CN115510737A - Method and device for calculating ship collision risk and storage medium - Google Patents
Method and device for calculating ship collision risk and storage medium Download PDFInfo
- Publication number
- CN115510737A CN115510737A CN202211042282.2A CN202211042282A CN115510737A CN 115510737 A CN115510737 A CN 115510737A CN 202211042282 A CN202211042282 A CN 202211042282A CN 115510737 A CN115510737 A CN 115510737A
- Authority
- CN
- China
- Prior art keywords
- ship
- degrees
- meeting
- collision
- distance
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G3/00—Traffic control systems for marine craft
- G08G3/02—Anti-collision systems
Abstract
The invention provides a method, a device and a storage medium for calculating a collision risk of a ship, and relates to the technical field of navigation. The method comprises the following steps: acquiring target AIS data information of a ship, acquiring meeting situation information according to the target AIS data information, and calculating a ship collision parameter by using the meeting situation information; determining meeting situation data which accords with preset screening conditions of ship meeting situation data; then constructing a classification model of a support vector machine, and learning and classifying meeting situation data of the preselected samples by combining a preset collision avoidance rule to obtain a ship meeting direction map; and finally, constructing a BP neural network, optimizing and determining each layer of node parameters of the BP neural network by adopting a sparrow search algorithm, and predicting the collision risk value of the selected ship and the target ship according to the meeting direction bitmap. The scheme can judge the navigation state of each ship in advance, predict the possible collision risk of the ship, and has important significance for reducing the collision accident at sea and ensuring the navigation safety.
Description
Technical Field
The invention relates to the technical field of navigation, in particular to a method and a device for calculating a collision risk of a ship and a storage medium.
Background
The automatic ship identification system AIS is a novel ship collision avoidance system, is used for automatic response and identification between ships and between ship banks, and is widely applied to ship navigation. How to utilize and process AIS data, judge each boats and ships navigation state in advance, predict the possible collision risk of boats and ships, to reducing the marine collision accident, ensure navigation safety, have important meaning.
Disclosure of Invention
In order to solve the problems, the invention provides a method, a device and a storage medium for calculating the collision risk of a ship.
In order to achieve the above object, there is provided, in one aspect, a method of calculating a collision risk of a ship, including:
s1, acquiring target AIS data information of a ship, acquiring meeting situation information according to the target AIS data information, and calculating ship collision parameters by using the meeting situation information, wherein the target AIS data information comprises speed, direction and position, the meeting situation information comprises relative distance, relative direction and ship course crossing angle of the ship, and the collision parameters comprise the relative distance, relative speed, minimum meeting distance DCPA and minimum meeting time TCPA of the ship;
s2, screening meeting situation data of the ship meeting the screening conditions according to the preset screening conditions of the ship meeting situation data and the ship collision parameters, wherein the meeting situation data are determined by the meeting situation information;
s3, constructing a classification model of a ship support vector machine, and learning and classifying the meeting situation data of the preselected samples according to the classification model of the support vector machine and a preset collision avoidance rule to obtain a ship meeting orientation map;
and S4, constructing a BP neural network, optimizing and determining each layer of node parameters of the BP neural network by adopting a sparrow search algorithm, and predicting the collision risk value of the selected ship and the target ship according to the meeting direction bitmap, wherein input items of the BP neural network are the course of the target ship, the speed of the target ship, the distance between the two ships and the relative direction of the two ships, and output items are the collision risk value.
Further, the step of acquiring the target AIS data information of the ship in step S1 includes:
s11, eliminating abnormal data in the ship AIS original data;
s12, decoding the ship AIS data after the abnormal data are removed to obtain ship position information;
and S13, constructing a track of the ship in each time period according to the position information to obtain a ship track set, wherein the track of each time period comprises a starting point, an end point and a track characteristic point.
And S14, establishing a ship geodetic fixed coordinate system XOY, converting the information in the track set into the coordinate system XOY, and acquiring the target AIS data information, wherein the origin of the geodetic fixed coordinate system XOY is a preset point, the X axis points to the true east, and the Y axis points to the true north.
Further, the screening conditions of the ship meeting situation data preset in step S2 include:
setting one ship as a ship and the other ship as a target ship, and setting a ship-associated coordinate system xoy, wherein the ship-associated coordinate system xoy takes the center of mass of the ship as an origin o, the y axis points to the bow and is 0 degree of azimuth, the x axis points to the right transverse line of the ship,
the screening conditions of the encounter situation are set as follows: the two ships are motor ships, the directions of the two ships are opposite or nearly opposite, and the target ship is positioned in a sector azimuth in the range of 355-360 degrees or 0-5 degrees of the ship;
the overtaking situation screening condition is set as follows: the velocity vector component of the rear ship in the motion direction of the front ship exceeds the velocity of the front ship, the distance between the two ships is less than 4 nautical miles, and the target ship is positioned in a sector azimuth of 112.5-247.5 degrees of the ship;
the cross-meeting situation screening conditions are set as follows: the minimum meeting distance DCPA of the two ships is less than 2 nautical miles, the minimum meeting time TCPA is more than 0, and the target ship is located in a sector position of 5-112.5 degrees or 247.5-355 degrees of the ship.
