Yangtze river channel classification method based on ship feature clustering

Abstract

This invention discloses a Yangtze River waterway classification method based on ship feature clustering, comprising: dividing the Yangtze River waterway into multiple segments using an angle bisector method; assigning geographical semantic knowledge such as ports, prefecture-level cities, and provincial-level administrative regions to each segment using overlay analysis; calculating the navigation time of each type of ship within a square grid by dividing the grid, and calculating static traffic characteristics using a summation-based waterway mapping method; calculating relatively discrete traffic characteristics such as speed, acceleration, and turning rate using the mean method, and mapping them to waterway segments using the maximum value method; calculating flow characteristics using a summation method; selecting traffic characteristics with appropriate similarity based on hierarchical clustering analysis; and classifying the Yangtze River waterway using a combination of K-means clustering and sliding window methods. This invention can automatically classify waterway segments in the Yangtze River waterway based on ship traffic characteristics.

Description

Technical Field

[0001] This invention relates to the field of inland waterway transportation classification technology, and in particular to a Yangtze River waterway classification method based on ship feature clustering. Background Technology

[0002] The Yangtze River is a vital waterway connecting inland cities with major export hubs on the eastern coast of China. Therefore, in-depth research into the traffic characteristics and navigation patterns of vessels within the waterway is crucial for strengthening shipping traffic management, optimizing waterway planning, and improving shipping efficiency. The Automatic Identification System (AIS) collects a large amount of near real-time vessel behavior data, providing solid data support for analyzing the waterway's vessel traffic characteristics and thus offering a valid basis for the classification and grading of the Yangtze River waterway.

[0003] Chinese patent application CN119479369B discloses a method and apparatus for identifying waterway congestion using AIS big data. This method and apparatus can reflect the congestion situation during waterway passage by calculating accurate waterway congestion evaluation indicators, thereby improving navigation efficiency. However, this invention only considers the waiting time of ships within the waterway, ignoring the ship's own attributes and the geographical location characteristics of the waterway.

[0004] Inland waterway traffic research primarily focuses on vessel traffic flow within the waterways, starting from the spatiotemporal characteristics of vessel flow routes to grasp the overall traffic dynamics of inland waterways. Chinese patent application CN119849980A discloses a method for assessing the traffic status of inland waterway networks based on complex networks. This method utilizes the Space L method and graph theory to construct an inland waterway topology network and calculate network characteristic indicators. It selects an appropriate assessment model to quantify the weights of nodes and segments within the waterway network, thereby calculating the vessel capacity of nodes and segments within the network, assessing the traffic status of each node and segment, and finally evaluating the overall traffic status of the waterway network based on the constructed hierarchical calculation model. This invention, starting from complex network theory, comprehensively considers the importance of network structure and achieves a macroscopic understanding of the traffic status of the entire waterway network.

[0005] Due to the limitations imposed by the geographical conditions of the waterways, the traffic flow along inland waterways is relatively simple, mainly proceeding along the inland waterways. However, compared to open waterways, the navigation environment of inland waterways presents a higher degree of complexity.

[0006] Currently, research on waterway transportation mainly focuses on maritime shipping or specific regions, with a particular emphasis on vessel maneuverability. Overall research on inland waterways is relatively limited, especially regarding the systematic exploration of the correlation between the static and dynamic characteristics of vessels and their environmental features within inland waterways. As crucial hubs connecting inland areas and vital channels between inland and sea, inland waterways urgently require comprehensive transportation classification and analysis to optimize the allocation and utilization of waterway resources. Summary of the Invention

[0007] Purpose of the invention: The purpose of this invention is to provide a Yangtze River waterway classification method based on ship feature clustering. By applying the static and dynamic characteristics of ships, the method automatically classifies and grades the Yangtze River waterway sections according to the clustering method, thereby realizing the spatial distribution of the overall waterway transportation characteristics.

