A data acquisition and processing system for marine exploration missions

Abstract

This invention discloses a data acquisition and processing system for marine exploration missions, specifically relating to the field of data acquisition and transmission. It addresses the problem of transmitting multi-source observation data in heterogeneous link environments during existing marine exploration. The system employs a data acquisition module to collect and generate an event feature set containing data structure characteristics and information value density. A performance prediction module combines the performance parameters of each heterogeneous link with external environmental information to generate performance prediction curves. A matching and scoring module calculates the matching degree based on the event feature set and prediction curves and outputs a matching degree scoring matrix. A link allocation module generates a multi-link allocation table based on the matching degree score and information value density. A data transmission module slices the observation data according to the allocation table and assigns a global time index. A data integration module rearranges and merges the slices at the aggregation end according to the time index, generating complete observation data storage. This achieves an efficient, time-consistent, and resource-optimized marine exploration data transmission and processing flow.

Description

Technical Field

[0001] This invention relates to the field of data acquisition technology, and more specifically, to a data acquisition and processing system for marine exploration missions. Background Technology

[0002] In large-scale distributed marine observation arrays, observation nodes are typically distributed across different types of carriers, including unmanned buoys that float on the surface for extended periods, research vessels performing regional patrols, autonomous underwater vehicles (AUVs) with short-term stays, and base stations permanently deployed on the seabed. These nodes often maintain contact with the data aggregation center through various communication methods, such as narrowband satellite links, surface microwave communication, underwater acoustic communication, and short-range high-frequency wireless communication. Due to significant differences in the physical characteristics, available bandwidth, latency fluctuations, and available time windows of various links, data for the same observation event may need to be transmitted simultaneously through multiple heterogeneous links during a mission. Furthermore, the transmission rhythm, packet loss patterns, and reachability status of each link are inconsistent, leading to structural misalignments in arrival time, slice boundaries, and temporal consistency of cross-link data. Especially during sudden marine events (such as abnormal waves, submarine earthquakes, and instability of marine equipment), high-value data collected by observation nodes needs to be delivered to the analysis end in the shortest possible time. However, existing fixed-period batch reporting mechanisms cannot dynamically adjust reporting priorities based on the value density of the event, often resulting in high-value short-term data being delayed or mixed with low-value background data, reducing information utilization efficiency.

[0003] Furthermore, the lack of a unified triggering and alignment mechanism between different links makes it difficult to achieve accurate time synchronization when multimodal data of the same event is converged, affecting the effectiveness of subsequent correlation analysis and model training. In such a scenario, there is an urgent need for a data acquisition, reporting, and scheduling method that can cross heterogeneous links and is dynamically driven by event value, enabling the observation array to efficiently, accurately, and on-demand transmit key data under complex communication conditions at sea, ensuring the real-time performance of the mission and maximizing its scientific value. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a data acquisition and processing system for marine exploration missions to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A data acquisition and processing system for marine exploration missions includes:

[0007] The data acquisition module is used to collect observation data at the observation nodes and generate an event feature set containing data structure characteristics and information value density.

[0008] The performance prediction module is used to collect the performance parameters of each heterogeneous link and generate performance prediction curves for the future transmission window of the heterogeneous link by combining external environmental information.

[0009] The matching and scoring module is used to calculate the matching degree between the event feature set and the performance prediction curve, and output the matching degree scoring matrix between the observation data corresponding to the event and each link.

[0010] The link allocation module is used to generate a multi-link allocation table based on the matching degree score matrix and information value density.

[0011] The data transmission module is used to slice the observation data according to the multi-link allocation table, assign a globally consistent time index to each slice, and write the time index into the data packet header;

[0012] The data integration module is used at the data aggregation end to rearrange and merge slices arriving from multiple links based on time indexes, generate complete observation data, and store it.

[0013] In a preferred embodiment, the data acquisition module collects observation data at observation nodes and generates an event feature set containing data structure characteristics and information value density, specifically including:

[0014] In the current exploration mission, observation data is received through multi-source observation nodes;

[0015] The observation data packets from different sources are parsed into independent records according to the protocol header, and the records are synchronized based on the timestamp;

[0016] Perform structured analysis on the synchronized data to extract the sampling interval distribution, numerical change gradient, and data sequence length, and integrate them into structural parameters;

[0017] The structural parameters are quantized to generate a data structure feature vector.

[0018] Calculate the frequency factor, time decay factor, and cross-source correlation factor of the observed data, and generate information value density based on the preset factor coefficient combination.

[0019] The data structure feature vectors and information value density values ​​are encapsulated into an event feature set.

[0020] In a preferred embodiment, the calculation of the frequency factor, time decay factor, and cross-source correlation factor of the observed data, and the generation of information value density based on a preset combination of factor coefficients, specifically includes:

[0021] The frequency parameter is generated by the frequency of occurrence of the data type of the statistical observation data within the observation period of the current exploration task event;

[0022] The time decay factor is calculated based on the difference between the observation data reception time and the current system time.

[0023] Within the same time window of multiple observation nodes, the numerical relationship between the observation data of each node and the observation data of other nodes is compared to calculate the cross-source correlation factor.

[0024] After normalizing the frequency factor, time decay factor, and cross-source correlation factor, the preset factor coefficients are called to perform weighted comprehensive calculations and output the information value density.

[0025] In a preferred embodiment, the performance prediction module collects performance parameters of each heterogeneous link and generates performance prediction curves for the future transmission window of the heterogeneous links by combining external environmental information, specifically including:

[0026] Performance parameters are collected through the acquisition interfaces of each heterogeneous link. These parameters include real-time transmission rate, average round-trip delay, delay variance, available capacity, and available transmission time slot length.

[0027] Collect external environmental information, including sea state data, meteorological data, and node attitude data;

[0028] The performance parameters are aligned with the external environment information on a time reference to form a link input dataset on a unified time axis.

[0029] The link input dataset is input into the performance prediction model, which is based on sliding time window analysis and historical performance parameter regression calculation to obtain the start and end times of the future transmission window of the link and the corresponding performance prediction curve.

[0030] In a preferred embodiment, the matching and scoring module calculates the matching degree between the event feature set and the performance prediction curve, and outputs the matching degree scoring matrix between the observation data corresponding to the event and each link, specifically including:

[0031] Based on the sampling interval distribution of the event feature set and the data sequence length, a set of required intervals is generated within the target transmission window, and the continuous transmission length is labeled according to the numerical change gradient.