Further, the step of constructing a classification model of the support vector machine of the ship in step S3 includes:
S31, setting a known ship sample number set (x) i ,y i ) Wherein x belongs to R, y belongs to { -1,1}, i =1,2., N;
s32, selecting a proper kernel function K (x) i ,y i ) And an appropriate parameter C, converting the support vector machine problem of the nonlinear sample of the ship into an optimization problem:
wherein alpha is i Is a lagrange multiplier;
s33, solving the optimization problem to obtain
S34, selecting alpha * A positive component of 0 ≦ α * C is less than or equal to C, and a threshold value is calculated
S35, constructing a decision function
Further, the step S3 includes a step of determining an angle θ at which each azimuth area divided by the azimuth map is located and a distance relationship between each ship and a ship meeting a center point of the azimuth map in the azimuth area, where the ship meeting the azimuth map is located, according to the following relationship: 2 degrees < theta is less than or equal to 5 degrees, and the distance between the two ships is less than 6 nautical miles; theta is less than or equal to 45 degrees at an angle of 5 degrees, and the distance between the two ships is less than 6 nautical miles; theta is more than 45 degrees and less than or equal to 110 degrees, and the distance between the two ships is less than 3 nautical miles; the angle is more than or equal to 110 degrees and less than or equal to 200 degrees, and the distance between the two ships is less than 3 nautical miles; theta is less than or equal to 250 degrees at an angle of 200 degrees, and the distance between the two ships is less than 3 nautical miles; theta is more than 250 degrees and less than or equal to 310 degrees, and the distance between the two ships is less than 3 nautical miles; theta is more than 310 degrees and less than or equal to 355 degrees, and the distance between the two ships is less than 3 nautical miles; 355 degrees < theta is less than or equal to 360 degrees or 0 degrees < theta is less than or equal to 2 degrees, and the distance between the two ships is less than 3 nautical miles.
Further, the step of constructing the BP neural network in step S4 includes training the BP neural network, and performing error evaluation on a risk value output by the current BP neural training network and an expected output value when the BP neural network is trained, where the expected output value is obtained by a comprehensive evaluation method, and the method specifically includes:
s41, acquiring ship AIS original data in a ship collision accident in advance as a data source of a fuzzy comprehensive evaluation method, and processing according to the steps S1-S3;
s42, calculating the risk value of each ship in the ship collision accident by using a fuzzy comprehensive evaluation method, and taking the risk value as an expected output value Ex of the BP neural network;
s43, selecting the result of the mean square error evaluation prediction model
And when the error MSE meets a preset condition, finishing training, wherein X is a predicted value of the collision risk value of the current BP neural training network, and T is the number of collision accident samples.
Further, the step of calculating the risk value of each ship in the ship collision accident in step S42 includes:
is provided with
The value range is [0,1 ] for the risk membership degree of each parameter of the target ship i meeting the ship]，i＝1,2,...,n；a D 、a K 、a B 、a DCPA 、a TCPA The weight of the risk membership degree of each parameter is in the value range of [0,1]And a is a D +a K +a B +a DCPA +a TCPA =1 degree of risk of collision between own ship and target ship i
The above-mentioned
The calculation formula (2) includes:
Further, the step S4 of determining the node parameters of each layer of the BP neural network by using a sparrow search algorithm includes the following steps:
step1, initializing parameters of a sparrow search algorithm, wherein the initialized parameters comprise population scale, the number of discoverers, the number of sparrows for reconnaissance and early warning, the dimension of an objective function, the upper and lower bounds of an initial value, the maximum iteration times or solving precision;
step2: initializing a total group, generating an initial group, determining a fitness function, and selecting a network error as the fitness function f i ；
Step3, calculating individual fitness: calculating the fitness f of each sparrow i Selecting the current optimal fitness f g And its corresponding position x b And the current worst fitness f w And its corresponding position x w ；
Step4, position updating: selecting front p with excellent fitness Num Taking a sparrow as a finder and the rest as an enrollee, and then updating the positions of the finder and the enrollee; random selection of s from sparrow population Num Carrying out reconnaissance and early warning on sparrows and updating the positions of the sparrows;
step5, updating the fitness value: after one iteration is finished, the fitness value f of each sparrow is recalculated i And average fitness value f of sparrow population avg ；
Step6, updating the optimal position x of the whole population according to the current state of the sparrow population b And its fitness f g And the worst position x w And its fitness f w 。
The invention also provides a device for calculating the collision risk of a ship, which comprises a memory and a processor, wherein the memory stores at least one program, and the at least one program is executed by the processor to realize the method for calculating the collision risk of the ship.
A computer-readable storage medium, in which at least one program is stored, which at least one program is executed by a processor to implement the method of calculating a risk of collision of a vessel as claimed in any one of the above.
According to the technical scheme, the method has the following technical effects:
according to the technical scheme for calculating the ship collision risk provided by the embodiment of the invention, firstly, target AIS data information of a ship is obtained, meeting situation information is obtained according to the target AIS data information, and ship collision parameters are calculated by utilizing the meeting situation information; secondly, meeting situation data which accord with preset screening conditions of the ship meeting situation data are determined; then constructing a classification model of a support vector machine, and learning and classifying the meeting situation data by combining a preset collision avoidance rule to obtain a ship meeting direction map; and finally, constructing a BP neural network for predicting the collision risk value of the target ship and the ship in the meeting direction map. The scheme provides a meeting classification chart and collision risk degree of calculation make boats and ships can judge each boats and ships navigation state in advance at the navigation in-process, predicts the possible collision risk of boats and ships, to reducing marine collision accident, ensures navigation safety, has important meaning.