[0008] Technical solution: A Yangtze River waterway classification method based on ship feature clustering, comprising the following steps:

[0009] S1, Preprocess AIS data;

[0010] S2 divides the Yangtze River waterway into sections and assigns geographical semantic knowledge to each section;

[0011] S3 calculates the vessel traffic characteristics of the Yangtze River waterway, specifically including:

[0012] S31, the Yangtze River waterway is divided into multiple square grids, the navigation time of each type of vessel within the square grid is calculated, and the static characteristics of vessels in the waterway are calculated using the summation waterway mapping method;

[0013] S32, for traffic characteristics such as speed, acceleration, heading, and turning rate, the mean value of each grid is calculated, and the segment characteristic value is calculated using the channel mapping method with the maximum value; the flow characteristics are calculated using the summation method.

[0014] S4 employs a segment classification method combining K-means clustering and a sliding window approach to classify segments and output channel classification results; specifically including:

[0015] S41, Select appropriate ship traffic features with a similarity of less than 0.8;

[0016] S42. Based on the characteristics of ship traffic within the waterway segment, the waterway segment is classified using a combination of K-means clustering and sliding window, and the final waterway classification result is output.

[0017] Furthermore, the preprocessing steps for AIS data are as follows:

[0018] S11, remove data with a speed less than 2 sections or fewer than 100 trajectory records to obtain AIS data points. :

[0019] ,

[0020] in, For the number The One AIS data point; , , They are respectively The time, longitude, and latitude of the point; , , , These are the ship's type, dimensions, speed, and course.

[0021] S12, For missing ship size and type data, use the modulo value of the existing static data of the ship to replace the missing items;

[0022] Step 13 defines the process of a ship traveling from its starting point to its destination as a journey. Linear interpolation is used to interpolate AIS data points for the same trip, with an interpolation interval of 1 minute. Then, the... AIS data points for:

[0023] ,

[0024] in, for The current time value, for and The distance between two points; , These are the starting and ending points of each journey.

[0025] Furthermore, in step S2, the angle bisector method is used to divide the waterway into n consecutive segments. The dividing line is set as the angle bisector that passes through the inflection point of the waterway and coincides with the centerline. This ensures that the distance of each flight segment is the same;

[0026] The overlay analysis method is used to spatially overlay the spatial locations of each waterway segment with its surrounding ports, prefecture-level cities and provincial-level administrative regions, thereby endowing each waterway segment with geographical semantic knowledge.

[0027] Furthermore, the specific implementation process for calculating the static characteristics of vessels in the waterway is as follows:

[0028] The extracted Yangtze River channel was divided into multiple 200m square grids, representing each segment of the Yangtze River channel. It contains multiple square grids:

[0029] ,

[0030] in, This represents the m-th square grid; i = 1, 2, 3, ..., n;

[0031] The sailing time for each type of ship within the m-th square grid is calculated using the following expression:

[0032] ,

[0033] Where N represents the total number of AIS data points for a specific ship type within the square grid. ;

[0034] The navigation time for a certain type of vessel within the waterway segment The calculation expression is as follows:

[0035] ,

[0036] in, For the segment The total number of square grids within.

[0037] Furthermore, the dynamic traffic characteristics of speed, acceleration, heading, and turning rate are calculated using the averaging method as follows: the characteristic value of a certain square grid is the average value of the characteristic of all AIS data points within that grid, and the maximum value of the characteristic value of all square grids within a flight segment is selected as the dynamic traffic characteristic of that flight segment.

[0038] The vector method is used to count the number of ships entering and leaving each segment, and the traffic flow of each segment is calculated. This yields the total number of ships passing through each segment.

[0039] Furthermore, in step S41, the similarity between traffic features of each segment is calculated, and hierarchical clustering analysis is performed; suitable ship traffic features with a similarity of less than 0.8 are selected as the data basis for segment classification.

[0040] Furthermore, in step S42, the steps for obtaining the waterway classification are as follows:

[0041] Based on the vessel traffic characteristics of each segment, input the sliding window size and step size to traverse all segments;

[0042] Calculate the mean and standard deviation of the characteristic values ​​for all segments in each sliding window;

[0043] The K-means clustering algorithm is used to cluster the mean and standard deviation of all windows to identify the category of each window;

[0044] Merge adjacent and identical category segments to ensure that segments of each category do not overlap;

[0045] Output the waterway classification results, which divide the entire waterway into several non-overlapping cluster segments.

