ;

[0019] When the scanning radius is r and the minimum number of included objects is 2n, the important area set Area is obtained by scanning. imp ;

[0020] When the scanning radius is r / 2 and the minimum number of included objects is 2n, the core area set Area is obtained by scanning. cor ;

[0021] Then determine the current track point P now The location of the track point is calculated, and the abnormal level G of the track point is calculated. ads ;

[0022] If P now ∈Area cor , then G ads =1;

[0023] If P now ∈Area imp , then G ads =2;

[0024] If P now ∈Area com , then G ads =3;

[0025] If the above three conditions are not met, then G ads =4.

[0026] Furthermore, the calculation matrix of the position abnormal value, the motion trend abnormal value, the current motion abnormal value, the current behavior abnormal value and the method of dividing the corresponding abnormal level in S4 are:

[0027] The position outlier value is positively correlated with both the orientation level and the distance level;

[0028] The movement trend abnormal value is positively correlated with both the heading level and the speed level;

[0029] The current motion abnormal value is positively correlated with the position abnormal value and the motion trend abnormal value;

[0030] The current behavior abnormality value is positively correlated with the current movement abnormality level and the attribution abnormality level;

[0031] Furthermore, the position abnormality level, motion abnormality level, current motion abnormality level and current behavior abnormality level in S4 are positively correlated with the position abnormality value, motion trend abnormality value, current motion abnormality value and current behavior abnormality value, respectively.

[0032] Beneficial Effects

[0033] The regional warning method based on AIS civilian ship behavior prediction proposed by the present invention has the following beneficial effects compared with the prior art:

[0034] 1. The method effectively distinguishes different ships by comprehensively considering multiple factors, takes the threat level of ships as the basis for task allocation, and provides decision support for the coast guard and warships to take necessary preventive measures;

[0035] 2. The method uses the DBScan clustering algorithm and the target's historical route track points to obtain the target's belonging anomaly level, and obtains the target's position anomaly value, motion trend anomaly value, current motion anomaly value, and current behavior anomaly value in sequence according to the target's current state, and then combines the target's credit record to comprehensively generate the target's threat level. The present invention uses a comprehensive judgment of multiple factors to divide the threat level of ships, which not only effectively distinguishes different ships, but also uses the ship's threat level as an important basis for task allocation when it is necessary to drive away the ship, providing decision support for the coast guard and warships to take necessary preventive measures, and ensuring the safety of the navigation area and military secrets. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A flowchart of a regional warning method for predicting civilian ship behavior based on AIS according to the present invention;

[0037] Figure 2 A schematic diagram of track prediction clustering in a regional warning method for predicting civilian ship behavior based on AIS according to the present invention;

[0038] Figure 3 The present invention is a schematic diagram of track point extraction in a regional warning method for predicting civilian ship behavior based on AIS. DETAILED DESCRIPTION

[0039] The following is a further description of a regional warning method for predicting civilian ship behavior based on AIS and its specific implementation according to the present invention in conjunction with specific embodiments.

[0040] Example 1

[0041] like Figure 1 As shown, the AIS-based civilian ship behavior prediction and regional warning method disclosed in this embodiment includes the following steps:

[0042] S1 compiles and processes the historical binary AIS track information and stores it in the database track library table;

[0043] S2 enters the historical credit record document into the credit table of the database;

[0044] S3 selects a batch of current AIS targets, retrieves historical route track points from the track library table of the database according to the target's maritime mobile identification code MMSI, uses the DBScan clustering algorithm to calculate dense polygon areas, and calculates three types of area sets, namely core area set, important area set, and general area set, by setting different scanning radii and minimum number of included objects. Then, the location of the current track point is determined and the track point attribution anomaly level is calculated.

[0045] S4 calculates the shortest distance between the selected AIS target and the navigation area, and obtains the distance level according to the distance;

[0046] S5 determines whether the heading of the selected AIS target is toward the navigation area, and obtains the heading level according to the heading;

[0047] S6 obtains the speed level according to the speed of the selected AIS target;

[0048] S7 uses the matrix method to obtain the position anomaly value, motion trend anomaly value, current motion anomaly value and current behavior anomaly value of the AIS target, and divides the anomaly level of each anomaly value.