[0032] The target transmission window length is equal to the time axis range of the performance prediction curve;

[0033] A set of available intervals is generated based on the performance prediction curve, and the available capacity, available transmission slot length and expected queuing delay are marked.

[0034] Match the demand interval with the available interval, calculate the covered, partially covered and uncovered periods, and build a coverage record;

[0035] For periods of incomplete coverage, the data is discretized into micro-grooves with a preset fixed granularity. The capacity margin is calculated and negative margin micro-grooves are counted to construct a capacity margin record.

[0036] For the matched intervals, compare the delay range with the latest allowed start time of transmission, count the number of times the timing constraints are not met, and construct a timing constraint record;

[0037] The aforementioned coverage records, capacity reserve records, and time sequence constraint records are aggregated into a matching score matrix according to preset dictionary rules.

[0038] In a preferred embodiment, the link allocation module, specifically generating a multi-link allocation table based on the matching degree scoring matrix and information value density, includes:

[0039] Based on the information value density, the observation data in the current exploration task event are sorted in descending order of link allocation priority.

[0040] For the observation data with the highest information value density, select the priority link with the highest fit based on the matching degree score matrix, calculate the corresponding fragmentation ratio, and generate an initial allocation table containing link number, fragmentation ratio and allocation order;

[0041] After the allocation is executed, the remaining available capacity, available transmission time slot length and expected queuing delay performance parameters of all links are updated in real time, and a new performance prediction curve is generated based on the updated performance parameters.

[0042] The matching score matrix is ​​recalculated using the new performance prediction curve and the remaining unassigned observation data. Link selection and sharding ratio calculation are repeated until all observation data in the current exploration task event is assigned.

[0043] In a preferred embodiment, the data transmission module slices the observation data according to a multi-link allocation table, assigns a globally consistent time index to each slice, and writes the time index into the data packet header, specifically including:

[0044] Receive the multi-link allocation table and observation data, divide the observation data into multiple fragments according to the fragmentation ratio, generate a globally consistent time index for each fragment and write it into the fragment packet header;

[0045] According to the sending order in the multi-link allocation table, the fragments are loaded into the corresponding link sending buffers in sequence;

[0046] During the loading process, the consistency between the link identifier in the fragment header and the record in the allocation table is verified. After the verification is completed, the fragment set is marked as ready to be sent and the transmission is submitted.

[0047] In a preferred embodiment, the data integration module performs sequential rearrangement and merging of slices arriving from multiple links based on a time index at the data aggregation end to generate complete observation data and store it.

[0048] The data aggregation end receives fragmented data streams from each link, parses the global time index and link identifier in the header of each fragment packet, sorts all fragments by time index, and generates a sorted list.

[0049] The list is rearranged in order, and the fragments are extracted from the cache in sequence. Fragments with consecutive time indices are concatenated and merged to form a complete observation data sequence, which is then written to the specified data storage unit.

[0050] The technical effects and advantages of the data acquisition and processing system for marine exploration missions proposed in this invention are as follows:

[0051] This invention provides a data acquisition and processing system for marine exploration missions. By generating event feature sets containing data structure characteristics and information value density at observation nodes, it enables structured and quantitative expression of data when multi-source heterogeneous observation data is accessed. Furthermore, it prioritizes the transmission of high-value information based on information value priority. By generating performance prediction curves for future transmission windows using acquired link performance parameters and external environmental information, it can anticipate available link capacity, latency characteristics, and continuous time slot conditions, enabling pre-assessment of link status changes under complex sea conditions. Finally, by calculating the matching degree between the event feature set and the performance prediction curve, it establishes a demand range and... It can provide multi-dimensional mapping relationships for intervals, accurately identify insufficient coverage, capacity margin, and timing conflicts, and achieve precise matching of transmission resources and task requirements. On this basis, it dynamically generates a multi-link allocation table based on the matching degree scoring matrix and information value density, and updates the remaining performance of the links in real time after each round of allocation, realizing iterative optimization of allocation and significantly improving link utilization and data delivery timeliness. At the data aggregation end, it uses a global time index to rearrange and merge the slices arriving from multiple links, ensuring the integrity and timing correctness of cross-link data reconstruction, and improving the efficiency and reliability of multi-link data acquisition, transmission, and aggregation processing in marine exploration missions. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the structure of a data acquisition and processing system for marine exploration missions according to the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0054] Example 1, Figure 1 This invention provides a data acquisition and processing system for marine exploration missions, comprising:

[0055] The data acquisition module is used to collect observation data at the observation nodes and generate an event feature set containing data structure characteristics and information value density.

[0056] The performance prediction module is used to collect the performance parameters of each heterogeneous link and generate performance prediction curves for the future transmission window of the heterogeneous link by combining external environmental information.

[0057] The matching and scoring module is used to calculate the matching degree between the event feature set and the performance prediction curve, and output the matching degree scoring matrix between the observation data corresponding to the event and each link.

[0058] The link allocation module is used to generate a multi-link allocation table based on the matching degree score matrix and information value density.

[0059] The data transmission module is used to slice the observation data according to the multi-link allocation table, assign a globally consistent time index to each slice, and write the time index into the data packet header;

[0060] The data integration module is used at the data aggregation end to rearrange and merge slices arriving from multiple links based on time indexes, generate complete observation data, and store it.

[0061] In the data acquisition module, observation data is collected at the observation nodes to generate an event feature set containing data structure characteristics and information value density.

[0062] The multi-source observation nodes are data acquisition terminals deployed on offshore platforms, survey vessels, underwater detectors, or temporarily deployed buoys. During mission execution, the observation nodes continuously collect data at their respective geographical locations and depths according to the mission plan. The data collection types from different nodes may cover various observation information such as temperature, salinity, current velocity, wave height, seabed topography scanning, and chemical composition detection. After completing a certain batch of data collection, each observation node packages the data into a set of observation data packets and appends complete protocol header information to the packet header. The protocol header contains a unique identifier of the source node, the start and end times of the observation data collection, data type identifier, number of records, sampling frequency, data accuracy description, and necessary checksums. After receiving the data packets transmitted by the multi-source observation nodes, the receiving end first performs integrity checks on the protocol header of each data packet, such as checksum comparison, record count confirmation, and start and end time validity assessment. After confirming the correct structure of the protocol header and data body, the data body is split line by line according to the record count and record format indicated in the protocol header, so that each observation data becomes an independent record. After the data splitting is complete, synchronization is performed based on timestamps to ensure that data from different observation nodes are aligned on the same time base. The goal of synchronization is to arrange independent records from different nodes according to a unified time scale, so that observations at the same moment can be compared and combined in subsequent calculations. First, the acquisition timestamp in each record is read and converted to an absolute time format in the same time zone (e.g., year, month, day, hour, minute, second, millisecond based on UTC time), and the accuracy of the timestamps at each node is checked. If there are differences in sampling frequency or time accuracy between different nodes, a unified time step needs to be selected as the alignment standard, using 1 second, 10 seconds, or 100 milliseconds as the synchronization unit.