[0046] Furthermore, the flight segment classification method combining K-means clustering and sliding window includes the following steps:

[0047] S421, according to each waterway With all traffic features within the flight segment Enter the size of the sliding window. Step length and number of clusters ,in,

[0048] ,

[0049] ,

[0050] , , , Each segment The characteristics of the flight duration, flight speed, flight acceleration, and traffic flow;

[0051] S422, based on the size and step size of the sliding window, traverse all sections of the Yangtze River waterway and calculate the mean of each traffic characteristic for all windows. and standard deviation Then, the feature sequences obtained from all sliding windows... Data standardization is performed to ensure that each feature dimension has zero mean and unit variance; then, the standardized feature sequences are clustered using the K-means clustering algorithm, with a predetermined number of clusters. Assign a cluster label to each window ;

[0052] S423, iterate through the clustering results of all windows in segment order, when two adjacent windows belong to the same cluster label, and the previous window's end position... Starting position of the next window If there are overlapping or adjacent relationships, the ending position of the current cluster segment will be adjusted to... ; The clustered segments are obtained.

[0053] Compared with the prior art, the significant advantages of this invention are as follows:

[0054] 1. This invention uses angle bisectors to divide the Yangtze River waterway into multiple 2-kilometer segments, preparing for subsequent segment classification. Unlike traditional equidistant divisions, this invention is applicable not only to horizontal waterways but also to winding and extended waterways.

[0055] 2. This invention takes a specific section of the Yangtze River waterway as the research object to conduct overall waterway traffic analysis. Unlike traditional research methods that focus on ship positions and trajectories, this invention mainly focuses on specific sections of the waterway, reflecting the characteristics of the overall waterway through the spatial distribution of their traffic state features.

[0056] 3. This invention adopts a waterway segment classification method that combines K-means clustering and sliding window, which can flexibly classify and grade the Yangtze River waterway according to different sliding window sizes, step sizes and number of clusters, and obtain different classification results, thereby grasping the spatial distribution of the overall waterway transportation characteristics.

[0057] 4. This invention is not limited to a single traffic feature, but rather considers multiple dimensions, including not only static attribute information such as ship type, but also dynamic traffic feature information such as ship speed. In addition to classifying a certain type of traffic feature, it can also combine static and dynamic traffic feature information for comprehensive classification, enabling systematic classification and analysis of transportation in the Yangtze River waterway. This allows for automatic classification of the Yangtze River waterway from a macro perspective, providing data support for the daily management, ship scheduling, and safety monitoring of the Yangtze River waterway, and also providing a scientific basis for the future development planning of the Yangtze River waterway. Attached Figure Description

[0058] Figure 1 This is a flowchart of the Yangtze River waterway classification method based on ship feature clustering, as described in this invention.

[0059] Figure 2 This is a schematic diagram illustrating the irregular division of waterways into segments;

[0060] Figure 3 This is a schematic diagram of the spatial location relationships of flight segments;

[0061] Figure 4 It is a schematic diagram of ship traffic characteristics mapped onto a waterway segment;

[0062] Figure 5 This is a schematic diagram of the classification of Yangtze River waterway sections;

[0063] Figure 6 This is a schematic diagram showing the comprehensive transportation characteristics of the Yangtze River waterway section;

[0064] Figure 7 This is a map showing the 2023 classification results of Yangtze River waterways based on vessel size characteristics.

[0065] Figure 8 This is a map showing the results of the 2023 Yangtze River waterway segment classification based on the navigation and maneuvering characteristics of ships. Detailed Implementation

[0066] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0067] See Figure 1 A Yangtze River waterway classification method based on ship feature clustering specifically includes the following steps:

[0068] Step 1, AIS data processing;

[0069] Although AIS data contains a wealth of vessel navigation information, including static attributes such as vessel number, type, and length, as well as dynamic characteristics such as spatial location, speed, and status, it still requires a series of data processing steps before it can be directly applied to waterway traffic analysis. These steps include data cleaning, filtering, attribute completion, trip segmentation, and data interpolation.

