Abnormal ship trajectory detection method based on group behavior law recognition

Abstract

The disclosed embodiment provides a method for detecting abnormal ship trajectories based on the recognition of group behavior rules, which belongs to the field of data processing technology, and specifically includes: step 1, preprocessing the AIS base station data of the target ship; step 2, extracting the spatial attribute characteristics of the target ship according to the AIS base station data after the preprocessing operation and detecting the abnormal behavior of the spatial attribute characteristics accordingly; step 3, combining the spatial attribute characteristics and thematic attribute characteristics, and detecting and scoring the abnormal behavior of the target ship based on the isolation forest algorithm. Through the scheme disclosed in the present invention, the trajectory characteristics of ships with group regular behaviors in existing studies are sorted out in combination with actual geographical environment factors, and the scoring values ​​of the abnormal degree of trajectory points are obtained from the perspectives of spatial attributes and thematic attributes, so as to evaluate the ship movement pattern and further identify abnormal ship behaviors.

Description

Technical Field

[0001] The disclosed embodiments relate to the field of data processing technology, and in particular to a method for detecting abnormal ship trajectories based on the recognition of group behavior rules. Background Art

[0002] At present, maritime transportation is favored and developed rapidly due to its advantages such as large cargo volume, low energy consumption and low cost. In order to ensure navigation safety, AIS base stations came into being. The AIS system provides data support for maritime traffic analysis and modeling, and also enhances the feasibility of research on abnormal behavior detection of ships. Ship behavior is not the specific behavior of a certain ship, but refers to the mode and law of similar actions of a group of ships. It not only depends on human consciousness, thinking, decision-making and manipulation, but also is affected by the characteristics of the ship itself and the surrounding environment. Existing research has not effectively combined the characteristics of ship behavior to identify abnormal patterns. Since the occurrence of maritime accidents can usually be attributed to abnormal behavior of ships, this study abstracts the characteristics of ship behavior from three dimensions: spatial attributes, temporal attributes and thematic attributes of maritime accidents, obtains data form expression of abnormal behavior of ships, and detects abnormal behavior of groups of ships.

[0003] Most of the existing studies have reviewed the research on abnormal behavior of ships based on their respective fields and research methods. According to the literature, the research on abnormal behavior detection of ships can be divided into the following three categories according to the results: the detection result is only abnormal in spatial position, the detection result is abnormal in spatial position combined with thematic attributes, and the detection result is abnormal in specific situations.

[0004] The detection method whose detection result is spatial position anomaly only considers spatial attributes and is too single, and the accuracy of the machine learning method is affected by the set threshold size. The detection method whose detection result is spatial position combined with thematic attribute anomaly is difficult to effectively select and detect in combination and how to interpret the detection results. At the same time, due to the limitations of the selected detection method, the research still needs to improve the selection of thresholds and simplify the complexity of algorithms. The detection method whose detection result is specific situational anomaly only verifies certain factors and ignores other factors, which will also lead to errors.

[0005] It can be seen that there is an urgent need for an efficient, accurate and adaptable abnormal ship trajectory detection method based on the recognition of group behavior rules. Summary of the invention

[0006] In view of this, an embodiment of the present disclosure provides a method for detecting abnormal ship trajectories based on the recognition of group behavior rules, which at least partially solves the problems of poor detection efficiency, accuracy and adaptability in the prior art.

[0007] The present disclosure provides a method for detecting abnormal ship trajectories based on group behavior law recognition, including:

[0008] Step 1, pre-processing the AIS base station data of the target ship;

[0009] Step 2, extracting the spatial attribute characteristics of the target ship according to the AIS base station data after the preprocessing operation and detecting abnormal behavior of the spatial attribute characteristics accordingly;

[0010] Step 3: Combining the spatial attribute features and thematic attribute features, the abnormal behavior of the target ship is detected and scored based on the isolation forest algorithm.

