A ship target fusion recognition method and device

Abstract

The application discloses a ship target fusion recognition method and device, and the method comprises the following steps: acquiring a ship target historical recognition database; constructing a ship target fusion recognition model based on the ship target historical recognition database; updating the ship target fusion recognition model by using a collected ship target recognition dataset; and processing real-time collected ship target multi-source detection data by using the updated ship target fusion recognition model to obtain ship target fusion recognition result information.

Description

Technical Field

[0001] This invention relates to the fields of multi-source information fusion processing, big data analysis, and strategy optimization, and specifically to a method and apparatus for fusion identification of ship targets. Background Technology

[0002] In today's globalized world, the ocean serves as a vital passage connecting countries worldwide, carrying a vast amount of economic, military, and scientific research activities. Ships, as the primary carriers of these maritime activities, play an indispensable role in the accurate identification of targets in several key areas.

[0003] For example, in maritime traffic management, with the booming development of global trade, maritime transport is becoming increasingly busy, and the number of ships in ports and waterways is surging. Accurate identification of ship targets is crucial for rationally planning traffic routes, effectively scheduling ships, preventing collisions, and ensuring the safety and smooth flow of maritime traffic. For instance, in busy ports, by accurately identifying ship targets, traffic management departments can rationally arrange the order of ship entry and exit, improve port operational efficiency, and reduce congestion and accident risks. In the field of marine resource development, ship target identification helps to supervise the rational development and utilization of marine resources and protect the marine ecological environment.

[0004] Traditional ship target fusion identification methods exhibit numerous drawbacks when processing multi-source data. In feature extraction, they often fail to fully exploit the potential information within multi-source data, resulting in incomplete and superficial feature extraction. For example, for multi-source data such as optical, electromagnetic, and acoustic data, traditional methods may simply process each modality individually, failing to effectively integrate complementary information between them. This hinders the identification model's comprehensive understanding and accurate grasp of ship target features. In the data association stage, traditional methods suffer from inaccurate data association. Due to differences in time, space, and feature dimensions between data acquired by different sensors, traditional methods struggle to accurately establish correspondences between multi-source data, leading to erroneous data associations and reduced identification accuracy. For instance, in scenarios where multiple ships appear simultaneously, traditional methods may incorrectly associate multi-source data from different ships, resulting in misidentification of ship targets. Summary of the Invention

[0005] This invention primarily addresses the problem of how to efficiently and accurately identify relevant information about maritime target vessels. This invention discloses a fusion identification method and apparatus for vessel targets.

[0006] In a first aspect, the present invention discloses a fusion identification method for ship targets, comprising:

[0007] S1, Obtain the historical identification database of ship targets;

[0008] S2, Based on the historical identification database of ship targets, a ship target fusion identification model is constructed;

[0009] S3, update the ship target fusion recognition model using the collected ship target recognition dataset;

[0010] S4. Using the updated ship target fusion recognition model, the real-time collected multi-source detection data of ship targets is processed to obtain the fusion recognition result information of ship targets.

[0011] A second aspect of the present invention discloses a fusion identification device for ship targets, the device comprising:

[0012] Memory containing executable program code;

[0013] A processor coupled to the memory;

[0014] The processor calls the executable program code stored in the memory to execute the fusion identification method for ship targets.

[0015] In a third aspect, the present invention discloses a computer-readable storage medium storing computer instructions, which, when invoked by a computer, are used to execute the fusion identification method for ship targets.

[0016] In a fourth aspect of this invention, an information data processing terminal is disclosed, which is used to implement the aforementioned method for fusion identification of ship targets.

[0017] The beneficial effects of this invention are as follows:

[0018] This invention overcomes the limitations of single data sources in complex marine environments by integrating three types of multi-source information: optical detection data, electromagnetic detection data, and acoustic detection data. Optical data provides target appearance characteristics, electromagnetic data reflects the target's electromagnetic radiation characteristics, and acoustic data captures underwater acoustic features. These three types of data complement each other, effectively reducing the interference of weather, sea state, and other factors on individual data, providing a more comprehensive feature foundation for the integrated identification of ship targets, and improving the robustness of identification.

