Marine terrain multi-source data integration processing system
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 南通市万博测绘咨询有限公司
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-19
Smart Images
Figure CN122241004A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of marine surveying and mapping technology, and more specifically, to a system for integrating and processing multi-source marine topographic data. Background Technology
[0002] With the ocean's increasing importance in the global economy, resources, and environment, marine scientific research and marine resource development activities are flourishing. In this process, accurate marine topographic mapping has become a crucial foundational task, playing a vital role in marine geological structure research, marine engineering construction, and marine ecological protection. Accurate seabed topographic data can be used in resource development, environmental protection, underwater navigation, and seabed structure research. Existing marine exploration methods include airborne lidar measurement technology, shipborne sonar detection technology, and submersible seabed topographic measurement technology. While these technologies offer high accuracy, they require significant human, financial, material, and time resources and have numerous limitations, inevitably leading to errors in the data. Traditional single-source data sources in marine topographic mapping suffer from incomplete information and limited accuracy. Multi-source data fusion technology can integrate the advantages of different data sources, providing richer and more accurate information for marine topographic mapping. In-depth research on the application of multi-source data fusion technology in marine topographic mapping has significant theoretical and practical value. Therefore, there is an urgent need for a mapping system that can cope with complex marine topography, multiple data sources, and can comprehensively process objects with heterogeneous data types, taking into account the unique attributes of marine data such as the diversity of sources, structural differences, temporal correlation, and spatial resolution. Summary of the Invention
[0003] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a marine topographic multi-source data integration and processing system to solve one or more of the above-mentioned problems.
[0004] To achieve the above objectives, the present invention provides the following technical solution: A multi-source marine topography data integration and processing system, including those located within a local area network: The system includes: a field acquisition terminal for obtaining terrain information of the detection area; a data interaction terminal for collecting and storing all terrain information; a main management terminal for integrating and processing all information; a secondary management terminal for auxiliary information processing; a control and management terminal for realizing human-computer interaction; and extended early warning and distribution terminals. The early warning terminal is used for simple prevention and control alerts, triggering logic based on pre-stored data thresholds or keywords. The distribution terminal intervenes according to the instructions of the control and management terminal or adjusts and distributes the field acquisition terminals distributed in the sea area based on the data volume assessment. The main management terminal filters, stores, and stacks the pre-stored category-identified data to build a marine topography detection model; The secondary management terminal stores and categorizes all unparsed data, generates data anomaly alerts, and promptly reports them to the control management terminal.
[0005] Furthermore, the main management terminal receives datasets sent back by one or more field acquisition terminals, classifies them according to location, and marks them according to timestamps. If there are multiple datasets in the same category and with the same mark, each dataset corresponds to one volume count. Store datasets with existing category labels, and store unparsed data by location and time labeling to the secondary management terminal.
[0006] Furthermore, for the same category and the same label that only exists once, the data can be directly input into the existing dataset with a cycle of 3 to 7 days; For multiple occurrences of the same category and label, perform difference elimination after inputting an existing dataset, with a cycle of 2 to 5 days. The latest cycle data is updated based on the previous period's data; Using location as the quantitative factor and time as the variable, corresponding marine topography detection models are established by training the model using a training set for each region.
[0007] Furthermore, when there are multiple occurrences of the same category and the same label, the data discrepancies are eliminated according to the formula: Where α is the weighting coefficient, ΔX is the data after eliminating differences, and X i The original historical data is represented by X0, the mean of the input data is represented by δ, the standard deviation is represented by T, and the number of times the data is measured is represented by T.
[0008] Furthermore, the training set includes all data after eliminating discrepancies and random bias values; The bias values are updated using the Adam optimization algorithm with a learning rate of 0.001, the number of training iterations is set to 100, and the batch size is 64.
[0009] Furthermore, a blank node network is constructed, and after inputting it into the training set, it is mapped to form a basic node network. The loss value is calculated based on the mean squared error, and the loss value is directly compensated for the original nodes to form a bias network. If the loss function continues to decrease and the number of iterations has not reached 100, the bias value in the bias network is adjusted according to the Adam optimization algorithm. A new bias network is reconstructed based on the adjusted bias value until the loss function no longer decreases or the number of iterations reaches 100. The training ends, and the marine topography detection model is completed.
[0010] Furthermore, the source of unparsed datasets in the secondary management terminal was traced to the field acquisition terminal. If the volume count associated in the primary management terminal and the secondary management terminal is the same, then ignore the data in the secondary management terminal and build the marine topography detection model normally. If the volume counts associated with the primary management terminal and the secondary management terminal are inconsistent, the dataset of the current primary management terminal will be removed, the volume counts will be automatically discarded, and then the marine topography detection model will be built.
[0011] Furthermore, the allocation terminal prevents the entry of a corresponding number of field data collection terminals or directs their departure, ensuring that the volume of data collected in the same location area does not exceed nine times.
