A multi-ship collision risk prediction method

Abstract

The present invention belongs to the technical field of marine ship traffic safety, and specifically discloses a multi-ship collision risk prediction method. The method proposes a multi-ship collision risk prediction model integrating entropy weight method, TOPSIS and hierarchical analysis AHP, which evaluates the collision risk between multiple ships, and accurately calculates the key indicators between each pair of ships by carefully considering factors such as relative distance, relative speed, distance to the nearest approach point, time to the nearest approach point and heading angle difference, and determines the weight of each key indicator based on entropy weight, and obtains the risk index between multiple pairs of ships through decision analysis based on TOPSIS method, and further integrates the risk index of multiple pairs of ships based on AHP algorithm, which is conducive to comprehensive analysis of the mutual influence between each ship, predicting the overall collision risk profile of the target ship, and significantly improving the comprehensiveness and accuracy of marine ship traffic conflict risk assessment.

Description

Technical Field

[0001] The invention belongs to the technical field of marine vessel traffic safety, and in particular relates to a multi-ship collision risk prediction method. Background Art

[0002] In recent years, with the rapid increase in the number of ships, the waterway has been congested, the navigation environment has become more complex, and the risk of collision has increased. These risks are affected by many factors such as human operation, ship conditions, navigation environment and management. In the seaway, the safety management of ships is particularly important. Therefore, risk control of these ships is particularly important.

[0003] However, in order to deeply explore the characteristics, root causes and potential risks of collision accidents between ships, many scholars have analyzed marine ship accidents from different angles and proposed corresponding preventive measures.

[0004] For example, some studies have analyzed the spatial pattern and characteristics of global maritime accidents, summarized the laws of ship collision accidents, and proposed preventive measures from multiple dimensions (such as personnel, ships, environment, and management). Another study constructed a Bayesian network model by statistically analyzing the causes of fishing boat accidents between 2008 and 2018. The Bayesian network model can estimate the probability of fishing boat accidents in different scenarios, providing data support and probability prediction for the formulation of prevention strategies. In addition, there are studies that explore ship collision incidents in different sea areas and outline the characteristics and causes of collisions between ships.

[0005] Although the above methods have shown unique advantages in different models and practical applications, these methods are based on the data after the ship collision accident for analysis, and do not analyze the behavior of the ship. In addition, these methods have never fully considered the interaction between the ships when analyzing multi-ship encounter scenarios. They are limited to analyzing the direct impact of the relative motion between pairs of ships on the collision risk, ignoring the more complex interactions between ships in the multi-ship situation. Summary of the invention

[0006] The purpose of the present invention is to propose a multi-ship collision risk prediction method, which takes into account the navigation behavior between ships and performs correlation analysis on the risks between ships based on key indicators between ships; at the same time, when analyzing multi-ship encounter scenarios, the risk correlation between the ships is fully considered, which is conducive to accurately predicting the overall risk level of ships during navigation.

[0007] In order to achieve the above object, the present invention adopts the following technical scheme:

[0008] A multi-ship collision risk prediction method comprises the following steps:

[0009] Step 1. Obtain the AIS data information of ships in the sea channel, and pre-process the ship AIS data to obtain the ship's data information, including MMSI, timestamp, position, speed and heading;

[0010] Step 2. Based on the ship AIS data preprocessed in step 1, multiple encounter ships that interact with the target ship within the set time window are identified and extracted through spatial position matching and time series analysis to form ship trajectory pairs;

[0011] Among them, each set of ship trajectory pairs consists of the target ship and any one of the multiple encountering ships;

[0012] Step 3. Calculate five key indicators between the two ships in each pair of ship trajectories, including relative distance, relative speed, distance to the closest approach point, time to the closest approach point, and heading angle difference;

[0013] Step 4. Construct a multi-ship collision risk prediction model integrating entropy weight method, TOPSIS and hierarchical analysis AHP;

[0014] Firstly, the entropy weight method is used to assign weights to the five key indicators obtained in step 3; TOPSIS is used to predict the risk index between two encountering ships in each set of ship trajectory pairs, and the correlation analysis of risks between ships is performed;

[0015] Finally, based on the results of correlation analysis, the AHP algorithm is used to fit the risk index of each group of ship trajectory pairs to obtain the overall collision risk index between the target ship and multiple encountering ships, thereby realizing the multi-ship collision risk prediction of the target ship.