Drawings
Fig. 1 is a schematic flow chart of a method for calculating a collision risk of a ship according to an embodiment of the present invention;
FIG. 2 is a schematic view of a combination of a geodetic fixed coordinate system and a shipboard coordinate system in the method for calculating a collision risk of a ship according to an embodiment of the invention;
fig. 3 is a schematic diagram of an exemplary situation type identification model used in the method for calculating a collision risk of a ship according to an embodiment of the present invention;
FIG. 4 is a diagram of a ship encounter situation map obtained in a method for calculating a ship collision risk according to an embodiment of the present invention;
fig. 5 is a schematic flow chart of constructing a BP neural network in the method for calculating a collision risk of a ship according to an embodiment of the present invention;
fig. 6 is a schematic diagram of an euclidean space distance function involved in a basis function fitting output expression in constructing a BP neural network in the method for calculating a collision risk of a ship according to an embodiment of the present invention;
fig. 7 is a schematic diagram of an apparatus for calculating a collision risk of a ship according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The invention will now be further described with reference to the drawings and the detailed description.
The first embodiment is as follows:
fig. 1 is a schematic flow chart of a method for calculating a collision risk of a ship according to an embodiment of the present invention. As illustrated in fig. 1, the method of the embodiment of the present invention includes:
s1, acquiring target AIS data information of a ship, acquiring meeting situation information according to the target AIS data information, and calculating a ship collision parameter by using the meeting situation information. Specifically, the target AIS data information comprises speed, direction and position, the meeting situation information comprises relative distance, relative direction and ship course crossing angle of the ship, and the collision parameters comprise relative distance, relative speed, minimum meeting distance DCPA and minimum meeting time TCPA of the ship.
And S2, screening meeting situation data of the ship meeting the screening condition according to the meeting situation information and the ship collision parameter by using the preset screening condition of the ship meeting situation data.
And S3, constructing a classification model of the ship support vector machine, and learning and classifying the meeting situation data of the preselected samples meeting the screening conditions of the previous step according to the classification model of the support vector machine and a preset collision avoidance rule to obtain a ship meeting direction bitmap.
And S4, constructing a BP neural network, optimizing and determining each layer of node parameters of the BP neural network by adopting a sparrow search algorithm, and predicting the collision risk value of the target ship and the ship in the meeting orientation map, wherein the input items of the BP neural network are the course of the target ship, the speed of the target ship, the distance between the two ships and the relative orientation of the two ships, and the output items are the collision risk value.
Specifically, the method comprises the following steps:
firstly, target AIS data information of a ship is obtained, meeting situation information is obtained according to the target AIS data information, and ship collision parameters are calculated by utilizing the meeting situation information. Specifically, the target AIS data information comprises speed, direction and position, the meeting situation information comprises relative distance, relative direction and ship course crossing angle of the ship, and the collision parameters comprise relative distance, relative speed, minimum meeting distance DCPA and minimum meeting time TCPA of the ship.
Secondly, screening meeting situation data of the ship and the target ship according to the meeting situation information and the ship collision parameter by using preset screening conditions of the ship meeting situation data, wherein the meeting situation data accords with the screening conditions and is determined by the meeting situation information.
Fig. 2 is a schematic view of a combination of a geodetic fixed coordinate system and a shipboard coordinate system in the method for calculating a ship collision risk according to an embodiment of the present invention. Specifically, as shown in fig. 2, TC is the course of the ship, Q is the angle of the ship when the ship looks at the ship, and TB is the true azimuth of the ship when the ship looks at the ship; in a similar way, TB 1 Heading of his ship, Q 1 Viewing the angle, TB, of the ship for other ships 00 Looking at the true orientation of the vessel for other vessels. TC is related to TB, Q:
TB＝Q+TC。
the conversion relation between the ship-borne coordinates and the geodetic fixed coordinates is as follows: [ X, Y ]]＝[x,y]·A+[X 0 ,Y 0 ]Wherein [ x, y]Is the ship-to-ship coordinates of the ship,
[X 0 ,Y 0 ]coordinates of the current time of the ship in the geodetic fixed coordinate system。
Let the geodetic coordinate of the ship be (x) 0 ,y 0 ) Speed of voyage v 0 Course of C 0 (ii) a The geodetic coordinate of the target vessel is (x) R ,y R ) V speed of flight 1 Heading of C 1 Let Δ x = x R -x 0 ，Δy＝y R -y 0 Then, the first step is executed,
distance between two ships
Relative speed
Wherein the relative velocity on the X-axis
Relative velocity in the Y-axis
Relative course, α T Is the true orientation of the own ship relative to the target ship.
Specifically, fig. 3 is a schematic diagram of a typical situation type identification model used in the method for calculating a ship collision risk according to an embodiment of the present invention. As shown in fig. 3, other vessels from 0 ° to 5 ° or 355 ° to 360 ° right ahead of the bow are all opposite-meeting vessels, other vessels from 5 ° to 112.5 ° or 247.5 ° to 355 ° starboard are all cross-meeting vessels, and other vessels from 112.5 ° to 247.5 ° right behind the stem of the own vessel are all overtaking vessels.
According to the situation type identification model, screening conditions of situation data encountered by the ship are set, which are respectively as follows:
the screening conditions of the meeting situation are set as follows: the two ships are motor ships, the directions of the two ships are opposite or nearly opposite, and the target ship is positioned in a sector azimuth in the range of 355-360 degrees or 0-5 degrees of the ship; the overtaking situation screening condition is set as follows: the velocity vector component of the rear ship in the motion direction of the front ship exceeds the velocity of the front ship, the distance between the two ships is less than 4 nautical miles, and the target ship is positioned in a sector azimuth of 112.5-247.5 degrees of the ship; the cross-meeting situation screening conditions are set as follows: the minimum meeting distance DCPA of the two ships is less than 2 nautical miles, the minimum meeting time TCPA is more than 0, and the target ship is located in a sector position of 5-112.5 degrees or 247.5-355 degrees of the ship.