[0011] According to a specific implementation of the embodiment of the present disclosure, step 1 specifically includes:

[0012] Step 1.1, matching the trajectory points of the target ship with the map;

[0013] Step 1.2, delete the illogical data;

[0014] Step 1.3, cutting the trajectory of the target ship from the perspectives of space and time respectively;

[0015] Step 1.4, use the Kalman filter algorithm to smooth the trajectory data after cutting;

[0016] Step 1.5, according to the concavity and convexity of the trajectory points, the Lagrange first-order interpolation algorithm and the Lagrange second-order interpolation algorithm are respectively used to complete the trajectory line until the interval condition of the trajectory segment is met.

[0017] According to a specific implementation of the embodiment of the present disclosure, step 2 specifically includes:

[0018] Step 2.1, generating a channel network from the target trajectory point group based on main graph structure learning and tree connection strategy and calculating the best main graph and corresponding point set based on the main graph to form the spatial attribute feature;

[0019] Step 2.2, according to the optimal main image and the corresponding point set, the spatial attribute characteristic behavior of the target ship is detected based on the 3-sigma principle.

[0020] According to a specific implementation of the embodiment of the present disclosure, step 2.1 specifically includes:

[0021] Step 2.1.1, inversely mapping the target trajectory point group through a mapping mechanism to obtain an inverse mapping point group;

[0022] Step 2.1.2, learning an initial main graph from the target trajectory point group;

[0023] Step 2.1.3, initializing the inverse mapping point group and calculating the membership matrix between the inverse mapping point group and the target trajectory point group based on the inverse mapping point group;

[0024] Step 2.1.4, calculating a weight matrix of a graph when the latent space point set is represented by a graph according to the inverse mapping point group and the membership matrix;

[0025] Step 2.1.5, reversely solving the inverse mapping point group according to the membership matrix and the weight matrix to obtain an initial function;

[0026] Step 2.1.6, converting the initial function into an unconstrained quadratic programming problem to obtain an objective function;

[0027] Step 2.1.7, calculating the difference of the objective function between the graph before and after fitting based on the inverse mapping point group, the membership matrix and the initial main graph;

[0028] Step 2.1.8, iterate steps 2.1.3 to 2.1.7 in a loop until the difference between the objective functions is less than a threshold value and the loop ends;

[0029] Step 2.1.9, when the loop ends, the best main image and the corresponding point set are obtained to form the spatial attribute features.

[0030] According to a specific implementation of the embodiment of the present disclosure, step 2.2 specifically includes:

[0031] Step 2.2.1, based on the optimal main graph and the corresponding point set, using a normal probability density distribution function to calculate the probability that the route of the target ship belongs to the planned route;

[0032] Step 2.2.2, detecting abnormal behavior of the spatial attribute characteristics of the target ship based on the probability and the 3-sigma principle.

[0033] According to a specific implementation of the embodiment of the present disclosure, step 3 specifically includes:

[0034] Step 3.1, segmenting the waterway network in a clockwise direction with intersections as nodes based on ArcGIS tools, assigning unique identifiers to the segmented waterway network segments, and after completing the waterway segmentation, calculating the nearest waterway segment of each track point according to the minimum distance principle, and assigning it to the track point as an attribute;

[0035] Step 3.2, using the isolation forest algorithm to calculate the abnormality score of the input data group.

[0036] According to a specific implementation of the embodiment of the present disclosure, step 3.2 specifically includes:

[0037] Step 3.2.1, building a tree based on the input data set;

[0038] Step 3.2.2, randomly selecting a feature and a random cutting point, and cutting the feature based on the random cutting point;

[0039] Step 3.2.3, dividing the data space of the feature into two subspaces based on the random cutting point;

[0040] Step 3.2.4, looping and iterating step 3.2.2 to step 3.2.3 until a preset condition is met, wherein the preset condition includes any one of the following: the constructed isolated tree reaches the tree height, there is only one node on a certain node of the binary tree, and all sample features on the node are the same;

[0041] Step 3.2.5, repeat steps 3.2.2 to 3.2.4 to obtain a set of multiple trees to form an isolation forest;

[0042] Step 3.2.6, calculating the abnormality degree of the input data according to the isolation forest;

[0043] Step 3.2.7, normalize the value of the abnormality degree to obtain an abnormality score.