[0016] The present invention has the following advantages:

[0017] As described above, the present invention relates to a multi-ship collision risk prediction method, which constructs a multi-ship collision risk prediction model integrating entropy weight method, TOPSIS and hierarchical analysis AHP, wherein, in terms of model input, the present invention only needs to consider the navigation behavior between ships, such as five key indicators such as the relative distance between ships, relative speed, distance to the nearest approach point, time to the nearest approach point and heading angle difference, and determine the weight distribution of each key indicator, so as to quickly and accurately complete the comprehensive assessment of the risk of ship traffic collision, which not only improves the speed and efficiency of data processing, but also achieves a qualitative leap in accuracy. In addition, in the case of multi-ship encounter, the model proposed by the present invention comprehensively considers the key indicators such as the relative distance between ships, relative speed, distance to the nearest approach point, time to the nearest approach point and heading angle difference, and uses the entropy weight method to determine the weight distribution of each key indicator, and then uses TOPSIS to perform risk correlation analysis between ships for each key indicator with assigned weights, and finally, based on the correlation analysis and integrating the AHP algorithm, the correlation analysis between each pair of ships is fitted to evaluate the overall risk level of the ship during navigation, which helps to improve the safety of ship navigation. The method of the present invention shows significant advantages in multi-ship traffic collision risk assessment and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Flow chart of a method for predicting multi-ship collision risk in an embodiment of the present invention;

[0019] Figure 2 A schematic diagram of a scenario of a multi-ship collision risk prediction method in an embodiment of the present invention;

[0020] Figure 3 This is a schematic diagram of the correlation analysis of multi-ship collision risks in an embodiment of the present invention;

[0021] in, Figure 3 (a) shows the risk index changes of ships A and B, (b) shows the risk index changes of ships A and C, (c) shows the risk index changes of ships A and D, and (d) shows the risk index changes of ships A and E;

[0022] Figure 4 Schematic diagram of overall ship navigation risk assessment analysis in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0024] Example 1

[0025] This embodiment 1 describes a multi-ship collision risk prediction method, in which a multi-ship collision risk prediction model integrating entropy weight method, TOPSIS and hierarchical analysis AHP is proposed to evaluate the collision risk between multiple ships. The model accurately calculates the key indicators between each pair of ships by carefully considering factors such as relative distance, relative speed, distance to the nearest approach point, time to the nearest approach point and heading angle difference, and uses the entropy weight method to determine the weight distribution of each key indicator. Then, TOPSIS is used to analyze the risk correlation between ships for each key indicator with assigned weights. Finally, based on the correlation analysis and integrating the AHP algorithm, the correlation analysis between each pair of ships is fitted, the mutual influence between the ships is comprehensively analyzed, and the overall collision risk profile of the target ship is predicted, thereby improving the comprehensiveness and accuracy of the risk assessment of maritime ship traffic conflict.

[0026] like Figure 1 As shown, the multi-ship collision risk prediction method in this embodiment includes the following steps:

[0027] Step 1. Obtain the AIS data information of ships in the sea channel, and pre-process the ship AIS data to obtain the ship's data information, including MMSI, timestamp, position, speed and heading.

[0028] The raw AIS data collected in this embodiment may contain noise, repetitions or records with inconsistent formats. Therefore, it is necessary to clean the data, remove invalid or redundant information, and convert the data format to ensure consistency.

