Ship data verification method and system based on multivariate constraints, storage medium and electronic equipment

Abstract

The invention discloses a ship data verification method and system based on multivariate constraints, a storage medium and electronic equipment. The method comprises the following steps: constructing ship multi-tuple data according to ship information; calculating a multivariate constraint function of the ship multi-tuple data according to causal relevance and / or temporal relevance and / or spatial relevance among the ship multi-tuple data; identifying initial abnormal data in the data acquired by each sensor of the ship by adopting a data mutability rule; calculating an associated change reference value according to the data change condition of the multi-tuple data corresponding to the initial abnormal data in a preset time period; calculating an abnormal check value according to the multivariate constraint function and the associated change reference value; and correcting the initial abnormal data according to the multivariate constraint function and the associated change reference value to obtain verified data. The problems of high false alarm rate of data anomaly detection, poor real-time performance, insufficient model generalization ability and high time-space alignment difficulty in the existing intelligent ship technology are solved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent ship technology, and in particular relates to a ship data verification method, system, storage medium and electronic device based on multi-constraints. Background Technology

[0002] With the development of intelligent ships and autonomous navigation technologies, the demand for automated anomaly detection is increasing. Modern ships are equipped with various sensors and monitoring systems, including AIS (Automatic Identification System), radar, engine monitoring, and environmental sensors, to acquire a large amount of ship data for intelligent navigation and anomaly detection. However, current systems mainly rely on a single data source for anomaly detection, lack cross-validation mechanisms, and have a high false alarm rate.

[0003] To address the false alarm rate issue, current solutions mainly fall into two categories: multi-source fusion analysis algorithms and deep learning algorithms. Multi-source fusion analysis algorithms employ multi-sensor collaboration, such as fusing radar and AIS data. However, these algorithms heavily rely on manual feature extraction, are highly complex, and have poor adaptability. Furthermore, applying transfer learning to heterogeneous multi-source data presents challenges, resulting in insufficient model generalization ability. Additionally, the temporal synchronization and spatial alignment issues of data from different sensors increase the difficulty of analysis. While deep learning algorithms can improve anomaly detection accuracy, they require extensive labeled data for training and consume significant computational resources, leading to response delays and failing to meet the real-time early warning needs of emergency situations.

[0004] To address the issues of high false alarm rate, poor real-time performance, insufficient model generalization ability, and high difficulty in spatiotemporal alignment in ship data anomaly detection, a ship data verification method, system, storage medium, and electronic equipment based on multivariate constraints are proposed. Summary of the Invention

[0005] This invention proposes a ship data verification method, system, storage medium, and electronic device based on multivariate constraints, in order to at least solve the problems of high false alarm rate, poor real-time performance, insufficient model generalization ability, and high difficulty in spatiotemporal alignment in existing intelligent ship technologies.

[0006] According to an embodiment of the present invention, a ship data verification method based on multi-factor constraints is provided, comprising:

[0007] Construct multi-group data of ships based on ship information;

[0008] Calculate the multivariate constraint function for the ship tuple data based on the causal correlation and / or temporal correlation and / or spatial correlation among the ship tuple data;

[0009] The system employs data mutation rules to identify initial anomalous data in the data collected by various sensors on the ship; and calculates associated change reference values ​​based on the data changes of the tuple data corresponding to the initial anomalous data within a preset time period.

[0010] Calculate the anomaly check value based on the multivariate constraint function and the associated change reference value;

[0011] When the abnormal verification value is less than the preset verification threshold, the initial abnormal data is determined to be a misjudgment, and the initial abnormal data is corrected according to the multivariate constraint function and the associated change reference value to obtain the verified data; otherwise, the initial abnormal data is determined to be a correct judgment.

[0012] In a preferred embodiment, the construction of ship plural data based on ship information is a combination of plural data constructed based on any one or more of the following: ship fleet information, ship type information, ship sensor information, and ship data information. The fleet information includes any one or more of the following: ship number information, ship distribution information, fleet driving sequence information, and fleet driving route information. The ship type information includes any one or more of the following: ship displacement information, ship purpose information, and ship configuration information. The ship sensor information includes any one or more of the following: sensor category information, sensor model information, and sensor accuracy information. The ship data information includes any one or more of the following: ship data source, ship data format, and ship data sampling rate.

[0013] In a preferred embodiment, the step of calculating the multivariate constraint function of the ship multivariate data based on the causal correlation and / or temporal correlation and / or spatial correlation between the ship multivariate data includes the following steps:

[0014] The causal relationship influence function among multiple data elements is calculated based on the ship's fleet status, ship type, and the degree of influence of the causal relationship between ship sensors and ship data on each data element.

[0015] The time correlation influence function among multiple data elements is calculated based on the order of the ship's fleet, the ship type, and the degree of influence of the time correlation between ship sensors and ship data on each data element.

[0016] The spatial correlation influence function among multiple data elements is calculated based on the influence of the ship's fleet position, ship type, and the positional correlation between ship sensors and ship data on the degree of influence of each data element.