[0044] The abnormal ship trajectory detection scheme based on group behavior law recognition in the embodiment of the present disclosure includes: step 1, preprocessing the AIS base station data of the target ship; step 2, extracting the spatial attribute characteristics of the target ship according to the AIS base station data after the preprocessing operation and detecting the abnormal behavior of the spatial attribute characteristics accordingly; step 3, combining the spatial attribute characteristics and thematic attribute characteristics, and detecting and scoring the abnormal behavior of the target ship based on the isolation forest algorithm.

[0045] The beneficial effects of the embodiments of the present disclosure are as follows: through the scheme of the present disclosure, the trajectory characteristics of ships with group regular behaviors in existing studies are sorted out in combination with actual geographical environment factors, and the score value of the abnormality degree of the trajectory points is obtained from the perspective of spatial attributes and thematic attributes, so as to evaluate the ship movement pattern and further identify abnormal ship behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0047] Figure 1A flow chart of a method for detecting abnormal ship trajectories based on group behavior law recognition provided by an embodiment of the present disclosure;

[0048] Figure 2 A schematic diagram of a process for realizing abnormal ship trajectory detection based on group behavior law recognition provided by an embodiment of the present disclosure;

[0049] Figure 3 A schematic diagram of a waterway network detection result provided by an embodiment of the present disclosure;

[0050] Figure 4 A schematic diagram of a detection result of a ship behavior with abnormal spatial attributes provided by an embodiment of the present disclosure;

[0051] Figures 5 to 8 A schematic diagram of a detection result of ship behavior with abnormal thematic attributes provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0052] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0053] The following describes the embodiments of the present disclosure through specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0054] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described herein may be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein may be used to implement this device and / or practice this method.

[0055] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The drawings only show components related to the present disclosure rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0056] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0057] The embodiment of the present disclosure provides a method for detecting abnormal ship trajectories based on the recognition of group behavior rules, and the method can be applied to the ship accident warning process in marine transportation scenarios.

[0058] See also Figure 1 , is a flow chart of a method for detecting abnormal ship trajectories based on group behavior law recognition provided by an embodiment of the present disclosure. Figure 1 and Figure 2 As shown, the method mainly comprises the following steps:

[0059] Step 1, pre-processing the AIS base station data of the target ship;

[0060] Optionally, the step 1 specifically includes:

[0061] Step 1.1, matching the trajectory points of the target ship with the map;

[0062] Step 1.2, delete the illogical data;

[0063] Step 1.3, cutting the trajectory of the target ship from the perspectives of space and time respectively;

[0064] Step 1.4, use the Kalman filter algorithm to smooth the trajectory data after cutting;

[0065] Step 1.5, according to the concavity and convexity of the trajectory points, the Lagrange first-order interpolation algorithm and the Lagrange second-order interpolation algorithm are respectively used to complete the trajectory line until the interval condition of the trajectory segment is met.

[0066] In the specific implementation, first of all, the AIS data is preprocessed from four aspects: map matching, data cutting, data deletion, and data completion. Map matching is essentially the process of matching track points with maps; data deletion is to ensure the accuracy of the research, so it is necessary to delete illogical data; due to the randomness of the data obtained by the crawler, multiple voyages may be marked as one voyage, so it is necessary to cut the ship trajectory from the perspective of space and time; after completing the data cleaning and cutting, due to the turbulent marine environment, the shore-based satellite point drift is likely to be large, so the Kalman filter algorithm is used to smooth the ship trajectory data; for the data completion problem, the Lagrange first-order interpolation algorithm and the Lagrange second-order interpolation algorithm are used to complete the trajectory line according to the concavity of the trajectory points until the interval conditions of the trajectory segment are met.

