Unmanned surface vessel track fusion method and device

Abstract

The invention discloses a method and device for unmanned surface vessel track fusion, the method comprising: obtaining perception information of the unmanned surface vessel, the perception information comprising GPS data information and radar data information; performing data preprocessing on the radar data information to obtain target radar information; constructing a track association model, and using the track association model to perform track association between the GPS data information and the target radar information; constructing a fusion data weight distribution model, and using the fusion data weight distribution model to perform track fusion between the GPS data information and the target radar information associated with it; the invention realizes complementary advantages in perception of maritime targets through information fusion, improves matching calculation speed, saves calculation resources, and has higher credibility.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned surface vessels, and in particular to an unmanned surface vessel track fusion method and device. Background Art

[0002] USV (Unmanned Surface Vehicle) unmanned surface vessels are intelligent and advanced. In recent years, they have been widely used in many fields such as water search and rescue, hydrological and meteorological detection, etc. The real-time perception of surrounding maritime navigation targets by unmanned surface vessels during navigation at sea has become an important condition to ensure the safe navigation of USVs.

[0003] However, unmanned surface vessels have problems such as blind spots in the equipment itself, equipment data preprocessing, data format unification, and data fusion weight distribution. As a result, there is still no optimal solution for the data source, data preprocessing, and data fusion model during USV navigation. In order to improve the target perception effect of USV maritime navigation operations, the GPS and radar data commonly equipped on existing USVs are used as the data source. After data preprocessing, data correction, track association, data fusion and other stages of processing, the purpose of high-precision perception of USVs and surrounding navigation targets is achieved. Summary of the invention

[0004] In order to solve the problems in the prior art, the present invention provides a method and device for unmanned surface vehicle track fusion to achieve high-precision perception of USV and surrounding navigation targets, specifically:

[0005] In a first aspect, the present application provides a method for unmanned surface vessel track fusion, the method comprising:

[0006] Acquiring unmanned surface vessel perception information, wherein the perception information includes GPS data information and radar data information;

[0007] Performing data preprocessing on the radar data information to obtain target radar information;

[0008] Constructing a track association model, and using the track association model to perform track association between the GPS data information and the target radar information;

[0009] A fusion data weight distribution model is constructed, and the GPS data information and its associated target radar information are track fused using the fusion data weight distribution model.

[0010] In a second aspect, the present application provides an unmanned surface vessel track fusion device, the device comprising:

[0011] An information acquisition module is used to acquire the perception information of the unmanned surface vessel, wherein the perception information includes GPS data information and radar data information;

[0012] A data preprocessing module, used to perform data preprocessing on the radar data information to obtain target radar information;

[0013] A track association module, used to construct a track association model, and use the track association model to perform track association between the GPS data information and the target radar information;

[0014] The track fusion module is used to construct a fusion data weight distribution model, and use the fusion data weight distribution model to perform track fusion of the GPS data information and its associated target radar information.

[0015] The unmanned surface vessel track fusion method and device provided by the present invention have the following beneficial effects:

[0016] (1) Based on the fusion of GPS and radar data, the problems of unstable transmission frequency and return delay in GPS data, and the problems of sea clutter interference and short scanning distance in radar data are solved. The purpose of complementary advantages in the perception of maritime targets is achieved through the association and fusion of GPS data and radar data.

[0017] (2) Constructing a mathematical model of track association, combining coarse association with fine association, and specifically applying multi-factor Euclidean distance and Mahalanobis distance association algorithms to associate target tracks, so that the present application can provide multiple channels for identifying the same target track, improve the matching calculation speed, and save computing resources.

[0018] (3) Construct a fusion data weight allocation model based on the hierarchical analysis method and the entropy weight assignment method to avoid the problems existing in the single weight method, take into account the advantages of the hierarchical analysis method and the entropy method, and have higher credibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions and advantages of the embodiments of this specification or the prior art, the drawings required for use in the embodiments or the prior art descriptions are 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 creative work.

