A ship trajectory clustering method based on curve length distance

Abstract

The present invention discloses a ship trajectory clustering method based on curve length distance, which relates to the technical field of ship navigation trajectory analysis, and mainly includes the steps of: obtaining the normalized weighted length of the curve length between the terminal trajectory point in the current trajectory segment and the head trajectory point of its corresponding ship trajectory; according to the proportion of the weighted length of the current terminal trajectory point in the corresponding ship trajectory, obtaining the trajectory points at the same weighted length ratio in other ship trajectories as paired trajectory points by linear interpolation; calculating the comprehensive similarity measure according to the navigation data at the current terminal trajectory point and the corresponding paired trajectory point; obtaining the clustering results by DBSCAN clustering method; and extracting typical ship channels. The present invention utilizes the multidimensional data in the existing ship AIS to mine the physical information contained in the channel, and forms a trajectory cluster by clustering the ship trajectory to characterize the ship behavior characteristics, thereby realizing a trajectory clustering method with small difference interference and low complexity.

Description

Technical Field

[0001] The present invention relates to the technical field of ship navigation track analysis, and in particular to a ship track clustering method based on curve length distance. Background Art

[0002] With the globalization and integration of the economy, water transport has become one of the most important modes of comprehensive transportation systems at home and abroad. With the development of the shipping industry and the improvement of shipbuilding technology, ships are becoming larger, more specialized and faster, and the density of navigation in waters has also increased significantly. The waterway traffic environment has become more complex, which has put forward higher requirements on the navigation management capabilities of water area managers.

[0003] As a new digital navigation aid system and equipment, the Automatic Identification System (AIS) can not only capture the coordinate information of the continuous movement of ships, but also obtain dynamic information such as the speed and heading of ships at the time and space trajectory points by using the multi-dimensional navigation information in AIS data. Therefore, AIS data is very suitable for clustering research on ship trajectories.

[0004] The main purpose of trajectory segment clustering research is to take the entire trajectory or trajectory segment composed of continuous trajectory points as the clustering object. The algorithm divides the entire trajectory into different clusters by establishing a similarity measurement model between trajectory segments, and finally obtains relatively complete channel data through the clustering algorithm. Research on trajectory clustering has achieved certain results (such as DBSCAN density clustering), but the existing techniques for measuring similarity between trajectories have failed to make full use of prior knowledge. The commonly used trajectory similarity measurement methods are Hausdorff and Frechet. However, in practical applications, Hausdorff will select the maximum value of all distances, and outliers in the sequence may have a great impact on the performance of the clustering algorithm. Therefore, for ordered sequences, Hausdorff distance is not a suitable similarity measurement method. On the other hand, although Frechet can avoid the shortcomings of the Hausdorff method and maintain the sequence order of the data on the curve, Frechet has a high complexity in terms of computational time, and is prone to overflow in the algorithm stack or excessive recursive layers. It is not suitable for clustering analysis of long trajectory segments. Moreover, Frechet is very sensitive to noise, and outliers in the data will have a great impact on the clustering results. Summary of the invention

[0005] Based on the above problems, in order to reduce the impact of outliers on the ship trajectory clustering research, reduce the computational complexity and improve the computational efficiency, the present invention proposes a ship trajectory clustering method based on curve length distance. The ship trajectory consists of a number of trajectory segments formed between two trajectory points, including the steps of:

[0006] S1: Obtain each track segment in the ship track set and classify it as an unclassified track segment;

[0007] S2: Obtain the normalized weighted length of the curve between the end track point in the current track segment and the head track point of its corresponding ship track;

[0008] S3: according to the ratio of the weighted length of the current end track point in the corresponding ship track, the track points at the same weighted length ratio in other ship tracks are obtained as paired track points by linear interpolation method;

[0009] S4: Calculate the curve length distance according to the objective function at the current terminal trajectory point and the corresponding paired trajectory point, and calculate the comprehensive similarity measure in combination with the navigation data of the trajectory segment;

[0010] S5: Determine whether the comprehensive similarity measure is greater than a preset measure threshold. If so, obtain the trajectory segment within the preset neighborhood range of the trajectory segment corresponding to the current terminal trajectory point, and classify it as a clustered trajectory segment;

[0011] S6: Determine whether there is a newly added clustered trajectory segment. If so, return to and execute step S2 respectively according to each newly added clustered trajectory segment. If not, obtain the ship trajectory clustering result according to the current clustered trajectory segment.