[0067] Step 2, extracting the spatial attribute characteristics of the target ship according to the AIS base station data after the preprocessing operation and detecting abnormal behavior of the spatial attribute characteristics accordingly;

[0068] Furthermore, the step 2 specifically includes:

[0069] Step 2.1, generating a channel network from the target trajectory point group based on main graph structure learning and tree connection strategy and calculating the best main graph and corresponding point set based on the main graph to form the spatial attribute feature;

[0070] Step 2.2, according to the optimal main image and the corresponding point set, the spatial attribute characteristic behavior of the target ship is detected based on the 3-sigma principle.

[0071] Based on the above embodiment, step 2.1 specifically includes:

[0072] Step 2.1.1, inversely mapping the target trajectory point group through a mapping mechanism to obtain an inverse mapping point group;

[0073] Step 2.1.2, learning an initial main graph from the target trajectory point group;

[0074] Step 2.1.3, initializing the inverse mapping point group and calculating the membership matrix between the inverse mapping point group and the target trajectory point group based on the inverse mapping point group;

[0075] Step 2.1.4, calculating a weight matrix of a graph when the latent space point set is represented by a graph according to the inverse mapping point group and the membership matrix;

[0076] Step 2.1.5, reversely solving the inverse mapping point group according to the membership matrix and the weight matrix to obtain an initial function;

[0077] Step 2.1.6, converting the initial function into an unconstrained quadratic programming problem to obtain an objective function;

[0078] Step 2.1.7, calculating the difference of the objective function between the graph before and after fitting based on the inverse mapping point group, the membership matrix and the initial main graph;

[0079] Step 2.1.8, iterate steps 2.1.3 to 2.1.7 in a loop until the difference between the objective functions is less than a threshold value and the loop ends;

[0080] Step 2.1.9, when the loop ends, the best main image and the corresponding point set are obtained to form the spatial attribute features.

[0081] Furthermore, the step 2.2 specifically includes:

[0082] Step 2.2.1, based on the optimal main graph and the corresponding point set, using a normal probability density distribution function to calculate the probability that the route of the target ship belongs to the planned route;

[0083] Step 2.2.2, detecting abnormal behavior of the spatial attribute characteristics of the target ship based on the probability and the 3-sigma principle.

[0084] In the specific implementation, based on the AIS data after data cleaning, this study detects and evaluates abnormal ship behavior based on its spatial attribute characteristics and thematic attribute characteristics. First, the ship channel network is extracted to further explore its spatial attribute characteristics. The steps are as follows:

[0085] Based on the main graph structure learning and tree connection strategy, a road network G0 = (V0, E0, W0) is generated from the target trajectory point group, where V0 represents the nodes of the graph, E0 represents the edges of the graph, and W0 represents the connection matrix of the graph. The best fitting main graph can be expressed as G opt =(V opt ,E opt ,W opt ), the main steps are as follows:

[0086] Assume that X = {x1, x2, ... xN} represents the input spatial trajectory point data, and the mapping of X in the latent space is the point set Y = {y1, y2, ... yN}, and the point set Y is the node of the graph G = (V, E, W), which can fully express the data distribution characteristics of X. In order to discover the mapping relationship between the distribution characteristics of the two point sets, we need to learn the mapping mechanism H G , so that the point H after inverse mapping G (Y) also has a similar distribution to X. Next, we simplify H G (Y) is C.

[0087] Clearly based on the assumption of reverse embedding and the assumption of the fitting index calculation method, the problem of learning a main graph W from the data can be achieved by minimizing the objective function. The calculation formula is as follows:

[0088]

[0089] Among them, P represents the membership matrix, P i,k Indicates H G (y k ) represents x i The probability of . w is the weight matrix of the graph from the latent space.