Specifically, according to the 3 meeting situations, the AIS data of the ships with the meeting are classified according to the meeting, cross meeting and overtaking situations. The AIS track data of the ship is converted into relative motion data of the ship to the ship, and information such as the relative position, DCPA, TCPA and relative distance of the ship meeting the AIS track data is calculated.
Specifically, the processing of the ship AIS raw data acquired from the AIS base station includes:
eliminating abnormal data in the ship AIS original data; decoding the AIS data of the ship from which the abnormal data are removed, and matching the decoded AIS data according to a preset matching rule to obtain ship position information; according to the position information, constructing a track of the ship in each time period to obtain a ship track set, wherein the track of each time period comprises a starting point, an end point and a track characteristic point;
according to the coordinate conversion method, the information in the ship track set is converted into a geodetic fixed coordinate system XOY, and target AIS data information is obtained; acquiring meeting situation information including speed, direction, position and the like according to the target AIS data information; further, by using the method, the meeting situation information is processed, and the data of the collision parameters of the ship including the relative distance, the relative speed, the minimum meeting distance DCPA and the minimum meeting time TCPA are obtained through calculation.
And thirdly, constructing a classification model of the ship support vector machine, and learning and classifying the meeting situation data of the preselected samples meeting the screening conditions according to the classification model of the support vector machine and a preset collision avoidance rule to obtain a ship meeting direction map.
Specifically, a Support Vector Machine (SVM) was proposed in 1964, and was rapidly developed and a series of improvement and extension algorithms were derived after the 90 s of the twentieth century, and the SVM was applied to pattern recognition problems such as portrait recognition and text classification. The algorithmic idea behind the support vector machine is to assume that linear separable cannot be directly achieved at the original spatial sample points, and then use a non-linear transformation process to transform these points into a corresponding high-dimensional feature space. Then, various optimization algorithms are adopted to obtain the maximum classification interval, so that the sample points can be linearly separable in the high-dimensional space obtained by conversion. Wherein, there will be a part of sample points above the hyperplane of the maximum classification interval, and these sample points are the support vector points. The earliest proposals for SVM were to deal with the classification problem of linearly separable samples, assuming a data set of samples of (x) 1 ,y 1 )，(x 2 ,y 2 )，…，(x n ,y n ) X ∈ R, y ∈ { -1,1}, i =1,2. Wherein N is the number of samples, N is the dimension of the samples, x is the samples, and y is the sample category. The avoidance actions include port steering, starboard steering, and direction and speed maintenance. The text sets input parameters: the distance between two ships, the azimuth angle of the ship, the azimuth angle of the target ship, the navigational speed of the ship and the navigational speed of the target ship. The output parameter is the type of the meeting situation of the ship. Based on AIS data in the Fujian coastal region within three months selected as samples, an SVM training data sample set is constructed on the basis of meeting situation data extracted by the method in the foregoing. The SVM linearity problem can be transformed into the following equation.
Wherein, [ omega ] T x i +b]Representing the distance of point x from the hyperplane. For the nonlinear classification problem, when the sample data set is nonlinearly divisible in a low-dimensional space, the sample points are converted into a high-dimensional feature space through mapping conversion, and the sample data set (x) is divided linearly in the space 1 ,x 1 )，(x 2 ,x 2 )，…，(x n ,x n ) Mapping to (phi) 1 ,Φ 1 )，(Φ 2 ,Φ 2 )，…，(Φ n ,Φ n ). And then solving the generalized linear SVM classification decision function in a high-dimensional space. That is, a very complex classifier must be built in the low-dimensional space, the sample points are transformed to a high-dimensional space through mapping, and then the optimal hyperplane is solved in the new space. Wherein, the classifier of the high-dimensional space after conversion is simpler than that of the low-dimensional space. When introducing Lagrange multiplier alpha i Then the dual problem of the original problem is converted into the formula:
in the situation problem of ship, if phi (x) does not appear independently, if some function K (x) exists i ,x j )＝Φ(x i ) T Φ(x j ) When the function of this equation is called a kernel function, since the kernel function is easier to solve than Φ (x), the above equation can be converted into the following equation by converting the nonlinear vector machine into its dual problem:
the dual problem must have a solution, which is α = (α) 1 *，...α l *) T So that:
from the above analysis, a mathematical description of the encounter situation support vector machine for determining the ship encounter situation map can be established:
(1) let the known ship sample number set (x) i ,y i ) Wherein x belongs to R, y belongs to { -1,1}, i =1,2., N;
(2) selecting a proper kernel function K (x) i ,y i ) And an appropriate parameter C, converting the support vector machine problem of the nonlinear sample of the ship into an optimization problem:
wherein alpha is i Is a lagrange multiplier;
(3) solving the optimization problem to obtain
(4) Selecting a * A positive component of 0 ≦ a * C is less than or equal to C, and a threshold value is calculated
(5) Constructing a decision function
And respectively learning and classifying the ship encounter, cross encounter and overtaking data extracted based on the situation type identification model and the set data screening condition based on the constructed SVM classification model. And combining all SVM classification results with collision avoidance rules to obtain a ship meeting direction map.
In addition, for the convenience of use of the crew, all classification angles are rounded in combination with actual navigation conditions to obtain a final ship encounter azimuth map, as shown in fig. 4. Wherein, the distance parameters of each target ship in the same azimuth from the ship in the azimuth map are consistent and the collision avoidance measures can be taken are consistent, and the parameters are shown in table 1:
TABLE 1
| Azimuth area | Distance between two boats (NM) | Azimuth angle theta (°) |
| A | Less than 6NM | 2<θ≤5 |
| B | Less than 6NM | 5<θ≤45 |
| C | Less than 3NM | 45<θ≤110 |
| D | Less than 3NM | 110<θ≤200 |
| E | Less than 3NM | 200<θ≤250 |
| F | Less than 3NM | 250<θ≤310 |
| G | Less than 3NM | 310<θ≤355 |
| H | Less than 3NM | 355θ≤2 |
Illustratively, learning and classifying the meeting situation data extracted under different situation type identification models and data screening conditions set by the identification models, and the obtained meeting situation maps may have certain differences.
