A heterogeneous ship multi-track general segment feature extraction method

Abstract

The application discloses a kind of heterogeneous ship multi-track subsection general feature extraction method, the heterogeneous ship is the multiple ships of different ship type C and ship length L, the method comprises the following steps: determining AIS data acquisition sea area;Classify heterogeneous ship;General subsection is carried out to multi-track;Extract general subsection track feature;In turn select the track in track set, according to MMSI select the general subsection of belonging ship classification and its general subsection track feature.The application can simultaneously process multiple tracks of different ships, and can be used for trajectory compression or trajectory feature extraction.The application determines track feature points by scoring, ensuring the effectiveness of track feature extraction.In addition, other tracks can also be directly applied to the application, ensuring the generality of the track feature extraction algorithm.Compared with traditional feature extraction algorithms, the iteration and calculation time are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to a heterogeneous ship multi-track general segmented feature extraction method, belonging to the field of ocean science and technology and shipping artificial intelligence. BACKGROUND

[0002] With the continuous development of maritime trade and the application and development of the Automatic Identification System (AIS) technology, more and more researchers use AIS data collected by the Automatic Identification System to analyze maritime traffic problems.

[0003] With the popularity of the Automatic Identification System, AIS has become an important device for obtaining ship dynamic data, realizing information interaction between ships and shores, which also enables the ship supervisory department to obtain ship navigation data and process and analyze the data. Due to the characteristics of cross-time domain and cross-region in AIS data information collection of different ships, and the differences between ships, certain errors will be generated in subsequent AIS data track analysis. Among them, general feature extraction of the track is an effective solution, which not only ensures the accuracy of the track, but also improves the generalization of the track features. General feature extraction of the track is the main research direction of maritime traffic problem analysis, and extracting the required track features from massive AIS data for research has important significance for navigation safety, port construction, maritime management and other aspects.

[0004] Currently, domestic and foreign scholars mainly focus on single-track feature extraction and feature extraction of the entire track, and the subsequent ship track research also focuses on single-track research. Qi et al. proposed a ship track compression algorithm for identifying ship turning angles; Zhu et al. used the Douglas-Pok algorithm to extract ship track features; Gao et al. proposed an improved SlidingWindow online ship track compression algorithm; Zhang Yinhao proposed a ship track compression method based on threshold guided sampling; Jiang Yuling et al. proposed a ship track clustering algorithm based on DBSCAN, obtaining typical ship motion tracks. Currently, few scholars have conducted segmented research on heterogeneous ship multi-track general features. SUMMARY

[0005] In order to solve the problems in the prior art, the present application provides a heterogeneous ship multi-track general segmented feature extraction method, which can realize segmented general feature extraction of multiple ship tracks in a certain sea area, form a multi-dimensional feature matrix, ensure the accuracy of the track, and improve the generalization of the track features.

[0006] In order to achieve the above object, the technical scheme of the present application is as follows: a heterogeneous ship multi-track segmentation general feature extraction method, the heterogeneous ship is a plurality of ships of different ship types C and ship lengths L, the method comprises the following steps:

[0007] S1, determining an AIS data collection sea area

[0008] S11, taking a departure port as a starting point O and a destination port as an ending point D, and defining four points P1(x1, y1), P2(x2, y2), P3(x3, y3), and P4(x4, y4) of the sea area boundary by longitude and latitude.

[0009] S12, according to the longitude and latitude of the starting point, the ending point, and the boundary point, filtering a track set satisfying the sea area boundary condition in step S11 from the AIS data:

[0010] R={T1, T2, …, T i ,…,T n}

[0011] Wherein, R represents a track set satisfying the condition, and a track T i is composed of a series of AIS track points, and is represented as follows:

[0012] T i ={p i1 ,p i2 ,…,p ij ,…,p im}

[0013] Wherein, p ij is the jth track point in the ith track in the AIS data, represented by longitude x ij and latitude y ij ; i represents the track number, i=1, 2, …, n; j represents the number of track points on the same track, j=1, 2, …, m;

[0014] S2, classifying the heterogeneous ships

[0015] S21, according to the identification of the track T i in the AIS database, obtaining the ship mobile service identity code MMSI of each track.