[0012] Furthermore, the step S1 also includes the following steps:

[0013] S0: Screen and remove the trajectory points in the ship trajectory whose heading change rate or speed change rate is less than the preset change rate.

[0014] Furthermore, the comprehensive similarity metric is obtained by normalizing the curve similarity metric, the speed similarity metric and the heading similarity metric and then adding them according to preset weights.

[0015] Furthermore, in the step S2, the weighted length can be expressed by the following formula:

[0016]

[0017] Where i is the order of the end track point in the current track segment in the corresponding ship track, l i is the weighted length of the trajectory segment at the ith trajectory point, and the coordinates of the ith trajectory point are (ε i , σ i ), k is a constant greater than or equal to 1, ε|max| is the maximum absolute value of the coordinates in the ε coordinate axis of the ship trajectory corresponding to the current trajectory segment, σ |max| It is the maximum absolute value of the coordinate in the σ coordinate axis in the ship trajectory corresponding to the current trajectory segment.

[0018] Furthermore, the curve length distance is calculated according to the objective function at the current end trajectory point and the corresponding paired trajectory point, wherein the objective function can be expressed as the following formula:

[0019]

[0020]

[0021] In the formula, is the total weighted length of trajectory a, is the total weighted length of trajectory b, is the weighted length of the i-th trajectory point on trajectory a, is the weighted length of the paired trajectory point of the ith trajectory point in trajectory a in trajectory b, r i is the value of the objective function between the i-th trajectory point in trajectory a and the paired trajectory point in trajectory b, is the coordinate of the i-th trajectory point in trajectory a, is the coordinate of the paired trajectory point of the i-th trajectory point in trajectory b.

[0022] Furthermore, the acquisition of curve similarity according to the objective function can be expressed as the following formula:

[0023]

[0024]

[0025] Where N is the total number of trajectory points on the trajectory, is the one-way curve similarity measure from trajectory a to trajectory b, is the one-way curve similarity measure from trajectory b to trajectory a, d l (a,b) is the curve similarity measure between trajectory a and trajectory b in both directions.

[0026] Furthermore, the speed similarity metric can be expressed as follows:

[0027]

[0028] Where, d v is the speed similarity measure, v max (a,b) is the maximum speed difference between trajectory a and trajectory b, v min (a,b) is the minimum speed difference between trajectory a and trajectory b, v avg(a,b) is the average speed difference between trajectory a and trajectory b, v med (a, b) is the median speed difference between trajectory a and trajectory b.

[0029] Furthermore, the heading similarity metric can be expressed as the following formula:

[0030]

[0031] Where, d c is the heading similarity measure, c max (a,b) is the maximum heading difference between trajectory a and trajectory b, c min (a,b) is the minimum heading difference between trajectory a and trajectory b, c avg (a,b) is the average heading difference between trajectory a and trajectory b, c med (a, b) is the median heading difference between trajectory a and trajectory b.

[0032] Furthermore, after the step S6, the step of:

[0033] S7: extracting the cluster center track from the ship track clustering result, and using the cluster center track as an anchor point to extract the channel range of the fan-shaped scanning area perpendicular to the heading;

[0034] S8: Output the extraction results and use them as typical shipping routes.

[0035] Compared with the prior art, the present invention has at least the following beneficial effects:

[0036] (1) The ship trajectory clustering method based on curve length distance described in the present invention utilizes the multidimensional data in the existing ship AIS to mine the physical information contained in the waterway, and clusters the ship trajectories to form trajectory clusters to characterize the ship behavior characteristics. Based on this, a trajectory clustering method with small difference interference and low complexity can be realized, which can be better applied to regional ship violation trajectory monitoring;

[0037] (2) By screening and eliminating trajectory points based on heading and speed change rate, the track data can be compressed while retaining the original trajectory characteristics as much as possible, achieving a balance between integrity and simplicity;

[0038] (3) In addition to the curve similarity measurement of the trajectory, the impact of speed difference and heading difference on trajectory classification is also comprehensively considered, making the classification result more scientific and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A method step diagram of a ship trajectory clustering method based on curve length distance;

[0040] Figure 2 It is a schematic diagram of trajectory compression and division;

[0041] Figure 3 Schematic diagram of the working principle of curve similarity measurement;

[0042] Figure 4 This is a schematic diagram of DBSCAN clustering;

[0043] Figure 5 This is a schematic diagram of typical waterway extraction. DETAILED DESCRIPTION

[0044] The following are specific embodiments of the present invention and the accompanying drawings to further describe the technical solution of the present invention, but the present invention is not limited to these embodiments.

