Marine archaeological data fusion method based on multi-agent collaboration

Abstract

The invention discloses a marine archaeological data fusion method based on multi-agent collaboration, and belongs to the field of combination of a marine archaeological data processing technology and artificial intelligence. According to the method, a cooperative system formed by combining a plurality of execution agents deployed in a distributed mode and scheduling agents deployed in a centralized mode is constructed. The execution agent is responsible for collection, preprocessing and feature extraction of multi-source heterogeneous marine archaeological data, and the scheduling agent is responsible for collaborative scheduling of the execution agent, fusion decision of multi-source features and feedback optimization of a fusion result, so that full-process closed-loop processing including dynamic parameter adjustment is formed. According to the method, through deep collaboration between the intelligent agents and a feedback mechanism based on credibility evaluation, the method can be adaptive to complex marine environment interference, the fusion process is continuously optimized, and finally a high-credibility fusion result is output. According to the invention, the accuracy and reliability of data fusion and the overall efficiency of the system are effectively improved, and reliable technical support is provided for underwater ruin positioning and cultural relic identification.

Description

Technical Field

[0001] This invention relates to the field of marine archaeological data processing technology combined with artificial intelligence, specifically to a method for marine archaeological data fusion based on multi-agent collaboration. Background Technology

[0002] Marine archaeological data is characterized by its multi-source nature, heterogeneity, high noise levels, and significant incompleteness. Especially in dynamic marine environments, traditional data fusion methods struggle to achieve consistently reliable results. Existing technologies lack adaptive fusion mechanisms to address dynamic changes in the marine environment, and significant semantic and scale inconsistencies exist between multi-source data, leading to low reliability and poor practicality of the fusion results.

[0003] The lack of a closed-loop feedback and global optimization mechanism covering the entire process of data acquisition, processing, fusion, and evaluation prevents the system from self-correcting and continuously learning based on the real-time reliability of the fusion results, thus limiting its practicality and reliability in real and complex marine archaeological scenarios. With the development of agent technology, there is an urgent need for a multi-agent collaborative method for marine archaeological data fusion, introducing agents into data fusion tasks to address the shortcomings of existing technologies. Summary of the Invention

[0004] This invention addresses the challenges of significant environmental interference, strong data heterogeneity, and low reliability of fusion results in marine archaeological data fusion. It proposes a multi-agent collaborative fusion method, employing a "distributed execution + centralized scheduling" agent architecture to achieve closed-loop optimization of the entire process from data acquisition to result feedback. Specifically, this invention designs an adaptive parameter adjustment mechanism to address dynamic changes in the marine environment and introduces a reliability assessment system based on historical data and on-site feedback, thereby achieving highly reliable and accurate data fusion in complex marine archaeological scenarios.

[0005] The technical solution of this invention is as follows, mainly focusing on the multi-agent collaborative architecture, interaction process, and module design. Combining the specific implementation details and collaborative logic of each stage, it includes the following steps:

[0006] Step S1: Construct a multi-role customized multi-agent collaborative architecture. Based on the full-process requirements of marine archaeological data processing, rationally deploy five types of core agents: data acquisition agent, data preprocessing agent, feature extraction agent, fusion decision agent, and feedback optimization agent.

[0007] Each intelligent agent possesses independent data processing, parameter configuration, and communication capabilities, and establishes pre-defined collaborative logic with each other to ensure clear division of labor and efficient linkage. Each core intelligent agent achieves real-time information exchange and collaborative linkage through a unified communication module. This communication module adopts a standardized interaction protocol, which can flexibly adapt to the communication needs of different intelligent agents, enabling rapid transmission and parsing of status information, data information, and command information between intelligent agents.

[0008] Optional deployment of environmental perception agents is available, which are specifically designed to collect marine environmental parameters, including key environmental factors such as water flow velocity, underwater noise intensity, water turbidity, and water temperature. This provides data support for adjusting the working parameters and coordinating the operation of each agent, helping each agent to adaptively optimize its working mode according to the dynamic changes in the marine environment and reducing the impact of environmental interference on the data fusion effect.

[0009] Step S2: Based on the preset collaborative scheduling logic, the startup of each intelligent agent is triggered through collaborative scheduling. Among them, the data acquisition intelligent agent prioritizes the access and adaptation of various marine archaeological data acquisition devices, and simultaneously accesses various types of detection and import devices such as sonar acoustic detection devices, 4K underwater optical imaging devices, geological stratigraphy detection devices, and historical document data import devices, so as to realize the synchronous acquisition of multi-source heterogeneous marine archaeological data.