[0017] The influence function between multiple variables is calculated based on the causal relationship influence function and / or the temporal correlation influence function and / or the spatial correlation influence function between multiple variables, which is the multivariate constraint function for ship multivariate data.

[0018] In a preferred embodiment, the step of identifying initial anomalous data in the data collected by various sensors of the ship using data mutation rules includes:

[0019] The abnormal reference value of the data is calculated based on the instantaneous change of the data collected by the ship's various sensors and / or the average change within a preset time period and / or the degree of data change within a preset time period.

[0020] Initial abnormal data is identified based on the relationship between the abnormal reference value of the data and the preset data mutation threshold.

[0021] In a preferred embodiment, the step of calculating the correlation change reference value based on the data changes of the tuple data corresponding to the initial abnormal data within a preset time period includes:

[0022] Identify the tuple data corresponding to the initial abnormal data and obtain the data change information of each metadata within a preset time period;

[0023] The associated change reference value of each metadata is calculated based on the instantaneous change of each metadata and / or the average change within a preset time period and / or the degree of data change within a preset time period.

[0024] In a preferred embodiment, the step of calculating the anomaly check value based on the multivariate constraint function and the associated change reference value includes the following steps:

[0025] The multivariate constraint impact value is calculated based on the multivariate constraint function of the ship multivariate data and the reference value of the correlation change of each metadata.

[0026] Calculate the anomaly check value based on the influence value of multiple constraints.

[0027] In a preferred embodiment, the step of correcting the initial outlier data based on the multivariate constraint function and the correlation change reference value to obtain the verified data includes the following steps:

[0028] Calculate the change weight value based on the change pattern of the tuple data corresponding to the initial abnormal data within a preset time period;

[0029] Calculate the change value of the tuple data corresponding to the initial outlier data based on the multivariate constraint influence value and change weight value;

[0030] The verified data is calculated based on the changes in the data of the previous time step of the initial abnormal data and the corresponding tuple data of the initial abnormal data.

[0031] According to another embodiment of the present invention, a ship data verification system based on multi-factor constraints is provided, comprising:

[0032] The ship multi-data construction module is used to construct multi-data combinations based on one or more of the ship's fleet information, ship type information, ship sensor information, and ship data information.

[0033] The multivariate constraint function construction module is used to calculate the multivariate constraint function of the ship multivariate data based on the causal correlation and / or temporal correlation and / or spatial correlation between the ship multivariate data.

[0034] The initial abnormal data identification module is used to identify initial abnormal data in the data collected by various sensors of the ship according to the data mutation rules;

[0035] The correlation change identification module is used to calculate the correlation change reference value based on the data changes of the tuple data corresponding to the initial abnormal data within a preset time period.

[0036] The data verification module is used to calculate anomaly verification values ​​based on multivariate constraint functions and associated change reference values, identify misjudged initial abnormal data based on the anomaly verification values, and correct the initial abnormal data based on multivariate constraint functions and associated change reference values ​​to obtain verified data.

[0037] According to another embodiment of the present invention, a computer-readable storage medium is provided that stores a computer program for electronic data interchange, wherein the computer program causes a computer to execute the above-described ship data verification method based on multi-constraints.

[0038] According to another embodiment of the present invention, an electronic device is provided, comprising:

[0039] At least one processor;

[0040] and a memory communicatively connected to the at least one processor;

[0041] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the above-described ship data verification method based on multi-constraints.

[0042] The advantages of the ship data verification method, system, storage medium, and electronic device based on multi-constraints of the present invention are:

[0043] (1) Construct ship multivariate data and calculate the multivariate constraint function of ship multivariate data based on the causal correlation and / or temporal correlation and / or spatial correlation between ship multivariate data. Compared with traditional ship data anomaly detection algorithms, it can effectively classify massive and messy ship data according to different dimensions and construct cross-correlation models, realize the automatic correlation between ship multidimensional data, reduce the difficulty of data spatiotemporal synchronization, and facilitate the improvement of the efficiency of subsequent data anomaly detection.

[0044] (2) Calculate the associated change reference value based on the data change of the tuple data corresponding to the initial abnormal data within a preset time period, and calculate the abnormal verification value based on the multivariate constraint function and the associated change reference value. Compared with the traditional technical solution for ship data anomaly detection, it can effectively identify the changes of multidimensional data associated with a single data anomaly, thereby effectively verifying a single data anomaly and facilitating the reduction of the false alarm rate of subsequent data anomaly detection.