[0012] In summary, the present invention has the following beneficial effects: it classifies locations by region, sorts them by time, eliminates differences by calculating multi-source data, and then uses the calculated dataset for simple training to build a marine topographic detection model for the target location region, which is updated periodically. Compared with a single data source, the information is more comprehensive and the accuracy is higher. It also overcomes the differences caused by the differences in multi-source data, strengthens the temporal correlation, and improves the spatial resolution. Attached Figure Description
[0013] Figure 1 A system operation flowchart of one embodiment of the present invention; Figure 2 A flowchart illustrating the construction of a marine terrain model in one embodiment of the present invention; Figure 3 This is a schematic diagram of a node network in one embodiment of the present invention. Detailed Implementation Example
[0014] The following is in conjunction with the appendix Figure 1-3 The present invention will be described in further detail below.
[0015] Ocean topography multi-source data integration and processing system, such as Figure 1As shown, the system includes communication connections, field acquisition terminals within the same local area network (LAN), data exchange terminals, primary and secondary management terminals, control and management terminals, and extended functional early warning and distribution terminals. Field acquisition terminals are distributed across different target locations, categorized by spacing or depth. Typically, multiple terminals of the same or different types are deployed near a single observation point, such as floating terminals, submersible terminals, and contact terminals. These terminals have different functions and collect different types of data, though some overlap may exist. Field acquisition terminals within the same sea area can be considered a single target group. The system processes different data sources within this target group, and the field acquisition terminals present topographic information of the probed area in different data formats. Due to the unique characteristics of the sea, it is difficult to establish a complete LAN directly in most sea areas. Therefore, it is necessary to extend the network by using relay data exchange terminals to collect and store topographic information from all field acquisition terminals within the signal range, and then relay it out to complete the full coverage of the LAN. All information is ultimately integrated into a processing terminal comprised of a primary management terminal and secondary management terminals. The primary management terminal is responsible for classifying and processing data and building predictive models, while the secondary management terminals are responsible for classifying, labeling, and temporarily storing objectively existing data outside the required scope, playing a supplementary role. Extending from the processing terminal is the control and management terminal, primarily for human-computer interaction, including but not limited to database queries, terminal allocation, and manual system updates. As for the early warning terminal and the allocation terminal, these are intelligent modules derived from the processing terminal, autonomously making simple judgments and responses. For example, the early warning terminal can provide simple prevention and control alerts; if the collected data or predicted estimates exceed pre-set data thresholds in the system or directly trigger keywords to cause an alarm, it does not have processing capabilities and is only used for warning. Another example is the allocation terminal, which intervenes based on instructions from the control and management terminal or based on multi-source data assessment within the same location area, promptly adjusting and allocating on-site collection terminals within the sea area to avoid resource waste or long-term lack of sufficient analysis samples. The allocation terminal prevents the addition or relocation of a corresponding number of on-site collection terminals, ensuring that the number of such instances within the same location area does not exceed nine. That is, regardless of type or size, a maximum of nine field data acquisition terminals are allowed to exist simultaneously in the same location area. This is mainly related to some logical choices in the subsequent model building, and the number of nine terminals can basically cover the data acquisition needs of the entire area.
[0016] The main management terminal intercepts key information categories and filters, stores, and stacks the data according to pre-stored category identifiers, such as... Figure 2As shown, a marine topography survey model is constructed. The main management terminal receives datasets from one or more field acquisition terminals, classifies them by location, and labels them by timestamp. If multiple datasets exist within the same category and label, each dataset corresponds to one volumetric measurement. The main management terminal only stores datasets with existing category labels; unparsed data is stored to the secondary management terminal after being labeled by location and time. If only one volumetric measurement exists for the same category and label (i.e., only one field acquisition terminal exists in the same location area), the existing dataset is directly input with a cycle of 3-7 days. If multiple volumetric measurements exist for the same category and label (i.e., multiple field acquisition terminals exist in the same location area), the existing dataset is input with a cycle of 2-5 days, and data differences are eliminated using a formula: Where α is the weighting coefficient, ΔX is the data after eliminating differences, and X i The original historical data is used, where X0 is the mean of the input data, δ is the standard deviation, and T is the number of times the data is generated. The training set includes all data after eliminating discrepancies and random bias values. The training input data is input to the node, and the bias values are updated using the Adam optimization algorithm with a learning rate of 0.001. The number of training iterations is set to 100, and the batch size is 64. The format of the input data is particularly important for the training of subsequent node networks, directly affecting the accuracy of training and prediction. For example... Figure 3 As shown, based on the correlation between marine topography and gravity data, vertical deviation, gravity anomaly, and vertical gravity gradient are used as input vectors. The model's deep-sea values and slope at the 4×4 grid surrounding the training point (prism) and prediction point (triangle) are used as input data to assist training. The slope is the ratio of the seabed topographic height difference to the distance. Using location as the quantifier and time as the variable, a corresponding marine topography detection model is established after training with a training set by region. Subsequent updates are based on the bias node network of the previous period's data. Initially, a blank node network is built, which is then mapped after inputting into the training set to form a basic node network. The loss value is calculated based on the mean squared error, and the original nodes are directly compensated for the loss value to form a bias network. If the loss function continues to decrease and the number of iterations has not reached 100, the bias value in the bias network is adjusted according to the Adam optimization algorithm. A new bias network is then constructed based on the adjusted bias value until the loss function no longer decreases or the number of iterations reaches 100, at which point training ends, and the marine topography detection model is completed.