[0045] Embodiment 1

[0046] In order to solve the limitations of the existing trajectory similarity measurement methods, the present invention uses historical AIS data (including ship identification number, ship location, data upload time, heading, speed data) to obtain comprehensive similarity measurement data between trajectories by combining curve similarity measurement, heading similarity measurement and speed similarity measurement in both directions of the ship. Figure 1 As shown, a ship trajectory clustering method based on curve length distance includes the steps of:

[0047] S1: Obtain each track segment in the ship track set and classify it as an unclassified track segment;

[0048] S2: Obtain the normalized weighted length of the curve between the end track point in the current track segment and the head track point of its corresponding ship track;

[0049] S3: according to the ratio of the weighted length of the current end track point in the corresponding ship track, the track points at the same weighted length ratio in other ship tracks are obtained as paired track points by linear interpolation method;

[0050] S4: Calculate the curve length distance according to the objective function at the current terminal trajectory point and the corresponding paired trajectory point, and calculate the comprehensive similarity measure in combination with the navigation data of the trajectory segment;

[0051] S5: Determine whether the comprehensive similarity measure is greater than a preset measure threshold. If so, obtain the trajectory segment within the preset neighborhood range of the trajectory segment corresponding to the current terminal trajectory point, and classify it as a clustered trajectory segment;

[0052] S6: Determine whether there is a newly added clustered trajectory segment. If so, return to and execute step S2 respectively according to each newly added clustered trajectory segment. If not, obtain the ship trajectory clustering result according to the current clustered trajectory segment.

[0053] It should be noted that the trajectory segment described here is not composed of any two trajectory points in the ship trajectory, but is composed of two (or more) consecutive trajectory points. At the same time, if step S5 determines at the beginning that the similarity metric is less than the preset metric threshold, it is determined that there is no density-reachable object set in the trajectory.

[0054] In this example, considering that the Yangtze River estuary is one of the areas with the highest shipping density in my country, a bifurcated route at the Yangtze River estuary was selected as the research object. Some high-quality AIS data from July 31, 2017 to August 1, 2019 were selected to perform cluster analysis on ship trajectories.

[0055] Before starting the analysis, we must first delete some AIS data that exceeds reasonable values ​​(such as: the speed is negative or exceeds the maximum speed of the ship, the heading value is not within the range of 0-360 degrees, the longitude and latitude exceed reasonable values, etc.). Given that the frequency of uploading AIS data is too high, a lot of computing resources will be wasted in the process of trajectory clustering, and it needs to be compressed and divided. Trajectory compression and division need to ensure the feature integrity and simplicity of AIS data. Trajectory compression and division are as follows: Figure 2 As shown, P 1 -P 9 is the total amount of AIS data points of a track segment. If the original track P is retained 1 -P 9 As a training object, although all the features of the original trajectory are retained, it will consume a lot of computing resources. 1 -P 4 -P 7 -P 9 As training objects, although the computational complexity of trajectory clustering is reduced, too many features of the original trajectory are lost. Based on this, the trajectory compression and division method proposed in the present invention is: first, find trajectory points whose heading change rate and speed change rate are less than the preset change rate; then, remove these trajectory points. In this way, data compression can be achieved while retaining the original trajectory features as much as possible, achieving a balance between feature integrity and simplicity.