[0045] (3) When the initial abnormal data is determined to be a misjudgment, the initial abnormal data is corrected according to the multivariate constraint function and the associated change reference value to obtain the verified data. Compared with the traditional ship data anomaly detection technology, the abnormal data can be quickly and effectively corrected through the constraint relationship between multidimensional data associated with the abnormal data, eliminating data errors caused by sensor failure or transmission errors, ensuring the validity and time continuity of the data, and facilitating data retrieval and data processing for ship intelligent business. Attached Figure Description

[0046] Figure 1 This is a flowchart of a ship data verification method based on multi-constraints according to an embodiment of the present invention;

[0047] Figure 2 This is a flowchart of step S02 in an embodiment of the present invention;

[0048] Figure 3 This is a flowchart of step S03 in an embodiment of the present invention;

[0049] Figure 4 This is a flowchart of step S04 in an embodiment of the present invention;

[0050] Figure 5 This is a flowchart of step S05 in an embodiment of the present invention;

[0051] Figure 6 This is a flowchart of step S06 in an embodiment of the present invention;

[0052] Figure 7 This is an architecture diagram of a ship data verification system based on multi-constraints according to an embodiment of the present invention;

[0053] Figure 8This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0054] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0055] According to an embodiment of the present invention, a ship data verification method based on multi-factor constraints is provided, the flowchart of which is shown below. Figure 1 As shown, it includes:

[0056] Step S01: Construct ship plural data based on ship information;

[0057] Step S02: Calculate the multivariate constraint function of the ship multivariate data based on the causal correlation and / or temporal correlation and / or spatial correlation between the ship multivariate data;

[0058] Step S03: Identify initial abnormal data in the data collected by various sensors on the ship using data mutation rules;

[0059] Step S04: Calculate the correlation change reference value based on the data changes of the tuple data corresponding to the initial abnormal data within a preset time period;

[0060] Step S05: Calculate the anomaly check value based on the multivariate constraint function and the associated change reference value;

[0061] Step S06: When the abnormal verification value is less than the preset verification threshold, the initial abnormal data is determined to be a misjudgment. The initial abnormal data is corrected according to the multivariate constraint function and the associated change reference value to obtain the verified data; otherwise, the initial abnormal data is determined to be a correct judgment.

[0062] In a preferred embodiment, the construction of ship tuple data based on ship information is a combination of tuple data constructed based on any one or more of the following: ship fleet information, ship type information, ship sensor information, and ship data information. The fleet information includes any one or more of the following: ship number information, ship distribution information, fleet sailing order information, and fleet sailing route information. The ship type information includes any one or more of the following: ship displacement information, ship purpose information, and ship configuration information. The ship sensor information includes any one or more of the following: sensor type information, sensor model information, and sensor accuracy information. The ship data information includes any one or more of the following: ship data source, ship data format, and ship data sampling rate. In this embodiment, quaternion data is constructed.<a,b,c,d> Where 'a' represents the fleet information of the vessel, 'b' represents the vessel type information, 'c' represents the vessel sensor information, and 'd' represents the vessel data information. The fleet information mainly includes one or more of the following: vessel number information, vessel distribution information, fleet driving status information, fleet driving sequence information, and fleet driving route information. The vessel type information mainly includes one or more of the following: vessel displacement information, vessel purpose information, and vessel configuration information. The vessel sensor information mainly includes one or more of the following: sensor category information, sensor model information, and sensor accuracy information. The vessel data information mainly includes one or more of the following: vessel data source, vessel data format, and vessel data sampling rate.

[0063] In a preferred embodiment, step S02 involves calculating the multivariate constraint function of the ship tuple data based on the causal correlation and / or temporal correlation and / or spatial correlation between the ship tuple data, as shown in the flowchart below. Figure 2 As shown, the steps include:

[0064] Step S021: Calculate the causal relationship influence function between multiple data based on the ship's fleet status, ship type, and the degree of influence of the causal relationship between ship sensors and ship data on each metadata.

[0065] Step S022: Calculate the time correlation influence function between multiple data based on the order of the ship's fleet, the ship type, and the degree of influence of the time correlation between the ship's sensors and the ship's data on each data element;

[0066] Step S023: Calculate the spatial correlation influence function between multiple data based on the position of the ship in the fleet, the ship type, and the degree of influence of the positional correlation between the ship's sensors and the ship's data on each data element;

[0067] Step S024: Calculate the influence function between multiple variables based on the causal relationship influence function and / or the temporal correlation influence function and / or the spatial correlation influence function between multiple variables. This function is the multivariate constraint function for the ship multivariate data.

[0068] In this embodiment, the step of calculating the causal correlation influence function between multiple metadata based on the influence of the fleet's operating status, ship type, and the causal relationship between ship sensors and ship data on each metadata element includes: training a causal influence function of the fleet on other metadata based on changes in the fleet's operating status; training a causal influence function of the ship type on other metadata based on changes in the fleet's operating status, ship sensors, and ship data; training a causal influence function of ship sensors on other metadata based on changes in the fleet's operating status, ship type, and ship data; training a causal influence function of ship data on other metadata based on changes in the fleet's operating status, ship type, and ship sensors; and calculating the causal correlation influence function between multiple metadata elements based on the causal influence function of the fleet on other metadata and / or the causal influence function of ship type on other metadata and / or the causal influence function of ship sensors on other metadata and / or the causal influence function of ship data on other metadata. The causal correlation influence function between multiple metadata elements is represented by the function F(a,b,c,d).