[0063] After time synchronization of multi-source observation data, structured analysis is performed to extract structural parameters that characterize the data distribution and variation properties. Specifically, this involves: extracting the sampling interval distribution to analyze the time intervals between observation data records at each node. By calculating the time difference between adjacent records, the set of sampling intervals for the data sequence is obtained. The sampling interval distribution records not only the size of the intervals but also their frequency distribution, which directly reflects the temporal resolution and continuity of the data. Next, the numerical change gradient is extracted to analyze the rate of change of observed values ​​over time. By calculating the ratio of the numerical difference between two adjacent records to the time interval, the change gradient sequence can be obtained. For continuous data such as temperature and chemical concentration, a larger absolute value of the gradient indicates a more drastic change; for periodic data such as wave height and flow velocity, the positive and negative distributions of the gradients need to be statistically analyzed to determine the directionality of the change. Finally, the data sequence length is extracted, i.e., the total number of consecutive records of the dataset on the synchronization time axis is counted. A longer data sequence length means that the node has more sufficient observation coverage time in the task, resulting in higher data continuity and statistical reliability. After extracting the three structural parameters (sampling interval distribution, numerical gradient, and data sequence length), feature quantization is performed to generate numerical feature vectors that are easy to calculate and match. During quantization, the sampling interval distribution is converted into a set of values ​​containing statistical features such as mean, variance, and proportion; the numerical gradient is converted into a set of values ​​containing statistical features such as average gradient, maximum gradient, and gradient variance; and the data sequence length is directly used as a scalar feature. To facilitate the calculation of features of different dimensions on the same scale, these feature values ​​are normalized so that their range is uniformly between 0 and 1.

[0064] The frequency factor calculation begins at the start of the current task event and ends at the end of the data acquisition phase of that event. All observation records collected within this time period are included in the frequency statistics. During the statistical process, the processing end first classifies the records according to the data type identifier field in the data packet header, such as sea surface temperature, salinity, wave height, and current velocity. Each data type forms an independent counting channel, counting all records belonging to that type one by one. To ensure statistical accuracy, only records that have completed time synchronization are allowed to enter the statistical stage, preventing misclassification of the same type of data from different nodes at the same time as different events. Once the number of records for a certain data type is accumulated, this number is divided by the total duration of the observation period to obtain the frequency parameter, which is the number of occurrences of that data type per unit time. This parameter physically reflects the sampling density and information generation rate of that type of observation information. For example, within a certain observation period, the frequency parameter for sea surface wave height data is 2 records per second, while the frequency parameter for sea surface salinity data is 1 record per 5 seconds. Therefore, the real-time performance and data continuity of wave height data are significantly higher than those of salinity data.

[0065] The time-attenuation factor calculation process uses the current system time as the base time and calculates the difference between the timestamp of each observation record at the receiving end and its timestamp at the generating end to obtain the transmission delay of that data. The longer the delay, the greater the difference between the data reflecting the on-site state and the current actual state, and the lower its decision-making reference value. In implementation, the generating timestamp is first read from the data packet header. This timestamp is recorded when the observation node collects data and is generated based on the global synchronization time standard. Then, the receiving timestamp recorded at the receiving end is read, and the difference between the two is used to obtain the delay value. For easy and consistent comparison, the delay value is mapped to a normalized time-attenuation factor. The mapping function adopts a monotonically decreasing form, that is, the factor value is 1 when the delay is zero, and the factor decreases proportionally as the delay increases until it approaches zero. For example, with a maximum effective delay of 10 seconds, the time-attenuation factor for data with a delay of 5 seconds is 0.5, and the time-attenuation factor for data with a delay of more than 10 seconds is close to 0. This factor reflects the on-site representativeness of the data. The closer its value is to 1, the better the timeliness of the data; the closer it is to 0, the more serious the data lag.

[0066] The calculation of cross-source correlation factors aims to measure the consistency of observations of similar data from different observation nodes within the same time window. This indicator directly reflects the reliability and stability of the spatial distribution of observational data. The task cycle is divided into a series of fixed-length time windows, such as 1 second, 2 seconds, or 5 seconds, with the window length determined based on the task's real-time requirements and data sampling frequency. Then, within each time window, observations from different nodes with consistent data types are aggregated. During processing, observations from the target node are selected one by one and compared numerically with observations from other nodes within the same window. Comparison methods can include difference comparison, proportional difference comparison, or correlation coefficient calculation. For example, when comparing sea surface temperature data, the absolute value of the difference between the target node's data and other node data can be calculated and averaged. This average is then compared with the maximum reasonable difference range set by the task; the smaller the difference, the higher the correlation. After performing the above comparisons over the entire time window, the results are summarized into a cross-source association factor between 0 and 1 according to a certain proportion. The closer the value is to 1, the higher the consistency between the data of the target node and the observation results of other nodes, and the stronger the data substitutability; the closer the value is to 0, the stronger the data uniqueness.

[0067] After calculating the three factors, they need to be standardized to the same dimension for comprehensive calculation. Normalization typically employs linear scaling, mapping each factor value to between 0 and 1 to ensure comparability across different factors. For example, all frequency factors are linearly transformed based on their maximum and minimum values, the time-deterioration factor is kept within the 0-1 range, and cross-source correlation factors are normalized according to statistical results. The normalized factor set is then multiplied term by term by preset factor coefficients. The factor coefficients are determined based on the task objective; for instance, in tasks with high real-time requirements, the weight of the time-deterioration factor is significantly higher than other factors; in tasks with high multi-node consistency requirements, the weight of the cross-source correlation factor is correspondingly increased. The sum of the normalized factors multiplied by their corresponding weights yields the information value density. This value reflects the comprehensive value of the observed data in the task; a higher value indicates better overall performance in terms of real-time performance, stability, and consistency. Finally, this information value density is encapsulated with the previously calculated data structure feature vector to form a complete event feature set. During encapsulation, the event feature set uses data structure feature vectors as the main description field and information value density as the evaluation field to ensure that the structural features and value levels of the data can be used simultaneously in subsequent link matching and allocation stages, so as to achieve scientific and reasonable resource allocation and data transmission optimization.