[0070] Step 11, data cleaning and filtering;

[0071] Since this example primarily focuses on vessels navigating within a complete waterway, data with speeds less than 2 knots or fewer than 100 track records needs to be removed to ensure the selected vessels' track data is relatively complete. At this point, the AIS data points... Although it already contains high-quality ship characteristic information, further improvements in data quality and data mining are still needed.

[0072] (1)

[0073] in, For the number The One AIS data point; , , They are respectively The time, longitude, and latitude of the point; , , , These are the ship's type, size, current speed, and course.

[0074] Step 12, complete the attribute data;

[0075] In addition to the traditional focus on the dynamic characteristics of ships, this example will also focus on their static attribute characteristics. Therefore, it is particularly important to improve and supplement the static attribute information in AIS data. Specifically, for missing ship size and type data, these gaps are filled by supplementing other AIS data with the same MMSI (Maritime Mobile Service Identity). The specific approach is to use the modulo value of the existing static data of the ship to replace the missing items.

[0076] Step 13, route division and data interpolation;

[0077] For subsequent waterway studies, this example requires interpolating the navigation data of each vessel at 1-minute intervals. This interpolation assumes that the vessel's journey has already been divided into segments. The process of a vessel traveling from its starting point to its destination is defined as one journey. Only after the itinerary is divided can interpolation processing be performed on the navigation data for each itinerary. The key to dividing the itinerary lies in identifying the starting point of each itinerary. and the end point After the journey is divided, the study will use linear interpolation to interpolate AIS data points for the same journey, providing data support for the subsequent calculation of ship sailing time.

[0078] (2)

[0079] in, for The current time value, for and The distance between two points.

[0080] Step 2, segmentation of the Yangtze River waterway;

[0081] Due to the winding and elongated nature of the Yangtze River channel, and the varying traffic characteristics of different sections, this embodiment first divides the Yangtze River channel into segments to prepare for subsequent segment classification. To achieve nearly equal mileage divisions of the Yangtze River channel, this invention proposes a channel area division method based on angle bisectors. Furthermore, since only the main channel of the Yangtze River is continuously navigable, the segmentation of the Yangtze River channel in this embodiment only applies to the main channel and does not consider the tributaries. Figure 3 This illustrates the research area of ​​this embodiment.

[0082] Step 21, division of Yangtze River waterway sections;

[0083] First, the centerline of the Yangtze River channel was extracted, and channel turning points were generated at every 2km interval. The channel centerline was subsequently generated a second time. Based on the inflection points in the channel centerline and the secondary generated centerline, the angle bisector method is used to divide the entire Yangtze River channel into n consecutive segments. Each segment is approximately 2km long, such as Figure 2 As shown. The angle bisector division method sets the dividing line as the angle bisector that passes through the channel inflection point and coincides with the centerline, satisfying the following formula:

[0084] (3)

[0085] (4)

[0086] (5)

[0087] in, The connected lines represent the turning points of the waterway. The dividing line is generated by the angle bisector method; and , These are the intersection points generated by the dividing line and the waterway boundary.

[0088] Step 22, Geographic semantic association of flight segments;

[0089] Although the waterway has been divided into multiple 2km segments and the mileage position of each segment within the overall waterway is clear, the geographic semantic knowledge of each segment is still lacking. Therefore, this embodiment will further correlate the spatial locations of each segment to determine the provincial-level administrative region, prefecture-level city, and surrounding port locations of each segment. An overlay analysis method is used to spatially overlay the spatial locations of each segment with those of ports, prefecture-level cities, and provincial-level administrative regions, thereby assigning geographic semantic knowledge to each segment, such as... Figure 3 As shown.

[0090] Step 3, Calculation of waterway vessel traffic characteristics;

[0091] This embodiment employs different methods to calculate the characteristics of vessel traffic in the waterway. Static attributes such as vessel type and vessel size are calculated using static feature calculation methods, while dynamic attributes such as flow rate, speed, acceleration, heading, and turning rate are calculated using dynamic feature calculation methods.