[0069] The calculation of the time correlation influence function between multiple metadata based on the order of the ship's fleet, ship type, and the time correlation between ship sensors and ship data includes: training the time influence function of the ship's fleet on other metadata based on the time influence of the ship's fleet order on ship type, ship sensors, and ship data; training the time influence function of ship type on other metadata based on the time influence of ship type on the order of the ship's fleet, ship sensors, and ship data; training the time influence function of ship sensors on other metadata based on the time influence of ship sensors on the order of the ship's fleet, ship type, and ship data; training the time influence function of ship data on other metadata based on the time influence of ship data on the order of the ship's fleet, ship type, and ship sensors; and calculating the time correlation influence function between multiple metadata based on the time influence function of the ship's fleet and / or the time influence function of ship type and / or the time influence function of ship sensors and / or the time influence function of ship data on other metadata. The time correlation influence function between multiple metadata is represented by the function T(a,b,c,d).

[0070] The calculation of the spatial correlation influence function between multiple metadata based on the influence of the ship's fleet position, ship type, and the positional correlation between ship sensors and ship data on each metadata element includes: training the spatial influence function of the ship's fleet position on other metadata based on the spatial influence of the ship's fleet position on ship type, ship sensors, and ship data; training the spatial influence function of ship type on other metadata based on the spatial influence of ship type on the ship's fleet position, ship sensors, and ship data; training the spatial influence function of ship sensors on other metadata based on the spatial influence of ship sensors on the ship's fleet position, ship type, and ship data; training the spatial influence function of ship data on other metadata based on the spatial influence of ship data on the ship's fleet position, ship type, and ship sensors; and calculating the spatial correlation influence function between multiple metadata based on the spatial influence function of the ship's fleet position and / or the spatial influence function of ship type and / or the spatial influence function of ship sensors and / or the spatial influence function of ship data on other metadata. The spatial correlation influence function between multiple metadata is represented by the function S(a,b,c,d).

[0071] The multivariate constraint function, calculated based on the causal correlation influence function and / or the temporal correlation influence function and / or the spatial correlation influence function between multivariates, is obtained by jointly training the multivariate constraint function of the ship multivariate data with the causal correlation influence function and / or the temporal correlation influence function and / or the spatial correlation influence function between multivariates. The multivariate constraint function of the ship multivariate data is represented by the function H(a,b,c,d).

[0072] Examples A1 to A7 illustrate different implementation methods for calculating multivariate constraint functions, as follows:

[0073] Example A1: Calculate the multivariate constraint function based on the positive correlation between the causal relationship influence function and the multivariate constraint function.

[0074] Specifically, the impact values ​​of changes in the vessel's fleet status, vessel type, vessel sensor information, and vessel data on other metadata are obtained, and this data is used to train a causal correlation function F(a,b,c,d) for changes in vessel type, vessel sensors, and vessel data. Based on the positive correlation between the causal correlation function F(a,b,c,d) and the multivariate constraint function, the multivariate constraint function H(a,b,c,d) is calculated. In a preferred embodiment, the multivariate constraint function H(a,b,c,d) = w1·F(a,b,c,d) is calculated. w2 +w3, where w1 (w1>0), w2 (w2>0), and w3 are calculated coefficients obtained through prior training.

[0075] Example A2: Calculate the multivariate constraint function based on the positive correlation between the time correlation influence function and the multivariate constraint function.

[0076] Specifically, the temporal influence of the ship's fleet order, ship type, and ship sensor and data on other metadata is obtained (normalized using a preset temporal influence threshold), and a multivariate temporal correlation influence function T(a,b,c,d) is trained based on this. The multivariate constraint function H(a,b,c,d) is then calculated based on the positive correlation between the multivariate temporal correlation influence function T(a,b,c,d) and the multivariate constraint function. In a preferred embodiment, the multivariate constraint function H(a,b,c,d) is calculated as w4·T(a,b,c,d). w5 +w6, where w4 (w4>0), w5 (w5>0), and w6 are calculated coefficients obtained through prior training.

[0077] Example A3: Calculate the multivariate constraint function based on the positive correlation between the spatial correlation influence function and the multivariate constraint function.

[0078] Specifically, the spatial influence of the ship's fleet location, ship type, and ship sensor and data on other metadata is obtained (normalized using a preset spatial influence threshold), and a spatial correlation influence function S(a,b,c,d) among multiple variables is trained based on this. The multivariate constraint function H(a,b,c,d) is then calculated based on the positive correlation between the spatial correlation influence function S(a,b,c,d) and the multivariate constraint function. In a preferred embodiment, the multivariate constraint function H(a,b,c,d) is calculated as w7·S(a,b,c,d). w8 +w9, where w7 (w7>0), w8 (w8>0), and w9 are calculated coefficients obtained through prior training.

[0079] Example A4: Calculate the multivariate constraint function based on the positive correlation between the causal relationship influence function and the time correlation influence function between the multivariate components and the multivariate constraint function.