[0068] The performance prediction module collects the performance parameters of each heterogeneous link and combines them with external environmental information to generate a performance prediction curve for the future transmission window of the heterogeneous link.

[0069] In the communication phase of maritime exploration missions, different transmission links may originate from various heterogeneous technologies such as satellite communication, shortwave radio, microwave links, and cellular networks. These links differ significantly in hardware architecture, transmission media, and encoding methods, resulting in varying transmission performance patterns. Therefore, it is essential to acquire performance parameters reflecting the link's operational status in real time through a dedicated link acquisition interface to provide accurate data support for subsequent transmission optimization and mission scheduling. The acquisition interface is designed to communicate directly with the link's physical or data link layer to obtain key indicators such as real-time transmission rate, average round-trip delay, delay variance, available capacity, and available transmission time slot length. Among these, the real-time transmission rate reflects the link's effective throughput capacity at the current moment; the average round-trip delay represents the average time required for a data packet to travel from transmission to reception and back to acknowledgment, serving as a crucial indicator of link latency characteristics; delay variance reflects latency fluctuations, directly impacting data stream stability and jitter performance; available capacity refers to the link transmission resources that are currently unoccupied and readily usable; and available transmission time slot length represents the length of the time interval within which the link can be continuously used for transmission in the future. During the acquisition process, each performance parameter is obtained through the protocol status information of the underlying link or dedicated measurement instructions. For example, the transmission rate is calculated by the ratio of the cumulative number of transmitted bytes to the time interval, the round-trip delay can be obtained by sending test packets and measuring their acknowledgment time, and the delay variance is obtained by continuously sampling the delay and calculating its variance value.

[0070] In a marine environment, link performance depends not only on the technical specifications of the communication equipment itself but is also significantly affected by the external environment. Therefore, in addition to directly collecting link performance parameters, it is essential to simultaneously collect external environmental information that can influence the link status, including sea state data, meteorological data, and node attitude data. Sea state data encompasses information such as wave height, wave period, and ocean current speed and direction. These factors indirectly affect link performance by altering the spatial stability of the observation platform and influencing antenna pointing accuracy. Meteorological data includes indicators such as wind speed, wind direction, precipitation, temperature, humidity, and visibility. These conditions directly affect the propagation characteristics and attenuation of microwave and high-frequency links. Node attitude data, measured by sensors such as triaxial accelerometers and gyroscopes, reflects the platform's tilt, roll, and pitch states. These attitude changes can cause antenna pointing deviations, thus affecting link connection quality. After collecting the aforementioned external environmental information, it is necessary to align it with the link performance parameters using a unified time reference. Since the performance parameters and external environmental information may originate from different data acquisition devices, their timestamps may differ. Therefore, a unified time source for the mission must be used as a reference to map all data onto a unified time axis. If a large offset exists during the alignment process, it will be corrected using a time synchronization protocol (such as NTP or GPS timing signal).

[0071] After obtaining the aligned link input dataset, it is fed into a performance prediction model to predict the link's transmission capacity over a future period. This performance prediction model is constructed based on a combination of sliding time window analysis and historical performance parameter regression calculations, ensuring that it reflects both short-term fluctuations and long-term patterns. The core of the sliding time window analysis is to divide the link input dataset into several consecutive and overlapping time windows of fixed duration, such as a window of 30 seconds with a sliding step of 5 seconds. Within each window, statistical features are extracted for real-time transmission rate, mean round-trip delay, delay variance, available capacity, available transmission slot length, and external environmental information, including mean, variance, maximum, minimum, and gradient of change. These features reflect the stability and trend changes of the link state within that window. After the window analysis is completed, the model predicts future performance through regression calculations using historical performance parameters combined with historical environmental parameters. The regression calculation can be in the form of multiple linear regression or nonlinear regression, depending on the degree of nonlinearity between link characteristics and environmental parameters. Input variables include the statistical characteristics of the current window and multiple historical windows, as well as the external environmental characteristics corresponding to their time periods; the output variable is the predicted performance index value for the future transmission window. To improve prediction accuracy, the model incorporates historical data samples for parameter fitting during the training phase, ensuring high fit across different environmental conditions and link states. During model runtime, the window continuously scrolls during prediction; that is, every sliding step, the oldest window data is discarded and the latest window data is introduced to maintain the timeliness of the input data.

[0072] In the matching and scoring module, the matching degree between the event feature set and the performance prediction curve is calculated, and the matching degree scoring matrix between the observation data corresponding to the event and each link is output.

[0073] The event feature set records structured features such as the sampling interval distribution, data sequence length, and numerical change gradient of each set of observation data. These structured features are mapped to the target transmission window to form a set of demand intervals. The target transmission window length is strictly equal to the time axis coverage of the performance prediction curve, ensuring that demand and available transmission capacity are within the same time domain, avoiding scheduling deviations caused by time domain mismatch. In generating the demand interval set, the earliest possible transmission time and the latest required completion time of the data are first determined based on the sampling interval distribution. For example, if the data sampling intervals are uniformly distributed and the sequence length is relatively long (reaching 30% of the target window), the event needs to be continuously transmitted throughout the entire transmission window; conversely, if the sampling intervals are concentrated in a certain time period, the demand interval is mainly concentrated in that time period. Next, the data sequence length is mapped to the transmission demand, which can be in bytes or samples. This demand, combined with the transmission rate prediction results, can further derive the minimum required continuous transmission length. To reflect the importance and urgency of different demand intervals, each interval is labeled based on the numerical change gradient. Numerical gradient is an indicator that measures the amplitude and rate of change of data over time. Intervals with higher gradients typically contain data about sudden changes or critical events, and should be given higher transmission priority. During the annotation process, gradient values ​​are mapped to demand intensity levels, for example, divided into high, medium, and low levels, and the corresponding continuous transmission length is recorded. Ultimately, the demand interval set is structured data composed of multiple time periods, each containing start and end times, demand capacity, continuous transmission length, and demand intensity level. Below is a specific example:

[0074] Suppose a feature set of an event contains a set of observation data, with a sampling interval of once every 0.5 seconds, a total data sequence length of 1200 sampling points, and a total data volume of approximately 9.6MB (assuming each sampling point is 8KB). Calculation of the numerical gradient reveals a significant increase in gradient between sampling points 400 and 800, representing a critical observation segment with a gradient level of "high". In this embodiment, if the target transmission window is determined by the performance prediction curve time axis to be 0–60 seconds, mapping the total data to this window based on the sampling interval distribution shows that the data needs to be transmitted within the entire 60 seconds. The high-gradient segment requires continuous transmission for at least 20 seconds to ensure temporal integrity. The final set of demand intervals contains two types of records: the first type is low-gradient demand from 0 to 20 seconds (capacity 3.2MB, continuous length 10 seconds), the second type is high-gradient demand from 20 to 40 seconds (capacity 4.8MB, continuous length 20 seconds), and the third type is low-gradient demand from 40 to 60 seconds (capacity 1.6MB, continuous length 5 seconds), with the gradient level and continuous transmission length marked respectively.

[0075] In the performance prediction phase, performance prediction curves for each link within the future transmission window are obtained. These curves contain information such as transmission rate, mean round-trip delay, delay variance, and available transmission slot length over time. This time-series performance data is converted into a set of available intervals to clarify the actual transmission capacity that the link can provide in different time periods. Specifically, the performance prediction curves are divided according to continuous available transmission states. The division rule is based on the available transmission slot length; when the predicted available transmission slot length is greater than a set minimum continuous transmission time threshold, that period is considered a available interval. Within each available interval, the available capacity is calculated based on the integral result of the transmission rate prediction curve, i.e., the total amount of data transmission the link can complete within that interval, expressed in bytes or sample numbers. Simultaneously, the available transmission slot length is directly recorded as an interval attribute to facilitate determining whether it meets the continuous transmission length requirement of the demand interval during matching. The predicted values ​​of the mean round-trip delay and delay variance are used to estimate the expected queuing delay for that available interval. During calculation, the mean delay is used as the base delay, and the delay variance is mapped to an uncertainty range to obtain the interval value of the expected queuing delay. During the labeling process, ensure that each available interval has the following complete attributes: start and end times, available capacity, available transmission slot length, and expected queuing delay range. Below is a specific example:

[0076] Assuming the performance prediction curve shows that within a transmission window of 0–60 seconds, Link A has a transmission rate of approximately 2 Mbps from 0–15 seconds, a mean round-trip delay of 50 ms, a delay variance of 5 ms², and an available transmission time slot length of 15 seconds; from 20–40 seconds, the transmission rate is approximately 3 Mbps, the mean round-trip delay is 80 ms, the delay variance is 12 ms², and the available transmission time slot length is 15 seconds; from 45–55 seconds, the transmission rate is approximately 1 Mbps, the mean round-trip delay is 40 ms, the delay variance is 4 ms², and the available transmission time slot length is 10 seconds. Then, these time periods are divided into three available intervals based on the continuity of available time slots:

[0077] Interval 1 (0-15 seconds): Available capacity approximately 3.75MB, available transmission time slot length 15 seconds, estimated queuing delay range 45-55ms;

[0078] Interval 2 (20-40 seconds): Available capacity approximately 7.5MB, available transmission time slot length 15 seconds, estimated queuing delay range 74-86ms;

[0079] Interval 3 (45-55 seconds): Available capacity is approximately 1.25MB, available transmission time slot length is 10 seconds, and the expected queuing delay range is 38-42ms.

[0080] After constructing the demand interval set and the available interval set, timeline matching is required to determine the coverage of each demand interval within the available link resources. The matching process focuses on time overlap, comparing the start and end times of the demand interval with those of the available intervals, categorizing it into three scenarios: complete coverage, partial coverage, and no coverage. Complete coverage means the available interval completely encompasses the demand interval in time, and the available capacity is not less than the demand capacity, while the available transmission time slot length is not shorter than the continuous transmission length of the demand. In this case, the demand data can be transmitted completely within the interval without additional time fragmentation. Partial coverage means the available interval overlaps with the demand interval in time, but the available capacity is insufficient to meet all demands, or the available transmission time slot length is shorter than the continuous transmission length of the demand. In this case, part of the demand data can be transmitted, and the remainder needs to be supplemented in subsequent available intervals. No coverage means the demand interval has no time overlap with any available interval, or although there is time overlap, the available capacity is zero, meaning the demand for that period cannot be met under the current link conditions. After matching, the system records the coverage status of each demand interval, including the covered time range, coverage capacity, and coverage ratio. For partial coverage, the remaining uncovered demand capacity and the time range requiring additional scheduling are also marked; for uncovered demands, their total demand capacity and time information are recorded as input data for subsequent scheduling compensation. Matching results for all demand intervals are archived in chronological order and by link number to form a coverage record table. Below is a specific calculation example:

[0081] The previously obtained set of demand intervals is matched with the set of available intervals to identify the coverage of each demand period within the available intervals and form a coverage record. First, the three records in the demand interval set are aligned with the available interval set along the time axis: demand interval 1 is 0-20 seconds (3.2MB capacity, 10-second continuous length, low gradient level), demand interval 2 is 20-40 seconds (4.8MB capacity, 20-second continuous length, high gradient level), and demand interval 3 is 40-60 seconds (1.6MB capacity, 5-second continuous length, low gradient level); available interval 1 is 0-15 seconds (3.75MB capacity, 15-second continuous time slot), available interval 2 is 20-40 seconds (7.5MB capacity, 15-second continuous time slot), and available interval 3 is 45-55 seconds (1.25MB capacity, 10-second continuous time slot).