[0029] This embodiment preprocesses the AIS data in the sea channel, and the process is as follows:

[0030] Check the extracted AIS data type, ensure that the data type of each column (including longitude, latitude, speed, and heading) is correct, and handle abnormal values. For example, the longitude range is 0° to 180°, the latitude range is -90° to 90°, the speed range is 0-30kn, and the heading range is 0-360°. Check whether these data are beyond the reasonable range.

[0031] Handle duplicate values ​​by checking whether there are identical rows in the data and deleting them if they exist.

[0032] In addition, check whether there are missing values ​​in the data, such as latitude and longitude, heading, speed, etc. If there are few missing values, you can use interpolation or filling methods to process them. If there are many missing values, you can consider deleting the missing values.

[0033] After the data is cleaned and invalid or redundant information is removed, it is further standardized or normalized.

[0034] Since the units and magnitudes of the various parameters in the AIS data (ship's MMSI, timestamp, position, speed and heading) may be different, for example, the speed unit is knots, while the latitude and longitude are expressed in degrees, this will lead to deviations in data analysis and calculation. Therefore, the data needs to be standardized or normalized so that all parameters are at the same magnitude.

[0035] After the above preprocessing of the ship AIS data, the core parameters such as the unique identifier MMSI, precise timestamp, longitude and latitude position coordinates, instantaneous speed and heading angle of each ship are extracted by parsing the AIS message.

[0036] This step 1 is the prerequisite for subsequent analysis and ensures the accuracy and availability of the data.

[0037] Step 2. Based on the ship AIS data preprocessed in step 1, multiple encounter ships that interact with the target ship within the set time window are identified and extracted through spatial position matching and time series analysis to form ship trajectory pairs.

[0038] Each ship trajectory pair consists of a target ship and any one of the multiple encountering ships.

[0039] The specific steps of step 2 are:

[0040] Step 2.1. First, upload the ship AIS data processed in step 1 to the kepler online website to verify the ship trajectory data and visualize the ship trajectory to clearly show the proximity of the ships in spatial position and screen out the ship trajectory pairs that meet the target ship.

[0041] Upload the processed AIS data (including the timestamps, longitude and latitude, speed, and heading information of multiple ships encountered by the target ship) to the kepler online website to visualize the extracted ship trajectories. By comparing the trajectories of different ships, select the ships that overlap or approach the target ship's trajectory within the set time window;

[0042] The ship trajectory pairs formed by pairing these ships with the target ship one by one are used as the ship trajectory pairs that encounter the target ship.

[0043] Step 2.2. Based on the screened ship track pairs, with the help of the timestamp information in the AIS data, the data information of multiple ships that encounter the target ship within the set time window is accurately extracted.

[0044] These data information are organized in an orderly manner to form multiple groups of ship trajectory pairs. The data information of each group of ship trajectory pairs records in detail the dynamic changes of the encountering ships within the set time window.

[0045] In order to determine the meeting point between the target ship and multiple ships, a reasonable time window needs to be set. In this embodiment, the time window is set according to the timestamp in the AIS data. In this embodiment, the time window is set to 1400s, for example.

[0046] This time window is considered to be the time range in which two ships may be close to each other and interact with each other. This time window can be flexibly adjusted according to the busyness of the sea area, the speed of the ship and the accuracy requirements of risk assessment.

[0047] like Figure 2 A schematic diagram of a scenario of a multi-ship collision risk prediction method according to an embodiment of the present invention is shown. Figure 2 For example, if ship A is taken as the target ship, then ships B, C, D, and E are multiple ships that the target ship encounters.

[0048] The present invention adopts a multi-ship collision risk prediction model that integrates the entropy weight method, TOPSIS and hierarchical analysis AHP. When evaluating the multi-ship collision risk of ship A, the risk indexes of A and B, A and C, A and D, and A and E should be fully considered, so as to fit these four risks and obtain the overall collision risk index of A in the multi-ship encounter scenario.

[0049] In addition, the mutual influence between the ships when multiple ships meet in this embodiment is actually to analyze the risk index between A and B, the risk index between A and C, the risk index between A and D, and the risk index between A and E.