[0010] During the data acquisition process, the data acquisition agent receives marine environmental parameters transmitted in real time from the environmental perception agent. Through collaborative interaction logic among the agents, it dynamically adjusts the parameter thresholds for data filtering. The specific threshold adjustments are based on the impact of environmental parameters on data quality. For example, when the underwater noise intensity is too high, the filtering threshold for acoustic data is automatically increased to eliminate invalid data caused by noise interference; when the water turbidity is high, the filtering rules for optical data are optimized to retain valid imaging data.

[0011] The data acquisition agent performs preliminary formatting on the filtered valid data and transmits it to the data preprocessing agent in real time through a unified data interface. This enables automated collaboration between data acquisition and parameter adjustment, ensuring that the data transmitted to subsequent stages has basic validity and standardization, thus laying the foundation for subsequent preprocessing work.

[0012] Step S3: After receiving the valid data transmitted by the data acquisition agent, the data preprocessing agent first automatically identifies and classifies the data type, distinguishing different types of heterogeneous data such as acoustic data, optical data, geological data, and historical document data.

[0013] Subsequently, the built-in processing module performs personalized preprocessing operations for different types of data, including core operations such as denoising, normalization, missing value imputation, and heterogeneous data transformation. Specifically, denoising removes redundant and interfering information from various data types; normalization converts data of different magnitudes and units to a uniform magnitude; missing value imputation uses a method adapted to the characteristics of marine archaeological data to supplement missing data, avoiding the impact of missing data on subsequent processing; and heterogeneous data transformation standardizes the conversion of data in different formats, ensuring a unified processing foundation for all types of data.

[0014] During preprocessing, the data preprocessing agent maintains real-time collaborative interaction with the data acquisition agent and the feature extraction agent through a communication interface. On one hand, it receives data quality information transmitted by the data acquisition agent in real time. If the acquired data quality is found to be substandard, it promptly provides feedback to the data acquisition agent, triggering it to re-acquire data or adjust the filtering threshold. On the other hand, it receives feature extraction requirement information transmitted by the feature extraction agent and dynamically adjusts preprocessing parameters according to the focus of feature extraction to optimize the preprocessing effect and ensure that the preprocessed data is highly compatible with the requirements of subsequent feature extraction stages. After preprocessing is completed, the standardized data is accurately transmitted to the feature extraction agent through a unified data interface, ensuring smooth and seamless data connection between each stage.

[0015] In step S4, after receiving the standardized data transmitted by the data preprocessing agent through the data interface, the feature extraction agent automatically matches the corresponding feature extraction logic according to the data type, calls the built-in data feature extraction component, and completes the extraction and screening of core archaeological data features based on the core characteristics of different types of data. Specifically, acoustic data focuses on extracting core features such as the outline and distance of underwater targets; optical data focuses on extracting core features such as the morphology and texture of cultural relics; geological data focuses on extracting core features such as stratigraphic distribution and geological structure; and historical document data focuses on extracting core features such as textual records and geographical location related to the site.

[0016] During the feature extraction process, the intelligent agent automatically removes invalid and redundant feature data, retaining core features that have practical reference value for marine archaeology, thus reducing the interference of invalid features on subsequent fusion decisions. Simultaneously, through a multi-agent collaborative scheduling mechanism, multiple feature extraction agents can process different types of standardized data in parallel, significantly improving the efficiency of feature extraction.

[0017] Each feature extraction agent summarizes and organizes the extracted core feature data, encapsulates it according to a preset format specification, and transmits it to the fusion decision agent through a unified data interface. This provides data support for subsequent fusion decision-making, ensures the collaborative connection of each agent, and achieves seamless integration between the feature extraction stage and the fusion decision-making stage.

[0018] Step S5: The fusion decision-making agent, as the core scheduling node of the multi-agent collaborative architecture, coordinates the working status of each agent. It receives valid core feature data transmitted by each feature extraction agent in real time, and simultaneously acquires the latest marine environmental parameters transmitted by the environmental perception agent. Combining the information from both, it calls upon the built-in fusion decision-making component and weight calculation module to carry out fusion decision processing of multi-source core feature data.