[0056] For the trajectory points obtained after the above screening, the present invention proposes a curve similarity measurement method combining speed and heading. The working principle of the curve similarity measurement is as follows: Figure 3As shown in the figure, starting from the first data point of track a, there is a certain curve length between each point on the track and the first data point. After the curve length of the point is normalized by the coordinate axis, the weighted length of the point is obtained. The weighted length is used to calculate the proportion of each track point in the entire track. While maintaining the same proportion, linear interpolation can be used on another track b to find points with the same weighted length ratio, that is, paired track points. Then, the objective function is established according to the weighted length of each track point in the current ship track segment to quantify the distance between the track point and the paired track point, and the similarity measure between the two tracks is quantified by calculating the mean of these objective functions.

[0057] The following is an explanation of how to obtain the paired trajectory points using a formula.

[0058] In AIS data, the track consists of a set of ordered points, so there is no specific curve analytical description. In this case, the curve length of the track can be approximately expressed as:

[0059]

[0060] In general, if the x-axis and y-axis have different units and sizes, in order to prevent the larger unit from monopolizing the curve length, the curve length of the trajectory point needs to be normalized along the x-axis and y-axis. Here, it is assumed that there are N points on the trajectory, and the coordinates of the i-th point are (ε i , σ i ), the normalized curve length is defined as the weighted length, then the weighted lengths of each point in the trajectory are:

[0061]

[0062] Where i is the order of the end track point in the current track segment in the corresponding ship track, l i is the weighted length of the trajectory segment at the i-th trajectory point, k is a constant greater than or equal to 1, and ε |max| is the maximum absolute value of the coordinates in the ε coordinate axis of the ship trajectory corresponding to the current trajectory segment, σ |max| It is the maximum absolute value of the coordinate in the σ coordinate axis in the ship trajectory corresponding to the current trajectory segment.

[0063] At this time, suppose there are two trajectories a (the trajectory of the current ship) and b (the trajectory of other ships) with N trajectory points, and their total weighted lengths are and The weighted length of the i-th trajectory point on trajectory a is Then the weighted length of the corresponding position (paired trajectory point) of the i-th trajectory point on trajectory b should be:

[0064]

[0065] According to the weighted length of the i-th point on trajectory a on trajectory b, the coordinates of the paired trajectory points with the corresponding weighted length ratio of the i-th point on trajectory a on trajectory b can be obtained using one-dimensional linear interpolation.

[0066] Assume that the coordinates of the i-th trajectory point on trajectory a are Then the coordinates of the paired trajectory point of this trajectory point on trajectory b are The objective function between this set of corresponding trajectory points is:

[0067]

[0068] In the formula, r i is the value of the objective function between the i-th trajectory point in trajectory a and the paired trajectory point in trajectory b. After obtaining the objective function of each trajectory point, the values ​​of these objective functions are accumulated on trajectory a in turn, the average is calculated and the square root is taken to obtain the one-way curve similarity measure from trajectory a to trajectory b. Expressed as a formula:

[0069]

[0070] Then, the final bidirectional curve distance between trajectory a and trajectory b can be expressed as the following formula:

[0071]

[0072] In the formula, is the one-way curve similarity measure from trajectory b to trajectory a, d l (a,b) is the curve similarity measure between trajectory a and trajectory b in both directions.

[0073] As for the similarity measurement of speed, the present invention comprehensively considers the similarity measurement of speed from four aspects: maximum speed, minimum speed, average speed and median speed, namely:

[0074]

[0075] Where, d v is the speed similarity measure, v max (a,b) is the maximum speed difference between trajectory a and trajectory b, v min (a,b) is the minimum speed difference between trajectory a and trajectory b, v avg (a,b) is the average speed difference between trajectory a and trajectory b, v med (a, b) is the median speed difference between trajectory a and trajectory b.

[0076] Similarly, for the similarity measurement of headings, the present invention also comprehensively considers the heading similarity measurement from four aspects: maximum heading, minimum heading, average heading and median heading, that is:

[0077]

[0078] Where, d c is the heading similarity measure, c max (a,b) is the maximum heading difference between trajectory a and trajectory b, c min (a,b) is the minimum heading difference between trajectory a and trajectory b, c avg (a,b) is the average heading difference between trajectory a and trajectory b, c med (a, b) is the median heading difference between trajectory a and trajectory b.

[0079] After calculating the three similarity metrics, different weights need to be assigned to the three metrics to combine the three metrics. It should be noted that the weight assignment must satisfy the requirement that all weights are greater than 0 and less than 1, and the sum of the three is equal to 1.