[0050] Step 3. Calculate the five key indicators between the two ships in each pair of ship trajectories, including relative distance, relative speed, distance to the closest point of approach (DCPA), time to the closest point of approach (TCPA), and heading angle difference.

[0051] The relative distance is calculated using the Mercator algorithm.

[0052] The distance to the closest approach point is calculated as:

[0053] DCPA = D × sin(Cr - q)

[0054] Among them, D is the actual distance between the two ships, Cr is the relative heading, and q is the azimuth of the target ship relative to the own ship.

[0055] The time to the closest approach point is calculated as:

[0056] TCPA = DCPA / relative velocity (the component in the direction of the heading line of relative motion).

[0057] The magnitude of the relative speed can be calculated by the vector synthesis of the two ship speeds, the process is as follows:

[0058] First, decompose the velocities of the two ships into two components, one parallel to the direction of their connecting line and the other perpendicular to the direction of their connecting line; then, add these two components together (pay attention to the direction) to obtain the magnitude of the relative velocity.

[0059] The calculation process of the relative heading angle difference is as follows:

[0060] First, you need to know the heading of each ship, which is usually expressed in compass degrees, ranging from 0° (true north) to 360° (also true north, but expressed as a complete circle);

[0061] The angular difference between the two ships' headings is then calculated by subtracting the heading of one ship from the heading of the other ship.

[0062] The present invention can identify pairs of ships that may affect each other in a set time window through spatial position matching and time series analysis technology, including determining which ships are sailing in the same channel and tracking their motion trajectories, thereby forming a series of ship trajectory pairs. For each pair of ship trajectories, the relative motion parameters between them are calculated, including the relative distance between the two ships, relative speed, distance to the closest approach point, time to the closest approach point, and key indicators of heading angle difference, and the weight distribution of key indicators is determined, so as to conduct correlation analysis on risks between ships, and quickly and accurately complete a comprehensive assessment of the risk of collision based on ship traffic, improve the speed and efficiency of data processing, and achieve a qualitative leap in accuracy.

[0063] Step 4. Construct a multi-ship collision risk prediction model that integrates entropy weight method, TOPSIS and hierarchical analysis AHP.

[0064] Firstly, the entropy weight method is used to assign weights to the five key indicators obtained in step 3. TOPSIS is used to predict the risk index between the two encountering ships in each set of ship trajectory pairs, and the correlation analysis of risks between ships is carried out.

[0065] Finally, based on the results of correlation analysis, the AHP algorithm is used to fit the risk index of each group of ship trajectory pairs to obtain the overall collision risk index between the target ship and multiple encountering ships, thereby realizing the multi-ship collision risk prediction of the target ship.

[0066] Among them, the analytic hierarchy process, namely Analytic Hierarchy Process, is referred to as AHP algorithm.

[0067] The specific steps of step 4 are as follows:

[0068] Step 4.1. First, establish a risk matrix for the five key indicators calculated in step 3: relative distance, relative speed, distance to the closest approach point, time to the closest approach point, and heading angle difference.

[0069] set up For the The sample The value of the key indicator, therefore, the risk matrix is ​​defined as :

[0070] ;

[0071] In the formula, represents the ship navigation risk matrix, where Indicates the value of each key indicator. , , Indicates the number of samples for each key indicator, Indicates the number of key figures.

[0072] Step 4.2. Normalize the values ​​of the five key indicators.

[0073] The process of normalizing the values ​​of key indicators at each time point is as follows:

[0074] Extract the values ​​of each key indicator at each time point from the set time window and follow the steps below:

[0075] First, find the minimum and maximum values ​​of each group of key indicators at different times;

[0076] Secondly, the values ​​of each key indicator at each time point are processed as follows:

[0077] First, subtract the minimum value from all the values, then divide it by the difference between the maximum and minimum values, and finally add a very small number. The processed value is the normalized result of the key indicator at each time, which is recorded as .