[0019] During the fusion decision-making process, the fusion decision-making agent dynamically adjusts the fusion weight parameters of each feature data through collaborative linkage with other agents. Specifically, the weight calculation module receives the feature confidence scores (quantization range 0-1, higher scores indicate stronger feature confidence) from the feature extraction agent and the environmental interference coefficient (quantization range 0-1, larger coefficients indicate more severe environmental interference) from the environmental perception agent. It then dynamically calculates the fusion weight of each feature using a pre-configured weighting function: Weight = Basic Weight × Feature Confidence Score / (1 + Environmental Interference Coefficient). The basic weight can be pre-configured according to different marine archaeology scenarios (no specific value is required), the feature confidence score is automatically generated by the feature extraction agent based on its processing accuracy, and the environmental interference coefficient is calculated by the environmental perception agent based on collected noise, turbidity, and other parameters. This method achieves dynamic quantitative adjustment of the fusion weight, correcting data deviations caused by environmental interference. For example, when high water turbidity reduces the reliability of optical feature data, the confidence score of optical features decreases. Combined with an increased environmental interference coefficient, their fusion weight automatically decreases. Simultaneously, the fusion weights of acoustic and geological feature data are correspondingly increased to ensure the accuracy of the fusion decision results. After completing the fusion processing, the fusion decision agent outputs preliminary marine archaeological data fusion results, including preliminary location information of underwater sites and preliminary identification information of artifacts. The preliminary fusion results are then transmitted to the feedback optimization agent via a data interface for subsequent reliability verification and feedback optimization.

[0020] Step S6: After receiving the preliminary fusion results transmitted by the fusion decision-making agent, the feedback optimization agent performs a comprehensive credibility verification of the preliminary fusion results through the built-in credibility verification component and credibility calculation module. The specific implementation is as follows: The credibility calculation module calls the historical archaeological data stored in the parameter library and receives real-time data from the field feedback. It calculates the credibility value of the preliminary fusion results using a preset formula: Credibility = a × Historical Data Matching Degree + b × Field Data Matching Degree + y × Data Integrity, where a, b, and y are weighting coefficients (a + b + y = 1), which can be flexibly configured according to the actual marine archaeology scenario without limiting specific values. Historical data matching degree is the overlap ratio between the preliminary fusion results and historical data of similar sites in the parameter library; field data matching degree is the degree of fit between the preliminary fusion results and the field feedback data; and data integrity is the coverage ratio of effective information in the preliminary fusion results. All three are quantified as values ​​between 0 and 1.

[0021] During the verification process, the accuracy, completeness, and reliability of the preliminary fusion results are comprehensively evaluated based on preset credibility evaluation standards and preset thresholds. If the credibility is greater than or equal to the preset threshold, it indicates that the preliminary fusion results meet the practical application requirements of marine archaeology, and the final fusion result is output through the data interface. Simultaneously, the optimal collaborative parameters, environmental parameters, and final fusion results from this fusion process are stored in a parameter library for management, providing reference and support for subsequent similar marine archaeological data fusion tasks. If the credibility is less than the preset threshold, it indicates that the preliminary fusion results do not meet the application requirements, and feedback is sent back to each agent through the communication interface. Based on the problems discovered during the verification process (such as low matching degree of certain features or insufficient data completeness), the working parameters of each agent are collaboratively adjusted. For example, the screening threshold of the data acquisition agent, the preprocessing parameters of the data preprocessing agent, the feature screening rules of the feature extraction agent, and the fusion weight parameters of the fusion decision agent are adjusted. After adjustment, the process returns to step S2 to repeat the fusion process until the fusion result meets the credibility requirements, forming a closed-loop collaborative optimization logic to ensure that the final output fusion result has high reliability and practicality.

[0022] Furthermore, the multi-agent collaborative architecture in step S1 adopts a design pattern of "distributed deployment + centralized collaborative scheduling," specifically implemented as follows: The data acquisition agent, data preprocessing agent, and feature extraction agent are deployed in a distributed manner, running independently on different hardware nodes and achieving collaborative linkage through network communication. This deployment mode can flexibly allocate hardware resources according to the amount of data processed, avoiding overload of a single hardware node, while improving the fault tolerance of the entire method; the failure of a single agent does not affect the normal operation of other agents. The fusion decision agent and feedback optimization agent are deployed in a centralized manner, serving as the core of collaborative scheduling, responsible for the scheduling and management of each distributed agent, coordinating the work progress and collaborative logic of each agent, and ensuring the orderly progress of the entire fusion method.