[0080] At the same time, since the dimensions of these three metrics are different, they need to be normalized. The normalization formula is as follows:

[0081]

[0082] Among them, d is the metric before normalization in this type of metric, d min and d max is the minimum and maximum value of this type of metric before normalization, and d' is the metric after normalization. By assigning the normalized metric corresponding weights and summing them up, we can get the comprehensive similarity metric between trajectories, which is expressed as follows:

[0083] D dist (a, b) = w l d l ′+w v d v ′+w c d c '

[0084] Where D dist (a, b) is the comprehensive similarity measure between trajectory a and trajectory b, w l is the weight of the curve similarity measure, w v is the weight of the speed similarity measurement, w c is the weight of the heading similarity measure; d l ′ is the normalized curve similarity measure, d v ′ is the normalized speed similarity measure, dc ′ is the normalized heading similarity measure.

[0085] After obtaining the comprehensive similarity measure, the present invention uses DBSCAN clustering to perform clustering, such as Figure 4 As shown, that is, first determine whether the comprehensive similarity measure between the trajectory segments is greater than the preset measure threshold (regarding the setting of the preset measure threshold, due to the requirement of the DBSCAN algorithm for parameters, the parameter is set by subjective experience, and then the parameter is adjusted according to the specific clustering effect), that is, within the range formed by the preset measure threshold (β, which can be regarded as the distance range) around the trajectory point (black dot in the figure) (because the dimensions of different dimensions are different when the high-dimensional data is displayed as data in a two-dimensional sample space, the range circle formed is actually an irregular circle), whether the number of trajectory points other than the trajectory point is greater than the preset density threshold (the preset number of trajectory points in the circle), if so, the trajectory segment corresponding to the trajectory point in the circle is classified as a cluster trajectory segment. Then, according to the newly added cluster trajectory segment, return to step S2, and perform cluster analysis on the corresponding trajectory of the newly added trajectory segment, so as to continuously expand the number of cluster trajectory segments until the cluster trajectory segments no longer increase. After the cluster trajectory segments no longer increase, they can be output as clustering results.

[0086] After clustering the trajectory segments by the above method, different trajectory segments are divided into different clusters, but too many trajectory segments cannot quantitatively describe the overall motion law of the cluster. The typical trajectory of a ship is the best expression of the navigation behavior of the ship in the same cluster. Therefore, the present invention also proposes the following steps to further optimize the clustering results:

[0087] S7: extracting the cluster center track from the ship track clustering result, and using the cluster center track as an anchor point to extract the channel range of the fan-shaped scanning area perpendicular to the heading;

[0088] S8: Output the extraction results and use them as typical shipping routes.

[0089] like Figure 5 As shown, the cluster center trajectory is used as the anchor point (Pi) and a fan-shaped scanning area perpendicular to the heading is used to extract the internal features of the cluster results, which makes the internal features of the cluster results contained in the typical channel more complete and accurate, and prevents the data in the heading direction from affecting the typical channel.

[0090] In summary, the ship trajectory clustering method based on curve length distance described in the present invention utilizes the multidimensional data in the existing ship AIS to mine the physical information contained in the waterway, and clusters the ship trajectories to form trajectory clusters to characterize the ship behavior characteristics. Based on this, a trajectory clustering method with small difference interference and low complexity can be implemented, which can be better applied to regional ship violation trajectory monitoring.

[0091] By screening and eliminating trajectory points based on heading and speed change rates, the track data can be compressed while retaining the original track features as much as possible, achieving a balance between integrity and simplicity. In addition to the curve similarity measurement of the track, the impact of speed differences and heading differences on track classification is also comprehensively considered, making the classification results more scientific and reliable.