[0080] Specifically, the impact values ​​of changes in the ship's fleet status, ship type, ship sensor information, and ship data on other metadata are obtained, and this data is used to train a causal correlation function F(a,b,c,d) for changes in ship type, ship sensors, and ship data. The temporal influence of the ship's fleet order, ship type, ship sensors, and ship data on other metadata (normalized using a preset temporal influence threshold) is obtained, and this is used to train a temporal correlation function T(a,b,c,d) for multiple variables. Based on the positive correlation between the causal correlation function F(a,b,c,d), the temporal correlation function T(a,b,c,d), and the multiple constraint function, the multiple constraint function H(a,b,c,d) is calculated. In a preferred embodiment, the multiple constraint function H(a,b,c,d) is calculated as w10·F(a,b,c,d). w11 +w12·T(a,b,c,d) w13 +w14, where w10 (w10>0), w11 (w11>0), w12 (w12>0), w13 (w13>0), and w14 are pre-trained computational coefficients. In another preferred embodiment, the multivariate constraint function H(a,b,c,d)=w15·F(a,b,c,d) is calculated. w16 ·T(a,b,c,d) w17 +w18, where w15 (w15>0), w16 (w16>0), w17 (w17>0), and w18 are calculated coefficients obtained through prior training.

[0081] Example A5: Calculate the multivariate constraint function based on the positive correlation between the causal relationship influence function and the spatial correlation influence function between multivariate components and the multivariate constraint function.

[0082] Specifically, the impact values ​​of changes in the ship's fleet's navigation status, ship type, ship sensor information, and ship data on other metadata are obtained, and this data is used to train a causal correlation function F(a,b,c,d) for changes in ship type, ship sensors, and ship data. The spatial influence of the ship's fleet position, ship type, ship sensors, and ship data on other metadata (normalized using a preset spatial influence threshold) is obtained, and this is used to train a spatial correlation function S(a,b,c,d) for multiple variables. Based on the positive correlation between the causal correlation function F(a,b,c,d), the spatial correlation function S(a,b,c,d), and the multiple constraint function, the multiple constraint function H(a,b,c,d) is calculated. In a preferred embodiment, the multiple constraint function H(a,b,c,d) is calculated as w19·F(a,b,c,d). w20 +w21·S(a,b,c,d) w22 +w23, where w19 (w19>0), w20 (w20>0), w21 (w21>0), w22 (w22>0), and w23 are pre-trained calculation coefficients. In another preferred embodiment, the multivariate constraint function H(a,b,c,d)=w24·F(a,b,c,d) is calculated. w25 ·S(a,b,c,d) w26 +w27, where w24 (w24>0), w25 (w25>0), w26 (w26>0), and w27 are calculated coefficients obtained through prior training.

[0083] Example A6: Calculate the multivariate constraint function based on the positive correlation between the temporal correlation influence function and the spatial correlation influence function between multivariate variables and the multivariate constraint function.

[0084] Specifically, the temporal influence of the ship's fleet order, ship type, and ship sensor and data data on other metadata is obtained (normalized using a preset temporal influence threshold), and a temporal correlation influence function T(a,b,c,d) among multiple variables is trained based on this. The spatial influence of the ship's fleet position, ship type, and ship sensor and data data on other metadata is obtained (normalized using a preset spatial influence threshold), and a spatial correlation influence function S(a,b,c,d) among multiple variables is trained based on this. The multivariate constraint function H(a,b,c,d) is calculated based on the positive correlation between the temporal correlation influence function T(a,b,c,d), the spatial correlation influence function S(a,b,c,d), and the multivariate constraint function. In a preferred embodiment, the multivariate constraint function H(a,b,c,d) is calculated as w²⁸·T(a,b,c,d).w29 +w30·S(a,b,c,d) w31 +w32, where w28 (w28>0), w29 (w29>0), w30 (w30>0), w31 (w31>0), and w32 are pre-trained calculation coefficients. In another preferred embodiment, the multivariate constraint function H(a,b,c,d)=w33·T(a,b,c,d) is calculated. w34 ·S(a,b,c,d) w35 +w36, where w33 (w33>0), w34 (w34>0), w35 (w35>0), and w36 are calculated coefficients obtained through prior training.

[0085] Example A7: Calculate the multivariate constraint function based on the positive correlation between the causal relationship influence function, the temporal correlation influence function, the spatial correlation influence function, and the multivariate constraint function.

[0086] Specifically, the impact values ​​of changes in the vessel's fleet status, vessel type, vessel sensor information, and vessel data on other metadata are obtained, and this data is used to train a causal correlation function F(a,b,c,d) for changes in vessel type, vessel sensors, and vessel data. The temporal impact of the vessel's fleet order, vessel type, vessel sensors, and vessel data on other metadata is obtained (normalized using a preset temporal impact threshold), and this data is used to train a temporal correlation function T(a,b,c,d) for multiple variables. The process involves: acquiring the ship's fleet location, ship type, and the spatial influence of ship sensor and data on other metadata (normalized using a preset spatial influence threshold) and training a spatial correlation influence function S(a,b,c,d) between multiple variables; calculating the multivariate constraint function H(a,b,c,d) based on the positive correlation between the causal correlation influence function F(a,b,c,d), the temporal correlation influence function T(a,b,c,d), the spatial correlation influence function S(a,b,c,d), and the multivariate constraint function. In a preferred embodiment, the multivariate constraint function H(a,b,c,d) is calculated as w37·F(a,b,c,d). w38 +w39·T(a,b,c,d) w40 +w41·S(a,b,c,d) w42+w43, where w37 (w37>0), w38 (w38>0), w39 (w39>0), w40 (w40>0), w41 (w41>0), w42 (w42>0), and w43 are pre-trained calculated coefficients. In another preferred embodiment, the multivariate constraint function H(a,b,c,d)=w44·F(a,b,c,d) is calculated. w45 ·T(a,b,c,d) w46 ·S(a,b,c,d) w47 +w48, where w44 (w44>0), w45 (w45>0), w46 (w46>0), w47 (w47>0), and w48 are calculated coefficients obtained through prior training.