[0020] Figure 1 It is a flow chart of a method for fusion of tracks of unmanned surface vessels;

[0021] Figure 2 It is a rough track correlation flow chart for constructing a track correlation model;

[0022] Figure 3 It is a flow chart of track details in constructing track association model;

[0023] Figure 4 It is a schematic diagram of the design idea of ​​the unmanned surface vessel positioning data fusion framework;

[0024] Figure 5 It is a block diagram of the components of an unmanned surface vessel track fusion device;

[0025] Figure 6 It is a comparison chart of the radar target track, GPS target track, fusion target track and ideal target track obtained in the experiment. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of this specification to clearly and completely describe the technical solutions in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0027] This implementation provides a method for unmanned surface vessel track fusion, such as Figure 1 As shown, the method includes:

[0028] S102. Acquire the unmanned surface vessel perception information, wherein the perception information includes GPS data information and radar data information;

[0029] In practical applications, it can be:

[0030] Step 1.1 Receive GPS and radar data information through the serial port to obtain a data string;

[0031] Step 1.2: According to the sensor communication protocol, the received data string is checked and decoded to obtain the current speed v, longitude and latitude L, heading angle W and other motion status information of the unmanned surface vehicle.

[0032] S104. Perform data preprocessing on the radar data information to obtain target radar information;

[0033] Furthermore, the step S104 of performing data preprocessing on the radar data information to obtain target radar information includes:

[0034] S1042. A feasible method for extracting radar data information to obtain an intermediate data set may be:

[0035] The polar coordinates of the radar target position are converted to Cartesian coordinates through trigonometric functions. The pixel coordinates are used as the conversion basis in the image coordinate conversion, and the coordinate conversion is completed through formula (1).

[0036]

[0037] In formula (1), x r ,y r represents the horizontal and vertical coordinates in the Cartesian system; R represents the polar diameter, which is the maximum detection range of the radar; θ represents the polar angle.

[0038] The matching degree is determined by comparing the difference between the target template image and the sub-image of the whole image. The one with the smallest difference is the best match.

[0039] By traversing the target template image on the whole image, the absolute difference between the target template image and the sub-image in the whole image is calculated and compared. The minimum absolute difference is the optimal matching position.

[0040] The calculation model of the minimum error method can be expressed as formula (2).

[0041]

[0042] In formula (2), E(i, j) is the absolute difference, and the minimum value is the optimal matching position; the size of the target template image T is M×N; S is the overall image, S i,j For sub-graph.

[0043] The radar Cartesian coordinate image data is converted into a binary matrix, and the binary matrix of the target template map T is defined according to expert experience as formula (3).

[0044]

[0045] The template map matrix T is used to calculate the error on the overall map, locate the optimal matching position of the target and record it, mark it in the overall map data, and obtain the target coordinates in Cartesian coordinates. Using the longitude and latitude coordinates of the radar location and the scanning radius, through coordinate conversion and proportional calculation methods, the navigation data such as the longitude and latitude coordinates, direction, and distance of the target are obtained.

[0046] S1044. Eliminate data from the intermediate data set to obtain target radar information;

[0047] In detail, the target radar information is obtained. In practical applications, it can be specifically:

[0048] The longitude and latitude distance d1 is calculated by the longitude and latitude position of the target at the previous time point and the longitude and latitude position at the latest time point. As shown in equations (4) and (5), the distance between two points is calculated by longitude and latitude.

[0049] C=sin(latA)*sin(latB)*cos(lonA-lonB)+cos(latA)*cos(latB) (4)

[0050]

[0051] Where C represents the angle between the two target points and the center of the earth; lonA, latA, lonB, and latB represent the longitude and latitude of the target at point A and point B, respectively, and point B is the position point of the target on the navigation track at the next time; R represents the average distance from the equator, 6371.004 km; π represents pi; and D represents the calculated distance between the two points.