[0023] Each intelligent agent comprises four core components: a communication module, a data processing module, a parameter configuration module, and a fault self-diagnosis module. The communication module enables communication between intelligent agents and between intelligent agents and data acquisition devices, employing the TCP / IP communication protocol and adhering to unified communication standards to ensure the accuracy and efficiency of information exchange. The data processing module performs core data processing tasks for each intelligent agent, providing configurable processing interfaces and supporting the adaptation and processing of different data types. The parameter configuration module configures, modifies, and stores the working parameters of the intelligent agents, supporting real-time parameter adjustments. The fault self-diagnosis module monitors the operating status of the intelligent agents in real time, promptly issuing alarms upon detecting faults and activating redundant intelligent agents to prevent disruption to the normal operation of the entire collaborative system.

[0024] Optionally, the system can also include an information propagation optimization component. A graph neural network (GNN)-based component is a preferred implementation to further improve the collaborative efficiency between agents. Its specific application model is as follows: This component models each agent (executing agent, scheduling agent) and the standardized data blocks of each processing stage into two types of nodes. Agent nodes carry attributes such as their own function type and working state, while data block nodes carry attributes such as data type and data priority. A heterogeneous graph structure is constructed using the collaborative scheduling relationship between agents and the data flow relationship (such as data transmission and data processing requests) between agents and data blocks as edges. The component uses GNN to perform embedding representation learning on this heterogeneous graph, mining the potential correlation patterns between agents and data blocks, and thus optimizing the collaborative scheduling strategy and information interaction path. Specifically, the GNN model predicts the real-time data needs of each agent, pre-scheduling the corresponding data blocks to the cache of the requesting agent, avoiding invalid data transmission and duplicate data requests, thereby reducing data transmission overhead. Simultaneously, based on the agent association weights output by the GNN, the scheduling order of the scheduling agent to the executing agent is optimized, improving collaborative efficiency. This component does not require in-depth understanding of the specific algorithm details of GNN. It only needs to use conventional GNN models in this field (such as GAT and GCN) to achieve the above-mentioned functional adaptation, ensuring smooth collaboration between the agent and the data. Moreover, this component is an optional configuration and does not affect the implementation of the core functions of the system.

[0025] Furthermore, the multi-source heterogeneous marine archaeological data in step S2 is adapted to various marine archaeological exploration devices through the device access interface of the data acquisition agent. It supports flexible access of acoustic detection devices, optical imaging devices, geological exploration devices, and historical document data import devices, and has good device adaptability and scalability. Marine environmental parameters are collected through the sensor access interface of the environmental perception agent, including water flow velocity, underwater noise intensity, water turbidity, and water temperature. After collection, they are transmitted to the relevant agents through standardized data formats.

[0026] The filtering threshold adjustment of the data acquisition agent is achieved through a parameter collaborative adjustment mechanism. After the environmental perception agent transmits the collected environmental parameters to the data acquisition agent, the parameter configuration module of the data acquisition agent automatically calculates the threshold adjustment range, completing the dynamic adjustment of parameters without manual intervention and achieving automated collaboration.

[0027] Furthermore, the data preprocessing operation in step S3 is implemented through the built-in components of the data preprocessing agent, supporting functions such as noise reduction, normalization, missing value imputation, and heterogeneous data conversion. All preprocessing operations are implemented automatically, and the preprocessed data is uniformly converted into the JSON standardized data format, which facilitates data interaction and processing among various agents and improves data transmission and processing efficiency.

[0028] The data preprocessing agent interacts collaboratively with other agents through a unified interface. The data preprocessing agent receives data quality information transmitted by the data acquisition agent and feature extraction requirement information transmitted by the feature extraction agent in real time. It dynamically adjusts relevant preprocessing parameters through the parameter configuration module to ensure that the quality of preprocessed data is compatible with the processing requirements of subsequent feature extraction and fusion decision-making, thereby achieving collaborative optimization of each stage.

[0029] Furthermore, the feature extraction agent in step S4 supports feature extraction from multiple types of data. Its design adopts a modular approach, with independent components corresponding to the feature extraction functions of different types of data. The corresponding components can be flexibly activated according to actual data processing needs, and it has good scalability and maintainability.