[0087] In a preferred embodiment, step S03 involves identifying initial abnormal data in the data collected by various sensors on the ship using data mutation rules, as shown in the flowchart below. Figure 3 As shown, it includes:

[0088] Step S031: Calculate the abnormal reference value of the data based on the instantaneous change of the data collected by each sensor of the ship and / or the average change within a preset time period and / or the degree of data change within a preset time period;

[0089] Step S032: Identify initial abnormal data based on the relationship between the abnormal reference value of the data and the preset data mutation threshold.

[0090] In this embodiment, the calculation of the abnormal reference value of the data based on the instantaneous changes and / or the average changes and / or the degree of data change within a preset time period of the data collected by each of the ship's sensors comprises: calculating the abnormal reference value of the data based on the positive correlation between the instantaneous changes and the abnormal reference value of the data collected by each of the ship's sensors; calculating the abnormal reference value of the data based on the positive correlation between the average changes and the abnormal reference value of the data collected by each of the ship's sensors within a preset time period; calculating the abnormal reference value of the data based on the positive correlation between the degree of data change and the abnormal reference value of the data collected by each of the ship's sensors within a preset time period; and calculating the abnormal reference value of the data based on the instantaneous changes and the average changes within the preset time period of the data collected by each of the ship's sensors. The abnormal reference value of the data can be calculated based on any one of the following: the positive correlation between the abnormal reference value and the data; the positive correlation between the instantaneous changes in data collected by various ship sensors and the degree of data change within a preset time period; the positive correlation between the average changes in data collected by various ship sensors within a preset time period and the degree of data change within a preset time period; and the abnormal reference value of the data. The abnormal reference value of the data is represented by the variable x. Taking a specific ship as an example, this section uses instantaneous speed change to calculate the abnormal reference value of the data. At a certain moment, its speed is 10 knots, and the data source is the ship's AIS. The speed at the previous moment is 20 knots, so the instantaneous speed change is calculated to be 10 knots. Normalization is performed according to a preset instantaneous speed change threshold, resulting in an instantaneous speed change ∆v = 0.8. Based on the positive correlation between the instantaneous speed changes of data collected by various ship sensors and the abnormal reference value, the abnormal reference value x = g1·∆v is calculated. g2 +g3=1·0.8 1 +0=0.8, where g1, g2, and g3 are calculated coefficients obtained through pre-training, with g1=1, g2=1, and g3=0.

[0091] Based on the mission information of the fleet to which the ship belongs, the ship type, and the ship application scenario, a preset data mutation threshold is set. Here, the data mutation threshold is 0.6. The abnormal reference value of the data is x=0.8>0.6, so the speed data is identified as initial abnormal data.

[0092] In a preferred embodiment, step S04 involves calculating a correlation change reference value based on the data changes of the tuple data corresponding to the initial abnormal data within a preset time period. The flowchart is as follows: Figure 4 As shown, it includes:

[0093] Step S041: Identify the tuple data corresponding to the initial abnormal data and obtain the data change information of each metadata within a preset time period;

[0094] Step S042: Calculate the associated change reference value of each metadata based on the instantaneous change of each metadata and / or the average change within a preset time period and / or the degree of data change within a preset time period.

[0095] In this embodiment, the tuple data corresponding to the initial abnormal data is identified and the data change information of each metadata within a preset time period is obtained. Taking the speed abnormal data as an example, the tuple data corresponding to the initial speed abnormal data is identified, including the information of the fleet to which the ship belongs, the sensor information deployed by the ship, and the ship data information of the ship, and the amount of data change within the preset time period (here, 5 sampling periods) is obtained.

[0096] The calculation of the correlation change reference value for each metadata based on the instantaneous change and / or the average change and / or the degree of data change within a preset time period involves training the correlation between the instantaneous change and / or the average change and / or the degree of data change within a preset time period and the correlation change reference value for different metadata according to the rules of different metadata, and thereby obtaining the correlation change reference value y= for each metadata. y ,b y ,c y ,d y >

[0097] In a preferred embodiment, step S05, calculating the anomaly check value based on the multivariate constraint function and the associated change reference value, is illustrated in the flowchart below. Figure 5 As shown, the steps include:

[0098] Step S051: Calculate the multivariate constraint impact value based on the multivariate constraint function of the ship multivariate group data and the reference value of the correlation change of each metadata.