[0019] This invention introduces confidence level calculation. By analyzing the differences between newly collected data and similar data in historical databases, the reliability of the new data is assessed, and a weighted solution model is constructed based on the confidence level to update the fusion recognition model. This mechanism enables the model to dynamically adjust according to new data, making full use of high-reliability data to optimize the model while reducing interference from low-quality data. At the same time, by introducing trend information of data changes through first and second derivatives, the updated model can better adapt to the dynamic changes of ship characteristics, ensuring the stability of recognition accuracy in long-term use.

[0020] This invention achieves accurate and joint identification of ship target categories and threat levels through the synergistic effect of multi-source data fusion, two-dimensional information encoding, multi-regression model fusion verification, and dynamic update mechanism. Moreover, the model can be dynamically optimized with actual data, effectively meeting the needs of maritime traffic control, national defense security and other fields for highly accurate and adaptable ship identification technology. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention. Detailed Implementation

[0022] To better understand the content of this invention, an embodiment is provided here.

[0023] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention.

[0024] In a first aspect, the present invention discloses a fusion identification method for ship targets, comprising:

[0025] S1, Obtain the historical identification database of ship targets;

[0026] S2, Based on the historical identification database of ship targets, a ship target fusion identification model is constructed;

[0027] S3, update the ship target fusion recognition model using the collected ship target recognition dataset;

[0028] S4. Using the updated ship target fusion recognition model, the real-time collected multi-source detection data of ship targets is processed to obtain the fusion recognition result information of ship targets.

[0029] The ship target historical identification database includes a ship target multi-source detection dataset, a corresponding target category dataset, and a target threat level set; the ship target multi-source detection dataset includes ship target multi-source detection data; the target category dataset includes target category data; and the target threat level set includes target threat levels.

[0030] Each piece of multi-source detection data for a ship target in the aforementioned multi-source detection dataset has corresponding data in the target category dataset and the target threat level set, respectively.

[0031] The multi-source detection data of the ship target includes optical detection data, electromagnetic detection data and acoustic detection data;

[0032] The ship target fusion recognition model constructed based on the historical ship target recognition database includes:

[0033] S21, perform two-dimensional numerical encoding on the data corresponding to each ship target multi-source detection data in the target category dataset and the target threat level set in the ship target history identification database to obtain the corresponding target identification data;

[0034] S22, using the optical detection data, electromagnetic detection data, and acoustic detection data of the multi-source detection data of the ship target as three types of independent variables, and the target identification data as the dependent variable;

[0035] S23, Perform multi-type multiple regression analysis on the three types of independent and dependent variables to obtain a set of regression functions;

[0036] S24, perform fusion verification analysis on the set of regression functions to obtain the ship target fusion recognition model.

[0037] The two-dimensional numerical encoding can adopt prime number product encoding or binary interleaved encoding;

[0038] This invention uses two-dimensional numerical encoding to transform target category data and target threat level into unified target identification data, achieving a joint characterization of ship type and threat level. This encoding method integrates two types of key information into a holistic analysis object, avoiding the information fragmentation caused by processing category and threat level separately in traditional methods. It enables the model to learn the correlation features between the two simultaneously, providing a more direct basis for subsequent comprehensive decision-making and improving the practicality of the identification results.

[0039] The various types of multiple regression analysis include multiple linear regression, nonlinear regression, and fuzzy regression. The dependent variable and multiple independent variables are processed using the above methods to obtain the first regression function, the second regression function, and the third regression function, respectively.

[0040] The process of performing fusion verification analysis on the set of regression functions to obtain the ship target fusion recognition model includes:

[0041] S241, perform hypothesis testing on the distribution of each regression function in the set of regression functions to obtain a set of test indicators for each regression function; the set of test indicators includes the coefficient of determination, mean squared error, and F-value;

[0042] S242, For each regression function in the set of multiple regression models, a reasonableness test is performed to obtain the variance inflation factor of each regression function;

[0043] S243 utilizes the set of test indicators for all regression functions and the variance inflation factor to perform fusion calculations and obtain the ship target fusion identification model.