[0030] The collaborative interaction among the feature extraction agents follows a unified communication standard. They share data processing status information through communication modules, coordinate the adjustment of feature extraction parameters, avoid repetitive data processing operations, and improve the overall efficiency of collaborative operation. After feature extraction is completed, the effective core feature data is aggregated to the fusion decision agent through a standardized data interface to ensure the standardization and accuracy of data transmission.

[0031] Furthermore, the fusion decision-making agent in step S5, as the core of the collaborative scheduling, plays a crucial role in realizing the collaborative scheduling of various agents, configuring fusion parameters, and generating fusion results. The functions related to fusion decision-making are implemented through built-in components, without requiring in-depth understanding of algorithmic principles; the focus is on the collaborative scheduling and parameter adjustment logic.

[0032] The fusion decision-making agent and the collaborative operation of other agents are achieved through scheduling commands. The fusion decision-making agent receives status information and data transmitted from each agent in real time, and sends scheduling commands and parameter adjustment commands to each agent based on the data processing status. Each agent adjusts its own working parameters according to the commands, realizing dynamic optimization of the entire collaborative operation and ensuring the reliability of the fusion result.

[0033] Furthermore, the feedback optimization agent in step S6 plays a core role in verifying the credibility of the fusion results, adjusting parameters, and managing the parameter library, forming a closed-loop collaborative optimization logic. Credibility verification is achieved through a built-in verification component, comparing the initial fusion results with historical data and on-site feedback data to complete a credibility assessment. The parameter library is used for data storage and management, supporting the querying, retrieval, and updating of optimal collaborative parameters, providing parameter support for subsequent similar data fusion tasks. Reverse feedback adjustment is implemented through a unified communication interface, ensuring the rapid transmission and execution of parameter adjustment commands. Attached Figure Description

[0034] Figure 1 This is an overall flowchart of the marine archaeological data fusion method based on multi-agent collaboration of the present invention.

[0035] Figure 2 This is a diagram illustrating the multi-agent collaborative interaction of the present invention. Detailed Implementation

[0036] To further illustrate the implementation process, collaborative logic, and beneficial effects of the present invention, the following detailed description is provided in conjunction with specific embodiments. These embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention; they focus on describing the deployment, interaction process, and operation of each agent within the multi-agent collaborative architecture.

[0037] This embodiment applies to the data processing of an underwater ancient city site. The core requirement is to achieve multi-source heterogeneous data fusion for the site through a multi-agent collaborative model, thereby locating the site's extent and identifying the distribution of artifacts. This embodiment focuses on verifying the feasibility, stability, and practicality of the multi-agent collaborative architecture of this invention, demonstrating its innovative advantages.

[0038] The marine archaeological data fusion method based on multi-agent collaboration in this embodiment is implemented as follows:

[0039] Step S101: Establish a multi-agent collaborative mode and construct a multi-agent collaborative architecture. Based on the data processing requirements of the underwater ancient city site, deploy one environmental perception agent, three data acquisition agents, three data preprocessing agents, three feature extraction agents, one fusion decision agent, and one feedback optimization agent on the hardware device, and complete the installation and initialization configuration of each agent.

[0040] The centralized collaborative scheduling mechanism is activated, the communication and operational parameters of each agent are configured, and the historically optimal collaborative parameters from the parameter library are retrieved to complete the initialization configuration. The fault self-check function of each agent is activated to ensure normal operation of all agents. The GNN information propagation components are activated synchronously, and the association configuration between agents and data is completed, thus enabling the multi-agent collaborative mode to start.

[0041] In step S102, the sonar acoustic detection device, 4K underwater optical imaging device, geological stratum detection device and historical document data import device are connected through the device access interface of the data acquisition agent, and the data acquisition function is started. The three data acquisition agents are responsible for the acquisition of corresponding types of data. At the same time, the environmental perception agent is started to collect the current marine environmental parameters through the sensor access interface and transmit them to each data acquisition agent in real time.

[0042] Each data acquisition agent automatically adjusts the data filtering threshold based on the received environmental parameters through the parameter configuration module. After removing invalid data, the valid data is converted into a standardized JSON format and distributed in real time to the corresponding data preprocessing agent through a unified data interface, thereby achieving automated collaboration between data acquisition and parameter adjustment.