[0016] S22, according to the MMSI, obtaining the ship type C and the ship length L of each ship.

[0017] S23, taking the ship type V and the ship length L as attribute values, and using a classification algorithm to divide the heterogeneous ships into N classes, denoted as S1, S2, …, S N .

[0018] S3, general segmentation of multi-tracks

[0019] S31, Let k = 1

[0020] S32, Select S k Multiple tracks of the same type of vessel are used as the object, and track T is obtained from AIS data. i China Aviation Track Point p ij The corresponding speed v ij and heading d ij .

[0021] S33. Set the steering angle threshold θ, when the track point p ij If the difference Δd in the heading change between the two closest points is greater than the threshold θ, then the waypoint p ij The turning point is defined as the trajectory state of the area traversed by the turning point.

[0022] S34. Set the speed threshold δ, at the track point p ij If p ij If the difference Δv between the speed of the track point and the speed change between the two nearest points is less than the threshold δ, then the track point p ij The straight-ahead point is defined as the track state of the area traversed by the straight-ahead point.

[0023] S34. The area formed by other track points that have not been classified into track states is defined as other track states.

[0024] S35. The track segment containing the turning track state, the straight track state, and other track states is designated as the Sth segment. k Common sections for this type of ship.

[0025] S4. Extract general segmented track features

[0026] S41. Calculate the general segment in-segment track point p. ij The score is obtained by the following formula:

[0027] Score(p ij =InConnection(p ij )×OutConnection(p ij )

[0028] In the formula, InConnection(p ij The value () represents the connectivity importance of the waypoint within this general segment. It is obtained by dividing the number of wayspoints passing through this waypoint by the total number of wayspoints in this general segment, and is calculated using the following formula:

[0029]

[0030] In the formula N represents the total number of tracks, wherein N ij Q and J ij represent the value of 0 or 1, wherein 0 represents that the track point is not passed, and 1 represents that the track point is passed, and len(R) represents the total number of tracks of the current segment.

[0031] OutConnection(p ij ) represents the connection importance of the track point to other segments, and the value is obtained by taking the exponent of the result of dividing the total number of tracks passing through other segments by the number of tracks passing through only the track point, and the formula is as follows:

[0032]

[0033] In the formula, len(R) represents the total number of tracks passing through other segments, represents the total number of tracks passing through only the track point, wherein Q ij and J ij represent the value of 0 or 1.

[0034] S42, according to the descending order of the score, a certain proportion of the total number of track points in the segment is taken as a feature point to form a track feature of the segment.

[0035] S43, if k < N, then k = k + 1, and step S32 is returned.

[0036] S5, extracting multi-track features

[0037] S51, selecting a track T i in the track set R in turn, selecting a segment of a general segment according to the MMSI and the track feature of the general segment of the ship classification.

[0038] S52, according to the feature points in the track feature of the general segment, selecting the track point closest to the feature point in the actual track as the feature point of the track T i .

[0039] S53, the same number of feature points are extracted from the multiple tracks of the same type of ship to form the following feature matrix:

[0040]

[0041] In the formula, T' ij represents the jth track point of the ith track, wherein i = 1 … n, and j = 1 … M.

[0042] Further, in the step S3, the value of θ is 5-6°, and the value of δ is 2.5-3.5 kn / h.

[0043] Further, the classification algorithm in the step S23 comprises K-means method, K-Medoids method, hierarchical clustering method.

[0044] Further, the certain percentage in the step S23 is 10%-20%.

[0045] Compared with the prior art, the present application has the following advantages:

[0046] The present application can simultaneously process multiple tracks of different ships, and can be used for track compression or track feature extraction. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 is the determination of AIS data collection sea area map of the present application.