[0099] Step S052: Calculate the anomaly verification value based on the influence value of the multivariate constraints.

[0100] In this embodiment, the calculation of the multivariate constraint impact value based on the multivariate constraint function of the ship multivariate data and the correlation change reference value of each metadata element involves using the correlation change reference value of each metadata element... y ,b y ,c y ,d y ​​The impact value of each metadata change is calculated by substituting the reference value of each metadata change into the multivariate constraint function H(a,b,c,d). The difference between the impact value of each metadata change on other metadata and the reference value of the corresponding metadata change is calculated. The multivariate constraint impact value is calculated based on the average, maximum or minimum value of the corresponding difference of each metadata.

[0101] The abnormal verification value is calculated based on the positive correlation between the influence value of the multivariate constraints and the abnormal verification value. That is, the larger the average, maximum or minimum value of the corresponding difference of each metadata, the larger the abnormal verification value.

[0102] In a preferred embodiment, step S06, correcting the initial abnormal data based on the multivariate constraint function and the correlation change reference value to obtain the verified data, is illustrated in the flowchart below. Figure 6 As shown, the steps include:

[0103] Step S061: Calculate the change weight value based on the change pattern of the tuple data corresponding to the initial abnormal data within a preset time period;

[0104] Step S062: Calculate the change value of the tuple data corresponding to the initial abnormal data based on the influence value of the multivariate constraints and the change weight value;

[0105] Step S063: Calculate the verified data based on the data from the previous moment of the initial abnormal data and the change values ​​of the corresponding tuple data of the initial abnormal data.

[0106] In this embodiment, the calculation of the change weight value based on the change pattern of the tuple data corresponding to the initial abnormal data within a preset time period is based on the negative correlation between the number of consecutive changes, variance, or amount of consecutive changes of the tuple data corresponding to the initial abnormal data within the preset time period and the change weight value. The smaller the number of consecutive changes, variance, or amount of consecutive changes, the more stable the change of the tuple data is, the higher the confidence level of its data change, and the larger the change weight value.

[0107] The calculation of the change value of the tuple data corresponding to the initial abnormal data based on the influence value of the multivariate constraint and the change weight value is calculated by multiplying the influence value of the multivariate constraint and the change weight value obtained in the above embodiment. It should be noted that the sign of the change value of the tuple data corresponding to the initial abnormal data is distinguished by whether the data increases or decreases.

[0108] The process of calculating the verified data based on the change values ​​of the data before the initial abnormal data and the corresponding tuple data of the initial abnormal data is to add the change values ​​of the corresponding tuple data of the initial abnormal data to the data before the initial abnormal data to obtain the verified data, thus completing the verification of the abnormal data.

[0109] In another preferred embodiment, if the abnormal data is determined to be a misjudgment and data verification is performed, the verification process of the abnormal data is marked, and the abnormal data and the verified data are stored for secondary training. The secondary training method includes any one or a combination of simple incremental learning algorithms, weight-based incremental learning algorithms, neural network learning algorithms, machine learning algorithms, and large language model inference algorithms.

[0110] According to another embodiment of the present invention, a ship data verification system based on multi-factor constraints is provided, the structural schematic diagram of which is shown below. Figure 7 As shown, it includes:

[0111] The ship multi-data construction module is used to construct multi-data combinations based on one or more of the ship's fleet information, ship type information, ship sensor information, and ship data information.

[0112] The multivariate constraint function construction module is used to calculate the multivariate constraint function of the ship multivariate data based on the causal correlation and / or temporal correlation and / or spatial correlation between the ship multivariate data.

[0113] The initial abnormal data identification module is used to identify initial abnormal data in the data collected by various sensors of the ship according to the data mutation rules;

[0114] The correlation change identification module is used to calculate the correlation change reference value based on the data changes of the tuple data corresponding to the initial abnormal data within a preset time period.

[0115] The data verification module is used to calculate anomaly verification values ​​based on multivariate constraint functions and associated change reference values, identify misjudged initial abnormal data based on the anomaly verification values, and correct the initial abnormal data based on multivariate constraint functions and associated change reference values ​​to obtain verified data.

[0116] According to another embodiment of the present invention, a computer-readable storage medium is provided that stores a computer program for electronic data interchange, wherein the computer program causes a computer to execute the above-described ship data verification method based on multi-constraints.

[0117] According to another embodiment of the present invention, an electronic device is provided, the structural schematic diagram of which is shown below. Figure 8 As shown, it includes:

[0118] At least one processor;

[0119] and a memory communicatively connected to the at least one processor;

[0120] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the above-described ship data verification method based on multi-constraints.

[0121] Of course, those skilled in the art should recognize that the above embodiments are only used to illustrate the present invention and are not intended to limit the present invention. Any changes or modifications to the above embodiments that are within the scope of the present invention will fall within the protection scope of the present invention.