[0078] The formula for normalizing the values ​​of key indicators at each time point is as follows:

[0079] ;

[0080] in, This means finding the minimum value. This means finding the maximum value. is a number greater than 0 to prevent the result value from being 0. For example, the value is 0.0001.

[0081] Taking the relative distance index as an example, this embodiment extracts the trajectory of the ship within 1400 seconds, and obtains 1400 relative distance values. In order to normalize these values, the following steps need to be followed:

[0082] (1) Find the minimum value (min) and the maximum value (max) from these 1400 values;

[0083] (2) Perform the following processing on each relative distance value:

[0084] First, subtract the minimum value (min) of all the values, then divide it by the difference between the maximum value (max) and the minimum value (min) (that is, max - min), and finally add a very small number 0.0001.

[0085] The processed value is the normalized result of the relative distance index. The above normalization process is also used for the four key indicators of relative speed, distance to the closest approach point, time to the closest approach point and heading angle difference.

[0086] In this way, each indicator can get a normalized result, which will be used for subsequent analysis or comparison.

[0087] Step 4.3. Step 4.2 has obtained the normalized results of each key indicator (DCPA, TCPA, relative distance, relative speed and heading angle difference). Next, the entropy weight method will be used to calculate the weights of these indicators.

[0088] In this embodiment, the entropy weight method is used to calculate the weights of the five key indicators of relative distance, relative speed, distance to the nearest approach point, time to the nearest approach point, and heading angle difference based on the normalized data. .

[0089] Weight The calculation formula is as follows:

[0090] ;

[0091] in, Represents the value of each key indicator The result after normalization according to the processing method of step 4.2 is: Indicates the weight of risk indicator factors.

[0092] Next, the TOPSIS method is used for decision analysis to predict the risk index between two encountering ships in each set of ship trajectories. On this basis, an in-depth analysis of the correlation of risks between ships is conducted, as shown in steps 4.4-4.7.

[0093] Step 4.4. First, the values ​​of the five key indicators are dimensionlessly normalized to eliminate the impact of different dimensions on the results. Then, these real data are further normalized to convert them into a unified scale.

[0094] Then, combined with the weight values ​​of each key indicator calculated by the entropy weight method in step 4.3, the data normalized in step 4.4 is combined with the weights of the corresponding key indicators to obtain the weighted data. .

[0095] Finally, based on these dimensionless normalized and weighted data , construct the weighted matrix , the formula is as follows:

[0096] ;

[0097] .

[0098] Step 4.5. The weighting matrix obtained according to step 4.4 , confirm Optimal solution for weighted values ​​of key indicators and the worst solution , the optimal solution and the worst solution The expression is as follows:

[0099] , .

[0100] in In order to take the maximum value of the weighted value as the optimal solution, The minimum value of the weighted value is taken as the worst solution.

[0101] Step 4.6. Combine the weighted data from step 4.4 with the weighted data from step 4.5. Optimal solution for weighted values ​​of key indicators Calculate the optimal Euclidean distance to obtain the optimal solution At the same time, the weighted data in step 4.4 is combined with the weighted data in step 4.5. The worst solution for the weighted value of the key indicators of the class Calculate the worst Euclidean distance to get the worst solution .

[0102] Optimal Euclidean distance And the worst Euclidean distance The calculation formula is as follows:

[0103] , .

[0104] Step 4.7. Based on the optimal Euclidean distance and the worst Euclidean distance Conduct ship-related risk analysis, such as Figure 3 As shown, the risk index between two ships in each set of ship trajectory pairs is obtained, and the calculation formula is as follows:

[0105] ;

[0106] in, It represents the risk index of each group of ship trajectories at each time point in the set time window.

[0107] Specifically, under the time window (1400s) set in this embodiment, there will be multiple time points (for example, each second is regarded as a time point), and the risk index between two ships in a set of ship trajectory pairs will be calculated at each time point.