[0082] During matching, the demand interval 1 (0-20 seconds) is processed first: The available interval 1 (0-15 seconds) can completely cover the transmission demand for the first 15 seconds, and its capacity of 3.75MB is greater than the corresponding portion of the demand (approximately 2.4MB, calculated proportionally to time). Therefore, this segment is marked as fully covered. However, the last 5 seconds (15-20 seconds) of demand interval 1 have no corresponding time slots in the available interval, so this part is marked as uncovered and needs to be supplemented by subsequent links or time periods. Next, demand interval 2 (20-40 seconds, high gradient) is processed: The available interval 2 (20-40 seconds) is perfectly aligned in time, and its capacity of 7.5MB is greater than the demand capacity of 4.8MB. Simultaneously, a continuous time slot of 15 seconds can satisfy the first 15 seconds of continuous transmission within this 20-second interval. Although the remaining 5 seconds have time coverage, due to insufficient available transmission time slots, fragmented transmission is required. Therefore, the first 15 seconds are marked as fully covered, and the last 5 seconds are marked as partially covered (time margin is available but continuity is insufficient). Finally, the demand interval 3 (40-60 seconds, low gradient) is processed: the first 5 seconds (40-45 seconds) of this interval have no corresponding resources in the available interval and are marked as uncovered; the middle 10 seconds (45-55 seconds) can be covered by the available interval 3, with a capacity of 1.25MB, which is slightly less than the demand capacity of 1.6MB (gap of 0.35MB), so it is marked as partially covered; the last 5 seconds (55-60 seconds) also have no available interval and are marked as uncovered.

[0083] After the coverage record is generated, for the time periods marked as "partially covered" or "uncovered," the difference between available capacity and demanded capacity is further quantified to accurately schedule remaining bandwidth in subsequent link allocation optimization. In this embodiment, the incompletely covered time periods are discretized at a fixed granularity, for example, with a granularity of 1 second. The time period is divided into several micro-slots of equal length, each with an independent time index and capacity demand value. Taking the 15-20 second uncovered segment in demand interval 1 as an example, after discretization at a 1-second granularity, 5 micro-slots are obtained. The demand capacity of each micro-slot can be proportionally divided according to the total demand capacity. For example, the total demand capacity corresponding to these 5 seconds in 3.2MB is 0.8MB, so the demand of each micro-slot is 0.16MB. Then, based on the actual available capacity of this time period in the available interval (recorded as 0 if there is no available capacity), the capacity margin of each micro-slot is calculated using the formula: Capacity Margin = Available Capacity − Demand Capacity. If the result is negative, it indicates that there is a capacity gap in the micro-slot, which is recorded as a negative margin micro-slot. In this embodiment, there is no available capacity in the 15-20 second segment, therefore all five microslots have negative margins and a value of -0.16MB. Similarly, the same processing is performed on the partial coverage segment of demand interval 2 (35-40 seconds) and the partial coverage segment of demand interval 3 (45-55 seconds), and the number of microslots with negative margins and the total capacity gap are counted for each. Finally, the capacity margin data of all microslots are archived in chronological order and according to the corresponding demand interval number to form a capacity margin record. This record can accurately reflect the distribution and severity of bandwidth gaps during periods of incomplete link coverage.

[0084] After completing the capacity margin analysis, for the matched intervals in the coverage record, the transmission latency is evaluated to determine whether it meets the task's requirements for data arrival time. The timing constraint judgment is based on two key factors: first, the expected queuing latency range corresponding to the available interval; and second, the latest allowed start time for transmission in the interval (determined by the task timing plan, e.g., how many seconds after the window start point must transmission begin). In this embodiment, taking the first 15 seconds of the full coverage segment of demand interval 2 as an example, its available interval expected queuing latency range is 74–86 ms, and the latest allowed start time is the interval start point + 2 seconds. If the actual start time of transmission for this segment is the start point + 1.5 seconds, and the maximum queuing latency does not exceed the allowed latency threshold (e.g., 100 ms), then the segment is determined to meet the timing constraint; conversely, for example, in the partial coverage segment of demand interval 3 from 45 to 55 seconds, although there are available time slots, if transmission only begins at the 50th second, exceeding the latest allowed start time for this segment (assuming it is the start point + 3 seconds, i.e., 48 seconds), then it is recorded as a timing constraint violation. Similarly, if the expected queuing delay limit for a segment exceeds the task tolerance threshold, it is also considered a violation. The start time and delay range of each matched segment are compared one by one throughout the matching record. The number of times the timing constraints are not met is counted, and this, combined with information such as the segment number, reason for violation, and violation delay value, forms a timing constraint record. This record provides a timeliness constraint basis for subsequent link allocation strategies, ensuring that high-priority data is delivered on time.

[0085] After obtaining coverage records, capacity reserve records, and timing constraint records, the three types of data are aggregated according to preset dictionary rules to generate a matching score matrix. The preset dictionary rules are a set of fixed mapping relationships used to convert various record indicators into a unified scoring system. For example, in coverage status, "complete coverage" is assigned full marks, "partial coverage" is scored according to the coverage ratio, and "no coverage" is scored as 0 marks. In capacity reserve records, points are deducted based on the proportion of negative reserve micro-slots; the deduction ratio can be set according to the severity of the bandwidth gap, for example, if the negative reserve ratio exceeds 50%, the total score is halved. In timing constraint records, a fixed score is deducted for each timing violation or points are deducted according to the timeout magnitude. In this embodiment, the weight allocation of the dictionary rules is 50% for coverage records, 30% for capacity reserve records, and 20% for timing constraints. The final matching score matrix is ​​obtained by weighting these three factors. The rows of this matrix represent different demand intervals, the columns represent different availability intervals, and the matrix elements are the comprehensive matching scores of the demand interval within the available availability interval.

[0086] In the link allocation module, a multi-link allocation table is generated based on the matching degree score matrix and the information value density.

[0087] In the current exploration task, the obtained information value density data is first sorted according to the descending priority of link allocation, meaning that observation data with higher information value density values ​​are allocated earlier. Then, the generated matching score matrix is ​​used to compare the matching degree of the first sorted observation data with all available links. The link with the highest matching degree is selected as the priority allocation link for that data, and the segmentation ratio of this link is calculated based on the ratio of the required interval capacity to the available interval capacity. If the remaining capacity of the data still needs to be allocated after the priority allocation link is allocated, the segmentation ratio calculation continues among the remaining links in descending order of matching degree until the capacity allocation is complete. All allocation results are summarized to generate an initial allocation table, which records the link number, the segmentation ratio it undertakes, and the allocation order for each link.