[0044] The expression for the fusion calculation process is:

[0045]

[0046] in, This represents the j-th regression function in the set of regression functions. It is the ReLU activation function. , , and These represent the coefficient of determination, F-value, mean squared error, and variance inflation factor, respectively. These are the preset first constant factor, second constant factor, and third constant factor, respectively. This is the expression for the ship target fusion recognition model.

[0047] The expression used for fusion calculation in this invention weights and fuses multiple regression functions by integrating test indicators (including indicators reflecting goodness of fit, significance, and error). Simultaneously, it introduces activation functions and nonlinear transformations to fully leverage the advantages of different regression functions. Specifically, processing the indicators reflecting fit highlights the contribution of the better-performing regression function, while processing error and multicollinearity indicators suppresses the interference of poorly performing regression functions. The nonlinear transformation adapts to the complex relationships between the indicators. The resulting fusion identification model takes into account the characteristics of different regression methods, avoids the limitations of a single regression model, and improves its adaptability and identification accuracy for multi-source heterogeneous data.

[0048] This invention employs three different types of multiple regression analysis: multiple linear regression, nonlinear regression, and fuzzy regression. Regression functions are constructed for each, and a fusion test is then used to obtain a fusion recognition model. The combination of multiple regression methods can adapt to linear and nonlinear relationships between multi-source data, covering data characteristics in different scenarios. Furthermore, the fusion test, based on the coefficient of determination, mean squared error, F-value, and variance inflation factor, comprehensively evaluates the performance of each regression function. By weighted fusion, the advantages of each function are integrated, effectively avoiding the limitations of a single regression model and improving the model's accuracy and generalization ability.

[0049] The step of updating the ship target fusion recognition model using the collected ship target recognition dataset includes:

[0050] The collected ship target identification dataset includes ship target detection data, corresponding target category data, and target threat level; the ship target detection data includes optical detection data, electromagnetic detection data, and acoustic detection data.

[0051] S31, calculate the confidence level of the collected ship target recognition dataset to obtain the confidence level value of each ship target detection data in the collected ship target recognition dataset;

[0052] S32, using the collected ship target recognition dataset and confidence value, update the ship target fusion recognition model to obtain the updated ship target fusion recognition model.

[0053] The confidence calculation of the acquired ship target recognition dataset, to obtain the confidence value of each ship target detection data in the acquired ship target recognition dataset, includes:

[0054] S311, for each ship target detection data in the collected ship target identification dataset, according to its corresponding target category data and target threat level, obtain the multi-source detection data of ship targets with the same target category data and target threat level in the ship target history identification database;

[0055] S312, Subtract the corresponding category data from the ship target detection data and the corresponding ship target multi-source detection data to obtain the difference data for each category;

[0056] S313, calculate the variance of the difference data for all categories, evaluate the variance, and obtain the confidence value of the ship target detection data.

[0057] The data categories include optical detection data, electromagnetic detection data, and acoustic detection data.

[0058] The expression for calculating the confidence level information is:

[0059] ,

[0060] in, Let represent the second-order polynomial of a Chebyshev polynomial of the first kind. Let s represent a second-order Laguerre polynomial, where s is the confidence information. This represents the variance.

[0061] The expression used for confidence score calculation employs a specific polynomial to process the variance of the differential data, transforming the dispersion between data into a reasonable confidence score value. This polynomial possesses excellent numerical mapping properties, effectively compressing the impact of extreme variance values ​​while amplifying the differences between data of varying reliability. This ensures that the obtained confidence score accurately reflects the consistency between newly collected data and historical data (higher consistency equates to higher confidence, and vice versa), while also providing a clear reliability basis for subsequent model updates and avoiding interference from low-quality data.