[0052] The distance d2 is calculated from the average speed of the previous period and the time difference between the two target points. The error threshold is set to determine whether the difference between d1 and d2 is within the threshold range. If it is not within the threshold range, the latest time point is determined as a bad value and is eliminated.

[0053] S1046. Unify the data in the target radar information and the GPS data information in time and space.

[0054] Considering that the radar data update cycle is short, the route and working conditions are not prone to sudden changes, and every two points can be approximately regarded as uniform linear motion, the linear interpolation method is well applicable to interpolating GPS data.

[0055] Assume that at two times T1 and T2, the GPS target position information is (β1, γ1) and (β2, γ2) respectively. When T∈(T1, T2), the position information at time T can be calculated using formula (6).

[0056]

[0057]

[0058] There is a problem of inconsistency between radar and GPS data in space and coordinate system. Radar data is mainly based on the principle of electromagnetic wave reflection. The data is presented in polar coordinates, and the origin of the polar coordinates is the base station location. GPS location information is longitude and latitude information, which is the WGS84 coordinate system. Therefore, the coordinates of GPS targets and radar targets are converted to the same coordinate system. The radar base station and the GPS base station are in the same location, and the two coordinates are converted to a Cartesian coordinate system with the base station as the origin.

[0059] The Gauss-Krüger projection is used to transform the latitude and longitude coordinates into Cartesian coordinates, and the intersection of the central meridian and the equator is taken as the origin, as shown in formula (8).

[0060]

[0061] Where β and γ represent the longitude and latitude in the WGS84 coordinate system, respectively; x and y represent the horizontal and vertical coordinates in the Cartesian coordinate system, respectively.

[0062] S106. Constructing a track association model, and using the track association model to associate the GPS data information with the target radar information;

[0063] In a feasible implementation manner, a track association model is constructed, and the GPS data information is track-associated with the target radar information using the track association model, including:

[0064] S1062. Using a coarse correlation model, the GPS data information is coarsely correlated with the target radar information to obtain a coarse correlation database;

[0065] Specifically, the track coarse correlation can be:

[0066] Traverse the GPS longitude and latitude data and the radar-analyzed longitude and latitude data at a certain moment to calculate the distance one by one, and set the distance threshold to determine whether the rough association between the two points is successful.

[0067] If the calculated distance is greater than the threshold, it is judged as a point that cannot be associated. If the distance is less than the threshold, it is judged as a point that has been coarsely associated successfully. The location that has established a coarse association relationship is recorded in the coarse association database for subsequent track fine association. The track correlation flow chart is as follows: Figure 2 shown.

[0068] S1064. Utilize the fine association model to finely associate the GPS data information in the coarse association database with the target radar information, and obtain the GPS data information and the target radar information corresponding thereto.

[0069] Further, the GPS data information in the coarse association database is finely associated with the target radar information using the fine association model, including:

[0070] The GPS data information in the coarse association database and the target radar information are processed by using the multi-dimensional Euclidean distance and multi-dimensional Mahalanobis distance calculation methods.

[0071] like Figure 3 As shown, the track details can be:

[0072] In the detailed association stage, the Euclidean distance and Mahalanobis distance calculation based on statistical theory are considered as the measurement basis of track similarity, and the distance association algorithm is used to calculate and compare the distances of various data before and after the target navigation.

[0073] The multi-dimensional calculation of Euclidean distance is shown in formula (9).

[0074]

[0075] In formula (9), X and Y represent the vectors composed of the average speed, position distance, and heading of the GPS and radar target track, respectively. i ,y i Represents the same attribute data of different tracks.

[0076] The multi-dimensional calculation of Mahalanobis distance is shown in formula (10).

[0077]

[0078] Formula (10) represents the Mahalanobis distance from sample vector X to μ, where there are n sample vectors X 1 ~X n , S is its covariance matrix, and μ is its mean vector.