[0017] The secondary management terminal primarily stores and categorizes non-routine data filtered out by the primary management terminal. It also employs several different alarm logics to generate data anomaly alerts and promptly report them to the control management terminal. Specifically, it tracks the source of unparsed datasets from field acquisition terminals within the secondary management terminal. If the associated volume and frequency data are consistent between the primary and secondary management terminals, the secondary terminal's data is ignored, and the marine topography detection model is built normally. This indicates the unparsed data is caused by a normal, high-frequency phenomenon and is not abnormal. If the associated volume and frequency data are inconsistent between the primary and secondary management terminals, the dataset from the current primary management terminal is removed, and the volume and frequency data are automatically discarded before building the marine topography detection model. This indicates the unparsed data is a sudden event with questionable reproducibility and is not included in the model's training dataset. These data can be manually retrieved later, providing a traceable record. This system integrates and processes multi-source information within the region, effectively assisting professionals in making reasonable predictions and additions to marine topography, improving the efficiency of topography detection, and possessing significant practical value.
[0018] It should be noted that this specific embodiment is merely an explanation of the present invention and is not intended to limit the present invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but as long as they are within the scope of the claims of the present invention, they are protected by patent law.
Claims
1. A system for processing multi-source data of ocean topography, characterized in that, Including those within a local area network: The system includes: a field acquisition terminal for obtaining terrain information of the detection area; a data interaction terminal for collecting and storing all terrain information; a main management terminal for integrating and processing all information; a secondary management terminal for auxiliary information processing; a control and management terminal for realizing human-computer interaction; and extended early warning and distribution terminals. The early warning terminal is used for simple prevention and control alerts, triggering logic based on pre-stored data thresholds or keywords. The distribution terminal intervenes according to the instructions of the control and management terminal or adjusts and distributes the field acquisition terminals distributed in the sea area based on the data volume assessment. The main management terminal filters, stores, and stacks the pre-stored category-identified data to build a marine topography detection model; The secondary management terminal stores and categorizes all unparsed data, generates data anomaly alerts, and promptly reports them to the control management terminal.
2. The ocean topography multi-source data integrated processing system according to claim 1, characterized in that, The main management terminal receives datasets sent back by one or more field acquisition terminals, classifies them by location, and marks them by timestamp. If there are multiple datasets in the same category and with the same mark, each dataset corresponds to one volume count. Store datasets with existing category labels, and store unparsed data by location and time labeling to the secondary management terminal.
3. The ocean topography multi-source data integrated processing system according to claim 2, characterized in that, For data with the same category and label that appears only once, use an existing dataset with a cycle of 3 to 7 days. For multiple occurrences of the same category and label, perform difference elimination after inputting an existing dataset, with a cycle of 2 to 5 days. The latest cycle data is updated based on the previous period's data; Using location as the quantitative factor and time as the variable, corresponding marine topography detection models are established by training the model using a training set for each region.
4. The system for processing of multi-source data of marine terrain according to claim 3, wherein, When multiple quantities exist within the same category and under the same label, data discrepancies are eliminated using the following formula: wherein a is a weight coefficient, ΔX is the data after eliminating the difference, X i is the original historical data, X0 is the mean of the input data, δ is the standard deviation, and T is the volume number.
5. The marine topography multi-source data integration and processing system according to claim 4, characterized in that, The training set includes all data after eliminating differences and random bias values; The bias values are updated using the Adam optimization algorithm with a learning rate of 0.001, the number of training iterations is set to 100, and the batch size is 64.
6. The marine topography multi-source data integration and processing system according to claim 5, characterized in that, A blank node network is built, and after inputting it into the training set, it is mapped to form a basic node network. The loss value is calculated based on the mean squared error, and the loss value is directly compensated for the original nodes to form a bias network. If the loss function continues to decrease and the number of iterations has not reached 100, the bias value in the bias network is adjusted according to the Adam optimization algorithm. A new bias network is reconstructed based on the adjusted bias value until the loss function no longer decreases or the number of iterations reaches 100. The training ends, and the marine topography detection model is built.
7. The marine topography multi-source data integration and processing system according to claim 4, characterized in that, Tracking the source of unparsed datasets from field acquisition terminals in secondary management terminals. If the volume count associated in the primary management terminal and the secondary management terminal is the same, then ignore the data in the secondary management terminal and build the marine topography detection model normally. If the volume counts associated with the primary management terminal and the secondary management terminal are inconsistent, the dataset of the current primary management terminal will be removed, the volume counts will be automatically discarded, and then the marine topography detection model will be built.
8. The marine topography multi-source data integration and processing system according to claim 1, characterized in that, The allocation terminal prevents the entry of a corresponding number of field data collection terminals or directs their departure, ensuring that the volume of data collected in the same location area does not exceed nine times.