[0049] S8 searches for the selected AIS target in the credit database table of the database and obtains the historical credit rating based on the credit record;

[0050] S9 obtains the target threat value by using the matrix method according to the target threat value calculation model, and then obtains the target threat level by level classification.

[0051] In this embodiment, the ship data comes from the Automatic Identification System (AIS). When building the database, it is necessary to obtain the historical track information of the ship from the AIS. The AIS track information is binary data and needs to be compiled and processed. The AIS standard protocol is used to interpret the message, extract the data of each field, and store it in the database track library table, that is, convert each frame of the AIS message into a database record.

[0052] The historical credit records in step S2 are recorded by the coast guard or relevant departments and bound to the MMSI code of the ship. The records include those that have been determined by the coast guard or relevant departments to have eavesdropped, have entered controlled navigation areas, have been close to controlled navigation areas, have sailed around navigation areas, and have reported false navigation intentions, navigation status, nationality, and destinations. In addition, it also includes credit documents recorded in daily AIS monitoring and other types of credit information that can be obtained.

[0053] like Figure 2 After the historical data is entered into the database, it is necessary to predict the track of the ships in the target area and extract the characteristic laws of ship movement. In this embodiment, a clustering algorithm is used. Conventional trajectory clustering application objects are divided into two types:

[0054] (1) Clustering method that takes the entire trajectory as the target. This method has problems such as long trajectories, large time and space overhead, and loss of some similar trajectory sub-segments;

[0055] (2) The trajectory is divided into sub-segments and the clustering method is used as the target. This method may not be able to fully obtain the characteristics of the entire trajectory, but it can better grasp the characteristics of the trajectory sub-segments, and the characteristics of each sub-segment can also better describe the characteristics of the entire trajectory.

[0056] Therefore, the trajectory segmentation method is adopted to cluster the ship trajectories.

[0057] There is still a problem before performing the clustering algorithm, that is, the amount of AIS data to be processed may be relatively large. The dynamic information of the ship is usually broadcast once every 2 to 180 seconds. The specific time interval depends on the speed and the rate of change of heading. The huge amount of data will affect the clustering speed. Therefore, the ship trajectory points can be extracted, the threshold of the heading or speed change rate can be set, and the track points that meet the threshold can be extracted as clustering processing data.

[0058] like Figure 3 As shown in the figure, it is assumed that the trajectory points P1 to P8 are the location points collected by AIS, the solid line is the original trajectory, and the dotted line is the divided trajectory. At this time, if all points in P to P8 are selected as feature points, the trajectory accuracy will reach the highest, but the overhead time will be very large; if only some points P1-P5-P8 are classified as feature points, the trajectory will obtain better simplicity, but the characteristics of the original trajectory will be lost. Therefore, in order to achieve the best balance between the accuracy and simplicity of trajectory division (P2-P4-P6-P8), the trajectory is divided into two processes. The rate of change of heading or speed can be expressed by the following formula:

[0059]

[0060] By selecting an appropriate threshold, the size of the data to be processed can be greatly reduced without affecting the overall track information.

[0061] After obtaining the appropriate data to be processed, the DBScan cluster analysis method can be used to analyze the AIS historical data of civilian ship targets to obtain the track point attribution level. The method is as follows:

[0062] First, set the basic scanning radius r and the minimum number of included objects n, and then use DBScan to calculate the density:

[0063] When the scanning radius is r and the minimum number of included objects is n, the general area set Area is obtained by scanning. com ;

[0064] When the scanning radius is r and the minimum number of included objects is 2n, the important area set Area is obtained by scanning. imp ;

[0065] When the scanning radius is and the minimum number of included objects is 2n, the core area set Area is obtained by scanning. cor ;

[0066] Then determine the current track point P now The location of the track point is calculated, and the abnormal level G of the track point is calculated. ads ;

[0067] If P now ∈Area cor , then G ads =1;

[0068] If P now ∈Area imp , then G ads =2;

[0069] If P now ∈Area com , then G ads =3;

[0070] If the above three conditions are not met, then G ads =4.