And finally, constructing a BP neural network, optimizing and determining each layer of node parameters of the BP neural network by adopting a sparrow search algorithm, and predicting the collision risk value of the target ship and the ship in the meeting orientation diagram, wherein input items of the BP neural network are the course of the target ship, the speed of the target ship, the distance between the two ships and the relative orientation of the two ships, and output items are the collision risk value.
Fig. 5 is a schematic flow chart of constructing a BP neural network in the method for calculating a collision risk of a ship according to an embodiment of the present invention. The flow shown in fig. 5 includes: carrying out normalization processing on the training data of the BP neural network, wherein the normalization processing comprises elimination of abnormal values in AIS original data as an example; initializing BP algorithm parameters, including processing AIS original data to obtain intermediate data; determining input and output parameters; calculating node outputs of a hidden layer and an output layer; calculating training and calculating errors; if the preset error condition is met, the BP neural network training is finished and can be used for calculating the risk degree of the newly collected ship AIS data, otherwise, the hidden layer error is calculated, the weight is adjusted according to the error gradient, and the input and output parameters are re-determined for continuing the training.
Specifically, based on the above ship meeting situation identification model, a neural network is combined with a fuzzy comprehensive evaluation method to calculate the ship collision risk more objectively. And obtaining collision risk degree information such as the speed, the direction, the course, the positions of the two ships and the like of the ship based on the AIS information after the early-stage processing.
3000 samples in total are taken as training data sets of the BP neural network station, 1000 network test samples in total are taken as network test samples, AIS (automatic identification System) original data of each accident contained in the existing 2000 collision accidents, intermediate data obtained by calculation and CR (cognitive radio) obtained by a fuzzy comprehensive evaluation method are selected i The value is obtained.
Calculating the risk value CR of each ship in the ship collision accident by using a fuzzy comprehensive evaluation method i And using the risk value as an expected output value Ex of the BP neural network. And selecting mean square error
Evaluating a training result of the BP neural network model, and finishing training when error MSE meets a preset condition, wherein X is a predicted value of a collision risk value of the current BP neural training network, and T =2000 is the set AIS ship data quantity.
In the embodiment, 4 neural network input items such as the target ship course, the speed of the other ship, the relative distance between the two ships, the relative direction of the two ships and the like are selected. Then calculated by fuzzy comprehensive evaluation method i As the desired output value of the neural network.
Wherein the CR calculated by the fuzzy comprehensive evaluation method i The method comprises the following specific steps:
is provided with
The value range is [0,1 ] for the risk membership degree of each parameter of the target ship i meeting the ship]，i＝1,2,...,n；a D 、a K 、a B 、a DCPA 、a TCPA The weight of the risk membership degree of each parameter is in the value range of [0,1]And a is a D +a K +a B +a DCPA +a TCPA =1 degree of risk of collision between own ship and target ship i
The distance membership function
Membership function of speed ratio of two ships
Membership function of relative orientation
Membership function of DPCA
TCPA membership function
The calculation formulas of (A) and (B) are respectively as follows:
In this implementation, the BP neural network performs function approximation training, that is, the neural network is trained to be a nonlinear function capable of representing a mapping relationship between collision risk degrees of input data items and output items.
The basis function fitting output expression is:
A＝F l (||W-p||·B)＝radbas(||W-p||·B)
in the formula, radbas is the basis of the radial basis function, and usually the basis set has orthogonality. W-p is a euclidean spatial distance function as shown in fig. 6.
The distance function can be expressed as:
for different hidden layers, the input and output parameters have the same dimension, and the input vector is:
P＝[p1,p2,...,pn]
wherein n is the number of hidden layers of the neural network.
The activation function tansig can be calculated as follows:
the threshold function is:
B l ＝[B1,B2,...,Bn] l ，
then the ith hidden layer intermediate output vector can be defined as:
n l ＝K l +B l ，
K l is the connection weight vector between the input end and the threshold function, and l is the dimension;
the intermediate layer operation result is defined as:
A l ＝F l (K l +B l )，
according to the same iterative processing process, the weight and the threshold vector of each layer can be obtained, and the final output value is obtained:
A＝F i+1 (K 2 F i (K l +B l )+B i )，
and obtaining a smaller fitting error according to the neural network collision risk degree model to obtain a satisfactory effect, and then predicting the ship collision risk degree by combining the fuzzy comprehensive evaluation model and predicting the ship collision risk degree by combining the fuzzy comprehensive evaluation model.
In order to determine the parameters of the BP neural network, the embodiment of the invention introduces a sparrow search algorithm to perform optimization operation on the parameters of each layer of nodes of the typical BP neural network. The sparrow search algorithm can optimize the parameters of each layer of nodes of a typical BP neural network, so that the influence of training stagnation and local minimum value problems on a model prediction result is reduced. The method comprises the following specific steps:
step1: initializing parameters of a sparrow searching algorithm. The initialization parameters comprise population scale, the number of discoverers, the number of sparrows for reconnaissance and early warning, the dimension of a target function, the upper and lower bounds of an initial value, the maximum iteration number or the solving precision.
Step2: the total group is initialized. Initializing the population, including generating an initial population and determining a fitness function. In the model, the network error is selected as a fitness function f i . The embodiment initializes the population by using Tent chaotic sequence.