[0092] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0093] In addition, in the present invention, descriptions such as "first", "second", "one", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0094] In the present invention, unless otherwise clearly specified and limited, the terms "connection", "fixation", etc. should be understood in a broad sense. For example, "fixation" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0095] In addition, the technical solutions between the various embodiments of the present invention can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

Claims

1. A ship trajectory clustering method based on curve length distance. The ship trajectory consists of several trajectory segments between two trajectory points. It is characterized in that Includes steps: S1: Obtain each track segment in the ship track set and classify it as an unclassified track segment; S2: Obtain the normalized weighted length of the curve between the end track point in the current track segment and the head track point of its corresponding ship track; S3: according to the ratio of the weighted length of the current end track point in the corresponding ship track, the track points at the same weighted length ratio in other ship tracks are obtained as paired track points by linear interpolation method; S4: Calculate the curve length distance according to the objective function at the current terminal trajectory point and the corresponding paired trajectory point, and calculate the comprehensive similarity measure in combination with the navigation data of the trajectory segment; S5: Determine whether the comprehensive similarity measure is greater than a preset measure threshold. If so, obtain the trajectory segment within the preset neighborhood range of the trajectory segment corresponding to the current terminal trajectory point, and classify it as a clustered trajectory segment; S6: Determine whether there is a newly added clustered trajectory segment. If so, return to and execute step S2 respectively according to each newly added clustered trajectory segment. If not, obtain the ship trajectory clustering result according to the current clustered trajectory segment. The comprehensive similarity metric is obtained by normalizing the curve similarity metric, the speed similarity metric and the heading similarity metric and then adding them according to a preset weight; In S2, the weighted length can be expressed by the following formula: Where i is the order of the end track point in the current track segment in the corresponding ship track, l i is the weighted length of the trajectory segment at the ith trajectory point, and the coordinates of the ith trajectory point are (ε i , σ i ), k is a constant greater than or equal to 1, ε |max| is the maximum absolute value of the coordinates in the ε coordinate axis of the ship trajectory corresponding to the current trajectory segment, σ |max| is the maximum absolute value of the coordinates in the σ coordinate axis in the ship trajectory corresponding to the current trajectory segment; The curve length distance is calculated according to the objective function at the current end trajectory point and the corresponding paired trajectory point, wherein the objective function is expressed as the following formula: In the formula, is the total weighted length of trajectory a, is the total weighted length of trajectory b, is the weighted length of the i-th trajectory point on trajectory a, is the weighted length of the paired trajectory point of the ith trajectory point in trajectory a in trajectory b, r i is the value of the objective function between the i-th trajectory point in trajectory a and the paired trajectory point in trajectory b, is the coordinate of the i-th trajectory point in trajectory a, is the coordinate of the paired trajectory point of the i-th trajectory point in trajectory a in trajectory b; The objective function is used to obtain the curve similarity, which can be expressed as the following formula: Where N is the total number of trajectory points on the trajectory, is the one-way curve similarity measure from trajectory a to trajectory b, is the one-way curve similarity measure from trajectory b to trajectory a, d l (a,b) is the curve similarity measure between trajectory a and trajectory b in both directions.

2. A ship trajectory clustering method based on curve length distance as claimed in claim 1, It is characterized in that The step S1 also includes the following steps: S0: Screen and remove the trajectory points in the ship trajectory whose heading change rate or speed change rate is less than the preset change rate.

3. A ship trajectory clustering method based on curve length distance as claimed in claim 2, It is characterized in that The speed similarity measure can be expressed as the following formula: Where, d v is the speed similarity measure, v max (a,b) is the maximum speed difference between trajectory a and trajectory b, v min (a,b) is the minimum speed difference between trajectory a and trajectory b, v avg (a,b) is the average speed difference between trajectory a and trajectory b, v med (a, b) is the median speed difference between trajectory a and trajectory b.

4. A ship trajectory clustering method based on curve length distance as claimed in claim 3, It is characterized in that The heading similarity metric can be expressed as follows: Where, d c is the heading similarity measure, c max (a,b) is the maximum heading difference between trajectory a and trajectory b, c min (a,b) is the minimum heading difference between trajectory a and trajectory b, c avg (a,b) is the average heading difference between trajectory a and trajectory b, c med (a, b) is the median heading difference between trajectory a and trajectory b.

5. A ship trajectory clustering method based on curve length distance as claimed in claim 1, It is characterized in that After S6, the step further includes: S7: extracting the cluster center track from the ship track clustering result, and using the cluster center track as an anchor point to extract the channel range of the fan-shaped scanning area perpendicular to the heading; S8: Output the extraction results and use them as typical shipping routes.

Images