[0108] against Figure 2 The target ship A is calculated by calculating the risk index between A and B, the risk index between A and C, the risk index between A and D, and the risk index between A and E, where B, C, D, and E are all ships that encounter ship A.

[0109] The risk index of each time point in each set of ship trajectory pairs is visualized, such as Figure 3 shown.

[0110] Figure 3 (a) shows the changes in risk index of ships A and B, (b) shows the changes in risk index of ships A and C, (c) shows the changes in risk index of ships A and D, and (d) shows the changes in risk index of ships A and E.

[0111] Depend on Figure 3 It is not difficult to see that the risk index of each set of ship trajectory pairs is It changes continuously with the time. Next, the AHP algorithm is used to fit the risk of paired ship correlation and establish the overall collision risk of multiple ships.

[0112] Taking the set time window of 1400s as an example, within the set time window of 1400s, a risk index is output every second, so there are 1400 risk indexes between any two paired ships (such as ships A and B).

[0113] Calculate the distance between ship A and ship B , Ship A and Ship C , Ship A and Ship D , Ship A and Ship E Finally, a risk index matrix is ​​established based on the risk indices of these four groups of ship trajectory pairs. The specific process is shown in the following step 4.8.

[0114] Step 4.8. According to the risk index of each group of ship trajectories at different time points in the set time window, a risk index matrix for the target ship is established. In the risk index matrix, express;

[0115] The formula of the risk index matrix of the target ship is as follows:

[0116] ;

[0117] a represents the sample number of the ath risk index in each set time window (each risk index corresponds to a time point in the set time window), p represents the number of samples of the risk index of each group of ship trajectory pairs, b represents the bth group of ship trajectory pairs, and q represents the number of groups of ship trajectory pairs. , .

[0118] Risk Index Standardized processing to obtain , the formula is as follows:

[0119] ;

[0120] in, It represents the risk index between two ships in each group of ship trajectories at different time points in the set time window. The results obtained after standardization.

[0121] Step 4.9. Use the AHP algorithm to standardize the risk index The weight is calculated and the expression is as follows:

[0122] ;

[0123] In the formula, Represents the risk index of each group of ship trajectory pairs after standardization The weight of .

[0124] Step 4.10. Combine multiple risk indices between the target ship and other ships into a risk set , combined with the weight calculated in step 4.9 Calculate and obtain the overall collision risk index of the target ship , realizing the multi-ship collision risk prediction of the target ship.

[0125] Specifically, the collision risk set between the target ship and other ships is as follows:

[0126] ;

[0127] in, The matrix representing the risk composition between target ship A and other ships.

[0128] The overall collision risk index of the target ship is The calculation formula is as follows:

[0129] ;

[0130] like Figure 4 As shown, the overall collision risk index Constantly changing over time.

[0131] In this embodiment, in order to determine whether the target ship has a multi-ship collision risk in a multi-ship encounter scenario, a risk threshold needs to be set in advance. The value range of the risk threshold is [0, 1], for example, it is set to 0.8.

[0132] Depend on Figure 4 It can be seen that in the time period from 400s to 800s, the overall risk index of ship A is greater than 0.8, indicating that ship A is at risk of multi-ship collision during this time period, and multiple ships need to adjust their course and speed in time to avoid collision.

[0133] The multi-ship collision risk prediction model constructed by the present invention fully considers five key indicators, namely, the relative distance between paired ships, relative speed, distance to the nearest approach point, time to the nearest approach point, and heading angle difference, and uses the entropy weight method to allocate weights for the above five key indicators. Then, TOPSIS is used to analyze the risk correlation between paired ships, and the collision risk index between each pair of ships is accurately calculated. Finally, based on the correlation analysis, the AHP algorithm is integrated to further fit the risk correlation between ships, comprehensively analyze the mutual influence between ships, and evaluate the overall risk level of ships during navigation.

