Offshore signal thermodynamic diagram updating method based on big data analysis

Abstract

The invention discloses a sea signal thermodynamic diagram updating method based on big data analysis. The sea signal thermodynamic diagram updating method comprises the following steps of obtaining sea signal data and performing preprocessing to obtain first processing data; performing hot spot area detection on the first processing data by using a clustering algorithm to extract hot spot area information; inputting the hot spot region information into an abnormal signal identification model to output an abnormal identification result; generating a maritime signal thermodynamic diagram based on the abnormal recognition result and the hot spot region information, and realizing dynamic updating; and displaying the updated maritime signal thermodynamic diagram through a visual interface. Through fusion of clustering analysis and an anomaly detection technology, intelligent identification of hot spot area information and accurate positioning of abnormal signals are realized, real-time or periodic updating of a sea signal thermodynamic diagram is realized through combination of dynamic weight calculation and intelligent model prediction, signal distribution characteristics under complex sea conditions can be efficiently and accurately reflected, and the method has a good application prospect. And intelligent anomaly detection and early warning functions are supported.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and visualization, and particularly to a method for updating a maritime signal heat map based on big data analysis. Background Art

[0002] With the rapid development of the marine economy and the frequentization of shipping activities, the demand for real-time monitoring of marine information is increasing day by day. As a tool for intuitively displaying signal distribution and dynamic changes, the maritime signal heat map has important application values in the fields of marine transportation, fishery monitoring, emergency search and rescue, etc. Especially with the expansion of the 700MHz frequency band 5G network in the field of deep-sea communication applications (coverage in the area 80Km offshore), the dynamic analysis of maritime 5G signal coverage is particularly important.

[0003] However, the data sources in the existing technology are single, and it is impossible to comprehensively reflect the dynamic signal changes in the maritime area; the detection of hot spots is difficult, and it is difficult to accurately identify dynamic hot spots under complex sea conditions. In addition, the existing technology lacks the intelligent identification and early warning function for abnormal signals and lacks a maritime heat map update mechanism, and cannot meet the requirements of real-time and periodic updates. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to propose a method for updating a maritime signal heat map based on big data analysis, so as to efficiently reflect the signal distribution characteristics under complex sea conditions and support intelligent abnormal detection and early warning functions.

[0005] In order to achieve the above technical purpose, the technical solution adopted in the present application is: A method for updating a maritime signal heat map based on big data analysis, including:

[0006] Obtain maritime signal data, and preprocess the maritime signal data to obtain first processed data;

[0007] Use a clustering algorithm to detect hot spot areas for the first processed data to obtain hot spot area information;

[0008] Input the hot spot area information into an abnormal signal recognition model to obtain an abnormal recognition result;

[0009] Generate a maritime signal heat map according to the abnormal recognition result and the hot spot area information, and dynamically update the maritime signal heat map;

[0010] Display the maritime signal heat map through a visualization interface.

[0011] In some embodiments, obtaining maritime signal data includes:

[0012] Obtain multi-source data, which is configured as at least one of front-end device data, satellite communication system data, Automatic Identification System (AIS) data of ships, and Ship Traffic Service (STS) system data;

[0013] Perform data alignment on the multi-source data through interpolation. The data alignment includes time alignment and spatial alignment to obtain first-aligned data;

[0014] Convert the format of the first-aligned data to obtain second-aligned data;

[0015] Fuse the second-aligned data through the weighted average method to obtain maritime signal data, which is represented by formula (1). Formula (1) is as follows:

[0016]

[0017] In formula (1), S fu is the fused maritime signal data, ω i is the weight coefficient of the i-th second-aligned data, and S i is the i-th second-aligned data.

[0018] In some embodiments, preprocess the maritime signal data to obtain first-processed data, including:

[0019] Perform median filtering on the maritime signal data to obtain first denoised data;

[0020] Perform mean filtering on the first denoised data to obtain second denoised data;

[0021] Perform normalization on the second denoised data to obtain second-processed data;

[0022] Divide the second-processed data according to timestamps into time windows. The time window has a preset time length to obtain multiple time windows;

[0023] Extract features from each time window to obtain first feature information, which includes signal strength features and regional features;

[0024] Map and store the first feature information with the timestamps of the time windows to form a time feature column;

[0025] Concatenate the time feature column with the second-processed data to obtain first-processed data.

[0026] In some embodiments, use a clustering algorithm to detect hot spot areas in the first-processed data, and the obtained hot spot area information includes:

[0027] Initialize the parameters of the clustering algorithm, including the number of clusters K and the maximum number of iterations;

[0028] Perform clustering analysis on the first processed data until the maximum number of iterations is reached to obtain the initial hotspot distribution information;

[0029] Perform boundary optimization on the initial hotspot distribution information to obtain the final hotspot distribution information;

[0030] Use the final hotspot distribution information as the hotspot area information.

[0031] In some embodiments, performing boundary optimization on the initial hotspot distribution information to obtain the final hotspot distribution information includes:

[0032] The initial hotspot distribution information includes multiple data points. Calculate the local density value of each data point;

[0033] Calculate the minimum distance of each data point according to the local density value to obtain the density information of each data point;

[0034] Record the data point with the highest density value in the density information as the center of the first hotspot area. The center of the first hotspot area corresponds to a first hotspot area. Calculate the distance from the center of the first hotspot area to other data points and record it as the boundary distance;

[0035] Judge whether the boundary distance is within the threshold range of the preset boundary distance. If so, divide the data point into the first hotspot area;

[0036] And merge the first hotspot areas with overlapping boundaries to generate a second hotspot area;

[0037] Generate the final hotspot distribution information according to the second hotspot area.

[0038] In some embodiments, the abnormal signal is configured to be constructed by an LSTM model. Input the hotspot area information into the abnormal signal recognition model to obtain the abnormal recognition result, including:

[0039] Construct an LSTM model and use historical data as the sample data set to train the LSTM model. The loss function during the training of the LSTM model is the cross-entropy loss;

[0040] After the LSTM model is trained, input the hotspot area information into the abnormal signal recognition model to obtain the abnormal recognition result;

[0041] Trigger a warning for the abnormal recognition result.

[0042] In some embodiments, triggering a warning for the abnormal recognition result includes:

[0043] Set a warning threshold and trigger a warning when the abnormal signal strength exceeds the threshold;

[0044] Generate a warning message and display it through a visualization interface.

[0045] In some embodiments, generating a maritime signal heat map based on the anomaly recognition result and the hot spot area information, and dynamically updating the maritime signal heat map includes:

[0046] Mapping the hot spot area distribution data and the anomaly signal recognition result to the electronic chart grid to obtain a plurality of grid cells, and each grid cell corresponds to a geographical location coordinate;

[0047] Calculating the heat value of each grid cell according to the hot spot area information and the anomaly recognition result;

[0048] Defining that the heat values from the first level to the tenth level are distributed in a preset increasing trend;

[0049] Defining the first level to the tenth level according to the RGB colors to obtain the color levels corresponding to the first level to the tenth level respectively;

[0050] Classifying the heat values and converting the heat values into the saturation indication of the color levels to form a heat distribution layer;

[0051] Fusing the heat distribution layer with the map to obtain a maritime signal heat map;

[0052] Dynamically updating the maritime signal heat map according to a preset period.

[0053] In some embodiments, displaying the maritime signal heat map through a visualization interface includes:

[0054] Displaying the maritime signal heat map through Leaflet.js, supporting map zooming and panning;

[0055] Implementing the color gradient and interactive query of the heat map through Plotly.js.

[0056] In some embodiments, providing the maritime signal heat map to a third-party system through an API interface.

[0057] Adopting the above technical solution, compared with the prior art, the beneficial effects of the present invention are:

[0058] The present invention provides a method for updating a maritime signal heat map based on big data analysis, comprising the following steps: First, obtain maritime signal data and perform preprocessing to obtain first processed data; Then, use a clustering algorithm to perform hotspot area detection on the first processed data to extract hotspot area information; Next, input the hotspot area information into an abnormal signal recognition model to output an abnormal recognition result; Generate a maritime signal heat map based on the abnormal recognition result and the hotspot area information and achieve dynamic update; Finally, display the updated maritime signal heat map through a visualization interface. This method realizes the intelligent recognition of hotspot area information and the accurate positioning of abnormal signals by integrating clustering analysis and anomaly detection technologies, and at the same time establishes a dynamic update mechanism to ensure the real-time nature of the heat map, providing visualization analysis support for maritime signal monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention, and for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0060] Figure 1 It is a schematic diagram of steps S101 to S105 of the method for updating a maritime signal heat map described in the specific implementation manner;

[0061] Figure 2 It is a flowchart of the method for updating a maritime signal heat map described in the specific implementation manner. SPECIFIC IMPLEMENTATION MANNER

[0062] The following will further describe the present invention in detail with reference to the drawings and embodiments. It should be specifically noted that the following embodiments are only used to illustrate the present invention, but do not limit the scope of the present invention. Similarly, the following embodiments are only partial embodiments of the present invention rather than all embodiments, and all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0063] Please refer to Figures 1 to 2 , this embodiment provides a method for updating a maritime signal heat map based on big data analysis, comprising:

[0064] S101. Obtain maritime signal data and perform preprocessing on the maritime signal data to obtain first processed data;

[0065] S102. Use a clustering algorithm to perform hotspot area detection on the first processed data to obtain hotspot area information;

[0066] S103, inputting the hot spot area information into the abnormal signal recognition model to obtain an abnormal recognition result;

[0067] S104, generating a marine signal heat map according to the anomaly recognition result and the hot spot area information, and dynamically updating the marine signal heat map;

[0068] S105. Display the marine signal heat map through a visualization interface.

[0069] In step S101, preferably, the marine signal data is obtained from multiple source data such as Beidou + 5G communication and navigation fusion front-end equipment, satellite communication system, AIS (Automatic Identification System for Ships), VTS (Vessel Traffic Service System), etc. Preprocessing includes denoising, normalization and spatiotemporal feature extraction, and the first processed data is obtained through preprocessing to ensure data quality.

[0070] In step S102, a K-means clustering algorithm is used to perform cluster analysis on the first processed data to preliminarily determine the hotspot area, and a density peak detection algorithm is combined to optimize the hotspot area boundary to improve the detection accuracy.

[0071] In step S103, preferably, the abnormal signal recognition model is an LSTM model, and the LSTM deep learning model is used to train historical data, build an abnormal signal recognition model, and analyze real-time signals to identify potential abnormal signals and trigger an early warning mechanism.

[0072] In step S104, the detected hotspot area information and anomaly identification results are mapped to the electronic chart grid, and the dynamic weight is calculated according to the signal strength, and a heat map is generated based on the preset color scheme. The heat map supports real-time updates (such as automatic refresh every 5 minutes) or periodic batch updates (in hours or days).

[0073] In step S105, the heat map is displayed through the Web terminal or mobile terminal, supporting map zooming, hotspot annotation, abnormal query and warning setting. And the heat map data is provided to a third-party system through the API interface to achieve information sharing.

[0074] In this embodiment, by preprocessing the maritime signal data, the accuracy of subsequent analysis is effectively improved. The combined hotspot area detection method using the K-means clustering algorithm and the density peak detection algorithm not only retains the efficient characteristics of the clustering algorithm but also corrects the problem of fuzzy hotspot boundaries through the density optimization strategy, significantly improving the spatial accuracy of hotspot area recognition. Using the abnormal signal recognition module, it is possible to capture the temporal correlation based on the historical data characteristics, realize sensitive monitoring of dynamic signal changes, and enhance the timeliness and discrimination accuracy of abnormal signal early warning. According to the abnormal recognition results and hotspot area information, a maritime signal heat map is generated, visually presenting the signal intensity distribution and abnormal areas, and dynamically updating the maritime signal heat map, which not only meets the immediate response requirements of emergencies but also adapts to the stability requirements of regular monitoring. Finally, through visual display, multi-dimensional interaction of monitoring information is realized, providing expandable technical support for maritime situation awareness.

[0075] In some embodiments, obtaining the maritime signal data includes:

[0076] Obtaining multi-source data, where the multi-source data is configured as at least one of front-end device data, satellite communication system data, automatic identification system data of ships, and ship traffic service system data;

[0077] Performing data alignment on the multi-source data by the interpolation method, where the data alignment includes time alignment and spatial alignment, to obtain the first aligned data;

[0078] Converting the format of the first aligned data to obtain the second aligned data;

[0079] Fusing the second aligned data by the weighted average method to obtain the maritime signal data, which is represented by formula (1), and formula (1) is as follows:

[0080]

[0081] In formula (1), S fu is the fused maritime signal data, ω i is the weight coefficient of the i-th second aligned data, and S i is the i-th second aligned data.

[0082] In this embodiment, the front-end device data includes longitude, latitude, altitude, speed, heading, signal strength, PCI cell ID, signal-to-noise ratio SNR, frequency band number, network type, network latency, packet loss rate, bandwidth, etc.; the satellite communication system data is signal strength, coverage range, communication quality, etc.; the ship automatic identification system data includes the ship's MMSI (Maritime Mobile Service Identity), position, speed, heading, ship type, etc.; the ship traffic service system data includes the real-time position, track, traffic density of the ship, etc. Optionally, the front-end device collects data once every 5 seconds, and the satellite communication system collects data once every minute; the ship automatic identification system and the ship traffic service system collect data once every 10 seconds.

[0083] Optionally, use the numpy library of Python to perform spatial interpolation on the data to ensure uniform regional coverage.

[0084] In this embodiment, by integrating the front-end device data, satellite communication system data, ship automatic identification system data, and ship traffic service system data, a multi-dimensional maritime signal acquisition system is constructed, effectively covering various types of information such as ship dynamics, communication quality, and traffic situation. The interpolation method is used to perform time alignment and spatial alignment on multi-source data, solving the problem of time series misalignment caused by the different acquisition frequencies of the front-end device every 5 seconds, the satellite communication system every minute, and the ship automatic identification system and the ship traffic service system every 10 seconds. At the same time, spatial interpolation is used to make up for the missing data in local areas, ensuring the spatio-temporal continuity of the first aligned data. The heterogeneous data is unified into a standardized expression form through format conversion, eliminating the interference of data format differences on the fusion process, and generating the second aligned data. The weighted average method is used to fuse the second aligned data, and the weight coefficients are dynamically allocated through formulas, which not only compatible with communication parameters such as PCI cell ID, signal-to-noise ratio SNR, and frequency band number in the front-end device, but also comprehensively consider characteristics such as satellite communication coverage range, ship MMSI, and track density, enabling the final maritime signal data to balance the reliability and real-time performance of different data sources. Based on the numpy library of Python, efficient spatial interpolation operations are implemented, ensuring the efficiency of large-scale data processing and uniform regional coverage, providing a high-quality data basis for subsequent analysis.

[0085] In some embodiments, the maritime signal data is preprocessed to obtain the first processed data, including:

[0086] Perform median filtering on the maritime signal data to obtain the first denoised data;

[0087] Perform mean filtering on the first denoised data to obtain the second denoised data;

[0088] Perform normalization processing on the second denoised data to obtain the second processed data;

[0089] Divide the second processed data into time windows according to the timestamps. The time windows have a preset time length, and multiple time windows are obtained;

[0090] Extract features from each time window to obtain first feature information, which includes signal strength features and regional features;

[0091] Map and store the first feature information and the timestamps of the time windows to form a time feature column;

[0092] Concatenate the time feature column and the second processed data to obtain the first processed data.

[0093] In this embodiment, for the processing of outliers in time series data, median filtering is used to remove noise points and retain the main trend of the data, obtaining first denoised data. Median filtering is represented by the following formula:

[0094] y i = median(x′ i-k ,x′ i-l ,…,x′ i+k );

[0095] where y i is the filtered value of the i-th data point (i.e., the first denoised data), and x ∑ i-k and x ∑ i+k represent the data of k windows before and after the current data point respectively.

[0096] Use mean filtering to perform denoising processing on the first denoised data to smooth the data and reduce high-frequency noise, obtaining second denoised data. Mean filtering is represented by the following formula:

[0097]

[0098] where y i is the filtered value of the i-th data point (i.e., the second denoised data), and x″ i+k represents the data of k windows before and after the current data point.

[0099] Perform normalization processing to standardize the second denoised data of different scales to the same range, and the range is preferably [0, 1]. Normalization processing is represented by the following formula:

[0100]

[0101] where z i is the normalized data point (i.e., the second processed data), and x iis the original data point (i.e., the second denoised data), u is the mean of the data set, and σ is the standard deviation of the data set.

[0102] Preferably, the preset time length is 5 (i.e., 5 data points before and after).

[0103] The regional feature is the longitude and latitude coordinates.

[0104] The signal strength feature is represented by the following formula:

[0105] Mean:

[0106]

[0107] Maximum:

[0108] max(S) = max(S1, S2, …, S n );

[0109] Minimum:

[0110] min(S) = min(S1, S2, …, S n );

[0111] Variance:

[0112]

[0113] The regional feature is represented by the following formula:

[0114] Center point coordinates:

[0115] center(Lon, Lat) = (mean(Lon), mean(Lat));

[0116] Coverage range:

[0117] range(Lon, Lat) = (max(Lon) - min(Lon), max(Lat) - min(Lat));

[0118] Feature extraction is performed on each time window to obtain the first feature information, which can be understood as:

[0119] Create the following feature columns for each time window: signal strength mean column mean(S), signal strength maximum column max(S), signal strength minimum column min(S), signal strength variance column var(S), center point longitude column center(Lon), center point latitude column center(Lat), longitude coverage range column range(Lon), and latitude coverage range column range(Lat).

[0120] Fill the signal strength features and regional feature values extracted within each time window into the corresponding feature columns. For example, for time window W1, fill the value of mean(S) into the mean(S) column; fill the value of max(S) into the max(S) column; fill the value of center(Lon) into the center(Lon) column; fill the value of center(Lat) into the center(Lat) column.

[0121] Merge the second processed data with the feature columns according to time windows to generate the first processed data. The merged data format is:

[0122] Data preprocessed =[Time, Lon, Lat, S, mean(S), max(S), min(S), val(S), center(Lon), center(Lat), range(Lon), range(Lat)];

[0123] where Time is the timestamp, Lon is the longitude, Lat is the latitude, and S is the signal strength.

[0124] Optionally, use the pandas library in Python to clean and standardize the multi-source data, remove invalid data points, and fill in missing values.

[0125] In this embodiment, a cascaded denoising strategy of median filtering and mean filtering is adopted to effectively eliminate impulse noise and high-frequency random interference in the original marine signal data. The median filtering suppresses the interference of outliers on the data trend through the calculation method of taking the median value within a sliding window to obtain the first denoised data; the mean filtering further smooths the first denoised data to reduce the remaining fine fluctuations and generate the second denoised data. The second denoised data with different dimensions is mapped to the [0,1] interval by using a normalization formula based on the mean and standard deviation to eliminate the scale differences of parameters such as multi-source signal strength, longitude, and latitude, forming the second processed data. The second processed data is divided into time series through a preset time window, and combined with the extraction mechanism of signal strength features and regional features, to dynamically capture the signal strength distribution law and spatial coverage change characteristics within each time window, generating the first feature information. The feature columns after timestamp mapping are spliced with the second processed data according to the window to construct the structured data Data preprocessed , which not only retains the original time, longitude, latitude, and signal strength fields, but also integrates statistical features and spatial features to enhance the data expression dimension. Through the method provided in this embodiment, it is possible to achieve invalid data cleaning and missing value filling, ensure the robustness of the preprocessing process, and provide the first processed data with high integrity, low noise, and rich features for subsequent clustering analysis and anomaly recognition.

[0126] In some embodiments, a clustering algorithm is used to detect hot spot regions for the first processed data, and the obtained hot spot region information includes:

[0127] Initialize the parameters of the clustering algorithm, including the number of clusters K and the maximum number of iterations;

[0128] Perform clustering analysis on the first processed data until the maximum number of iterations is reached to obtain the initial hot spot distribution information;

[0129] Optimize the boundary of the initial hot spot distribution information to obtain the final hot spot distribution information;

[0130] Use the final hot spot distribution information as the hot spot region information.

[0131] In this embodiment, the K-means clustering algorithm is used to perform clustering analysis on the first processed data. The historical signal data is clustered into several clusters using the K-means algorithm to form a preliminary cluster division. Each cluster represents a potential hot spot region. The number of clusters K is determined by the elbow method. By calculating the SSE (sum of squared errors) under different K values, the K value at which the SSE decreases significantly more slowly is selected.

[0132] Recalculate the position of the cluster center based on the data points in the current cluster, and continuously repeat the distance calculation and cluster assignment process until the centroid position no longer changes or reaches the preset maximum number of iterations, and output the hot spot region distribution map after clustering. The calculation process is as follows:

[0133] The K-means algorithm is used to calculate the within-cluster distance to measure the similarity between two data points. The distance calculation (Euclidean distance) formula is as follows:

[0134]

[0135] where d(i,j) represents the distance between the i-th and j-th data points, γ i,l represents the l-th eigenvalue of the i-th data point, γ j,l represents the l-th eigenvalue of the j-th data point.

[0136] Cluster center update is adopted to calculate the new center of each cluster, which reflects the central position of the data points within the cluster. The cluster center update formula is as follows:

[0137]

[0138] where c m is the new clustering center of the m-th cluster, N m is the number of data points in the m-th cluster, γ i is the i-th data point.

[0139] In this embodiment, the K-means clustering algorithm is used to perform multi-dimensional analysis on the first processed data, and the elbow method is used to dynamically determine the optimal number of clusters K value. Based on the inflection point characteristics of the sum of squared errors (SSE) under different K values, the deviation of cluster division caused by subjective setting is avoided, and the accuracy of hotspot area detection is improved. The Euclidean distance formula is used to calculate the similarity between data points, and features such as the mean value of signal strength, variance, longitude and latitude of the center point, and coverage range are comprehensively considered to ensure that the initial cluster division can reflect the correlation between signal strength and spatial distribution. By iteratively executing the distance calculation and cluster center update formula, the centroid position of each cluster is continuously optimized until convergence or the maximum number of iterations is reached, generating the initial hotspot distribution information, effectively capturing the spatio-temporal aggregation characteristics of signal-dense areas. Combining the boundary optimization algorithm to refine the geographical coverage range of the initial clusters, solving the problem of insufficient spatial accuracy of hotspot areas caused by the fuzzy boundary of the clustering algorithm, and forming the final hotspot distribution information with a clear description of the longitude and latitude range. Based on the cluster division results generated by clustering historical signal data, quantifiable attributes such as the mean value of signal strength and coverage radius are assigned to each hotspot area, and the finally output hotspot area distribution map can intuitively display the high-frequency active areas of ship communication, providing data support for maritime signal resource scheduling.

[0140] In some embodiments, boundary optimization is performed on the initial hotspot distribution information to obtain the final hotspot distribution information, including:

[0141] The initial hotspot distribution information includes multiple data points, and the local density value of each data point is calculated;

[0142] According to the local density value, the minimum distance of each data point is calculated to obtain the density information of each data point;

[0143] The data point with the highest density value in the density information is recorded as the center of the first hotspot area. The first hotspot area center corresponds to a first hotspot area, and the distance from the first hotspot area center to other data points is calculated and recorded as the boundary distance;

[0144] It is judged whether the boundary distance is within the threshold range of the preset boundary distance. If so, the data point is divided into the first hotspot area;

[0145] In addition, the first hotspot areas with overlapping boundaries are merged to generate a second hotspot area;

[0146] The final hotspot distribution information is generated according to the second hotspot area.

[0147] In this embodiment, the Gaussian kernel function is used to calculate the local density of each data point to reflect the density degree of the data point. The calculation formula is as follows:

[0148]

[0149] Among them, q is the bandwidth parameter, usually set to 1 / 10 of the average distance between data points.

[0150] Calculate the minimum distance from each data point to a data point with higher density to identify density peak points. The minimum distance δ of each data point i is calculated by the following formula:

[0151] δ i = min j:ρj＞ρi d(i, j);

[0152] By calculating the minimum distance δ of each data point i , density peak points are determined and the boundaries of the hotspot areas are optimized. Specifically, select data points with both relatively high local density ρ i and relatively high minimum distance δ i as density peak points. That is, by calculating the product of ρ i and δ i (ρ i ×δ i ), select the top K data points with the largest product values as density peak points to ensure the accuracy of the hotspot areas.

[0153] For each density peak point, calculate the distance from it to the surrounding data points. According to the local density ρ i and the minimum distance δ i , determine the boundary threshold, and classify the data points with a distance less than the boundary threshold into the first hotspot area to optimize the boundaries of the hotspot areas.

[0154] Calculate the centroid of the overlapping area, merge the overlapping areas into a new hotspot area (i.e., the second hotspot area), and recalculate its boundary. By merging the hotspot areas with overlapping boundaries, ensure the continuity and integrity of the hotspot areas.

[0155] In this embodiment, the local density value of each data point is calculated by the Gaussian kernel function. Combining the reasonable setting of the bandwidth parameter, the density degree of the data distribution is accurately quantified, avoiding regional division errors caused by density estimation deviations. Based on the dual screening mechanism of the local density ρ i and the minimum distance δ i , select density peak points as the centers of the first hotspot areas by the product, effectively identify the core areas with both significant signal strength and spatial distribution, and improve the positioning accuracy of the hotspot areas. Using the minimum distance δ iDynamically determine the boundary threshold, and classify the data points whose boundary distances meet the preset threshold range into the corresponding first hot regions, solving the problem of fuzzy initial clustering boundaries and realizing the refined adjustment of the hot region boundaries. By calculating the centroid of the overlapping regions and merging them to generate the second hot region, the redundant segmentation caused by the boundary overlap of adjacent regions is eliminated, ensuring the continuity and integrity of the final hot spot distribution information in space. Based on the density peak points, iteratively optimize the boundary and region merging strategies, taking into account the independence of the hot regions and the overall coverage range. The finally output second hot region not only retains the statistical characteristics of the initial distribution but also has a clear geographical boundary description, providing a highly reliable spatial decision-making basis for maritime signal resource management.

[0156] In some embodiments, the abnormal signal is configured to be constructed by an LSTM model. The hot region information is input into the abnormal signal recognition model, and the obtained abnormal recognition results include:

[0157] Construct an LSTM model and use historical data as the sample data set to train the LSTM model. The loss function during the training process of the LSTM model is the cross-entropy loss;

[0158] After the LSTM model is trained, input the hot region information into the abnormal signal recognition model to obtain the abnormal recognition results;

[0159] Trigger a warning for the abnormal recognition results.

[0160] In this embodiment, historical data (such as data in the past 30 days) is used for training to ensure the generalization ability of the model. The data is divided into a training set (80%) and a validation set (20%).

[0161] The loss function during the training process is the cross-entropy loss, and the formula is:

[0162]

[0163] where y′ i is the true label (0 represents normal signal, 1 represents abnormal signal), is the model prediction value, and N is the number of samples.

[0164] Use the Adam optimizer for model training, and set the learning rate to 0.001 to improve the training efficiency of the model.

[0165] Adopt the early stopping method to prevent overfitting. Stop training when the validation set loss does not decrease for 5 consecutive times to ensure the robustness of the model.

[0166] Determine which information needs to be retained from the state of the previous time step through anomaly signal recognition. The closer the value is to 1, the more information is retained. The formula for anomaly signal recognition is as follows:

[0167] f t = σ1(W f β t + U f h t-1 + b f );

[0168] Among them, f t is the output of the forget gate at the t-th time step, β t is the input data at the t-th time step, h t-1 is the hidden state at the (t - 1)-th time step, W f is the weight matrix of the forget gate, U f is the recurrent weight matrix of the forget gate, b f is the bias term of the forget gate, and σ1 is the sigmoid function with an output range between [0, 1].

[0169] Generate new memory content through the candidate state and update it by combining the information controlled by the forget gate. The formula for the candidate state is as follows:

[0170]

[0171] Among them, is the candidate state at the t-th time step, W c is the weight matrix of the candidate state, U c is the recurrent weight matrix of the candidate state, b c is the bias term of the candidate state.

[0172] The input gate determines which new information needs to be stored in the current state. The closer the value is to 1, the more new information is introduced. The formula for the input gate is as follows:

[0173] i t = σ1(W i β t + U i h t-1 + b i );

[0174] Among them, i t is the output of the input gate at the t-th time step.

[0175] Through the final state calculation, integrate forgetting and new memories to form a new state that reflects the important information at the current time step. The formula for the final state calculation is as follows:

[0176]

[0177] Among them, c t is the final state at the t-th time step, and f t is the output of the forget gate, i t is the output of the input gate, and c t-1 is the state at the (t - 1)-th time step. ⊙ represents element-wise multiplication.

[0178] The output gate determines which information in the current state needs to be passed to the next time step. The formula for the output gate is as follows:

[0179] o t = σ1(W o β t + U o h t-1 + b o );

[0180] Among them, o t is the output of the output gate at the t-th time step.

[0181] The final hidden state model passes the important information in the current state to the next time step for subsequent processing. The formula for the final hidden state is as follows:

[0182] h t = o t ⊙ c t ;

[0183] Among them, h t is the final hidden state at the t-th time step.

[0184] The abnormal signal recognition method of this embodiment constructs a time series feature extraction framework through an LSTM model, utilizes the collaborative mechanism of the forget gate, input gate, and output gate to dynamically capture the time series dependence characteristics of hot spot area information. Among them, the forget gate controls the retention degree of historical information through the sigmoid function, the candidate state and the input gate jointly screen new features, and the calculation of the final state realizes the precise fusion of long-term and short-term memories. The cross-entropy loss function is used to quantify the model prediction error, combined with the Adam optimizer to improve the training efficiency. The risk of overfitting is effectively suppressed through the division of the training set and the validation set and the early stopping method, ensuring the generalization ability of the model to historical data. Based on the hidden state regulated by the output gate, the time series features are passed to the subsequent time steps, enhancing the continuity of abnormal signal recognition. The finally output abnormal recognition result realizes the real-time monitoring of marine signal anomalies through the early warning trigger mechanism, providing a high-timeliness decision-making basis for dynamically adjusting communication resources.

[0185] In some embodiments, triggering an early warning for the abnormal recognition result includes:

[0186] Setting an early warning threshold, and triggering an early warning when the abnormal signal strength exceeds the threshold;

[0187] Generate warning information and display it through a visual interface.

[0188] In this embodiment, the warning information includes the time, location, and signal strength of the abnormal signal.

[0189] The warning trigger mechanism of this embodiment accurately captures the event of abnormal signal strength exceeding the limit by setting a dynamic warning threshold, ensuring the timeliness and accuracy of warning triggering; the generated warning information includes three elements: the time, location, and strength of the abnormal signal, providing a multi-dimensional description of abnormal features and facilitating the quick positioning of the problem source; combined with the visual interface to intuitively display the warning information, reducing the complexity of manual analysis and improving the efficiency of maritime signal monitoring. The quantitative correlation mechanism between the warning threshold and the abnormal signal strength effectively balances the risks of false alarms and missed alarms, ensuring the robustness of the warning system in complex environments.

[0190] Please refer to Figure 2 , in some embodiments, generating a maritime signal heat map based on the abnormal recognition result and the hot spot area information, and dynamically updating the maritime signal heat map includes:

[0191] Mapping the hot spot area distribution data and the abnormal signal recognition result to the electronic chart grid to obtain a plurality of grid cells, and each grid cell corresponds to a geographical location coordinate;

[0192] Calculating the heat value of each grid cell according to the hot spot area information and the abnormal recognition result;

[0193] Defining that the heat values from the first level to the tenth level show a preset upward trend;

[0194] Defining the first level to the tenth level according to RGB colors to obtain the color levels corresponding to the first level to the tenth level respectively;

[0195] Classifying the heat values and converting the heat values into saturation indicators of color levels to form a heat distribution layer;

[0196] Fusing the heat distribution layer with the map to obtain a maritime signal heat map;

[0197] Dynamically updating the maritime signal heat map according to a preset period.

[0198] In this embodiment, the maritime area is divided into latitude and longitude grids (i.e., electronic chart grids) to ensure the coverage and accuracy of the heat map. Each grid cell corresponds to a geographical location coordinate. According to the analysis of the signal intensity of real-time signal data and historical data, the heat value of each grid is calculated, and the signal intensity distribution is represented by color gradient to ensure the accuracy of the heat map. And the distribution data of the detected hot spot areas and the results of abnormal signal recognition are mapped to the electronic chart grids. The dynamic update function of the heat map is realized to ensure automatic refreshing every 5 minutes or periodic batch update in units of hours or days.

[0199] The formula for calculating the heat value of each grid cell is:

[0200]

[0201] where S k is the intensity of the k-th signal point, and w k is the dynamic weight (such as distance weight).

[0202] For the abnormal signal area, an additional abnormal weight coefficient α (such as α = 1.5) is added to the heat value, and the formula is:

[0203]

[0204] Adding an abnormal weight coefficient to the abnormal signal area to highlight the abnormal signal.

[0205] The signal intensity distribution from the first level to the tenth level is represented by color gradient for easy intuitive understanding by users. Preferably, the abnormal signal area is highlighted with a special mark (such as a flashing effect) for quick identification by users.

[0206] The method for generating a maritime signal heat map in this embodiment has the following beneficial effects: By mapping the hot spot area distribution data and the abnormal signal recognition results to the electronic chart grid divided by longitude and latitude, each grid cell corresponds to an accurate geographical location coordinate. Combining with the heat value calculation formula to quantify the signal intensity distribution, it ensures the objectivity and data adaptability of the heat value calculation; An abnormal weight coefficient is introduced for the abnormal signal area to enhance the heat value of the abnormal area, and special markings are used to achieve visual highlighting of the abnormal signal, improving the user recognition efficiency. The RGB color is used to define the heat value levels from the first level to the tenth level, and a heat distribution layer is constructed through color gradient and saturation indication, intuitively showing the continuous change trend of the signal intensity from low to high, reducing the complexity of data interpretation. The fusion of the heat distribution layer and the map generates a maritime signal heat map, combined with a preset periodic dynamic update mechanism, ensuring the real-time and continuity of the signal status information. The synergistic effect of the dynamic weight and the abnormal weight optimizes the accuracy of the heat value spatial distribution representation, and the longitude and latitude grid division ensures the consistency of the heat map coverage range and geographical positioning, providing high-precision and multi-dimensional visual decision support for maritime communication resource scheduling.

[0207] In some embodiments, displaying the maritime signal heat map through a visualization interface includes:

[0208] Displaying the maritime signal heat map through Leaflet.js, supporting map zooming and panning;

[0209] Implementing color gradient and interactive query of the heat map through Plotly.js.

[0210] In this embodiment, interactive query, multi-dimensional analysis and abnormal alarm functions are provided. Tools such as Leaflet or Plotly are used to embed the heat map into the Web interface. The heat map is displayed through the Web side or the mobile terminal, supporting map zooming (including zooming from the global view to the local sea area), hot spot marking (that is, clicking on the hot spot area to display detailed information such as signal intensity and the number of ships), abnormal query (that is, inputting the time range or area to query the abnormal signal record) and early warning setting, providing user-defined functions such as color scheme adjustment and early warning threshold setting (such as triggering an early warning when the signal intensity is lower than -90dBm).

[0211] This embodiment uses Leaflet.js to achieve seamless integration of marine signal heat maps and electronic maps, supports multi-level zooming and panning operations from global views to local sea areas, and ensures accurate positioning of geographic information; combined with Plotly.js dynamic rendering color gradient effect, it intuitively presents the signal strength distribution trend, and improves data readability through interactive query functions. The cross-platform adaptability of the Web and mobile terminals meets the monitoring needs of multiple scenarios. The abnormal query module supports retrieval of historical records by time range or region. The warning setting function allows users to customize signal strength thresholds and color schemes, enhancing system flexibility and human-computer interaction efficiency.

[0212] In some embodiments, the marine signal heat map is provided to a third-party system through an API interface.

[0213] In this embodiment, the marine signal heat map is pushed to a third-party system (such as a maritime search and rescue center, fishery regulatory department) in real time through an API interface in a standardized data format to ensure the cross-platform compatibility and transmission efficiency of the heat map data; the third-party system can seamlessly integrate the heat map information, call real-time or historical signal distribution data based on its own business needs, and realize collaborative decision-making such as search and rescue route planning and dynamic adjustment of fishery regulatory areas; the information sharing mechanism strengthens the collaboration capabilities of multiple departments, improves the accuracy of maritime emergency response and resource scheduling, and avoids the problem of data islands.

[0214] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0215] The present invention improves data integrity and quality through multi-source data fusion and preprocessing, combines optimized clustering algorithms to accurately identify hot spot area boundaries, uses LSTM models to achieve reliable detection of abnormal signals, and intuitively displays signal distribution changes through dynamically updated heat maps. The time-space alignment and weighted fusion of multi-source data ensure the comprehensiveness and consistency of data sources; the double filtering and feature extraction in the preprocessing stage effectively remove noise and enhance data features, thereby improving the accuracy of subsequent analysis. The hot spot area division is optimized through density analysis and boundary optimization algorithms to improve the accuracy of regional detection; the LSTM model is trained based on historical data to enhance the reliability of abnormal signal identification and the timeliness of early warning. The dynamic update mechanism combines hierarchical color mapping to convert thermal values into intuitive visualization effects, supporting real-time monitoring; the interactive visualization tool improves user operation experience and analysis efficiency, while achieving seamless docking with third-party systems through the API interface to expand the scope of application. The overall method effectively improves the real-time, accuracy and decision-making support capabilities of marine signal monitoring.

[0216] In addition, in each embodiment of the present invention, each functional unit may be integrated into a processing unit, may exist physically alone for each unit, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0217] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods in each embodiment of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0218] The above are only partial embodiments of the present invention, and thus do not limit the protection scope of the present invention. Any equivalent device or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A method for updating a maritime signal heat map based on big data analysis, characterized in that, Including: Obtain maritime signal data, and preprocess the maritime signal data to obtain first processed data; Use a clustering algorithm to detect hot spot areas for the first processed data to obtain hot spot area information; Input the hot spot area information into an abnormal signal recognition model to obtain an abnormal recognition result; Generate a maritime signal heat map based on the abnormal recognition result and the hot spot area information, and dynamically update the maritime signal heat map; Display the maritime signal heat map through a visualization interface.

2. The method for updating the maritime signal heat map based on big data analysis according to claim 1, wherein Obtaining maritime signal data includes: Obtain multi-source data, where the multi-source data is configured as at least one of front-end device data, satellite communication system data, Automatic Identification System (AIS) data of ships, and Vessel Traffic Service (VTS) system data; Perform data alignment on the multi-source data through interpolation. The data alignment includes time alignment and spatial alignment to obtain first aligned data; Convert the format of the first aligned data to obtain second aligned data; Fuse the second aligned data through the weighted average method to obtain the maritime signal data, which is represented by formula (1), and the formula (1) is as follows: In formula (1), S fu is the fused offshore signal data, ω i is the weight coefficient of the i-th second alignment data, and S i is the i-th second alignment data.

3. The method for updating the maritime signal heat map based on big data analysis according to claim 1, characterized in that, Preprocessing the maritime signal data to obtain first processed data includes: Perform median filtering on the maritime signal data to obtain first denoised data; Perform mean filtering on the first denoised data to obtain second denoised data; Perform normalization processing on the second denoised data to obtain second processed data; Divide the second processed data into time windows according to timestamps. The time window has a preset time length to obtain multiple time windows; Extract features for each time window to obtain first feature information, where the first feature information includes signal strength features and regional features; Map and store the first feature information and the timestamps of the time windows to form a time feature column; Concatenate the time feature column and the second processed data to obtain the first processed data.

4. The method for updating the maritime signal heat map based on big data analysis according to claim 1, characterized in that, Using a clustering algorithm to detect hot spot areas for the first processed data to obtain hot spot area information includes: Initialize the parameters of the clustering algorithm, including the number of clusters K and the maximum number of iterations; Perform clustering analysis on the first processed data until the maximum number of iterations is reached to obtain initial hot spot distribution information; Optimize the boundaries of the initial hot spot distribution information to obtain final hot spot distribution information; Use the final hot spot distribution information as the hot spot area information.

5. The method for updating the maritime signal heat map based on big data analysis according to claim 4, characterized in that, Optimizing the boundaries of the initial hot spot distribution information to obtain final hot spot distribution information includes: The initial hot spot distribution information includes multiple data points. Calculate the local density value of each data point; Calculate the minimum distance of each data point according to the local density value to obtain the density information of each data point; Denote the data point with the highest density value in the density information as the center of the first hot spot area. The center of the first hot spot area corresponds to a first hot spot area. Calculate the distance from the center of the first hot spot area to other data points and denote it as the boundary distance; Determine whether the boundary distance is within the threshold range of the preset boundary distance. If so, divide the data point into the first hot spot area; In addition, merge the first hot spot areas with overlapping boundaries to generate a second hot spot area; Generate the final hot spot distribution information according to the second hot spot area.

6. The method for updating the maritime signal heat map based on big data analysis according to claim 1, characterized in that, The abnormal signal is configured to be constructed by an LSTM model. The hot spot area information is input into the abnormal signal recognition model, and the abnormal recognition results include: Construct an LSTM model and use historical data as a sample data set to train the LSTM model. The loss function during the training process of the LSTM model is cross-entropy loss; After the LSTM model is trained, input the hot spot area information into the abnormal signal recognition model to obtain the abnormal recognition results; Trigger a warning for the abnormal recognition results.

7. The method for updating the marine signal heat map based on big data analysis according to claim 6, characterized in that, Triggering a warning for the abnormal recognition results includes: Set a warning threshold, and trigger a warning when the abnormal signal strength exceeds the threshold; Generate a warning message and display it through a visualization interface.

8. The method for updating the maritime signal heat map based on big data analysis according to claim 1, characterized in that, Generate a maritime signal heat map according to the abnormal recognition results and the hot spot area information, and dynamically update the maritime signal heat map, including: Map the hot spot area distribution data and the abnormal signal recognition results to the electronic chart grid to obtain a plurality of grid cells, and each grid cell corresponds to a geographical location coordinate; Calculate the heat value of each grid cell according to the hot spot area information and the abnormal recognition results; Define that the heat values from the first level to the tenth level show a preset trend of increasing distribution; Define the first level to the tenth level according to RGB colors to obtain the color levels corresponding to the first level to the tenth level respectively; Classify the heat values and convert the heat values into saturation indicators of color levels to form a heat distribution layer; Fuse the heat distribution layer with the map to obtain a maritime signal heat map; Dynamically update the maritime signal heat map according to a preset period.

9. The method for updating the maritime signal heat map based on big data analysis according to claim 1, wherein Display the maritime signal heat map through a visualization interface, including: Display the maritime signal heat map through Leaflet.js, supporting map zooming and panning; Implement color gradient and interactive query of the heat map through Plotly.js.

10. The method for updating the maritime signal heat map based on big data analysis according to claim 1, characterized in that, Provide the maritime signal heat map to a third-party system through an API interface.

Images