Step3: and calculating the individual fitness. Calculating the fitness f of each sparrow i Selecting the current optimal fitness f g And its corresponding position x b And the current worst fitness f w And its corresponding position x w 。
Step4: and (4) updating the position. Selecting front p with excellent fitness Num The sparrows are treated as discoverers and the rest are treated as joiners, and then the locations of the discoverers and joiners are updated. Random selection of s from sparrow population Num And (5) carrying out reconnaissance and early warning on the sparrows and updating the positions of the sparrows.
Step 5: and updating the fitness value. After one iteration is finished, recalculating the fitness value f of each sparrow i And average fitness value f of sparrow population avg . When f is i ≤f avg If the result is better than that of the individuals before the variation, the individuals after the variation are used for replacing the individuals before the variation, otherwise, the original individuals are kept unchanged. When f is i ＞f avg And if the performance of the disturbed individual is better, the disturbed individual is used for replacing the individual before disturbance, otherwise, the original individual is kept unchanged.
Step6: and (6) judging. According to the current state of the sparrow population, updating the optimal position x experienced by the whole population b And its adaptationDegree f g And the worst position x w And its fitness f w 。
The meeting azimuth map of the ship is constructed through the method, the collision risk value of the ship and other target ships is obtained through BP neural network prediction, and then collision avoidance operation is carried out according to the specific direction and the collision risk of the target ships and the preset collision avoidance rules.
In the method of the embodiment, target AIS data information of a ship is obtained firstly, meeting situation information is obtained according to the target AIS data information, and ship collision parameters are calculated by utilizing the meeting situation information; secondly, meeting situation data which accord with preset screening conditions of the ship meeting situation data are determined; then constructing a classification model of a support vector machine, and learning and classifying the meeting situation data by combining a preset collision avoidance rule to obtain a ship meeting direction map; and finally, constructing a BP neural network for predicting the collision risk value of the target ship and the ship in the meeting direction diagram. In the process of ship navigation, the navigation state of each ship can be judged in advance, the possible collision risk of the ship can be predicted, and the method has important significance for reducing the collision accident at sea and ensuring the navigation safety.
The second embodiment:
the present invention also provides an apparatus for calculating a ship collision risk, as shown in fig. 7, the apparatus includes a processor 701, a memory 702, a bus 703, and a computer program stored in the memory 702 and executable on the processor 701, the processor 701 includes one or more processing cores, the memory 702 is connected to the processor 301 through the bus 703, the memory 702 is used for storing program instructions, and the processor implements the steps in the above-mentioned method embodiment of the first embodiment of the present invention when executing the computer program.
Further, as an executable solution, the device for calculating the collision risk of the ship may be a computer unit, and the computer unit may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer unit may include, but is not limited to, a processor, a memory. It will be appreciated by those skilled in the art that the above-described constituent structures of the computer unit are merely examples of the computer unit, and do not constitute a limitation of the computer unit, and may include more or less components than those described above, or combine some components, or different components. For example, the computer unit may further include an input/output device, a network access device, a bus, and the like, which is not limited in this embodiment of the present invention.
Further, as an executable solution, the Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, a discrete hardware component, and the like. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like, the processor being the control center for the computer unit, various interfaces and lines connecting the various parts of the overall computer unit.
The memory may be used to store the computer programs and/or modules, and the processor may implement the various functions of the computer unit by running or executing the computer programs and/or modules stored in the memory, as well as by invoking data stored in the memory. The memory can mainly comprise a program storage area and a data storage area, wherein the program storage area can store an operating system and an application program required by at least one function; the storage data area may store data created according to the use of the mobile phone, and the like. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), at least one magnetic disk storage device, a Flash memory device, or other volatile solid state storage device.
Example three:
the present invention also provides a computer-readable storage medium, which stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned method of an embodiment of the present invention.
The computer unit integrated module/unit, if implemented in the form of a software functional unit and sold or used as a separate product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the flow of the method according to the embodiments of the present invention may also be implemented by a computer program, which may be stored in a computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method embodiments may be implemented. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, usb disk, removable hard disk, magnetic disk, optical disk, computer Memory, read-Only Memory (ROM), random Access Memory (RAM), software distribution medium, and the like. It should be noted that the computer readable medium may contain content that is appropriately increased or decreased as required by legislation and patent practice in the jurisdiction.
While the invention has been particularly shown and described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. A method of calculating a risk of collision for a vessel, comprising:
s1, acquiring target AIS data information of a ship, acquiring meeting situation information according to the target AIS data information, and calculating ship collision parameters by using the meeting situation information, wherein the target AIS data information comprises speed, direction and position, the meeting situation information comprises relative distance, relative direction and ship course crossing angle of the ship, and the collision parameters comprise the relative distance, relative speed, minimum meeting distance DCPA and minimum meeting time TCPA of the ship;
s2, screening meeting situation data of the ship meeting the screening condition according to the meeting situation information and the ship collision parameter by using the preset screening condition of the ship meeting situation data, wherein the meeting situation data is determined by the meeting situation information;
s3, constructing a classification model of the ship support vector machine, and learning and classifying the meeting situation data of the preselected sample according to the classification model of the support vector machine and a preset collision avoidance rule to obtain a ship meeting direction map;
and S4, constructing a BP neural network, optimizing and determining each layer of node parameters of the BP neural network by adopting a sparrow search algorithm, and predicting the collision risk value of the selected ship and the target ship according to the meeting direction bitmap, wherein input items of the BP neural network are the course of the target ship, the speed of the target ship, the distance between the two ships and the relative direction of the two ships, and output items are the collision risk value.