[0092] Step 31, Calculation of static characteristics of vessels in the waterway;

[0093] Based on the ship's sailing time, the static characteristics of ships in the waterway are calculated using a grid method. The specific calculation process is as follows:

[0094] Step 311, Channel Grid Division: Divide the extracted Yangtze River channel into multiple 200m square grids, representing each segment of the Yangtze River channel. It contains multiple square grids.

[0095] (6)

[0096] in, Let i represent the m-th square grid; i = 1, 2, 3, ..., n.

[0097] Step 312, Calculation of ship sailing time: First, calculate the sailing time of each type of ship within the m-th square grid. The percentage of this type of vessel in the total vessel duration is the percentage of that vessel type. This is based on the processed AIS data for each vessel during each journey. The interval is 1 minute, therefore, the total number of AIS data for a certain ship type within each square grid is... This refers to the sailing time of this type of vessel within the m-th square grid area. .

[0098] (7)

[0099] Where N represents the total number of AIS data for a certain ship type within the square grid.

[0100] Step 313, Static traffic feature mapping of waterway segments: Since this embodiment uses 2km segments within the Yangtze River waterway as the basis for mapping, Since this is a research unit, appropriate methods are needed to assign the information values ​​of all square grids within each small segment to that segment. Various mapping methods can be used, such as taking the average, mode, or median, etc. Figure 4 As shown. For the static attribute information of ships, the summation method is selected, that is, the navigation time of a certain ship type within this waterway segment. The total value for all its grids.

[0101] (8)

[0102] in, For the segment The total number of square grids within.

[0103] Step 32, Calculation of dynamic characteristics of vessels in the waterway;

[0104] For the calculation of the dynamic characteristics of vessels in the waterway, the relatively discrete traffic characteristics such as speed, acceleration, heading, and turning rate are also calculated using a grid-based method. First, the extracted Yangtze River waterway is divided into multiple 200m square grids. However, the specific calculations differ slightly from those for static features. Taking velocity features as an example, the specific calculation process is as follows:

[0105] Step 321, Dynamic Feature Value Calculation: For the dynamic features of the ship, the average method is used to represent the dynamic feature value of the square grid region. Taking the ship's speed feature as an example, the ship's speed feature of the grid... This is the average velocity value of all AIS data points within the square grid.

[0106] (9)

[0107] Step 322, Dynamic traffic feature mapping of the waterway segment: Considering the large differences in the dynamic features of ships within the waterway, the maximum value of all square grid feature values ​​within the waterway segment is selected as the dynamic feature of the waterway segment. This can effectively avoid the interference of the dynamic features of the shore area on the overall result.

[0108] (10)

[0109] in, Indicates each flight segment The characteristics of ship speed.

[0110] Step 323: However, considering the characteristic of traffic flow having flow attributes, a vector method is used to specifically calculate the traffic flow for each flight segment.

[0111] (11)

[0112] in, This indicates the traffic flow for each flight segment; This indicates the ship's journey within a given segment of the voyage.

[0113] Step 4: Classification of Yangtze River waterway sections;

[0114] Step 41, Selection of vessel traffic characteristics in the Yangtze River waterway;

[0115] Choose different ship traffic characteristics, such as Figure 5 As shown, the classification results for the Yangtze River waterway differ. Classifying by different vessel sizes allows for the categorization of vessel size distribution along the Yangtze River waterway, thus revealing the vessel transport capacity of different sections. Selecting too many similar features may negatively impact the overall classification results; therefore, this embodiment first performs hierarchical clustering analysis on each traffic feature. The specific process involves: first, calculating the similarity of each traffic feature; then, performing hierarchical clustering; and finally, avoiding the simultaneous selection of traffic features with excessively high similarity. Specifically, traffic features with similarity values ​​higher than 0.8 are considered too similar and should be avoided to prevent affecting the waterway classification results.