[0088] After completing an allocation and distributing the data slicing task to each link, the link performance parameters need to be updated immediately to reflect the real-time impact of the allocation execution on the link status. In this embodiment, the performance parameters include remaining available capacity, available transmission slot length, and estimated queuing delay. Remaining available capacity is obtained by subtracting the allocated capacity from the original available capacity. For example, if link 1 originally had 2.5MB of available capacity, the remaining capacity after allocation is 0MB; if link 2 originally had 1.8MB of capacity, the remaining capacity after allocation is 0.3MB. The available transmission slot length is calculated based on the difference between the link's transmission window time and the occupied slots. For example, if link 2 has occupied 15 seconds out of a total of 20 slots, then the remaining available slots are 5 seconds. The estimated queuing delay is calculated based on the allocated queue length and the current link transmission rate. For example, if link 2 has 0.3MB of data queued in the remaining 5 seconds of available slots, and the transmission rate is 0.5MB / s, then the estimated queuing delay is 0.6 seconds. After updating the performance parameters, these real-time data are input into the performance prediction model. The model generates a new performance prediction curve based on the sliding time window and historical regression data. The prediction includes the trend of available capacity changes, the continuity of available time slots, and the trend of queuing delay changes in the link in future time periods.

[0089] After the new performance prediction curve is generated, its matching degree is recalculated with the remaining unallocated observation data to generate a new matching degree score matrix. In this embodiment, the matching degree calculation is still based on a comprehensive score of coverage, capacity margin, and temporal constraints, and the score result directly affects the link selection order. Based on the new matching result, the fragmentation ratio is calculated again, and the result is appended to the allocation table. After each allocation, the process of updating performance parameters and generating prediction curves is repeated to dynamically couple link status with allocation strategy. This loop continues until all observation data in the current exploration task event has been allocated to links and is in a ready-to-send state. In this embodiment, through this real-time closed-loop allocation strategy, optimal utilization of data transmission resources in a multi-link heterogeneous environment is achieved, while ensuring that high information value density data is given priority to obtain the optimal link and reasonable fragmentation ratio, thereby maximizing the transmission efficiency and reliability of observation data in marine exploration tasks.

[0090] In the data transmission module, the observation data is sliced ​​according to the multi-link allocation table, and a globally consistent time index is assigned to each slice. The time index is then written into the data packet header.

[0091] After obtaining the multi-link allocation table and corresponding observation data, the link number, fragmentation ratio, and allocation order of each record in the allocation table are read item by item to form a set of control parameters for fragmentation. Simultaneously with fragmentation, to ensure that subsequent concurrent transmissions across multiple links can be merged in the correct order at the data aggregation end, a globally consistent time index is generated for each fragment. In this embodiment, the time index uses a 64-bit integer structure, with the first 32 bits representing the second-level timestamp of the event trigger time, and the last 32 bits representing the fragment's sequence number within that event, incrementing from 0. The generated time index is directly written into the packet header of each fragment. The packet header structure also includes a link identifier field, a fragment length field, and a data type identifier field. In this way, all fragments, after being distributed to different links, can still be unambiguously identified in their global order at the aggregation end, ensuring the continuity and integrity of data reassembly. The entire process is completed after generating the fragments and time index at once, avoiding the delay risk caused by recalculating the index during transmission.

[0092] After the fragments and time indexes are generated, the allocation tasks are processed one by one according to the sending order recorded in the multi-link allocation table, and the corresponding fragment data is loaded into the sending buffer of the corresponding link. In this embodiment, the loading strategy of the link sending buffer is a sequential write mode, that is, fragments are added to the link buffer queue in the order of the allocation table to ensure that the link end sends data in the predetermined order. During the loading process, the link identifier field in the fragment header must be strictly verified to confirm that it is completely consistent with the link number recorded in the allocation table to prevent cross-link data mistransmission due to identification errors. The verification steps include reading the link identifier field in the packet header and comparing it with the number of the current target link; if they match, the fragment is written to the buffer and the loaded data volume counter of the link is updated; if they do not match, the fragment is immediately marked as abnormal and the loading process of the fragment is stopped to prevent it from entering the erroneous link queue. After all fragments are loaded, an integrity verification is performed on the buffer of each link to confirm that the fragment set of this batch is all in place and in the correct order. After confirming that everything is correct, the entire fragment set is marked as ready to be sent and submitted to the link transmission control unit to enter the formal transmission phase.

[0093] In the data integration module, the slices arriving from multiple links are rearranged and merged based on the time index at the data aggregation end to generate complete observation data and store it.

[0094] When data aggregation begins processing, it first receives fragmented data streams from multiple links simultaneously. Each fragment arrives according to its actual transmission delay, and there is a possibility of out-of-order arrival. To restore the order of the original observation data during subsequent merging, the aggregation end parses the fragment packet header upon receiving the fragment, reading the global time index and link identifier fields. The global time index is used to mark the unique sequential position of the fragment within the event, consisting of the event trigger timestamp and the fragment sequence number, accurately reflecting the fragment's positional relationship in the original data. The link identifier field indicates the link number from which the fragment originated, used for subsequent transmission path statistics and anomaly analysis. After parsing, the aggregation end sorts all fragments in ascending order according to the global time index, forming a rearranged list. Each row of this list records the time index of a fragment, the corresponding link identifier, and a pointer to the fragment's physical location in the buffer, ensuring that the sorting result can directly locate the actual data position in the buffer. For example, when the total number of event fragments is 10, the sorting result will accurately indicate the order and storage address of fragments numbered 0 to 9, thereby ensuring the correctness and efficiency of the subsequent splicing process.

[0095] After generating the rearranged list, the aggregation end retrieves the corresponding data fragments from the cache sequentially according to the list order. During retrieval, the time index of the current fragment is compared with the time index of the previous fragment. If they are consecutive in sequence number, the current fragment is directly appended to the end of the existing data sequence. If they are not consecutive, it indicates that a fragment is missing or has not arrived. In this case, the current fragment is temporarily stored, and the merging is performed only after the missing fragment arrives. The completed data sequence is immediately written to the designated data storage unit. In this embodiment, the storage unit is a high-reliability storage array, which can be managed in a catalog manner according to event number and retains link identifiers and splicing records for easy traceability and quality verification later. Through the above process, the aggregation end can not only accurately recover the original data in the case of out-of-order arrival of multiple links, but also provide orderly, complete, and traceable observation data results for subsequent data analysis and archiving.

[0096] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0097] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0098] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0099] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0100] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0101] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0102] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0103] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0104] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0105] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A data acquisition and processing system for marine exploration missions, characterized in that, include: The data acquisition module is used to collect observation data at the observation nodes and generate an event feature set containing data structure characteristics and information value density. The performance prediction module is used to collect the performance parameters of each heterogeneous link and generate performance prediction curves for the future transmission window of the heterogeneous link by combining external environmental information. The matching and scoring module is used to calculate the matching degree between the event feature set and the performance prediction curve, and output the matching degree scoring matrix between the observation data corresponding to the event and each link. The link allocation module is used to generate a multi-link allocation table based on the matching degree score matrix and information value density. The data transmission module is used to slice the observation data according to the multi-link allocation table, assign a globally consistent time index to each slice, and write the time index into the data packet header; The data integration module is used at the data aggregation end to rearrange and merge slices arriving from multiple links based on time index, generate complete observation data and store it. The matching and scoring module calculates the matching degree between the event feature set and the performance prediction curve, and outputs the matching degree scoring matrix between the observation data corresponding to the event and each link. Specifically, this includes: Based on the sampling interval distribution of the event feature set and the data sequence length, a set of required intervals is generated within the target transmission window, and the continuous transmission length is labeled according to the numerical change gradient. The target transmission window length is equal to the time axis range of the performance prediction curve; A set of available intervals is generated based on the performance prediction curve, and the available capacity, available transmission slot length and expected queuing delay are marked. Match the demand interval with the available interval, calculate the covered, partially covered and uncovered periods, and build a coverage record; For periods of incomplete coverage, the data is discretized into micro-grooves with a preset fixed granularity. The capacity margin is calculated and negative margin micro-grooves are counted to construct a capacity margin record. For the matched intervals, compare the delay range with the latest allowed start time of transmission, count the number of times the timing constraints are not met, and construct a timing constraint record; The aforementioned coverage records, capacity reserve records, and time sequence constraint records are aggregated into a matching score matrix according to preset dictionary rules.

2. The data acquisition and processing system for marine exploration missions according to claim 1, characterized in that, In the data acquisition module, the acquisition of observation data at the observation nodes and the generation of an event feature set containing data structure characteristics and information value density specifically include: In the current exploration mission, observation data is received through multi-source observation nodes; The observation data packets from different sources are parsed into independent records according to the protocol header, and the records are synchronized based on the timestamp; Perform structured analysis on the synchronized data to extract the sampling interval distribution, numerical change gradient, and data sequence length, and integrate them into structural parameters; The structural parameters are quantized to generate a data structure feature vector. Calculate the frequency factor, time decay factor, and cross-source correlation factor of the observed data, and generate information value density based on the preset factor coefficient combination. The data structure feature vectors and information value density values ​​are encapsulated into an event feature set.

3. A data acquisition and processing system for marine exploration missions according to claim 2, characterized in that, The calculation of the frequency factor, time decay factor, and cross-source correlation factor of the observed data, and the generation of information value density based on the preset factor coefficient combination calculation, specifically includes: The frequency factor is generated by the frequency of occurrence of the data type of statistical observation data within the observation period of the current exploration task event; The time decay factor is calculated based on the difference between the observation data reception time and the current system time. Within the same time window of multiple observation nodes, the numerical relationship between the observation data of each node and the similar observation data of other nodes is compared to calculate the cross-source correlation factor. After normalizing the frequency factor, time decay factor, and cross-source correlation factor, the preset factor coefficients are called to perform weighted comprehensive calculations and output the information value density.

4. A data acquisition and processing system for marine exploration missions according to claim 1, characterized in that, The performance prediction module collects performance parameters of each heterogeneous link and generates performance prediction curves for the future transmission window of the heterogeneous links by combining them with external environmental information. Specifically, this includes: Performance parameters are collected through the acquisition interfaces of each heterogeneous link. These parameters include real-time transmission rate, average round-trip delay, delay variance, available capacity, and available transmission time slot length. Collect external environmental information, including sea state data, meteorological data, and node attitude data; The performance parameters are aligned with the external environment information on a time reference to form a link input dataset on a unified time axis. The link input dataset is input into the performance prediction model, which is based on sliding time window analysis and historical performance parameter regression calculation to obtain the start and end times of the future transmission window of the link and the corresponding performance prediction curve.

5. A data acquisition and processing system for marine exploration missions according to claim 1, characterized in that, The link allocation module generates a multi-link allocation table based on the matching degree scoring matrix and information value density, specifically including: Based on the information value density, the observation data in the current exploration task event are sorted in descending order of link allocation priority. For the observation data with the highest information value density, select the priority link with the highest fit based on the matching degree score matrix, calculate the corresponding fragmentation ratio, and generate an initial allocation table containing link number, fragmentation ratio and allocation order; After the allocation is executed, the remaining available capacity, available transmission time slot length and expected queuing delay performance parameters of all links are updated in real time, and a new performance prediction curve is generated based on the updated performance parameters. The matching score matrix is ​​recalculated using the new performance prediction curve and the remaining unassigned observation data. Link selection and sharding ratio calculation are repeated until all observation data in the current exploration task event is assigned.

6. A data acquisition and processing system for marine exploration missions according to claim 1, characterized in that, In the data transmission module, the observation data is sliced ​​according to the multi-link allocation table, and a globally consistent time index is assigned to each slice. Writing the time index into the data packet header specifically includes: Receive the multi-link allocation table and observation data, divide the observation data into multiple fragments according to the fragmentation ratio, generate a globally consistent time index for each fragment and write it into the fragment packet header; According to the sending order in the multi-link allocation table, the fragments are loaded into the corresponding link sending buffers in sequence; During the loading process, the consistency between the link identifier in the fragment header and the record in the allocation table is verified. After the verification is completed, the fragment set is marked as ready to be sent and the transmission is submitted.

7. A data acquisition and processing system for marine exploration missions according to claim 1, characterized in that, In the data integration module, at the data aggregation end, the slices arriving from multiple links are rearranged and merged based on the time index to generate complete observation data and store it. The data aggregation end receives fragmented data streams from each link, parses the global time index and link identifier in the header of each fragment packet, sorts all fragments by time index, and generates a sorted list. The list is rearranged in order, and the fragments are extracted from the cache in sequence. Fragments with consecutive time indices are concatenated and merged to form a complete observation data sequence, which is then written to the specified data storage unit.

Images