Claims

1. A method for verifying ship data based on multi-factor constraints, characterized in that, include: Construct multi-group data of ships based on ship information; Calculate the multivariate constraint function for the ship tuple data based on the causal correlation and / or temporal correlation and / or spatial correlation among the ship tuple data; Data mutation rules are used to identify initial abnormal data in the data collected by various sensors on the ship. Calculate the correlation change reference value based on the data changes of the tuple data corresponding to the initial abnormal data within a preset time period; calculate the anomaly check value based on the tuple constraint function and the correlation change reference value; When the abnormal verification value is less than the preset verification threshold, the initial abnormal data is determined to be a misjudgment, and the initial abnormal data is corrected according to the multivariate constraint function and the associated change reference value to obtain the verified data; otherwise, the initial abnormal data is determined to be a correct judgment.

2. The ship data verification method based on multi-factor constraints according to claim 1, characterized in that, The construction of ship plural data based on ship information is a combination of multiple data elements constructed based on any one or more of the following: ship fleet information, ship type information, ship sensor information, and ship data information. The fleet information includes any one or more of the following: ship number information, ship distribution information, fleet driving sequence information, and fleet driving route information. The ship type information includes any one or more of the following: ship displacement information, ship purpose information, and ship configuration information. The ship sensor information includes any one or more of the following: sensor category information, sensor model information, and sensor accuracy information. The ship data information includes any one or more of the following: ship data source, ship data format, and ship data sampling rate.

3. The ship data verification method based on multi-constraints according to claim 1, characterized in that, The step of calculating the multivariate constraint function for ship multivariate data based on the causal correlation and / or temporal correlation and / or spatial correlation between ship multivariate data includes the following steps: The causal relationship influence function among multiple data elements is calculated based on the ship's fleet status, ship type, and the degree of influence of the causal relationship between ship sensors and ship data on each data element. The time correlation influence function among multiple data elements is calculated based on the order of the ship's fleet, the ship type, and the degree of influence of the time correlation between ship sensors and ship data on each data element. The spatial correlation influence function among multiple data elements is calculated based on the influence of the ship's fleet position, ship type, and the positional correlation between ship sensors and ship data on the degree of influence of each data element. The influence function between multiple variables is calculated based on the causal relationship influence function and / or the temporal correlation influence function and / or the spatial correlation influence function between multiple variables, which is the multivariate constraint function for ship multivariate data.

4. The ship data verification method based on multi-factor constraints according to claim 1, characterized in that, The method of identifying initial abnormal data in the data collected by various sensors of the ship using data mutation rules includes: The abnormal reference value of the data is calculated based on the instantaneous change of the data collected by the ship's various sensors and / or the average change within a preset time period and / or the degree of data change within a preset time period. Initial abnormal data is identified based on the relationship between the abnormal reference value of the data and the preset data mutation threshold.

5. The ship data verification method based on multi-factor constraints according to claim 1, characterized in that, The step of calculating the correlation change reference value based on the data changes of the tuple data corresponding to the initial abnormal data within a preset time period includes: Identify the tuple data corresponding to the initial abnormal data and obtain the data change information of each metadata within a preset time period; The associated change reference value of each metadata is calculated based on the instantaneous change of each metadata and / or the average change within a preset time period and / or the degree of data change within a preset time period.

6. The ship data verification method based on multi-factor constraints according to claim 1, characterized in that, The calculation of the anomaly verification value based on the multivariate constraint function and the associated change reference value includes the following steps: The multivariate constraint impact value is calculated based on the multivariate constraint function of the ship multivariate data and the reference value of the correlation change of each metadata. Calculate the anomaly check value based on the influence value of multiple constraints.

7. The ship data verification method based on multi-factor constraints according to claim 6, characterized in that, The process of correcting initial outlier data based on multivariate constraint functions and correlation change reference values ​​to obtain validated data includes the following steps: Calculate the change weight value based on the change pattern of the tuple data corresponding to the initial abnormal data within a preset time period; Calculate the change value of the tuple data corresponding to the initial outlier data based on the multivariate constraint influence value and change weight value; The verified data is calculated based on the changes in the data of the previous time step of the initial abnormal data and the corresponding tuple data of the initial abnormal data.

8. A ship data verification system based on multi-factor constraints, characterized in that, include: The ship multi-data construction module is used to construct multi-data combinations based on one or more of the ship's fleet information, ship type information, ship sensor information, and ship data information. The multivariate constraint function construction module is used to calculate the multivariate constraint function of the ship multivariate data based on the causal correlation and / or temporal correlation and / or spatial correlation between the ship multivariate data. The initial abnormal data identification module is used to identify initial abnormal data in the data collected by various sensors of the ship according to the data mutation rules; The correlation change identification module is used to calculate the correlation change reference value based on the data changes of the tuple data corresponding to the initial abnormal data within a preset time period. The data verification module is used to calculate anomaly verification values ​​based on multivariate constraint functions and associated change reference values, identify misjudged initial abnormal data based on the anomaly verification values, and correct the initial abnormal data based on multivariate constraint functions and associated change reference values ​​to obtain verified data.

9. A computer-readable storage medium storing a computer program for electronic data interchange, wherein, The computer program causes the computer to perform the method as described in any one of claims 1-7.

10. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1-7.

Images