[0116] Step 42, segment classification combining K-means clustering and sliding window;

[0117] First, a sliding window method is used, based on the input sliding window size. and step length The Yangtze River waterway is divided into multiple sliding windows. Calculate the mean value of the feature information for each flight segment in the window. and standard deviation Then, the K-means clustering algorithm is used to cluster all windows based on their mean and standard deviation, thereby identifying the category of each window. Finally, for windows of the same category, overlapping regions are merged to ensure that segments of each category do not overlap, such as... Figure 6 As shown.

[0118] Step 421, input traffic characteristics and parameters;

[0119] After calculating and selecting traffic characteristics, input the traffic characteristics of each segment within the required waterway. and After performing sensitivity analysis on each parameter, select the parameter with strong robustness and input the size of the sliding window. Step length and number of clusters .

[0120] (12)

[0121] (13)

[0122] in, , , , Each segment Information on flight duration, flight speed, flight acceleration, and traffic flow; This represents the set of all traffic features within a flight segment; This indicates the input waterway information.

[0123] Step 422: Slide window traversal and K-means clustering classification of flight segments;

[0124] Based on the size of the input sliding window With step size Traverse all sections of the Yangtze River waterway and calculate the mean of each traffic characteristic for all windows. and standard deviation Then, the feature sequences obtained from all sliding windows... Data standardization is performed to ensure that each feature dimension has zero mean and unit variance. Then, the standardized feature sequences are clustered using the K-means clustering algorithm, with a predetermined number of clusters selected. Assign a cluster label to each window .

[0125] Step 423: Merge adjacent and identical category segments and output the results;

[0126] Since the desired outcome is the overall channel classification, not the segment window classification results, adjacent and identical segment classifications need to be merged. Specifically, this example iterates through the clustering results of all windows in segment order. When two adjacent windows belong to the same cluster label, and the previous window's ending position... Starting position of the next window When there is overlap or adjoining relationship (i.e.) Then adjust the ending position of the current cluster segment to... The final output is the waterway classification result, which divides the entire waterway into several non-overlapping cluster segments, i.e. ,in,

[0127] .

[0128] By applying AIS data from the Yangtze River waterway in 2023 to the method of this invention, different classification results for the Yangtze River waterway in 2023 can be obtained, such as... Figure 7 and Figure 8 The classification results of vessel size and maneuverability in the Yangtze River waterway in 2023 are presented respectively, which demonstrates the feasibility and superiority of the method of the present invention. Figure 8 middle, Indicates traffic flow. Indicates the ship's position. Indicates the rate of change of the ship's turning direction. This indicates the characteristics of ship acceleration.

[0129] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A Yangtze River waterway classification method based on ship feature clustering, characterized in that, Includes the following steps: S1, Preprocess AIS data; S2 divides the Yangtze River waterway into sections and assigns geographical semantic knowledge to each section; S3 calculates the vessel traffic characteristics of the Yangtze River waterway, specifically including: S31, the Yangtze River waterway is divided into multiple square grids, the navigation time of each type of vessel within the square grid is calculated, and the static characteristics of vessels in the waterway are calculated using the summation waterway mapping method; S32, for traffic characteristics such as speed, acceleration, heading, and turning rate, the mean value of each grid is calculated, and the segment characteristic value is calculated using the channel mapping method with the maximum value; the flow characteristics are calculated using the summation method. S4 employs a segment classification method combining K-means clustering and a sliding window approach to classify segments and output channel classification results; specifically including: S41, Select appropriate ship traffic features with a similarity of less than 0.8; S42. Based on the characteristics of ship traffic within the waterway segment, the waterway segment is classified using a combination of K-means clustering and sliding window, and the final waterway classification result is output.

2. The Yangtze River waterway classification method based on ship feature clustering according to claim 1, characterized in that, The steps for preprocessing AIS data are as follows: S11, remove data with a speed less than 2 sections or fewer than 100 trajectory records to obtain AIS data points. : in, For the number The One AIS data point; , , They are respectively The time, longitude, and latitude of the point; , , , These are the ship's type, dimensions, speed, and course. S12, For missing ship size and type data, use the modulo value of the existing static data of the ship to replace the missing items; Step 13 defines the process of a ship traveling from its starting point to its destination as a journey. Linear interpolation is used to interpolate AIS data points for the same trip, with an interpolation interval of 1 minute. Then, the... AIS data points for: ， in, for The current time value, for and The distance between two points; , These are the starting and ending points of each journey.