[0079] Vector X i and X j The Mahalanobis distance between them can be expressed as formula (11).

[0080]

[0081] Two conditions need to be met in the Mahalanobis distance calculation process: the total sample size is greater than the sample dimension; the inverse matrix of the covariance matrix exists.

[0082] Therefore, in the track detailed association stage, the conditions that the Mahalanobis distance needs to meet are prioritized, and the Euclidean distance association algorithm is used as a substitute calculation scheme when the Mahalanobis distance association algorithm is not satisfied.

[0083] S108. Construct a fusion data weight distribution model, and use the fusion data weight distribution model to perform track fusion on the GPS data information and its associated target radar information.

[0084] In practical applications, the analytic hierarchy process (AHP) and entropy weighting method are used to solve the weights of the evaluation indexes of the factors affecting the same target track.

[0085] The first-level evaluation indicators are divided into GPS and radar reliability; the second-level evaluation indicators are divided into GPS speed V1, longitude and latitude L1, and heading C1 data values ​​describing the same target track and radar speed V2, longitude and latitude L2, and heading C2 data values ​​describing the same target track.

[0086] Specifically, a fusion data weight distribution model is constructed, and the GPS data information and the target radar information associated therewith are track-fused using the fusion data weight distribution model, including:

[0087] S1082. Determine the weight of the evaluation index of the track influencing factor according to the hierarchical analysis method to obtain the first weight;

[0088] Furthermore, to obtain the first weight vector, a feasible processing method is:

[0089] The steps for AHP to determine the weights of U(V), U(L), and U(C) are as follows:

[0090] (1) Construct a judgment matrix. Take U(V), U(L), and U(C) as the criteria of the judgment matrix. Different criteria have different weights in the subjective impressions of different decision makers. ij Constitute the AU judgment matrix P.

[0091]

[0092] (2) Calculate the importance ranking. According to the judgment matrix, find its maximum characteristic root λ max The corresponding eigenvector w. The equation is as follows:

[0093] P w =λ max w (13)

[0094] The required eigenvector w is normalized to obtain W = (W 1 ,W 2 ,W 3 ), W is the importance ranking of each evaluation factor, that is, the weight distribution.

[0095] (3) In order to verify whether the obtained weight distribution is reasonable, it is also necessary to perform a consistency test on the judgment matrix. The test uses the formula:

[0096] CR=CI / RI (14)

[0097] In the formula, CR is the random consistency ratio of the judgment matrix; CI is the general consistency index of the judgment matrix. It is given by the following formula:

[0098] CI=(λ max -n) / (n-1) (15)

[0099] RI is the average random consistency index of the judgment matrix, and n is the order of the judgment matrix P.

[0100] When the CR of the judgment matrix P is less than 0.1 or λ max When , P is considered to have satisfactory consistency, otherwise the elements in P need to be adjusted to make it have satisfactory consistency.

[0101] After obtaining a matrix P with satisfactory consistency, w=(w 1 ,w 2 ,w 3 ) are the weight values ​​obtained by the hierarchical analysis method.

[0102] S1084. Determine the weight of the evaluation index of the track influencing factor according to the entropy weight assignment method to obtain a second weight;

[0103] Furthermore, the second weight vector is obtained, which in practical applications can be:

[0104] The steps for determining the weights of U(V), U(L), and U(C) using the entropy weighting method are as follows:

[0105] (1) Construct a numerical matrix. Take U(V), U(L), and U(C) as the numerical matrix criteria. ij Make up the AX numerical matrix W.

[0106]

[0107] Select n samples and m indicators for each sensor, then x ij is the value of the jth indicator of the i-th sample (i=1, 2…, n; j=1, 2,…, m);

[0108] (2) Index normalization: homogenization of heterogeneous indicators. The specific method is as follows:

[0109] Positive indicators:

[0110]

[0111] Negative indicators:

[0112]

[0113] Among them, minx j is the minimum value of the jth index, maxx j is the maximum value of the jth index.