[0071] After obtaining the abnormal level of the target's track points through behavior prediction, the target's distance level, orientation level and speed level are obtained according to the target's current state. The corresponding level evaluation table is shown in Table 1-3:

[0072] Table 1 Distance level assignment table

[0073] Distance Classification Evaluation Criteria Distance level assignment Already in the flight area The target is currently inside the navigation area 5 Very close to the flight area The target is within 0 to 10 nautical miles from the navigation area 4 Very close to the flight area The target is within 10 to 50 nautical miles from the navigation area 3 Close to the flight area The target is within 50 to 100 nautical miles of the navigation area 2 Far from the navigation area The target is more than 100 nautical miles away from the navigation area 1

[0074] Table 2 Orientation level assignment table

[0075] Orientation classification Evaluation Criteria Towards level assignment Through the navigation area The extended ray of the heading passes through the navigation area 3 Approaching area The target is close to the navigation area, but the heading ray does not pass through the navigation area 2 Stay away from the navigation area Overall away from the navigation area 1

[0076] Table 3 Speed ​​level assignment table

[0077] Speed ​​classification Evaluation Criteria Speed ​​rating Ultra-high speed Target speed: 30 knots or more 5 high speed The target speed is between 15 knots and 30 knots. 4 Medium Target speed is between 8 knots and 15 knots 3 Low speed The target speed is between 2 knots and 8 knots. 2 Ultra-low speed Target speed is within 2 knots 1

[0078] The target position anomaly value and motion trend anomaly value are obtained through the above distance level, direction level and speed level, and each anomaly value is classified into an anomaly level, and the current motion anomaly value and current behavior anomaly level are obtained. The methods for obtaining each anomaly value and the corresponding anomaly level classification standards are shown in Table 4-11:

[0079] Table 4 Position outlier calculation matrix

[0080]

[0081] Table 5 Principles for classifying location anomalies

[0082] Location Outliers 1～3 4～8 9～10 11～13 14～15 Location anomaly level 1 2 3 4 5

[0083] Table 6 Movement trend outlier calculation matrix

[0084]

[0085] Table 7 Principles for classifying abnormal movement trends

[0086] Motion Trend Outliers 1～3 4～8 9～10 11～13 14～15 Abnormal level of movement trend 1 2 3 4 5

[0087] Table 8 Current motion outlier calculation matrix

[0088]

[0089] Table 9 Current principles for the classification of movement abnormalities

[0090] Current motion abnormality 1～5 6～10 11～15 16～20 21～25 Current movement abnormality level 1 2 3 4 5

[0091] Table 10 Current behavior outlier calculation matrix

[0092]

[0093]

[0094] Table 11 Current principles for classifying abnormal behaviors

[0095] Current behavior outliers 1～4 5～8 9～13 14～17 18～20 Current level of abnormal behavior 1 2 3 4 5

[0096] The selected AIS target is searched in the credit library table of the database, and the historical credit anomaly level is generated according to the credit record and the evaluation criteria in Table 12.

[0097] Table 12 Historical credit abnormality level assignment table

[0098]

[0099] The target threat value and target threat level of the selected AIS target are obtained through the current behavior anomaly level and the historical credit anomaly level. The specific standards are shown in Table 13:

[0100] Table 13 Target threat value calculation matrix

[0101]

[0102] According to the different target threat values ​​calculated, the target threat level is obtained by using the level classification method, and is divided according to Table 14:

[0103] Table 14 Current principles for classifying abnormal behaviors

[0104] Target Threat Value 1～5 6～9 10～14 15～17 18～20 Target Threat Level 1 2 3 4 5

[0105] The method of this embodiment extracts various factors that reflect the abnormal behavior of civilian ships, and establishes evaluation criteria and quantitative indicators, making the originally complex and difficult to achieve civilian ship behavior prediction and regional warning simple and feasible. In actual use, the prediction results are very accurate, and suspicious targets are discovered many times, providing decision-making support for the coast guard and warships to take necessary preventive measures, thereby ensuring the safety of the navigation area and military secrets.