[0090] First, initialize C = X. Based on the initialization C, we use it to calculate the membership matrix between the mapped point set C and the original point set X, that is, use H G (y k ) represents x i The probability P is calculated as follows:

[0091]

[0092] Based on the obtained C and P, C and P are used to calculate the weight matrix W of the graph when the latent space point set Y is represented by a graph opt Considering that the graph to be solved is sparse, in order to avoid multiple edges on the same road, this study uses the MST model instead. Therefore, the objective function of the reverse embedding hypothesis can be solved by the Krustal algorithm. The cost function calculation formula of the Krustal algorithm is as follows:

[0093]

[0094] Based on the obtained P and W opt Solve C in reverse. The objective function can be rewritten as follows:

[0095]

[0096] Where L is the Laplace matrix and ∧ is the matrix of the graph.

[0097] Through transformation, formula (4) is converted into an unconstrained quadratic programming problem, as shown in the following formula:

[0098] C=XP(2γ -1 L+∧) -1 (5)

[0099] Finally, based on the calculated C, P, and W, the difference in the objective function between the graphs before and after fitting is calculated until the following conditions are met:

[0100] abs(F(i+1)-F(i)) <e (5)

[0101] Where e is the set threshold, and F(i) is the objective function value at the i-th iteration in formula (1).

[0102] The loop iterates from 2.1.3 to 2.1.7 until the threshold condition e is met and the loop ends.

[0103] Finally, we get the best main graph W that satisfies the fitting opt The point set C and the best main graph.

[0104] It can be assumed that given a ship traveling on its planned route, the lateral distance between the ship and the route should meet the requirements of normal distribution, that is, the probability that the ship belongs to the route can be calculated using the normal probability density distribution function. Therefore, we detect the deviation behavior of the ship based on the 3-sigma principle.

[0105] When the data satisfies the normal distribution, we can write X~N(μ,δ 2 ). If μ=0 and δ=1, then the normal distribution can be called the standard normal distribution, and its simplified formula is as follows:

[0106]

[0107] The three-sigma rule of thumb states that "almost all" values ​​are within the range of plus or minus three standard deviations from the mean, which means that in experiments, a 99.7% probability can be considered "almost certain". Therefore, three times the standard deviation can be used as the threshold to detect abnormalities in the spatial attribute characteristics of ships.

[0108] Step 3: Combining the spatial attribute features and thematic attribute features, the abnormal behavior of the target ship is detected and scored based on the isolation forest algorithm.

[0109] Based on the above embodiment, step 3 specifically includes:

[0110] Step 3.1, segmenting the waterway network in a clockwise direction with intersections as nodes based on ArcGIS tools, assigning unique identifiers to the segmented waterway network segments, and after completing the waterway segmentation, calculating the nearest waterway segment of each track point according to the minimum distance principle, and assigning it to the track point as an attribute;

[0111] Step 3.2, using the isolation forest algorithm to calculate the abnormality score of the input data group.

[0112] Furthermore, the step 3.2 specifically includes:

[0113] Step 3.2.1, building a tree based on the input data set;

[0114] Step 3.2.2, randomly selecting a feature and a random cutting point, and cutting the feature based on the random cutting point;

[0115] Step 3.2.3, dividing the data space of the feature into two subspaces based on the random cutting point;

[0116] Step 3.2.4, looping and iterating step 3.2.2 to step 3.2.3 until a preset condition is met, wherein the preset condition includes any one of the following: the constructed isolated tree reaches the tree height, there is only one node on a certain node of the binary tree, and all sample features on the node are the same;

[0117] Step 3.2.5, repeat steps 3.2.2 to 3.2.4 to obtain a set of multiple trees to form an isolation forest;

[0118] Step 3.2.6, calculating the abnormality degree of the input data according to the isolation forest;

[0119] Step 3.2.7, normalize the value of the abnormality degree to obtain an abnormality score.

[0120] In the specific implementation, this study uses ArcGIS tools to segment the waterway network in a clockwise direction with intersections as nodes. Finally, a unique identifier is assigned to the segmented waterway network segment. After the waterway segmentation is completed, the nearest waterway segment of each track point is calculated according to the minimum distance principle and assigned to the track point as an attribute.