[0114] (3) Calculate the characteristic weight of the i-th group of data for the j-th indicator

[0115]

[0116] where i = 1, ..., n; j = 1, ..., m

[0117] (4) Calculate the entropy value of the jth item

[0118]

[0119] Where k = 1 / ln(n) > 0, if x ij The value of is 0, so replace it with 0.00001 in the calculation and calculate the value of ln(0.00001). j ≥0.

[0120] (5) Normalization to determine the weight of each indicator

[0121]

[0122] Where m is e j The number of, w=(w 1 ,w2 ,w 3 ) are the weight values ​​obtained by the entropy weighting method.

[0123] S1086. Combine the first weight and the second weight to obtain a combined weight;

[0124] The characteristic vector w obtained by AHP and entropy weighting method in the above steps is the indicator weight. The indicator weight sets of the two are respectively set as and The combined weights are obtained.

[0125] The indicator weight vector obtained by the hierarchical analysis method is:

[0126] The index weight vector obtained by the entropy weighting method is:

[0127] Among them, i is the i-th indicator in the first-level evaluation indicator, m is the number of secondary evaluation indicators under the i-th indicator in the first-level evaluation indicator, and the combined weight value is calculated using the hierarchical analysis method and entropy weighting method. The comprehensive weight of the indicator to be evaluated is calculated by the following formula:

[0128]

[0129] Among them, α is the coefficient for determining the combined weight, 0≤α≤1, and α is generally 0.5. When α is 1, it means that the indicator weight value is determined by the hierarchical analysis method; when α is 0, it means that the indicator weight value is determined by the entropy weighting method, thus obtaining Table 1:

[0130]

[0131] A=(A 1 ,A 2 ,A 3 ) can be used as the weight coefficient of U←{U(V),U(L),U(C)}, and the weight distribution of each component forms a data fusion strategy.

[0132] S1088. Using the combined weight, the GPS data information is track-fused with the target radar information associated with it.

[0133] The vector matrix of the speed V, longitude and latitude L, and heading C data values ​​describing the same target track using GPS and radar is multiplied by the combined weight matrix to calculate the fused x and y track values ​​S respectively. ij and S′ ij The formula is as follows:

[0134] S ij =S ix +S jx (twenty two)

[0135] S′ ij =S iy +S jy (twenty three)

[0136] Where i represents the track measured by the GPS sensor and j represents the track measured by the radar.

[0137] This embodiment provides a design idea of ​​USV unmanned surface vessel positioning data fusion framework based on the characteristics of GPS and radar data, and gives Figure 4 :

[0138] (1) In the target sea area, the navigation target USV transmits its own GPS data information to the shore-based end and then sends it to the USV receiving end; the radar data information is obtained through the USV ship-borne navigation radar data link;

[0139] (2) In the data preprocessing stage, radar data information, specifically radar image data, is processed by coordinate conversion, target matching, distance and azimuth conversion, longitude and latitude conversion, and further data elimination, time unification, and space unification.

[0140] (3) Construct a mathematical model for track association and use multi-factor Euclidean distance and Mahalanobis distance association algorithms to associate target tracks;

[0141] (4) The navigation data is fused using a weighted data fusion algorithm based on the analytic hierarchy process and entropy weight assignment method.

[0142] This embodiment provides an unmanned surface vessel track fusion device, such as Figure 5 As shown, the device comprises:

[0143] An information acquisition module 502 is used to acquire unmanned surface vessel perception information, wherein the perception information includes GPS data information and radar data information;

[0144] A data preprocessing module 504 is used to perform data preprocessing on the radar data information to obtain target radar information;

[0145] A track association module 506 is used to construct a track association model, and use the track association model to perform track association between the GPS data information and the target radar information;

[0146] The track fusion module 508 is used to construct a fusion data weight distribution model, and use the fusion data weight distribution model to perform track fusion on the GPS data information and the target radar information associated with it.