[0043] In step S103, after each data preprocessing agent receives valid data of the corresponding type, it automatically starts the built-in preprocessing component to perform personalized preprocessing operations for different types of data, completes data denoising, normalization and format conversion, and generates standardized data. During this process, the data preprocessing agent receives feature extraction requirement information from the feature extraction agent in real time through the communication interface. When it finds that the optical data preprocessing quality does not meet the requirements, it automatically adjusts the preprocessing parameters to improve the data preprocessing quality. After the preprocessing is completed, the standardized data is transmitted to the feature extraction agent through the data interface.

[0044] In step S104, after receiving standardized data, each feature extraction agent automatically activates the corresponding feature extraction component according to the data type to complete the extraction and screening of core archaeological features and remove invalid feature data. Each feature extraction agent shares data processing status information through a communication interface and coordinates to adjust relevant feature extraction parameters to avoid duplicate data processing. After extraction, the effective core feature data is summarized and transmitted to the fusion decision agent through a unified data interface.

[0045] In step S105, the fusion decision-making agent, acting as the core of the collaborative scheduling, receives all valid core feature data and environmental parameters, activates the built-in fusion decision component, and achieves collaborative linkage with each agent through the communication interface. It dynamically adjusts the fusion weight parameters of each feature data, corrects data deviations caused by environmental interference, generates preliminary marine archaeological data fusion results, and transmits them to the feedback optimization agent through the data interface.

[0046] Step S106: After receiving the preliminary fusion result, the feedback optimization agent starts the built-in credibility verification component to compare and analyze the preliminary fusion result with the historical data and on-site feedback data of the underwater ancient city site, and calculate the credibility of the fusion result.

[0047] Meanwhile, the feedback optimization agent stores the optimal collaborative parameters, environmental parameters, and final fusion results from this fusion process into a parameter library for management, completing parameter updates and backups, and providing parameter support for subsequent similar marine archaeological data processing tasks.

[0048] Step S107: After the fusion task is completed, the feedback optimization agent sends instructions to each agent through the collaborative scheduling mechanism, disables redundant functions, saves all relevant data and operation logs of this task, and forms a complete task record; each agent enters standby state, waiting for the start of the next fusion task, thus completing all data fusion processing work in this embodiment.

[0049] The above embodiments are preferred implementations of the present invention, but the implementation of the present invention is not limited to the above embodiments. Any architectural adjustments, collaborative logic modifications, or module optimizations made without departing from the innovative spirit and principle of the present invention should be considered equivalent substitutions and are included within the protection scope of the present invention.

Claims

1. A method for fusing marine archaeological data based on multi-agent collaboration, characterized in that, This method is applied to the fusion of multi-source heterogeneous archaeological data in complex marine environments. It constructs a collaborative processing system composed of multiple software intelligent agent modules. The system adopts an architecture combining multiple distributed execution agents with at least one centralized scheduling agent. The method uses the execution agents to complete the acquisition, preprocessing, and feature extraction of marine archaeological data, and the scheduling agent to coordinate the execution agents, make fusion decisions for multi-source features, and optimize the fusion results, forming a closed-loop processing flow that includes dynamic parameter adjustment, until a final fusion result meeting preset reliability requirements is output. Both the execution agents and the scheduling agent are containerized software modules running on virtualized or physical hardware nodes.

2. The method as described in claim 1, characterized in that, In the fusion decision-making process, the scheduling agent responsible for fusion decision-making dynamically calculates the fusion weight based on the confidence score of each feature data and the current environmental interference coefficient using the following formula: Weight = Base weight × Feature confidence score / (1 + Environmental interference coefficient); The environmental interference coefficient is calculated in real time based on underwater noise intensity and water turbidity.

3. The method as described in claim 1, characterized in that, The feedback optimization specifically includes: comparing the preliminary fusion results with historical archaeological databases or on-site feedback data, and calculating the credibility using the following formula: Credibility = a × Historical data matching degree + b × On-site data matching degree + y × Data integrity; If the credibility is lower than a preset threshold, a parameter adjustment instruction is sent to the executing agent and the scheduling agent, and the fusion process is restarted.

4. The method as described in claim 1, characterized in that, The system also includes a parameter library for storing and managing the optimal collaborative parameters, environmental parameters, and fusion results from previous fusion tasks, and for providing initial parameter support for new marine archaeological data fusion tasks.

Images