[0121] Existing abnormal behaviors usually give a result of 0-1, that is, the result is abnormal or non-abnormal, but does not give a score for the abnormality of the target behavior. Therefore, we use the isolation forest algorithm to obtain the abnormality score of the input data group. The main steps are as follows:

[0122] Assume L = {l1,l2,…l N} is the input ship feature combination, where l i Represents the combined value of multiple features of a trajectory point. Assume that T is a tree in an isolation forest, and its nodes have two situations: one is that there is no node below, and the other is that the node has two child nodes (T l ,T r ), where T l Represents the left child node, T r Represents the right child node.

[0123] Randomly select a feature q and a random cutting point p, and cut the feature q based on p, where the p value should be between the maximum and minimum values ​​of the specified data dimension.

[0124] The data space of feature q is divided into two subspaces based on the cutting point: data with a specified dimension value less than p is placed in the left node of the current node, and data greater than p is placed in the right node of the current node.

[0125] The loop iterates from 3.2.2 to 3.2.3 until one of the following conditions is met: (i) The constructed isolation tree reaches the tree height. (ii) There is only one node at a certain node of the binary tree. (iii) All sample features at the node are the same.

[0126] By repeating operations 3.2.2 to 3.2.4 multiple times, we can obtain a set of multiple trees, that is, the isolation forest iForest = {T1, T2, ... T k}.

[0127] The number of times an outlier is split from a data set indicates the degree of abnormality of the data. Let h(x) be the path length of a data point on a tree along the corresponding condition to the child node, that is, the degree of abnormality of the data. In a data set containing n data points, the average path length of a data point in the isolation forest is represented by c(n), and its calculation formula is as follows:

[0128]

[0129] Where H(i) is the harmonic number, which can be estimated as ln(i)+ε, where ε is the Euler constant.

[0130] Since c(n) is used to standardize the path h(x), the calculation formula after normalizing the outliers is as follows:

[0131]

[0132] Where s represents the anomaly score obtained after training with n data. E(h(x)) is the average value of the isolation tree h(x).

[0133] The abnormal ship trajectory detection method based on the recognition of group behavior rules provided in this embodiment sorts out the trajectory characteristics of ships with group regular behaviors in existing studies by combining actual geographical environment factors, and obtains the score value of the abnormality degree of trajectory points from the perspective of spatial attributes and thematic attributes, so as to evaluate the ship movement pattern and further identify abnormal ship behavior.

[0134] The present solution will be described below in conjunction with a specific embodiment. The navigation data of all ships in Zhejiang Province within a period of one month from October 25, 2021 to November 24, 2021, including fishing vessels, cargo ships, oil tankers, passenger ships and engineering ships, will be used to illustrate the specific implementation of the present invention. The specific implementation steps of the present invention to solve the relevant problems will be specifically described below in conjunction with this example:

[0135] 1) Channel network detection. For AIS data, the maximum number of iterations is 20, γ is 1, and σ is 0.05. The method in step 2.1 is used to detect the channel network. The results are as follows: Figure 3 shown.

[0136] 2) Detection of ship trajectories with abnormal spatial attributes. Using three times the standard deviation as the detection threshold, the method in step 2.2 is used to detect abnormal ship behavior based on AIS data. The detection results are as follows: Figure 4 shown.

[0137] 3) Ship track detection with thematic attribute anomalies. For AIS data, using the method proposed in step 3, the number of subsampling is set to 256 and the number of trees is set to 100. The detection results are as follows: Figures 5 to 8 As shown, the dotted box is the abnormal result of the detection.

[0138] It should be understood that various parts of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof.