2. The method of claim 1, wherein the step of obtaining target AIS data information for the vessel in step S1 comprises:
s11, eliminating abnormal data in the ship AIS original data;
s12, decoding the ship AIS data after the abnormal data are removed to obtain ship position information;
s13, constructing a track of the ship in each time period according to the position information to obtain a ship track set, wherein the track of each time period comprises a starting point, an end point and a track characteristic point;
and S14, establishing a ship geodetic fixed coordinate system XOY, converting the information in the track set into the coordinate system XOY, and acquiring the target AIS data information, wherein the origin of the geodetic fixed coordinate system XOY is a preset point, the X axis points to the true east, and the Y axis points to the true north.
3. The method according to claim 1, wherein the screening condition of the ship encounter situation data preset in step S2 comprises:
setting one ship as a ship and the other ship as a target ship, and setting a ship-associated coordinate system xoy, wherein the ship-associated coordinate system xoy takes the center of mass of the ship as an origin o, the y axis points to the bow and is 0 degree of azimuth, the x axis points to the right transverse line of the ship,
the screening conditions of the meeting situation are set as follows: the two ships are motor ships, the directions of the two ships are opposite or nearly opposite, and the target ship is positioned in a sector azimuth in the range of 355-360 degrees or 0-5 degrees of the ship;
the overtaking situation screening condition is set as follows: the velocity vector component of the rear ship in the motion direction of the front ship exceeds the velocity of the front ship, the distance between the two ships is less than 4 nautical miles, and the target ship is positioned in a sector azimuth of 112.5-247.5 degrees of the ship;
the cross-meeting situation screening conditions are set as follows: the minimum meeting distance DCPA of the two ships is less than 2 nautical miles, the minimum meeting time TCPA is more than 0, and the target ship is located in a sector position of 5-112.5 degrees or 247.5-355 degrees of the ship.
4. The method of claim 1, wherein the step of constructing a support vector machine classification model of the ship in step S3 comprises:
s31, setting a known ship sample number set (x) i ,y i ) Wherein x belongs to R, y belongs to { -1,1}, i =1,2., N;
s32, selecting a proper kernel function K (x) i ,y i ) And an appropriate parameter C, converting the support vector machine problem of the nonlinear sample of the ship into an optimization problem:
wherein alpha is i Is a lagrange multiplier;
s33, solving the optimization problem to obtain
S34, selecting alpha * A positive component of 0. Ltoreq. Alpha * C is less than or equal to C, and a threshold value is calculated
S35, constructing a decision function
5. The method of claim 1, wherein the step S3 of relating the angle θ at which each azimuth area divided by the azimuth map is located to the distance between each ship and the ship at the center point of the azimuth map comprises: 2 degrees < theta is less than or equal to 5 degrees, and the distance between the two ships is less than 6 nautical miles; theta is more than 5 degrees and less than or equal to 45 degrees, and the distance between the two ships is less than 6 nautical miles; theta is more than 45 degrees and less than or equal to 110 degrees, and the distance between the two ships is less than 3 nautical miles; the angle is more than or equal to 110 degrees and less than or equal to 200 degrees, and the distance between the two ships is less than 3 nautical miles; theta is less than or equal to 250 degrees at an angle of 200 degrees, and the distance between the two ships is less than 3 nautical miles; theta is less than or equal to 310 degrees at 250 degrees, and the distance between the two ships is less than 3 nautical miles; theta is more than 310 degrees and less than or equal to 355 degrees, and the distance between the two ships is less than 3 nautical miles; 355 degrees < theta is less than or equal to 360 degrees or 0 degrees < theta is less than or equal to 2 degrees, and the distance between the two ships is less than 3 nautical miles.
6. The method according to claim 1, wherein the step of constructing the BP neural network in step S4 includes training the BP neural network, and performing error evaluation on a risk value output by the current BP neural training network and an expected output value when the BP neural network is trained, wherein the expected output value is obtained by a comprehensive evaluation method, and specifically includes:
s41, acquiring ship AIS original data in a ship collision accident in advance as a data source of a fuzzy comprehensive evaluation method, and processing according to the steps S1-S3;
s42, calculating the risk value of each ship in the ship collision accident by using a fuzzy comprehensive evaluation method, and taking the risk value as an expected output value Ex of the BP neural network;
s43, selecting the result of the mean square error evaluation prediction model
And when the error MSE meets a preset condition, finishing training, wherein X is a predicted value of the collision risk value of the current BP neural training network, and T is the set AIS ship data number.
7. The method of claim 6, wherein the step of calculating the risk value of each ship in the ship collision accident in step S42 comprises:
is provided with
The value range is [0,1 ] for the risk membership degree of each parameter of the target ship i meeting the ship]，i＝1,2,...,n；a D 、a K 、a B 、a DCPA 、a TCPA The weight of the risk membership degree of each parameter is in the value range of [0,1]And a is a D +a K +a B +a DCPA +a TCPA =1 degree of risk of collision between own ship and target ship i
The above-mentioned
The calculation formula (2) includes:
8. The method according to claim 1, wherein the step S4 of optimizing and determining the node parameters of each layer of the BP neural network by using a sparrow search algorithm comprises the following steps:
step1, initializing parameters of a sparrow search algorithm, wherein the initialized parameters comprise population scale, the number of discoverers, the number of sparrows for reconnaissance and early warning, the dimension of an objective function, the upper and lower bounds of an initial value, the maximum iteration times or solving precision;
step2: initializing a total group, generating an initial group, determining a fitness function, and selecting a network error as the fitness function f i ；
Step3, calculating individual fitness: calculating the fitness f of each sparrow i Selecting the current optimal fitness f g And its corresponding position x b And the current worst fitness f w And its corresponding position x w ；
Step4, position updating: selecting front p with excellent fitness Num Taking sparrows as discoverers and the rest as joiners, and then updating the positions of the discoverers and the joiners; random selection of s from sparrow population Num Carrying out reconnaissance and early warning on sparrows and updating the positions of the sparrows;
step5, updating the fitness value: after one iteration is finished, the fitness value f of each sparrow is recalculated i And average fitness value f of sparrow population avg ；
Step6, updating the optimal position x of the whole population according to the current state of the sparrow population b And its fitness f g And the worst position x w And its fitness f w 。
9. An apparatus for calculating a collision risk of a vessel, the apparatus comprising a memory and a processor, the memory storing at least one program, the at least one program being executable by the processor to perform the method of calculating a collision risk of a vessel according to any one of claims 1 to 8.