3. The Yangtze River waterway classification method based on ship feature clustering according to claim 1, characterized in that, In step S2, the angle bisector method is used to divide the waterway into n consecutive segments. The dividing line is set as the angle bisector that passes through the inflection point of the waterway and coincides with the centerline. This ensures that the distance of each flight segment is the same; The overlay analysis method is used to spatially overlay the spatial locations of each waterway segment with its surrounding ports, prefecture-level cities and provincial-level administrative regions, thereby endowing each waterway segment with geographical semantic knowledge.

4. The Yangtze River waterway classification method based on ship feature clustering according to claim 1, characterized in that, The specific implementation process for calculating the static characteristics of vessels in a waterway is as follows: The extracted Yangtze River channel was divided into multiple 200m square grids, representing each segment of the Yangtze River channel. It contains multiple square grids: ， in, This represents the m-th square grid; i = 1, 2, 3, ..., n; The sailing time for each type of ship within the m-th square grid is calculated using the following expression: ， Where N represents the total number of AIS data points for a specific ship type within the square grid. ; The navigation time for a certain type of vessel within the waterway segment The calculation expression is as follows: ， in, For the segment The total number of square grids within.

5. The Yangtze River waterway classification method based on ship feature clustering according to claim 4, characterized in that, The dynamic traffic characteristics of speed, acceleration, heading, and turning rate are calculated using the averaging method as follows: the characteristic value of a certain square grid is the average value of the characteristic of all AIS data points within that grid, and the maximum value of the characteristic of all square grids within a flight segment is selected as the dynamic traffic characteristic of that flight segment. The vector method is used to count the number of ships entering and leaving each segment, and the traffic flow of each segment is calculated. This yields the total number of ships passing through each segment.

6. The Yangtze River waterway classification method based on ship feature clustering according to claim 1, characterized in that, In step S41, the similarity between traffic features of each segment is calculated, and hierarchical clustering analysis is performed; suitable ship traffic features with a similarity of less than 0.8 are selected as the data basis for segment classification.

7. The Yangtze River waterway classification method based on ship feature clustering according to claim 1, characterized in that, In step S42, the steps to obtain the waterway classification are as follows: Based on the vessel traffic characteristics of each segment, input the sliding window size and step size to traverse all segments; Calculate the mean and standard deviation of the characteristic values ​​for all segments in each sliding window; The K-means clustering algorithm is used to cluster the mean and standard deviation of all windows to identify the category of each window; Merge adjacent and identical category segments to ensure that segments of each category do not overlap; Output the waterway classification results, which divide the entire waterway into several non-overlapping cluster segments.

8. The Yangtze River waterway classification method based on ship feature clustering according to claim 1, characterized in that, The flight segment classification method combining K-means clustering and sliding window includes the following steps: S421, according to each waterway With all traffic features within the flight segment Enter the size of the sliding window. Step length and number of clusters ,in, ， ， , , , Each segment The characteristics of the flight duration, flight speed, flight acceleration, and traffic flow; S422, based on the size and step size of the sliding window, traverse all sections of the Yangtze River waterway and calculate the mean of each traffic characteristic for all windows. and standard deviation Then, the feature sequences obtained from all sliding windows... Data standardization is performed to ensure that each feature dimension has zero mean and unit variance; then, the standardized feature sequences are clustered using the K-means clustering algorithm, with a predetermined number of clusters. Assign a cluster label to each window ; S423, iterate through the clustering results of all windows in segment order, when two adjacent windows belong to the same cluster label, and the previous window's end position... Starting position of the next window If there are overlapping or adjacent relationships, the ending position of the current cluster segment will be adjusted to... ; The clustered segments are obtained.

Images