[0134] The method of the present invention focuses on improving the level of ship traffic safety in maritime areas. By deeply mining and analyzing historical automatic identification system (AIS) data, the ship trajectory pairs are accurately extracted according to the ship MMSI information, and a multi-ship collision risk model is established to analyze the collision risks between multiple ships, providing solid data support and decision-making basis for strengthening maritime traffic safety management.

[0135] Of course, the above description is only a preferred embodiment of the present invention, and the present invention is not limited to the above embodiments. It should be noted that all equivalent substitutions and obvious deformation forms made by any technician familiar with the field under the guidance of this specification fall within the essential scope of this specification and should be protected by the present invention.

Claims

1. A multi-ship collision risk prediction method, characterized in that: The steps include: Step 1. Obtain the AIS data information of ships in the sea channel, and pre-process the ship AIS data to obtain the ship's data information, including MMSI, timestamp, position, speed and heading; Step 2. Based on the ship AIS data preprocessed in step 1, multiple encounter ships that interact with the target ship within the set time window are identified and extracted through spatial position matching and time series analysis to form ship trajectory pairs; Among them, each set of ship trajectory pairs consists of the target ship and any one of the multiple encountering ships; Step 3. Calculate five key indicators between two ships in each pair of ship trajectories, including relative distance, relative speed, distance to the closest approach point, time to the closest approach point, and heading angle difference; Step 4. Construct a multi-ship collision risk prediction model integrating entropy weight method, TOPSIS and hierarchical analysis AHP; Firstly, the entropy weight method is used to assign weights to the five key indicators obtained in step 3; TOPSIS is used to predict the risk index between two encountering ships in each set of ship trajectory pairs, and the correlation analysis of risks between ships is performed; Finally, based on the correlation analysis results, the risk index of each group of ship trajectory pairs is fitted using the AHP algorithm to obtain the overall collision risk index between the target ship and multiple encountering ships, thus realizing the multi-ship collision risk prediction of the target ship. The step 4 is specifically as follows: Step 4.

1. First, establish a risk matrix for the five key indicators of relative distance, relative speed, distance to the closest approach point, time to the closest approach point, and heading angle difference calculated in step 3; Let x ij is the value of the jth key indicator of the ith sample, so the risk matrix is ​​defined as X = [x ij ] n×m : In the formula, X represents the ship navigation risk matrix, x ij represents the value of each key indicator, i=1,,n, j=1,...,m, n represents the number of samples of each key indicator, and m represents the number of key indicators; Step 4.

2. Normalize the values ​​of the five key indicators; Step 4.

3. Based on the normalized data in step 4.2, the entropy weight method is used to calculate the weights w of the five key indicators: relative distance, relative speed, distance to the nearest approach point, time to the nearest approach point, and heading angle difference. j ; In step 4.3, the weight w j The calculation formula is as follows: Where n represents the number of samples for each key indicator, m represents the number of key indicators; y ij Represents the result of normalizing the value of each key indicator, w j Indicates the weight of risk indicator factors; Step 4.

4. First, calculate the values ​​x of the five key indicators in step 4.

1. ij Perform dimensionless normalization processing; Then, combined with the weight values ​​of each key indicator calculated by the entropy weight method, the data normalized in step 4.4 is combined with the weights of the key indicators to obtain the weighted data Finally, based on the weighted data Construct a weighted matrix In step 4.4, the weighted data And the weight matrix The calculation formulas are as follows: Step 4.

5. According to the weighting matrix Determine the optimal solution Z for the weighted value of the jth key indicator + j and the worst solution Z - j ; In step 4.5, the optimal solution Z of the weighted value of the j-th key indicator + j and the worst solution Z - j The expression is as follows: WITH + j =max(from * 1j ,With * 2j ,…With * ij ),WITH - j =min(of * 1j ,With * 2j ,…With * ij ); Among them, max() takes the maximum value of the weighted value as the optimal solution, and min() takes the minimum value of the weighted value as the worst solution; Step 4.