10. A computer-readable storage medium, in which at least one program is stored, which at least one program is executed by a processor to carry out the method of calculating a risk of collision of a vessel according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211042282.2A CN115510737A (en) | 2022-08-29 | 2022-08-29 | Method and device for calculating ship collision risk and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211042282.2A CN115510737A (en) | 2022-08-29 | 2022-08-29 | Method and device for calculating ship collision risk and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115510737A true CN115510737A (en) | 2022-12-23 |
Family
ID=84501889
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211042282.2A Pending CN115510737A (en) | 2022-08-29 | 2022-08-29 | Method and device for calculating ship collision risk and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115510737A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116107328A (en) * | 2023-02-09 | 2023-05-12 | 陕西科技大学 | Optimal automatic obstacle avoidance method for ornithopter based on improved genetic algorithm |
| CN116993167A (en) * | 2023-09-27 | 2023-11-03 | 交通运输部水运科学研究所 | Real-time risk judging method and system in production process |
| CN117556172A (en) * | 2024-01-11 | 2024-02-13 | 青岛哈尔滨工程大学创新发展中心 | Ship maneuvering motion prediction model construction method and ship maneuvering motion prediction method |
| CN117826824A (en) * | 2024-01-18 | 2024-04-05 | 交通运输部水运科学研究所 | Ship self-adaptive autonomous navigation decision-making method and device and electronic equipment |
-
2022
- 2022-08-29 CN CN202211042282.2A patent/CN115510737A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116107328A (en) * | 2023-02-09 | 2023-05-12 | 陕西科技大学 | Optimal automatic obstacle avoidance method for ornithopter based on improved genetic algorithm |
| CN116993167A (en) * | 2023-09-27 | 2023-11-03 | 交通运输部水运科学研究所 | Real-time risk judging method and system in production process |
| CN116993167B (en) * | 2023-09-27 | 2023-12-15 | 交通运输部水运科学研究所 | Real-time risk judging method and system in production process |
| CN117556172A (en) * | 2024-01-11 | 2024-02-13 | 青岛哈尔滨工程大学创新发展中心 | Ship maneuvering motion prediction model construction method and ship maneuvering motion prediction method |
| CN117556172B (en) * | 2024-01-11 | 2024-04-19 | 青岛哈尔滨工程大学创新发展中心 | Ship maneuvering motion prediction model construction method and ship maneuvering motion prediction method |
| CN117826824A (en) * | 2024-01-18 | 2024-04-05 | 交通运输部水运科学研究所 | Ship self-adaptive autonomous navigation decision-making method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115510737A (en) | Method and device for calculating ship collision risk and storage medium | |
| Kaulbersch et al. | A cartesian b-spline vehicle model for extended object tracking | |
| Moreno-Salinas et al. | Modelling of a surface marine vehicle with kernel ridge regression confidence machine | |
| CN113128826B (en) | Navigation risk prediction method, navigation risk prediction device and storage medium | |
| CN111861155B (en) | Ship collision risk detection method, system, computer device and storage medium | |
| He et al. | Nonparametric modeling of ship maneuvering motion based on self-designed fully connected neural network | |
| Mittendorf et al. | Sea state identification using machine learning—A comparative study based on in-service data from a container vessel | |
| Mousavi et al. | A new support vector finder method, based on triangular calculations and K-means clustering | |
| CN114565861B (en) | Airborne downward-looking target image positioning method based on probability statistical differential homoembryo set matching | |
| Raj et al. | One-shot learning-based SAR ship classification using new hybrid Siamese network | |
| Yasukawa et al. | Investigation on the rudder force of a ship in large drifting conditions with the MMG model | |
| Yang et al. | A sea-sky-line detection method based on Gaussian mixture models and image texture features | |
| Zhu et al. | A circle grid-based approach for obstacle avoidance motion planning of unmanned surface vehicles | |
| Wu et al. | Sailing status recognition to enhance safety awareness and path routing for a commuter ferry | |
| CN114445456B (en) | Data-driven intelligent maneuvering target tracking method and device based on partial model | |
| CN111325094A (en) | High-resolution range profile-based ship type identification method and system | |
| US20240127585A1 (en) | Method for determining the encoder architecture of a neural network | |
| CN115511853A (en) | Remote sensing ship detection and identification method based on direction variable characteristics | |
| Oneto et al. | Crash stop maneuvering performance prediction: a data-driven solution for safety and collision avoidance | |
| Dinç et al. | Identification of hydrodynamic coefficients of AUV in the presence of measurement biases | |
| Shen et al. | Ship dynamics model identification based on Semblance least square support vector machine | |
| Luo et al. | An unsupervised learning based simplification on ship motion model and its verification | |
| Yu et al. | Virtual adversarial active learning | |
| US20230419640A1 (en) | Object classification based on spatially discriminated parts | |
| Atalar et al. | Ship Classification Using an Image Dataset |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |