UUV trajectory data enhancement method, system and device based on time sequence fusion

Abstract

The invention discloses a UUV trajectory data enhancement method, system and device based on time sequence fusion, and the method comprises the steps: obtaining the position data, time index data and navigation data of an underwater unmanned vehicle, and analyzing the position data to obtain three-dimensional distance data; performing normalization processing on the time index data and the three-dimensional distance data to obtain an initial time sequence data set, performing distance analysis on the initial time sequence data set to obtain initial distance data, and performing preprocessing to obtain a time sequence data set; establishing a time series data clustering model to analyze the time series data set to obtain a clustering time series data set; performing iterative updating through a particle swarm optimization algorithm and a fitness function to obtain an updated time sequence data set; and performing fitting correction on the updated time sequence data set to obtain a corrected time sequence data set. According to the method, the time index data is combined, updating is carried out through the particle swarm optimization algorithm, and then the accuracy of trajectory data simulation of the unmanned underwater vehicle is improved through fitting correction.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a UUV trajectory data enhancement method, system and device based on time series fusion. Background Art

[0002] With the increasing strategic value of the ocean, the research and application prospects of UUV (unmanned undersea vehicle) have received extensive attention. Whether it is deep-sea exploration, underwater rescue, or marine scientific research, it all depends on the UUV's ability to accurately navigate and effectively record its trajectory. However, due to the complexity and unpredictability of the marine environment, UUVs often face the problem of data loss or error accumulation when performing tasks. In addition, in the underwater environment, signal propagation is limited, such as significant absorption and scattering effects, resulting in short communication distance, low bandwidth, and signal susceptibility to interference. The risk of data packet loss is high, which not only affects the efficiency of task execution, but also may lead to the loss of important information data; inaccurate trajectory data recording not only causes the UUV to deviate from the predetermined path, increase energy consumption, but may also require repeated detection, reducing mission efficiency and economy; trajectory deviation increases the risk of collision in the complex environment of the seabed, causing damage to equipment or even mission failure; trajectory deviation also increases the difficulty and risk of recovery in emergency situations, affecting the accuracy of scientific research data.

[0003] As an effective data processing method, clustering algorithm has attracted attention in the study of UUV trajectory data. At present, commonly used clustering algorithms include DBSCAN algorithm, K-Means algorithm and Gaussian mixture model. Among them, DBSCAN algorithm can identify abnormal points in trajectory data and is suitable for clusters with arbitrary shapes; although K-Means algorithm is simple and easy to use, it is sensitive to the initial center point, resulting in misclassification; Gaussian mixture model realizes data distribution analysis by assuming the data generation process, thereby improving the accuracy of anomaly detection. The minimum residual positioning algorithm achieves the purpose of optimizing the positioning result by constructing an objective function to minimize the distance error between the measured target position and the known reference point. However, due to environmental noise and measurement errors, directly solving the problem through the minimum residual will result in local optimal solution or large error.

[0004] By combining clustering algorithms with positioning algorithms, significant progress has been made in identifying anomalies in UUV trajectory data, improving the sensitivity and accuracy of anomaly detection. In the process of analyzing UUV trajectory data, the time series data of data points plays a key role, but the existing algorithms may have calculation errors because they do not take into account the order of time series; in addition, traditional clustering algorithms calculate the distance between all data points, resulting in excessive calculations. Therefore, the accuracy of existing algorithms and the detection of outliers are far from meeting the needs of actual engineering. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a UUV trajectory data enhancement method, system and device based on time series fusion.

[0006] In order to solve the above technical problems, the present invention is solved by the following technical solutions:

[0007] A UUV trajectory data enhancement method based on time series fusion includes the following steps:

[0008] Obtaining the position data, time index data and navigation data of the underwater unmanned vehicle, analyzing the position data, and obtaining three-dimensional distance data;

[0009] Normalizing the time index data and the three-dimensional distance data respectively, and obtaining an initial time series data set based on the normalized time index data and the three-dimensional distance data;

[0010] Calculate the initial distance data between each pair of initial time series data in the initial time series data set, and preprocess the initial time series data set by using the initial distance data to obtain a time series data set;

[0011] Establish a time series data clustering model, and use the time series data clustering model to remove abnormal data in the time series data set to obtain a clustered time series data set;

[0012] Based on the navigation data and the initial distance data, the fitness function of the particle swarm optimization algorithm is constructed, the position data and the navigation data are used as particles in the particle swarm optimization algorithm, and the local optimal value and the global optimal value of the particles are updated through the fitness function. Based on the global optimal value of the particles, the clustering time series data set is updated to obtain the updated time series data set;

[0013] The updated time series data set is subjected to data fitting correction to obtain a corrected time series data set, that is, an enhanced underwater unmanned vehicle trajectory data set.

[0014] As an implementable method, the analysis based on the position data to obtain the three-dimensional distance data includes the following steps:

[0015] The location data includes longitude data, latitude data and depth data. Based on the longitude data and latitude data, the longitude difference and latitude difference are obtained, and combined with the radius of the earth, the surface distance data is obtained, which is expressed as follows:

[0016]

[0017] The depth difference is obtained based on the depth data in the position data. By analyzing the depth difference and the surface distance data, the three-dimensional distance data is obtained, which is expressed as follows:

[0018]

[0019] Among them, d represents the surface distance data, r represents the radius of the earth, Δφ represents the latitude difference, φ 1 ,φ 2 represents latitude data, Δλ represents longitude difference, d 3D represents three-dimensional distance data, and Δh represents the depth difference.

[0020] As an implementation method, normalizing the time index data and the three-dimensional distance data respectively, and obtaining an initial time series data set based on the normalized time index data and the three-dimensional distance data, includes the following steps:

[0021] The maximum and minimum values ​​of the three-dimensional distance data are obtained, and the three-dimensional distance data is normalized based on the maximum and minimum values ​​to obtain the normalized three-dimensional distance data, which is expressed as follows:

[0022]

[0023] Get the maximum and minimum values ​​of the time index data, normalize the time index data based on the maximum and minimum values, and obtain the normalized time index data, which is expressed as follows:

[0024]

[0025] The normalized three-dimensional distance data and time index data are converted to obtain the initial time series data set, which is expressed as follows:

[0026]

[0027] in, represents the normalized three-dimensional distance data of the i-th node, d 3 i D represents the i-th three-dimensional distance data, min(D) represents the minimum value of the three-dimensional distance data set, and max(D) represents the maximum value of the three-dimensional distance data set. represents the normalized time index data of the ith time, ti represents the i-th time index data, min(T) represents the minimum value of the time index data, max(T) represents the maximum value of the time index data, DT represents the initial time series data set, and n represents the number of data.

[0028] As an implementable method, the preprocessing of the initial time series data set by using the initial distance data includes the following steps:

[0029] The spacing of the initial time series data in the initial time series data set is analyzed to obtain the initial distance data, which is expressed as follows:

[0030]

[0031] A range threshold is set to remove the initial time series data whose initial distance data does not meet the range threshold in the initial time series data set, and data analysis is performed on the initial time series data set to remove redundant data in the initial time series data set to obtain a time series data set;

[0032] Among them, D(t s ,t d ) represents the initial distance data, represents the i-th normalized time index data, represents the jth normalized time index data, represents the normalized three-dimensional distance data of the i-th element, Represents the j-th normalized three-dimensional distance data.

[0033] As an implementable method, the method of removing abnormal data from a time series data set by using a time series data clustering model includes the following steps:

[0034] A distance threshold is preset, and the distance between the time series data at the current moment and the time series data at the previous moment in the time series data set is obtained to obtain the adjacent time series distance. If the adjacent time series distance is less than the distance threshold, the time series data at the current moment and the time series data at the previous moment are classified into the same classification cluster;

[0035] If the adjacent time series distance is greater than the distance threshold, the time series data at the current moment is set as a new classification cluster until the time series data set is traversed and the classification of the time series data set is completed to obtain the classification cluster of the time series data set;

[0036] A data volume threshold is preset, and the classification clusters in the classification clusters of the time series data set whose number of time series data is less than the data volume threshold are eliminated to obtain a clustered time series data set.

[0037] As an implementable method, the fitness function is obtained by the following steps:

[0038] The navigation data of the underwater unmanned vehicle includes speed data and heading data, which are expressed as follows:

[0039] V i nos =V i +v 1i

[0040]

[0041] Based on the speed data and heading data, combined with the initial distance data and the position data of the underwater unmanned vehicle, a fitness function is established, which is expressed as follows:

[0042]

[0043] Among them, V i nos Indicates the speed data, V i Represents the true value of the speed data, v 1i represents the measurement noise of the speed data, Indicates heading data, S i represents the true value of the heading data, ν 2i represents the measurement noise of the heading data, f(X) represents the fitness function, represents the nth initial distance data, ζ represents the weighting coefficient used to adjust the heading influence, N represents the number of data, and X represents the position data.

[0044] As an implementable method, the method updates the local optimal value of the particle and the global optimal value of the particle by using the fitness function, and updates the clustered time series data set based on the global optimal value of the particle to obtain an updated time series data set, including the following steps:

[0045] The position data and speed data are used as the particle position and particle speed in the particle swarm optimization algorithm, and the initial global optimal value is obtained based on the particle position, which is expressed as follows:

[0046]

[0047] Based on the particle velocity at the current moment, the particle state at the same moment is obtained, which is expressed as follows:

[0048]

[0049] Get the fitness function of the particle state at the current moment, analyze it in combination with the fitness function of the particle state obtained in the previous iteration, and then iterate to get the local optimal value of the particle until the number of iterations is greater than the number of particles, and get the global optimal value of the particle. The global optimal value of the particle includes the particle speed and particle position, which is expressed as follows:

[0050]

[0051] The clustering time series data set is updated based on the particle position in the global optimal value of the particle to obtain an updated time series data set, wherein the updated time series data includes updated longitude data, updated latitude data and updated depth data, which are expressed as follows:

[0052]

[0053] in, represents the initial global optimal value, represents the fitness function of the particle position of the mth particle at the first iteration, M represents the number of particles, represents the particle velocity of the mth particle at the wth iteration, w represents the number of iterations, c 1 、c 2 represents a positive coefficient, ξ and η represent random pseudo numbers, represents the local optimal value of the mth particle at the w-1th iteration, represents the global optimal value of the mth particle at the w-1th iteration, represents the state of the mth particle at the w-1th iteration, represents the state of the mth particle at the wth iteration, The particle velocity of the mth particle at the w-1th iteration, represents the local optimal value of the particle, represents the global optimal value of the particle, represents the fitness function of the global optimal value of the mth particle at the w-1th iteration, Indicates updating time series data. Indicates updating longitude data. Indicates updating latitude data. Indicates updating depth data, and t indicates the current time.

[0054] As an implementable method, performing data fitting correction on the updated time series data set includes the following steps:

[0055] A time threshold is preset, and the updated time series data corresponding to the time threshold is selected. Based on the data sampling time, the fitting order and the fitting coefficient, the updated time series data corresponding to the time threshold is fitted by polynomial fitting to obtain the fitting result, which is expressed as follows:

[0056]

[0057] Based on the fitting results, the fitting relationship is converted into a matrix form through the data sampling time matrix and the fitting coefficient matrix, which is expressed as follows:

[0058]

[0059] Analyze the relationship between the time threshold and the fitting order. If the time threshold and the fitting order meet the fitting constraints, the fitting coefficient matrix is ​​obtained by solving, where the constraints are expressed as follows:

[0060] h≥k+1

[0061] The fitting coefficient matrix is ​​expressed as follows:

[0062]

[0063] Based on the fitting coefficients in the fitting coefficient matrix, the updated time series data at the current moment is corrected to obtain a corrected time series data set, which is expressed as follows:

[0064]

[0065] in, represents the k-order data sampling time of the h-th updated time series data, α k Indicates the k-order fitting coefficient for updating longitude data, β k represents the k-order fitting coefficient of the updated latitude data, γ k Indicates the k-order fitting coefficient of the updated depth data, k represents the fitting order, t represents the current time, Indicates the hth updated longitude data, represents the hth updated latitude data, represents the hth updated depth data, t represents the data sampling time matrix, A represents the fitting coefficient matrix for updating longitude data, B represents the fitting coefficient matrix for updating latitude data, and C represents the fitting coefficient matrix for updating depth data. Represents the corrected longitude data in the corrected time series data, Represents the corrected latitude data in the corrected time series data, represents the corrected depth data in the corrected time series data, t k represents the moment of the k-th order fitting coefficient.

[0066] A UUV trajectory data enhancement system based on time series fusion, including a data acquisition module, a normalization processing module, a preprocessing module, an abnormality elimination module, a data update module and a fitting correction module;

[0067] The data acquisition module acquires the position data, time index data and navigation data of the underwater unmanned vehicle, analyzes the position data and obtains three-dimensional distance data;

[0068] The normalization processing module performs normalization processing on the time index data and the three-dimensional distance data respectively, and obtains an initial time series data set based on the normalized time index data and the three-dimensional distance data;

[0069] The preprocessing module calculates initial distance data between each pair of initial time series data in the initial time series data set, and preprocesses the initial time series data set by using the initial distance data to obtain a time series data set;

[0070] The abnormal elimination module establishes a time series data clustering model, and eliminates abnormal data in the time series data set through the time series data clustering model to obtain a clustered time series data set;

[0071] The data updating module constructs a fitness function of a particle swarm optimization algorithm based on the navigation data and the initial distance data, uses the position data and the navigation data as particles in the particle swarm optimization algorithm, and updates the local optimal value and the global optimal value of the particles through the fitness function, and updates the clustering time series data set based on the global optimal value of the particles to obtain an updated time series data set;

[0072] The fitting correction module performs data fitting correction on the updated time series data set to obtain a corrected time series data set, that is, an enhanced underwater unmanned vehicle trajectory data set.

[0073] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following method is implemented:

[0074] Obtaining the position data, time index data and navigation data of the underwater unmanned vehicle, analyzing the position data, and obtaining three-dimensional distance data;

[0075] Normalizing the time index data and the three-dimensional distance data respectively, and obtaining an initial time series data set based on the normalized time index data and the three-dimensional distance data;

[0076] Calculate the initial distance data between each pair of initial time series data in the initial time series data set, and preprocess the initial time series data set by using the initial distance data to obtain a time series data set;

[0077] Establish a time series data clustering model, and use the time series data clustering model to remove abnormal data in the time series data set to obtain a clustered time series data set;

[0078] Based on the navigation data and the initial distance data, the fitness function of the particle swarm optimization algorithm is constructed, the position data and the navigation data are used as particles in the particle swarm optimization algorithm, and the local optimal value and the global optimal value of the particles are updated through the fitness function. Based on the global optimal value of the particles, the clustering time series data set is updated to obtain the updated time series data set;

[0079] The updated time series data set is subjected to data fitting correction to obtain a corrected time series data set, that is, an enhanced underwater unmanned vehicle trajectory data set.

[0080] A UUV trajectory data enhancement device based on time series fusion includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the following method is implemented:

[0081] Obtaining the position data, time index data and navigation data of the underwater unmanned vehicle, analyzing the position data, and obtaining three-dimensional distance data;

[0082] Normalizing the time index data and the three-dimensional distance data respectively, and obtaining an initial time series data set based on the normalized time index data and the three-dimensional distance data;

[0083] Calculate the initial distance data between each pair of initial time series data in the initial time series data set, and preprocess the initial time series data set by using the initial distance data to obtain a time series data set;

[0084] Establish a time series data clustering model, and use the time series data clustering model to remove abnormal data in the time series data set to obtain a clustered time series data set;

[0085] Based on the navigation data and the initial distance data, the fitness function of the particle swarm optimization algorithm is constructed, the position data and the navigation data are used as particles in the particle swarm optimization algorithm, and the local optimal value and the global optimal value of the particles are updated through the fitness function. Based on the global optimal value of the particles, the clustering time series data set is updated to obtain the updated time series data set;

[0086] The updated time series data set is subjected to data fitting correction to obtain a corrected time series data set, that is, an enhanced underwater unmanned vehicle trajectory data set.

[0087] The present invention has significant technical effects due to the adoption of the above technical solution:

[0088] The present invention adds time index data to the data analysis process, considers the order of time series, normalizes the Euclidean distance between the position data of underwater unmanned vehicles and the time index, classifies the normalized results, and corrects and optimizes them through the particle swarm optimization algorithm to further improve the accuracy of the electronic trajectory of underwater unmanned vehicles. Solving the abnormal problems faced by the data collected by underwater unmanned vehicles is crucial to the subsequent resource development and environmental monitoring tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0090] Figure 1 It is a schematic flow diagram of the method of the present invention;

[0091] Figure 2 It is an overall schematic diagram of the system of the present invention;

[0092] Figure 3 It is a schematic diagram of the underwater unmanned vehicle and the hardware it carries;

[0093] Figure 4 It is a schematic diagram of data collection of the present invention;

[0094] Figure 5 It is a schematic diagram of the abnormal data processing flow of the present invention;

[0095] Figure 6 It is a schematic diagram of the motion trajectory of the underwater unmanned vehicle of the present invention;

[0096] Figure 7 It is a schematic diagram of longitude data and abnormal data points;

[0097] Figure 8 It is a schematic diagram of latitude data and abnormal data points;

[0098] Fig. 9 It is a schematic diagram of depth data and abnormal data points;

[0099] Fig.10 This is a schematic diagram of the comparison of trajectory data enhancement results;

[0100] Fig.11 3 is a schematic diagram of the comparison of the root mean square error of the method of the present invention. DETAILED DESCRIPTION

[0101] The present invention is further described in detail below in conjunction with embodiments. The following embodiments are for explanation of the present invention but the present invention is not limited to the following embodiments.

[0102] Embodiment 1:

[0103] A UUV trajectory data enhancement method based on time series fusion, such as Figure 1 As shown, the following steps are included:

[0104] S100, obtaining position data, time index data and navigation data of the underwater unmanned vehicle, analyzing the position data, and obtaining three-dimensional distance data;

[0105] S200, normalizing the time index data and the three-dimensional distance data respectively, and obtaining an initial time series data set based on the normalized time index data and the three-dimensional distance data;

[0106] S300, calculating initial distance data between each pair of initial time series data in the initial time series data set, and preprocessing the initial time series data set by using the initial distance data to obtain a time series data set;

[0107] S400, establishing a time series data clustering model, and removing abnormal data in the time series data set through the time series data clustering model to obtain a clustered time series data set;

[0108] S500, based on the navigation data and the initial distance data, construct a fitness function of the particle swarm optimization algorithm, use the position data and the navigation data as particles in the particle swarm optimization algorithm, and update the local optimal value and the global optimal value of the particles through the fitness function, and update the clustering time series data set based on the global optimal value of the particles to obtain an updated time series data set;

[0109] S600: Perform data fitting correction on the updated time series data set to obtain a corrected time series data set, that is, an enhanced underwater unmanned vehicle trajectory data set.

[0110] The present invention introduces time index data when analyzing the data of underwater unmanned vehicles, considers the time sequence between data, avoids the problem that the distance between abnormal points and normal points is reachable, but the actual time interval is far, and solves the problem of huge calculation amount of clustering algorithm based on the time continuity of data points in the same classification cluster, improves the accuracy and performance of clustering algorithm, and realizes accurate elimination of abnormal data points in underwater unmanned vehicles. The trajectory data of underwater unmanned vehicles after elimination is fitted and corrected by particle swarm optimization algorithm and minimum residual positioning algorithm, so as to achieve the purpose of improving the accuracy of electronic trajectory of underwater unmanned vehicles.

[0111] In the vast blueprint of the earth, the ocean is a key component, covering up to 70% of the earth's surface. With the increasing importance of marine resource development and environmental protection, underwater unmanned vehicles, as a key marine engineering equipment, not only bear the responsibility of marine exploration and monitoring, but also are responsible for important data collection and analysis. The important resources of current marine research are collected by sensors carried by underwater unmanned vehicles. Therefore, it is particularly important to ensure the accuracy of data collected by underwater unmanned vehicles during underwater movement. However, due to the complexity and unpredictability of the underwater environment, such as strong underwater currents, multilateral seabed topography, and electromagnetic interference, the data collected by underwater unmanned vehicles often face abnormal problems, such as data loss or error accumulation, and underwater communication problems will increase the risk of data packet loss. The above problems make the trajectory data of underwater unmanned vehicles prone to loss and errors.

[0112] The accuracy of trajectory data has an important impact on the efficiency and safety of underwater unmanned vehicle mission execution. The inaccuracy of trajectory data will not only cause the underwater unmanned vehicle to deviate from the predetermined trajectory, increase energy consumption, but also may require repeated detection. More seriously, the deviation of trajectory data increases the risk of collision in complex seabed environments, which may cause equipment damage and mission failure. In addition, it may increase the difficulty and risk of recovery in emergency situations, affect the accuracy of scientific research data, and reduce the reliability of research results. In military applications, the accuracy of trajectory data is related to the effectiveness of tactical deployment. Trajectory errors cause tactical tasks to fail to achieve the expected results, such as mistakenly entering the enemy's monitoring area or failing to accurately reach the target area for reconnaissance or mine laying. Therefore, effective processing of trajectory data is crucial for subsequent tasks such as resource development, environmental monitoring and tactical deployment, and has important practical and strategic significance.

[0113] In the study of anomalies in trajectory data, clustering algorithm is an effective data processing method. It divides the data set into multiple classification clusters and ensures that the data similarity within the same classification cluster is high, while the data differences between different classification clusters are obvious. It can effectively identify data and improve the overall quality of trajectory data. Currently, commonly used clustering algorithms include DBSCAN clustering model, K-Means clustering model and Gaussian mixture model. Among them, DBSCAN clustering model can process noisy data and effectively identify anomalies in trajectories, especially suitable for clusters with arbitrary shapes; K-Means clustering model is simple and easy to use, but it is sensitive to the selection of initial center points, which may cause anomalies to be misclassified; Gaussian mixture model can make a more detailed analysis of data distribution by assuming the data generation process, thereby improving the accuracy of anomaly detection.

[0114] The minimum residual positioning algorithm is a technology based on minimizing the residual between the estimated position and the actual observed position. It is widely used in navigation, communication, and geographic mapping. The basic principle of this algorithm is to optimize the positioning result by constructing an objective function to minimize the distance error between the measured target position and the known reference point. For example, the positioning calculation in the GPS system is usually based on this algorithm to process the distance measurement between the satellite and the receiver. In wireless sensor networks, the distance between nodes is usually estimated by the received signal strength, and the minimum residual algorithm is used to calculate the actual position of the node. In indoor positioning, due to the complex signal environment, a lot of noise is introduced. The minimum residual algorithm improves the positioning accuracy by minimizing the residual caused by noise. However, the traditional minimum residual positioning algorithm also faces many challenges. Due to environmental noise and measurement errors, direct solution may lead to local optimal solutions or large errors. In order to deal with these problems, researchers have proposed a variety of improved methods, such as the weighted minimum residual method, which assigns different weights to different measurements to reduce the impact of data with large measurement errors on the final result. In addition, Kalman filtering and particle filtering techniques are often used in combination with the minimum residual algorithm to optimize positioning accuracy and reduce noise interference. However, in some application scenarios, especially the multipath effect in complex indoor environments, the signal error will be further aggravated, which makes it particularly important to improve the traditional algorithm.

[0115] In recent years, the combination of clustering algorithm and minimum residual positioning algorithm has provided a new perspective for underwater unmanned vehicle trajectory data enhancement. By learning the low-dimensional representation of trajectory data, the anomalies in the trajectory data can be effectively identified. Researchers have proposed a variety of improved algorithms to improve the sensitivity and accuracy of abnormal trajectory data detection, such as hybrid methods that combine time series information and spatial features. However, the accuracy of current methods and the performance of anomaly detection are far from meeting the needs of actual engineering.

[0116] This embodiment provides a schematic diagram of an underwater unmanned vehicle and its onboard hardware, such as Figure 3 As shown in Table 1, GPS is a global positioning system used to obtain the longitude and latitude data of the underwater unmanned vehicle; the pressure sensor obtains the depth data of the underwater unmanned vehicle by measuring the change in water pressure; DVL is a Doppler velocimeter used to measure the speed of the underwater unmanned vehicle relative to the water flow and obtain the heading data and speed data; AHRS is an attitude reference system. The relevant parameters of the underwater unmanned vehicle are shown in Table 1.

[0117] Table 1 shows the relevant parameters of the underwater unmanned vehicle

[0118]

[0119] This embodiment provides a schematic diagram of data collection, such as Figure 4As shown in the figure, during the data collection process, a buoy-type hydroacoustic communication relay station is used to realize the data transmission of the underwater unmanned vehicle. In the underwater environment, the underwater unmanned vehicle collects data through hydroacoustic signals and sends the information to the buoy relay station on the surface. The relay station is equipped with hydroacoustic transducers and hydroacoustic electronic equipment, which are responsible for converting the received hydroacoustic signals into electrical signals. The processed data is transmitted to the shore-based control service platform in the form of radio waves through a digital radio station.

[0120] In this embodiment, the position data, time index data and navigation data of the underwater unmanned vehicle are obtained, wherein the position data includes longitude data and latitude data, and the navigation data includes speed data and heading data, as shown in Table 2, which is an example of data collected by the underwater unmanned vehicle numbered 001.

[0121] Table 2 shows the data collected by the underwater unmanned vehicle No. 001

[0122]

[0123] The location data is analyzed, normalized in combination with the time index data, and the normalized data is preprocessed to obtain a time series data set, including the following steps:

[0124] Step 1: Based on the longitude data and latitude data, the longitude difference and latitude difference are obtained. Combined with the radius of the earth, the surface distance data is calculated using the Haversing formula, which is expressed as follows:

[0125]

[0126] Step 2: Based on the depth data in the location data, the depth difference is obtained. The depth difference is taken into consideration and analyzed in combination with the surface distance data to obtain three-dimensional distance data. The three-dimensional distance data is a comprehensive representation of longitude data, latitude data and depth data. In order to realize the time series analysis of trajectory data, the time index data is used as a new dimension. In order to eliminate the dimensional difference between the three-dimensional distance data and the time index data, normalization processing is required. The three-dimensional distance data is calculated by the following formula:

[0127]

[0128] Step 3: Obtain the maximum and minimum values ​​of the three-dimensional distance data, normalize the three-dimensional distance data based on the maximum and minimum values, and obtain the normalized three-dimensional distance data; obtain the maximum and minimum values ​​of the time index data, normalize the time index data based on the maximum and minimum values, and obtain the normalized time index data, which is expressed as follows:

[0129]

[0130]

[0131] Step 4: Combine the normalized three-dimensional distance data and time index data and perform data conversion to obtain an initial time series data set, and calculate the distance between each pair of initial time series data in the initial time series data set to obtain initial distance data, where the initial time series data set is represented as follows:

[0132]

[0133] The initial distance data is calculated by the following formula:

[0134]

[0135] Step 5: Based on the initial distance data, set a range threshold, remove the initial time series data whose initial distance data does not meet the range threshold in the initial time series data set, remove the redundant data in the initial time series data set, reduce data redundancy, and obtain the time series data set;

[0136] Among them, d represents the surface distance data, r represents the radius of the earth, Δφ represents the latitude difference, φ 1 ,φ 2 represents latitude data, Δλ represents longitude difference, d 3D represents three-dimensional distance data, Δh represents the depth difference, represents the normalized three-dimensional distance data of the i-th element, represents the i-th three-dimensional distance data, min(D) represents the minimum value of the three-dimensional distance data set, and max(D) represents the maximum value of the three-dimensional distance data set. represents the normalized time index data of the ith time, t i represents the i-th time index data, min(T) represents the minimum value of the time index data, max(T) represents the maximum value of the time index data, DT represents the initial time series data set, n represents the number of data, D(t s ,t d ) represents the initial distance data, represents the jth normalized time index data, Represents the j-th normalized three-dimensional distance data.

[0137] In this embodiment, the time series data set is clustered based on the DBSCAN clustering algorithm. The core concept of the DBSCAN clustering algorithm is to classify closely connected points into the same cluster based on the density characteristics between samples, that is, to perform category division based on density differences, to ensure that sample points in the same cluster are closely connected, and to eliminate abnormal data in the time series data set based on the clustering results to obtain a clustered time series data set.

[0138] During the implementation of the DBSCAN algorithm, a point P in the data set is randomly selected, and then a determination is made as to whether P is a core point. If P is a core point, all points that are density-connected to P are searched throughout the data set and are considered part of the same classification cluster. Next, the points in this cluster will be used to continuously search for new density-reachable points until no new points can be added to the classification cluster, thus completing a complete clustering. The algorithm then continues to search for new points that have not yet been classified, and repeats the same process to construct the next classification cluster. After this series of iterations, the clustering process eventually ends, and those points that are not classified into any classification cluster are marked as outliers or noise points. The flowchart for exception handling of a time series data set based on the DBSCAN clustering algorithm in this embodiment is shown in the figure. Figure 5 As shown, the following steps are included:

[0139] Step 1: Starting from the first point in the time series data set, classify it into the first classification cluster C_1;

[0140] Step 2: Preset the distance threshold eps. For the time series data at the current moment in the time series data set, if the distance between the time series data at the previous moment and the time series data at the previous moment does not exceed the preset threshold, the current time series data and the time series data at the previous moment are classified into the same classification cluster, otherwise a new classification cluster is created until all the time series data in the time series data set are classified;

[0141] Step 3: After the classification is completed, the data volume threshold MinPts is preset, and the number of data points in each classification cluster is calculated. If the number of data points in a classification cluster is less than the data volume threshold, the time series data in the classification cluster is regarded as abnormal data, and the abnormal data in the time series data set is removed to obtain the clustered time series data set.

[0142] A fitness function is established based on the position data and the navigation data, and the speed data in the position data and the navigation data are used as particles in the particle swarm optimization algorithm. The particle swarm optimization algorithm is used to iteratively update the local optimal value of the particle until the number of iterations is greater than the number of particles, and the global optimal value of the particle is obtained. The clustered time series data set is updated based on the obtained global optimal value of the particle to obtain an updated time series data set, including the following steps:

[0143] Step 1: The navigation data of the underwater unmanned vehicle includes speed data and heading data, which are expressed as follows:

[0144] V i nos =V i +v 1i

[0145]

[0146] Step 2: By combining the time index data, speed data and heading data of data sampling, the initial distance data and the position data of the underwater unmanned vehicle, a fitness function is established, which is expressed as follows:

[0147]

[0148] If there is no noise in the data acquisition process, that is, At this point, the particle's estimate is consistent with the target position, which is the purpose of this method, as shown below:

[0149]

[0150] Step 3: Use the position data and speed data as particles in the particle swarm optimization algorithm, and obtain the initial global optimal value of the particle swarm optimization algorithm, which is expressed as follows:

[0151]

[0152] Step 4: Get the particle velocity at the current moment, and based on the particle velocity at the current moment, get the particle position at the same moment, as shown below:

[0153]

[0154] Step 5: Get the fitness function of the particle position at the current moment, analyze it in combination with the fitness function of the particle position obtained in the previous iteration, perform iterative updates based on the analysis results, and obtain the local optimal value of the particle, until the number of iterations is greater than the number of particles, and obtain the global optimal value of the particle, where the global optimal value of the particle includes the particle position and particle velocity. The process of updating the local optimal value of the particle and the global optimal value of the particle is expressed as follows:

[0155]

[0156] Step 6: Based on the global optimal value of the particle, the clustering time series data set is updated to obtain an updated time series data set, wherein the updated time series data includes updated longitude data, updated latitude data, and updated depth data, which are expressed as follows:

[0157]

[0158] Among them, V i nos Indicates the speed data, V i Represents the true value of the speed data, v 1i represents the measurement noise of the speed data, Indicates heading data, S i represents the true value of the heading data, ν 2i represents the measurement noise of the heading data, f(X) represents the fitness function, represents the nth initial distance data, ζ represents the weighting coefficient used to adjust the heading influence, N represents the number of position data, X represents the position data, represents the initial global optimal value, represents the fitness function of the particle position of the mth particle at the first iteration, M represents the number of particles, represents the particle velocity of the mth particle at the wth iteration, w represents the number of iterations, c 1 、c 2 represents a positive coefficient, ξ and η represent random pseudo numbers, represents the local optimal value of the mth particle at the w-1th iteration, represents the global optimal value of the mth particle at the w-1th iteration, represents the position of the mth particle at the w-1th iteration, represents the position of the mth particle at the wth iteration, The particle velocity of the mth particle at the w-1th iteration, represents the local optimal value of the particle, represents the global optimal value of the particle, represents the fitness function of the global optimal value of the mth particle at the w-1th iteration, Indicates updating time series data. Indicates updating longitude data. Indicates updating latitude data. Indicates updating depth data, and t indicates the current time.

[0159] In general, the predicted trajectory will fluctuate around the real trajectory. In order to reduce data errors, the update time series data set is fitted and corrected by a polynomial fitting method. In this embodiment, a time threshold is preset, and the update time series data within the time period corresponding to the time threshold in the update time series data set is selected, and the selected update time series data is fitted, including the following steps:

[0160] Step 1: Based on the data sampling time, fitting order and fitting coefficient, the updated time series data in the time period corresponding to the time threshold is fitted by polynomial fitting to obtain the fitting result, which is expressed as follows:

[0161]

[0162] Step 2: Based on the fitting results, the fitting relationship of the updated time series data is converted into a matrix form through the data sampling time matrix and the fitting coefficient matrix, which is expressed as follows:

[0163]

[0164] Step 3: Analyze the relationship between the time threshold and the fitting order. If the time threshold and the fitting order meet the fitting constraints, the fitting coefficient matrix is ​​obtained by solving. If the time threshold and the fitting order do not meet the fitting constraints, no fitting is required. The constraints are expressed as follows:

[0165] h≥k+1

[0166] The fitting coefficient matrix is ​​expressed as follows:

[0167]

[0168] Step 4: After the fitting coefficients are obtained by solving, the updated time series data at the current moment is corrected based on the fitting coefficients to obtain a corrected time series data set, which is expressed as follows:

[0169]

[0170]

[0171] in, represents the k-order data sampling time of the h-th updated time series data, α k Indicates the k-order fitting coefficient for updating longitude data, β k represents the k-order fitting coefficient of the updated latitude data, γ k Indicates the k-order fitting coefficient of the updated depth data, k represents the fitting order, t represents the current time, Indicates the hth updated longitude data, represents the hth updated latitude data, represents the hth updated depth data, t represents the data sampling time matrix, A represents the fitting coefficient matrix for updating longitude data, B represents the fitting coefficient matrix for updating latitude data, and C represents the fitting coefficient matrix for updating depth data. Represents the corrected longitude data in the corrected time series data, Represents the corrected latitude data in the corrected time series data, represents the corrected depth data in the corrected time series data, t k represents the moment of the k-th order fitting coefficient.

[0172] This embodiment simulates a real underwater scene in a local environment, simulates three different turning rates to coordinate turns, and takes the data collected by an underwater unmanned vehicle with a turning rate of 0.052 rad / s as an example to visualize its navigation path trajectory, such as Figure 6 As shown, the blue curve represents the true trajectory of the underwater unmanned vehicle, and the red points represent abnormal data points.

[0173] Since there are many data points and they are densely distributed, in order to more clearly display the characteristics of the abnormal points, such as Figure 7 represents the distribution of abnormal data points in longitude data, Figure 8 represents the distribution of abnormal data points in latitude data, Fig. 9 The blue curve represents the real trajectory of the underwater unmanned vehicle, the red points represent the abnormal data points, and the horizontal axis in the figure represents the order of data records. At the same time, this embodiment provides a comparative schematic diagram of the present invention, the DBSCAN clustering algorithm and the real trajectory, as shown in FIG. Fig.10 As shown, the blue curve represents the real trajectory of the underwater unmanned vehicle, the red curve represents the trajectory data obtained by the method of the present invention, and the black curve represents the trajectory data obtained by the DBSCAN clustering algorithm. It can be seen from the figure that the basic DBSCAN clustering algorithm can no longer meet the requirements of high-precision trajectory fitting of underwater unmanned vehicles, while the method of the present invention can perform better data enhancement and fitting on the trajectory data.

[0174] In order to better verify the performance of the method of the present invention, this embodiment provides a schematic diagram of the root mean square error of the trajectory data after data enhancement by the method of the present invention and the trajectory data after the DBSCAN clustering algorithm, as shown in FIG. Fig.11 As shown, the red curve represents the root mean square error of the DBSCAN clustering algorithm, and the blue curve represents the root mean square error of the method of the present invention; at the same time, to prove the importance of time index data in the data enhancement process, as shown in Table 3, MRLA represents the method of the present invention for data enhancement without combining time index data, TF_MRLA represents the method of the present invention, X represents longitude data, Y represents latitude data, and Z represents depth data. It can be seen from the errors of the three dimensions of longitude data, latitude data and depth data that the fitting error of the method of the present invention for data enhancement combined with time index data is significantly reduced, that is, the fitting result is more accurate.

[0175] Table 3 shows the error comparison

[0176]

[0177] Embodiment 2:

[0178] A UUV trajectory data enhancement system based on time series fusion, such as Figure 2 As shown, it includes a data acquisition module 100, a normalization processing module 200, a preprocessing module 300, an abnormality elimination module 400, a data update module 500 and a fitting correction module 600;

[0179] The data acquisition module 100 acquires the position data, time index data and navigation data of the underwater unmanned vehicle, analyzes the position data and obtains three-dimensional distance data;

[0180] The normalization processing module 200 performs normalization processing on the time index data and the three-dimensional distance data respectively, and obtains an initial time series data set based on the normalized time index data and the three-dimensional distance data;

[0181] The preprocessing module 300 calculates initial distance data between each pair of initial time series data in the initial time series data set, and preprocesses the initial time series data set by using the initial distance data to obtain a time series data set;

[0182] The abnormal elimination module 400 establishes a time series data clustering model, and eliminates abnormal data in the time series data set through the time series data clustering model to obtain a clustered time series data set;

[0183] The data updating module 500 constructs a fitness function of a particle swarm optimization algorithm based on the navigation data and the initial distance data, uses the position data and the navigation data as particles in the particle swarm optimization algorithm, and updates the local optimal value and the global optimal value of the particles through the fitness function, and updates the clustered time series data set based on the global optimal value of the particles to obtain an updated time series data set;

[0184] The fitting correction module 600 performs data fitting correction on the updated time series data set to obtain a corrected time series data set, that is, an enhanced underwater unmanned vehicle trajectory data set.

[0185] Various changes and modifications can be made without departing from the spirit and scope of the present invention, and all equivalent technical solutions also belong to the scope of the present invention.

[0186] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0187] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0188] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0189] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0190] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0191] It should be noted that:

[0192] The "one embodiment" or "embodiment" mentioned in the specification means that the specific features, structures or characteristics described in conjunction with the embodiment are included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment" or "embodiment" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0193] In addition, it should be noted that the shapes and names of the parts and components of the specific embodiments described in this specification may be different. Any equivalent or simple changes made based on the structure, features and principles described in the patent concept of the present invention are included in the protection scope of the patent of the present invention. The technicians in the technical field of the present invention can make various modifications or supplements to the specific embodiments described or replace them in a similar manner, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A UUV trajectory data enhancement method based on time series fusion, characterized in that: The following steps are involved: Obtaining the position data, time index data and navigation data of the underwater unmanned vehicle, analyzing the position data, and obtaining three-dimensional distance data; Normalizing the time index data and the three-dimensional distance data respectively, and obtaining an initial time series data set based on the normalized time index data and the three-dimensional distance data; Calculate the initial distance data between each pair of initial time series data in the initial time series data set, and preprocess the initial time series data set by using the initial distance data to obtain a time series data set; Establish a time series data clustering model, and use the time series data clustering model to remove abnormal data in the time series data set to obtain a clustered time series data set; Based on the navigation data and the initial distance data, the fitness function of the particle swarm optimization algorithm is constructed, the position data and the navigation data are used as particles in the particle swarm optimization algorithm, and the local optimal value and the global optimal value of the particles are updated through the fitness function. Based on the global optimal value of the particles, the clustering time series data set is updated to obtain the updated time series data set; The updated time series data set is subjected to data fitting correction to obtain a corrected time series data set, that is, an enhanced underwater unmanned vehicle trajectory data set.

2. The UUV trajectory data enhancement method based on time series fusion according to claim 1 is characterized in that: The method of analyzing the position data to obtain the three-dimensional distance data includes the following steps: The location data includes longitude data, latitude data and depth data. Based on the longitude data and latitude data, the longitude difference and latitude difference are obtained, and combined with the radius of the earth, the surface distance data is obtained, which is expressed as follows: The depth difference is obtained based on the depth data in the position data. By analyzing the depth difference and the surface distance data, the three-dimensional distance data is obtained, which is expressed as follows: Among them, d represents the surface distance data, r represents the radius of the earth, Δφ represents the latitude difference, φ1 and φ2 represent the latitude data, Δλ represents the longitude difference, d 3D represents three-dimensional distance data, and Δh represents the depth difference.

3. The UUV trajectory data enhancement method based on time series fusion according to claim 1 is characterized in that: The normalizing of the time index data and the three-dimensional distance data respectively, and obtaining the initial time series data set based on the normalized time index data and the three-dimensional distance data, comprises the following steps: The maximum and minimum values ​​of the three-dimensional distance data are obtained, and the three-dimensional distance data is normalized based on the maximum and minimum values ​​to obtain the normalized three-dimensional distance data, which is expressed as follows: Get the maximum and minimum values ​​of the time index data, normalize the time index data based on the maximum and minimum values, and obtain the normalized time index data, which is expressed as follows: The normalized three-dimensional distance data and time index data are converted to obtain the initial time series data set, which is expressed as follows: in, represents the normalized three-dimensional distance data of the i-th element, represents the i-th three-dimensional distance data, min(D) represents the minimum value of the three-dimensional distance data set, and max(D) represents the maximum value of the three-dimensional distance data set. represents the normalized time index data of the ith time, t i represents the i-th time index data, min(T) represents the minimum value of the time index data, max(T) represents the maximum value of the time index data, DT represents the initial time series data set, and n represents the number of data.

4. The UUV trajectory data enhancement method based on time series fusion according to claim 1 is characterized in that: The preprocessing of the initial time series data set by using the initial distance data comprises the following steps: The spacing of the initial time series data in the initial time series data set is analyzed to obtain the initial distance data, which is expressed as follows: A range threshold is set to remove the initial time series data whose initial distance data does not meet the range threshold in the initial time series data set, and data analysis is performed on the initial time series data set to remove redundant data in the initial time series data set to obtain a time series data set; Among them, D(t s ,t d ) represents the initial distance data, represents the i-th normalized time index data, represents the jth normalized time index data, represents the normalized three-dimensional distance data of the i-th element, Represents the j-th normalized three-dimensional distance data.

5. The UUV trajectory data enhancement method based on time series fusion according to claim 1 is characterized in that: The method of removing abnormal data from a time series data set by using a time series data clustering model includes the following steps: A distance threshold is preset, and the distance between the time series data at the current moment and the time series data at the previous moment in the time series data set is obtained to obtain the adjacent time series distance. If the adjacent time series distance is less than the distance threshold, the time series data at the current moment and the time series data at the previous moment are classified into the same classification cluster; If the adjacent time series distance is greater than the distance threshold, the time series data at the current moment is set as a new classification cluster until the time series data set is traversed and the classification of the time series data set is completed to obtain the classification cluster of the time series data set; A data volume threshold is preset, and the classification clusters in the classification clusters of the time series data set whose number of time series data is less than the data volume threshold are eliminated to obtain a clustered time series data set.

6. The UUV trajectory data enhancement method based on time series fusion according to claim 1 is characterized in that: The fitness function is obtained by following the steps below: The navigation data of the underwater unmanned vehicle includes speed data and heading data, which are expressed as follows: Based on the speed data and heading data, combined with the initial distance data and the position data of the underwater unmanned vehicle, a fitness function is established, which is expressed as follows: in, Indicates the speed data, V i Represents the true value of the speed data, v 1i represents the measurement noise of the speed data, Indicates heading data, S i represents the true value of the heading data, ν 2i represents the measurement noise of the heading data, f(X) represents the fitness function, represents the nth initial distance data, ζ represents the weighting coefficient used to adjust the heading influence, N represents the number of data, and X represents the position data.

7. The UUV trajectory data enhancement method based on time series fusion according to claim 1 is characterized in that: The method of updating the local optimal value of the particle and the global optimal value of the particle by using the fitness function, and updating the clustered time series data set based on the global optimal value of the particle to obtain an updated time series data set includes the following steps: The position data and speed data are used as the particle position and particle speed in the particle swarm optimization algorithm, and the initial global optimal value is obtained based on the particle position, which is expressed as follows: Based on the particle velocity at the current moment, the particle state at the same moment is obtained, which is expressed as follows: Get the fitness function of the particle state at the current moment, analyze it in combination with the fitness function of the particle state obtained in the previous iteration, and then iterate to get the local optimal value of the particle until the number of iterations is greater than the number of particles, and get the global optimal value of the particle. The global optimal value of the particle includes the particle speed and particle position, which is expressed as follows: The clustering time series data set is updated based on the particle position in the global optimal value of the particle to obtain an updated time series data set, wherein the updated time series data includes updated longitude data, updated latitude data and updated depth data, which are expressed as follows: in, represents the initial global optimal value, represents the fitness function of the particle position of the mth particle at the first iteration, M represents the number of particles, represents the particle velocity of the mth particle at the wth iteration, w represents the number of iterations, c1 and c2 represent positive coefficients, ξ and η represent random pseudo numbers, represents the local optimal value of the mth particle at the w-1th iteration, represents the global optimal value of the mth particle at the w-1th iteration, represents the state of the mth particle at the w-1th iteration, represents the state of the mth particle at the wth iteration, The particle velocity of the mth particle at the w-1th iteration, represents the local optimal value of the particle, represents the global optimal value of the particle, represents the fitness function of the global optimal value of the mth particle at the w-1th iteration, Indicates updating time series data. Indicates updating longitude data. Indicates updating latitude data. Indicates updating depth data, and t indicates the current time.

8. The UUV trajectory data enhancement method based on time series fusion according to claim 1 is characterized in that: The step of performing data fitting correction on the updated time series data set comprises the following steps: A time threshold is preset, and the updated time series data corresponding to the time threshold is selected. Based on the data sampling time, the fitting order and the fitting coefficient, the updated time series data corresponding to the time threshold is fitted by polynomial fitting to obtain the fitting result, which is expressed as follows: Based on the fitting results, the fitting relationship is converted into a matrix form through the data sampling time matrix and the fitting coefficient matrix, which is expressed as follows: Analyze the relationship between the time threshold and the fitting order. If the time threshold and the fitting order meet the fitting constraints, the fitting coefficient matrix is ​​obtained by solving, where the constraints are expressed as follows: h≥k+1 The fitting coefficient matrix is ​​expressed as follows: Based on the fitting coefficients in the fitting coefficient matrix, the updated time series data at the current moment is corrected to obtain a corrected time series data set, which is expressed as follows: in, represents the k-order data sampling time of the h-th updated time series data, α k Indicates the k-order fitting coefficient for updating longitude data, β k represents the k-order fitting coefficient of the updated latitude data, γ k Indicates the k-order fitting coefficient of the updated depth data, k represents the fitting order, t represents the current time, Indicates the hth updated longitude data, represents the hth updated latitude data, represents the hth updated depth data, t represents the data sampling time matrix, A represents the fitting coefficient matrix for updating longitude data, B represents the fitting coefficient matrix for updating latitude data, and C represents the fitting coefficient matrix for updating depth data. Represents the corrected longitude data in the corrected time series data, Represents the corrected latitude data in the corrected time series data, represents the corrected depth data in the corrected time series data, t k represents the moment of the k-th order fitting coefficient.

9. A UUV trajectory data enhancement system based on time series fusion, characterized in that: It includes data acquisition module, normalization processing module, preprocessing module, abnormal elimination module, data update module and fitting correction module; The data acquisition module acquires the position data, time index data and navigation data of the underwater unmanned vehicle, analyzes the position data and obtains three-dimensional distance data; The normalization processing module performs normalization processing on the time index data and the three-dimensional distance data respectively, and obtains an initial time series data set based on the normalized time index data and the three-dimensional distance data; The preprocessing module calculates initial distance data between each pair of initial time series data in the initial time series data set, and preprocesses the initial time series data set by using the initial distance data to obtain a time series data set; The abnormal elimination module establishes a time series data clustering model, and eliminates abnormal data in the time series data set through the time series data clustering model to obtain a clustered time series data set; The data updating module constructs a fitness function of a particle swarm optimization algorithm based on the navigation data and the initial distance data, uses the position data and the navigation data as particles in the particle swarm optimization algorithm, and updates the local optimal value and the global optimal value of the particles through the fitness function, and updates the clustering time series data set based on the global optimal value of the particles to obtain an updated time series data set; The fitting correction module performs data fitting correction on the updated time series data set to obtain a corrected time series data set, that is, an enhanced underwater unmanned vehicle trajectory data set.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

11. A UUV trajectory data enhancement device based on time series fusion, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

Images