6. Add the weighted data from step 4.4 The optimal solution Z of the weighted value of the j-th key indicator + j Calculate the optimal Euclidean distance to obtain the optimal solution At the same time, the weighted data in step 4.4 The worst solution Z with the weighted value of the j-th key indicator - j Calculate the worst Euclidean distance to get the worst solution In step 4.6, the optimal Euclidean distance And the worst Euclidean distance The calculation formula is as follows: Step 4.

7. Based on the optimal Euclidean distance and the worst Euclidean distance The ship correlation risk analysis is performed to obtain the risk index between two ships in each pair of ship trajectories. The calculation formula is as follows: Among them, CRI i It is the risk index at each time point in the set time window; Step 4.

8. According to the risk index of each group of ship trajectories at different time points in the set time window, a risk index matrix for the target ship is established, where CRI is used for convenience. i Using CRI in the Risk Index Matrix iab express; The formula of the risk index matrix of the target ship is as follows: Where a represents the sample number of the ath risk index in each set time window, p represents the number of samples of the risk index of each group of ship trajectory pairs, b represents the bth group of ship trajectory pairs, and q represents the number of groups of ship trajectory pairs; CRI iab Standardization process to obtain CRI * iab , the formula is as follows: Among them, CRI * iab represents the result obtained by normalizing the risk index between two ships at different time points in the set time window for each group of ship trajectories, a=1,2,…p, b=1,2,…q; Step 4.

9. Use the AHP algorithm to standardize the risk index CRI * iab The weight is calculated and the expression is as follows: Where N ib Represents the risk index CRI of each group of ship trajectory pairs after standardization * iab The weight of Step 4.

10. Combine multiple risk indices between the target ship and other ships into a risk set CRI xy , combined with the weight N calculated in step 4.9 ib Calculate and obtain the overall collision risk index CRI of the target ship + ; In step 4.10, the collision risk between the target ship and other ships is as follows: Among them, CRI xy represents the set of risks between target ship A and other ships; The overall collision risk index (CRI) of the target ship + The calculation formula is as follows: Among them, the overall collision risk index CRI + Constantly changing over time.

2. The multi-ship collision risk prediction method according to claim 1, characterized in that: The step 2 is specifically as follows: Step 2.

1. First, upload the ship AIS data processed in step 1 to the kepler online website to verify the ship trajectory data, visualize the ship trajectory, clearly show the proximity of the ships in spatial position, and screen out the ship trajectory pairs that meet the target ship; Step 2.

2. Based on the screened ship track pairs, with the help of the timestamp information in the AIS data, the data information of multiple ships that encounter the target ship in the set time window is accurately extracted.

3. The multi-ship collision risk prediction method according to claim 2, characterized in that: The step 2.1 is specifically as follows: The processed AIS data information is uploaded to the kepler online website to visualize the extracted ship tracks. By comparing the tracks of different ships, the ships that overlap or approach the target ship's track within the set time window are screened out; Within the set time window, the latitude, longitude, speed, and heading information of multiple ships that encounter the target ship are extracted, and their timestamps and position information are used to construct the navigation track of each encountering ship; The ship trajectory pairs formed by pairing these ships with the target ship one by one are used as the ship trajectory pairs that encounter the target ship.

4. The multi-ship collision risk prediction method according to claim 1, characterized in that: In step 4.2, the process of normalizing the values ​​of the key indicators at each time point is as follows: Extract the values ​​of each key indicator at each time point from the set time window and follow the steps below: First, find the minimum and maximum values ​​of each group of key indicators at different times; Secondly, the values ​​of each key indicator at each time point are processed as follows: First, subtract the minimum value from all the values, then divide it by the difference between the maximum and minimum values, and finally add a very small number. The processed value is the normalized value of the key indicator at each time, which is recorded as y ij ; The formula for normalizing the values ​​of key indicators at each time point is as follows: Among them, min[] means finding the minimum value, max[] means finding the maximum value, and σ is a number greater than 0.

Images