[0062] The collected ship target identification dataset includes ship target detection data, corresponding target category data, and target threat level; the ship target detection data includes optical detection data, electromagnetic detection data, and acoustic detection data.

[0063] The process of updating the ship target fusion recognition model using the acquired ship target recognition dataset and confidence values ​​to obtain an updated ship target fusion recognition model includes:

[0064] Based on the collected ship target identification dataset and confidence values, a weighted solution model is constructed.

[0065] The expression for the weight calculation model is:

[0066] ,

[0067]

[0068] in, and These represent the detection data of the i-th ship target in the acquired ship target recognition dataset and its corresponding confidence value. and The weighting coefficients to be solved are: and These are the first and second derivatives of the expression for the ship target fusion recognition model, respectively. To obtain multi-source detection data of the i-th ship target in the acquired ship target identification dataset, based on its corresponding target category data and target threat level, multi-source detection data of ship targets with the same target category data and target threat level in the ship target historical identification database are obtained. Then, the difference value between the obtained i-th ship target detection data and its corresponding multi-source detection data is calculated. To obtain the corresponding target identification data, two-dimensional numerical encoding is performed on the target category data and target threat level corresponding to the i-th ship target detection data, where N is the total number of ship target detection data.

[0069] Solving the weight calculation model yields the following results: and ;

[0070] Based on the solution and The updated ship target fusion recognition model is constructed, and its expression is:

[0071]

[0072] in, y This refers to the fusion identification results of the ship targets output by the model.

[0073] The expression used to solve the weighting model can be optimized by constructing an optimization objective that includes confidence level, first-order derivative, and second-order derivative, thus obtaining reasonable weighting coefficients. The introduction of confidence level gives higher weight to highly reliable data in the optimization process, ensuring the dominant role of high-quality data. The first and second derivatives reflect the model's sensitivity to data changes and the rate of change, respectively. By incorporating them into the optimization objective, the solved coefficients can adapt to the dynamic trends of the data, avoiding bias caused by over-reliance on static data during model updates, and ensuring the rationality of the weighting coefficients and the scientific nature of the model updates.

[0074] The expression for the ship target fusion recognition model used for updating processing adds first-order and second-order derivative terms to the original fusion model, and combines them with the weighting coefficients obtained from the solution, enabling the model to dynamically adjust with new data. The first-order derivative term captures the instantaneous trend of data changes, while the second-order derivative term reflects the acceleration of change. The combination of the two allows the model not only to adapt to the characteristics of the current data but also to predict the direction of data change, effectively improving the model's adaptability to dynamic changes in ship target characteristics and ensuring the stability of recognition accuracy in long-term use.

[0075] The weight calculation model can be solved using a particle filter algorithm or a numerical iterative optimization algorithm.

[0076] The difference value is a vector composed of the difference values ​​of the three data types.

[0077] In all embodiments of the present invention, the variables involved in all computational expressions or mathematical functions have been dimensionlessized before computation.

[0078] In all embodiments of the present invention, the values ​​of the independent variables in the input of all computational expressions or mathematical functions meet the reasonable requirements of the input range of the computational expressions or mathematical functions, and can ensure that the computational expressions or mathematical functions can be calculated smoothly without violating physical laws or mathematical rules.

[0079] A second aspect of the present invention discloses a fusion identification device for ship targets, the device comprising:

[0080] Memory containing executable program code;

[0081] A processor coupled to the memory;

[0082] The processor calls the executable program code stored in the memory to execute the fusion identification method for ship targets.

[0083] In a third aspect, the present invention discloses a computer-readable storage medium storing computer instructions, which, when invoked by a computer, are used to execute the fusion identification method for ship targets.

[0084] In a fourth aspect of this invention, an information data processing terminal is disclosed, which is used to implement the aforementioned method for fusion identification of ship targets.