[0048] Figure 2 is the heterogeneous ship classification map of the present application.

[0049] Figure 3 is the flowchart of the track segmentation algorithm of the present application.

[0050] Figure 4 is the flowchart of the track segmentation algorithm of the present application.

[0051] Figure 5 is the multi-track general segmentation map of the present application.

[0052] Figure 6 is the general segmentation track feature extraction map of the present application.

[0053] Figure 7 is the multi-track feature extraction map of the present application. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical scheme and advantages of the embodiments clearer, the technical scheme of the embodiments will be described clearly and completely in conjunction with the drawings in the implementation of the present application. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0055] As shown in the figure, a heterogeneous ship multi-track general segmentation feature extraction method comprises the following steps. Figures 1-7

[0056] S1, determine the AIS data collection sea area. As shown in the figure, Figure 1 ​As shown, based on the four points defining the sea area boundary using latitude and longitude, a set of tracks R = {T1, T2, ..., T...} that meet the conditions is selected from AIS data. i ,…,T n}. Among them, track T i ={p i1 ,p i2 ,…,p ij ,…,p im It consists of a series of AIS waypoints, where T i This is a sequence of waypoints in AIS data, composed of latitude and longitude x ij and y ij express.

[0057] S2. Classify heterogeneous vessels. For example... Figure 2 As shown, the ship type (C) and length (L) of each ship are obtained according to MMSI, and the heterogeneous ships are divided into N categories using the ship type and length as attribute values.

[0058] S3. Perform common segmentation on multiple tracks. For example... Figures 3-5 As shown, take a class of ships S k Multiple flight paths are used as objects to obtain flight path T from AIS data. i Midpoint p ij The corresponding speed v ij and heading d ij First, set a steering angle threshold θ, when the waypoint p ij If the difference Δd in the heading change between the two closest points is greater than the threshold θ, then the waypoint p ij As a turning point, the track state of the area where the turning point is located is defined as the turning track, such as... Figure 3 As shown. Next, a speed threshold δ is set at the track point p. ij If p ij If the difference Δv between the speed of the current point and the speed change between the two nearest points is less than the threshold δ, then the track in the area where the straight-ahead point is located is defined as a straight-ahead track. Figure 4 As shown in the diagram. The track segments formed by other undivided waypoints are defined as other tracks. Finally, the track segmentation diagram is shown in the diagram below. Figure 5 As shown.

[0059] S4. General segmented track feature extraction. For example... Figure 6 As shown, the score (p) of the track points within the segment is calculated. ij Based on the scores, the points are sorted in descending order, and a certain percentage (e.g., 10%-20%) of the total track points within each segment are taken as feature points to form general segmented track features.

[0060] S5. Multi-track feature extraction. For example... Figure 7As shown, the track T in the track set R is selected in turn i , the general segment of the classification of the ship to which the MMSI belongs and the segment track characteristics thereof are selected. According to the feature points in the segment track characteristics, the point in the actual track closest to the feature point is selected as the feature point of the track T i .

[0061] The technical content and technical features of the present application have been disclosed above, however, those skilled in the art can make various substitutions and modifications based on the teachings and disclosures of the present application without departing from the spirit of the present application, therefore, the protection scope of the present application should not be limited to the disclosed content of the embodiments, but should include various substitutions and modifications without departing from the present application, and is covered by the claims of the present application.