[0106] The technical means disclosed in the scheme of the present invention are not limited to the technical means disclosed in the above-mentioned implementation mode, but also include technical schemes composed of any combination of the above technical features.

Claims

1. A regional warning method based on AIS civilian ship behavior prediction, characterized in that: The steps include: S1 builds a database, compiles and processes historical binary AIS track information and stores it in the database to obtain a track library table, and enters historical credit record documents into the database to obtain a credit library table; S2 selects a batch of current AIS targets, retrieves the historical route track points of the targets in the database track library table according to the target's maritime mobile identification code MMSI, uses the DBScan clustering algorithm to calculate dense polygon areas, and calculates three types of area sets, namely core area set, important area set, and general area set, by setting different scanning radii and minimum number of included objects. Then, the location of the current track point is determined and the track point attribution anomaly level is calculated. S3 calculates the distance between the selected AIS target and the set of areas with the shortest distance therefrom, and obtains a distance level according to the distance; Determine whether the heading of the selected AIS target is toward the area set with the shortest distance to it, and obtain the heading level according to the heading; According to the speed of the selected AIS target, the speed level is obtained; S4 uses a matrix method to obtain the position anomaly value, motion trend anomaly value, current motion anomaly value and current behavior anomaly value of the target according to the attribute anomaly level, distance level, direction level and speed level of the selected AIS target, and then obtains the position anomaly level, motion trend anomaly level, current motion anomaly level and current behavior anomaly level according to the size of each anomaly value; the method for obtaining the calculation matrix of the position anomaly value, motion trend anomaly value, current motion anomaly value and current behavior anomaly value and the corresponding anomaly level division is as follows: The position outlier value is positively correlated with both the orientation level and the distance level; The movement trend abnormal value is positively correlated with both the heading level and the speed level; The current motion abnormality value is positively correlated with both the position abnormality level and the motion trend abnormality level; The current behavior abnormality value is positively correlated with the current movement abnormality level and the attribution abnormality level; The position abnormality level, movement trend abnormality level, current movement abnormality level and current behavior abnormality level are positively correlated with the position abnormality value, movement trend abnormality value, current movement abnormality value and current behavior abnormality value respectively; S5 searches for the selected AIS target in the credit library table, and obtains the historical credit rating of the selected AIS target based on the historical credit records in the credit library table; S6 obtains the target threat value by using the matrix method according to the current behavior anomaly level and historical credit level of the selected AIS target, obtains the threat level of the selected AIS target according to the threat value, and determines whether to execute the alarm task according to the threat level.

2. According to claim 1, a regional warning method based on AIS civilian ship behavior prediction is characterized by: In S1, the binary AIS information recorded in history is compiled and processed, and the AIS standard protocol is used to interpret the message, and the data of each field is extracted and stored in the database track library table, that is, each frame of AIS message is converted into a database record.

3. The regional warning method based on AIS civilian ship behavior prediction according to claim 1 is characterized by: The historical credit record documents include credit documents recorded in previous tasks and credit documents recorded in daily AIS monitoring.

4. The regional warning method based on AIS civilian ship behavior prediction according to claim 1 is characterized by: In S2, a batch of current AIS targets are selected, and the historical route track points are retrieved from the track library table of the database according to the MMSI of the target. The DBScan clustering algorithm is used to calculate the dense polygon area. The specific method is as follows: Set the basic scanning radius r and the minimum number of included objects n, and use DBScan to calculate the density: When the scanning radius is r and the minimum number of included objects is n, the general area set is obtained by scanning. ; When the scanning radius is r and the minimum number of included objects is 2n, the scanned area is a set of important areas. ; When the scanning radius is r / 2 and the minimum number of included objects is 2n, the core area set is obtained by scanning. ; Then determine the current track point Position attribution, calculate the abnormal level of track point attribution ; if ,but ; if ,but ; if ,but ; If the above three conditions are not met, then .

Images