[0085] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A fusion identification method for ship targets, characterized in that, include: S1, Obtain the historical identification database of ship targets; S2, Based on the aforementioned historical ship target identification database, a ship target fusion identification model is constructed, including: S21, perform two-dimensional numerical encoding on the data corresponding to each ship target multi-source detection data in the target category dataset and the target threat level set in the ship target history identification database to obtain the corresponding target identification data; S22, using the optical detection data, electromagnetic detection data, and acoustic detection data of the multi-source detection data of the ship target as three types of independent variables, and the target identification data as the dependent variable; S23, Perform multi-type multiple regression analysis on the three types of independent and dependent variables to obtain a set of regression functions; S24, perform fusion verification analysis on the set of regression functions to obtain the ship target fusion recognition model; The multivariate regression analysis includes multiple linear regression, nonlinear regression and fuzzy regression. The dependent variable and multiple independent variables are processed using the above methods to obtain the first regression function, the second regression function and the third regression function, respectively. S3, update the ship target fusion recognition model using the collected ship target recognition dataset; S4. Using the updated ship target fusion recognition model, the real-time collected multi-source detection data of ship targets is processed to obtain the fusion recognition result information of ship targets. The ship target historical identification database includes a ship target multi-source detection dataset, a corresponding target category dataset, and a target threat level set; the ship target multi-source detection dataset includes ship target multi-source detection data; the target category dataset includes target category data; and the target threat level set includes target threat levels. Each ship target multi-source detection data in the ship target multi-source detection dataset has corresponding data in the target category dataset and the target threat level set, respectively. The multi-source detection data for ship targets includes optical detection data, electromagnetic detection data, and acoustic detection data.

2. The fusion identification method for ship targets as described in claim 1, characterized in that, The process of performing fusion verification analysis on the set of regression functions to obtain the ship target fusion recognition model includes: S241, perform hypothesis testing on each regression function in the set of regression functions to obtain a set of test indicators for each regression function; the set of test indicators includes the coefficient of determination, mean squared error, and F-value; S242, For each regression function in the set of regression functions, a rationality test is performed to obtain the variance inflation factor of each regression function; S243, using the set of test indicators and variance inflation factor of all regression functions, the set of regression functions is fused and calculated to obtain the ship target fusion identification model.

3. The fusion identification method for ship targets as described in claim 2, characterized in that, The expression for the fusion calculation process is: in, This represents the j-th regression function in the set of regression functions. It is the ReLU activation function. , , and These represent the coefficient of determination, F-value, mean squared error, and variance inflation factor, respectively. These are the preset first constant factor, second constant factor, and third constant factor, respectively. This is the expression for the ship target fusion recognition model.

4. The fusion identification method for ship targets as described in claim 1, characterized in that, The step of updating the ship target fusion recognition model using the collected ship target recognition dataset includes: The collected ship target identification dataset includes ship target detection data, corresponding target category data, and target threat level; the ship target detection data includes optical detection data, electromagnetic detection data, and acoustic detection data. S31, calculate the confidence level of the collected ship target recognition dataset to obtain the confidence level value of each ship target detection data in the collected ship target recognition dataset; S32, using the collected ship target recognition dataset and confidence value, update the ship target fusion recognition model to obtain the updated ship target fusion recognition model.

5. The fusion identification method for ship targets as described in claim 4, characterized in that, The confidence calculation of the acquired ship target recognition dataset, to obtain the confidence value of each ship target detection data in the acquired ship target recognition dataset, includes: S311, for each ship target detection data in the collected ship target identification dataset, according to its corresponding target category data and target threat level, obtain the multi-source detection data of ship targets with the same target category data and target threat level in the ship target history identification database; S312, Subtract the corresponding category data from the ship target detection data and the corresponding ship target multi-source detection data to obtain the difference data for each category; S313, calculate the variance of the difference data for all categories, evaluate the variance, and obtain the confidence value of the ship target detection data.

6. A fusion identification device for ship targets, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the fusion identification method for ship targets as described in any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which, when invoked by a computer, are used to execute the fusion identification method for ship targets as described in any one of claims 1 to 5.

8. An information data processing terminal, characterized in that, The information data processing terminal is used to implement the fusion identification method for ship targets as described in any one of claims 1 to 5.

Images