Claims

1. A heterogeneous ship multi-track subsection universal feature extraction method, the heterogeneous ship being a plurality of ships of different ship types C and ship lengths L, characterized in that: The method comprises the following steps: S1, determining an AIS data collection sea area S11, taking the departure port as the starting point O and the arrival port as the ending point D, and defining four points P1(x1, y1), P2(x2, y2), P3(x3, y3), and P4(x4, y4) of the sea area boundary by longitude and latitude; S12, according to the longitude and latitude of the starting point, the ending point, and the boundary points, screening a track set satisfying the sea area boundary condition in step S11 from the AIS data: R = {T1, T2,..., T i n n} Where, R represents the track set satisfying the condition, the track T i is composed of a series of AIS track points, and is expressed as follows: T i = {p i1 , p i2 ,..., p ij ,..., p im} wherein p ij is the jth track point in the ith track in the AIS data, expressed in longitude x ij and latitude y ij ; i represents the track number, i = 1, 2, …, n; j represents the track point number on the same track, j = 1, 2, …, m. S2, classifying heterogeneous ships S21, according to the track T i In the AIS database, the identification of each track of the ship water mobile communication service identification code MMSI; S22, according to the MMSI, obtaining the ship type C and the ship length L of each ship; S23, with ship type C and ship length L as attribute values, use a classification algorithm to divide the heterogeneous ships into N classes, denoted as S1, S2, …, SN N ; S3, performing general segmentation on multiple tracks S31, let k = 1 S32, select S k A plurality of ship-like tracks as objects, track T is obtained from AIS data i Mid-course point p ij Corresponding speed v ij And the heading d ij ; S33, set a steering angle threshold value θ, when the track point p ij The difference Δd between the heading changes between the two closest points is greater than the threshold value θ, and the track point p ij is a steering point, and the track state of the area passed through by the steering point is defined as a steering track state; S34, set a threshold of speed δ, in the case of Δd≤θ of the track point p ij , if the difference Δv between the speed of p ij and the speed change between the nearest two points is less than the threshold δ, the track point p ij is a straight point, and the track state of the area passed by the straight point is defined as a straight track state; S34, the region composed of other track points in the state of other tracks is defined as the state of other tracks; S35. The track segment containing the turning track state, the straight track state, and other track states is designated as the Sth segment. k Common sections for this type of vessel; S4, extracting general segmented track features S41, calculate the score of the track point p in the general segment ij by the following formula: Score(p ij ) = InConnection(p ij ) x OutConnection(p ij ) In the formula, InConnection(p ij The value () indicates the connectivity importance of the waypoint within this general segment. It is obtained by dividing the number of wayspoints passing through this waypoint by the total number of wayspoints in this general segment, and is calculated using the following formula: In the formula represents the total number of tracks of the jth track point, wherein N ij , wherein 0 represents that the track point is not passed through, and 1 represents that the track point is passed through, and len(R) represents the total number of tracks of the present universal segment. OutConnection(p ij ) denotes the connection importance of the track point to other generic segments, and its value is obtained by taking the exponential of the result of dividing the number of all tracks passing through other generic segments by the number of tracks passing through only the track point, and the formula is as follows: wherein represents the total number of tracks through other generic segments, represents the total number of tracks through only this waypoint, wherein Q ij and J ij have a value of 0 or 1; S42, according to the descending order of the score, taking a certain proportion of the total number of track points in the general segmented track as the feature points to form the general segmented track features; S43, if k < N, let k = k + 1, and return to step S32; S5, extracting multiple track features S51, select the track T in the track set R in turn i According to the MMSI, select the general segment of the classification of the ship to which the MMSI belongs and the general segment track features of the general segment. S52, according to the feature points in the general segmented track feature, select the track point closest to the feature point in the actual track as the track T i of the feature point; S53, the same type of ship has the same number of feature points extracted from multiple tracks to form the following feature matrix: where T' is the time of the jth track point of the ith track. ij represents the jth track point of the ith track.

2. The heterogeneous ship multi-track segmentation general feature extraction method according to claim 1, characterized in that: In the step S3, θ takes a value of 5-6°; and δ takes a value of 2.5-3.5 km / h.

3. The heterogeneous ship multi-track segmentation general feature extraction method according to claim 1, characterized in that: The classification algorithm in the step S23 comprises the K-means method, the K-Medoids method, and the hierarchical clustering method.

4. The heterogeneous ship multi-track segmentation general feature extraction method according to claim 1, characterized in that: The certain proportion in the step S23 is 10%-20%.

Images