[0147] The embodiments of the present application are described by experiments:

[0148] The raw data obtained from the experimental test were processed, and the target GPS track, radar track and data fusion track were compared and analyzed. The due north direction was taken as the positive direction of the y-axis, and the due east direction was taken as the positive direction of the x-axis. The following results were obtained: Figure 6 The target flight track is shown.

[0149] Figure 6 It can be seen that the radar target track fluctuates greatly; the GPS target track is relatively stable, but sometimes there will be temporary data loss; the fusion target track is stable and reliable, basically consistent with the target speed and direction maintaining track, and has higher credibility.

[0150] The unmanned surface vessel track fusion method and device provided by the present invention have the following beneficial effects:

[0151] (1) Based on the fusion of GPS and radar data, the problems of unstable transmission frequency and return delay in GPS data, and the problems of sea clutter interference and short scanning distance in radar data are solved. The purpose of complementary advantages in the perception of maritime targets is achieved through the association and fusion of GPS data and radar data.

[0152] (2) Constructing a mathematical model of track association, combining coarse association with fine association, and specifically applying multi-factor Euclidean distance and Mahalanobis distance association algorithms to associate target tracks, so that the present application can provide multiple channels for identifying the same target track, improve the matching calculation speed, and save computing resources.

[0153] (3) Construct a fusion data weight allocation model based on the hierarchical analysis method and the entropy weight assignment method to avoid the problems existing in the single weight method, take into account the advantages of the hierarchical analysis method and the entropy method, and have higher credibility.

[0154] It should be further noted that the data information in the implementation manner of the present invention can be coordinate information, or longitude and latitude information, or a combination of speed, position distance, and heading angle, or a combination of speed, longitude and latitude, and heading, etc., depending on the specific application.

[0155] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention. Multiple implementation methods can be combined and applied if they are feasible.

Claims

1. A method for fusion of unmanned surface vessel tracks, characterized in that: The method comprises: Acquiring unmanned surface vessel perception information, wherein the perception information includes GPS data information and radar data information; Performing data preprocessing on the radar data information to obtain target radar information; Constructing a track association model, and using the track association model to perform track association between the GPS data information and the target radar information; Constructing a fusion data weight distribution model, and using the fusion data weight distribution model to perform track fusion of the GPS data information and its associated target radar information; Constructing a track association model, and using the track association model to perform track association between the GPS data information and the target radar information, including: Using a coarse correlation model, the GPS data information is coarsely correlated with the target radar information to obtain a coarse correlation database; Using a fine association model, finely associate the GPS data information in the coarse association database with the target radar information to obtain the GPS data information and the target radar information corresponding to the GPS data information; Using the fine association model, finely associating the GPS data information in the coarse association database with the target radar information includes: The GPS data information in the rough association database and the target radar information are processed by using multi-dimensional Euclidean distance and multi-dimensional Mahalanobis distance calculation methods. The step of constructing a fusion data weight distribution model and using the fusion data weight distribution model to perform track fusion of the GPS data information and the target radar information associated therewith comprises: Determine the weights of the evaluation indexes of the factors affecting the track according to the analytic hierarchy process and obtain the first weight; Determine the weight of the evaluation index of the track influencing factor according to the entropy weight assignment method to obtain the second weight; Combining the first weight and the second weight to obtain a combined weight; By using the combined weight, the GPS data information and its associated target radar information are track fused. Determining the weights of the evaluation indexes of the track influencing factors according to the hierarchical analysis method to obtain the first weight vector includes: Construct a judgment matrix; Determining a weight distribution vector according to the eigenvector corresponding to the maximum eigenroot in the judgment matrix; When the judgment matrix has satisfactory consistency, the weight allocation vector is the first weight vector, Determining the weight of the track influencing factor evaluation index according to the entropy weight assignment method to obtain the second weight vector includes: constructing a numerical matrix, wherein the numerical matrix includes a plurality of numerical values; Normalizing the values ​​in the numerical matrix to obtain positive indicators and negative indicators corresponding to each value; According to the positive index and the negative index, a characteristic weight of a corresponding value is obtained; According to the characteristic weight, the entropy value of the corresponding indicator is determined; The entropy value of each indicator is normalized to obtain the second weight vector.