[0139] The above is only a specific implementation of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present disclosure should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for detecting abnormal ship trajectories based on group behavior law recognition, characterized in that: include: Step 1, pre-processing the AIS base station data of the target ship; Step 2, extracting the spatial attribute characteristics of the target ship according to the AIS base station data after the preprocessing operation and detecting abnormal behavior of the spatial attribute characteristics accordingly; The step 2 specifically includes: Step 2.1, generating a channel network from the target trajectory point group based on main graph structure learning and tree connection strategy and calculating the best main graph and corresponding point set based on the main graph to form the spatial attribute feature; The step 2.1 specifically includes: Step 2.1.1, inversely mapping the target trajectory point group through a mapping mechanism to obtain an inverse mapping point group; Step 2.1.2, learning an initial main graph from the target trajectory point group; Step 2.1.3, initializing the inverse mapping point group and calculating the membership matrix between the inverse mapping point group and the target trajectory point group based on the inverse mapping point group; Step 2.1.4, calculating a weight matrix of a graph when the latent space point set is represented by a graph according to the inverse mapping point group and the membership matrix; Step 2.1.5, reversely solving the inverse mapping point group according to the membership matrix and the weight matrix to obtain an initial function; Step 2.1.6, converting the initial function into an unconstrained quadratic programming problem to obtain an objective function; Step 2.1.7, calculating the difference of the objective function between the graph before and after fitting based on the inverse mapping point group, the membership matrix and the initial main graph; Step 2.1.8, iterate steps 2.1.3 to 2.1.7 in a loop until the difference between the objective functions is less than a threshold value and the loop ends; Step 2.1.9, when the loop ends, the best main image and the corresponding point set are obtained to form the spatial attribute feature; Step 2.2, according to the optimal main image and the corresponding point set, the spatial attribute characteristic behavior of the target ship is detected based on the 3-sigma principle; Step 3, combining the spatial attribute features and thematic attribute features, and detecting and scoring the abnormal behavior of the target ship based on the isolation forest algorithm; The step 3 specifically includes: Step 3.1, segment the waterway network in a clockwise direction with the intersection as the node based on ArcGIS tools, assign unique identifiers to the segmented waterway network segments, and after the waterway segmentation is completed, calculate the nearest waterway segment of each track point according to the minimum distance principle, and assign it to the track point as an attribute; Step 3.2, using the isolation forest algorithm to calculate the abnormality score of the input data group; The step 3.2 specifically includes: Step 3.2.1, building a tree based on the input data set; Step 3.2.2, randomly selecting a feature and a random cutting point, and cutting the feature based on the random cutting point; Step 3.2.3, dividing the data space of the feature into two subspaces based on the random cutting point; Step 3.2.4, looping and iterating step 3.2.2 to step 3.2.3 until a preset condition is met, wherein the preset condition includes any one of the following: the constructed isolated tree reaches the tree height, there is only one node on a certain node of the binary tree, and all sample features on the node are the same; Step 3.2.5, repeat steps 3.2.2 to 3.2.4 to obtain a set of multiple trees to form an isolation forest; Step 3.2.6, calculating the abnormality degree of the input data according to the isolation forest; Step 3.2.7, normalize the value of the abnormality degree to obtain an abnormality score.

2. The method according to claim 1, characterized in that , the step 1 specifically includes: Step 1.1, matching the trajectory points of the target ship with the map; Step 1.2, delete the illogical data; Step 1.3, cutting the trajectory of the target ship from the perspectives of space and time respectively; Step 1.4, use the Kalman filter algorithm to smooth the trajectory data after cutting; Step 1.5, according to the concavity and convexity of the trajectory points, the Lagrange first-order interpolation algorithm and the Lagrange second-order interpolation algorithm are respectively used to complete the trajectory line until the interval condition of the trajectory segment is met.

3. The method according to claim 1, characterized in that , the step 2.2 specifically includes: Step 2.2.1, based on the optimal main graph and the corresponding point set, using a normal probability density distribution function to calculate the probability that the route of the target ship belongs to the planned route; Step 2.2.2, detecting abnormal behavior of the spatial attribute characteristics of the target ship based on the probability and the 3-sigma principle.

Images