2. The unmanned surface vessel track fusion method according to claim 1, characterized in that: The performing data preprocessing on the radar data information to obtain target radar information includes: Extract radar data information to obtain an intermediate data set; Performing data elimination on the intermediate data set to obtain target radar information; The data in the target radar information and the GPS data information are unified in time and space.

3. The unmanned surface vessel track fusion method according to claim 2, characterized in that: The step of extracting radar data information to obtain an intermediate data set includes: Performing Cartesian coordinate transformation on the radar data information to obtain a transformed data set; According to the conversion data set, the whole image corresponding to the radar data information is divided into multiple sub-images; Matching the multiple sub-images with the target template image and selecting the best matching sub-image; An intermediate data set is determined according to the optimal matching subgraph.

4. The unmanned surface vessel track fusion method according to claim 2, characterized in that: The step of removing data from the intermediate data set to obtain target radar information includes: Determine a first distance according to the first moment coordinates and the second moment coordinates in the intermediate data set; Get the time difference between the first moment and the second moment; Obtaining a second distance according to the average speed in the first period of time in the intermediate data set and the time difference; Obtaining a difference between the first distance and the second distance; When the difference is within a set threshold range, the coordinate at the second moment is determined to be a bad value, and the bad value is eliminated; According to the bad value determination method, all data in the intermediate data set are traversed to obtain target radar information.

5. An unmanned surface vessel track fusion device, characterized in that: The device comprises: An information acquisition module is used to acquire the perception information of the unmanned surface vessel, wherein the perception information includes GPS data information and radar data information; A data preprocessing module, used to perform data preprocessing on the radar data information to obtain target radar information; A track association module is used to construct a track association model, and use the track association model to perform track association between the GPS data information and the target radar information, including: using a coarse association model to perform coarse association between the GPS data information and the target radar information to obtain a coarse association database; using a fine association model to perform fine association between the GPS data information in the coarse association database and the target radar information to obtain the GPS data information and the target radar information corresponding to the GPS data information; using the fine association model to perform fine association between the GPS data information in the coarse association database and the target radar information, including: using a multi-dimensional Euclidean distance and a multi-dimensional Mahalanobis distance calculation method to process the GPS data information in the coarse association database and the target radar information; A track fusion module is used to construct a fusion data weight distribution model, and use the fusion data weight distribution model to perform track fusion on the GPS data information and its associated target radar information, including: determining the weight of the track influencing factor evaluation index according to the hierarchical analysis method to obtain a first weight; determining the weight of the track influencing factor evaluation index according to the entropy weight assignment method to obtain a second weight; combining the first weight with the second weight to obtain a combined weight; using the combined weight to perform track fusion on the GPS data information and its associated target radar information, the determining the weight of the track influencing factor evaluation index according to the hierarchical analysis method to obtain a first weight vector, including: constructing a judgment matrix; determining a weight distribution vector according to the eigenvector corresponding to the maximum eigenroot in the judgment matrix; when the judgment matrix has satisfactory consistency, the weight distribution vector is the first weight vector, Determining the weight of the track influencing factor evaluation index according to the entropy weight assignment method to obtain the second weight vector includes: Constructing a numerical matrix, the numerical matrix includes multiple numerical values; normalizing the numerical values ​​in the numerical matrix to obtain positive indicators and negative indicators corresponding to each numerical value; obtaining the characteristic weight of the corresponding numerical value according to the positive indicators and negative indicators; determining the entropy value of the corresponding indicator according to the characteristic weight; normalizing the entropy value of each indicator to obtain the second weight vector.

Images