Real-time track fusion algorithm and system based on radar and AIS and application

Abstract

The invention discloses a real-time track fusion algorithm and system based on a radar and an AIS (Automatic Identification System) and application, and particularly relates to the technical field of shipping monitoring and management, which is used for carrying out batch elimination processing on targets subjected to batch elimination. And for the associated tracks, association judgment is performed on the associated radar tracks and AIS tracks in the latest time window in a manner of sliding the time window, and if the number of unassociated points exceeds a misassociation threshold, it is judged that the associated tracks are misassociated, and de-batching is performed on the associated tracks. The method comprises the following steps: calculating the correlation degree of a radar track to be correlated and an AI track to be correlated in pairs to obtain a correlation matrix, setting a coarse correlation time window, and performing correlation degree calculation on track points in the time window; and merging the regional associated tracks. And on the basis of an incidence matrix obtained by coarse correlation, through a KM algorithm, a radar and AI target combination with the highest correlation degree is screened out to serve as a correlation track.

Description

Technical Field

[0001] The present invention relates to the technical field of shipping monitoring and management, and in particular to an algorithm, system and application of real-time track fusion based on radar and AIS. Background Art

[0002] With the improvement of productivity and the development of shipping, the number of ships at sea is increasing, and maritime traffic is becoming more and more busy. There is an urgent need to monitor and manage the operation of ships at sea in real time. Common monitoring methods include monitoring based on radar detection data and monitoring based on ship automatic identification system (AIS) reporting data. However, although radar can autonomously detect surrounding ships and form a panoramic traffic situation in the area, with high real-time characteristics, it has its own shortcomings such as low accuracy, easy to be affected by weather and sea conditions, and easy to be affected by hull obstruction. AIS equipment has the advantages of being able to automatically exchange navigation information between ships, rich information, and high positioning accuracy. However, the data reporting frequency of AIS is slow, which cannot meet the needs of real-time monitoring. Moreover, not all ships are equipped with AIS equipment, and there are also behaviors such as artificial closure of AIS and counterfeiting of AIS.

[0003] At present, traditional fusion methods, and more research based on fusion methods, do not address the following three types of problems in real-time fusion monitoring activities:

[0004] 1) How to associate targets, fuse tracks, and batch fused targets in real time when targets are discovered;

[0005] 2) As time goes by and the number of data points used for fusion increases, how to judge whether there is a wrong association and whether it is necessary to unbatch and re-associate and fuse;

[0006] 3) How to cancel a batch if the target disappears or the target track data cannot be received for a long time? Summary of the invention

[0007] To this end, the present invention provides an algorithm, system and application for real-time track fusion based on radar and AIS, which fuses the information of AIS and radar, thereby improving the position accuracy of the ship and the time accuracy of monitoring, and comprehensively grasping the real-time panoramic traffic situation in the area, so as to solve the problems raised in the background technology.

[0008] The present invention describes the fusion process of radar detection data and AIS reporting data in actual application from the perspective of real-time situation monitoring and the life cycle of the track, including data preprocessing (outlier processing, spatial alignment, temporal alignment), track association (coarse association, fine association, batch starting method, batch modification method, batch elimination method), track fusion (track fusion, data filtering), parameter configuration (to adapt to the fusion requirements of different scenarios), so as to adapt to the real-time fusion monitoring scenario requirements in actual applications.

[0009] The present invention provides the following technical solutions: an algorithm based on real-time track fusion of radar and AIS, monitoring based on radar detection data and monitoring based on data reported by the ship automatic identification system AIS, receiving track data through radar targets and AIS targets, regularizing the data format after receiving the radar track data, obtaining radar original track data and AIS original track data and performing data preprocessing on them,

[0010] Before association, check according to the received radar cancellation report or traverse all target tracks to determine whether there are targets that need to be cancelled, and cancel the cancelled targets, send a cancellation report, notify the situation display to cancel, and eliminate invalid targets in time;

[0011] Check the remaining batches of target tracks for misassociation:

[0012] For the associated target tracks, the KNN algorithm is used to perform association judgment on the associated radar tracks and AIS tracks in the latest time window in the form of a sliding time window; if the number of unassociated points exceeds the false association threshold, it is determined that the associated track has been misassociated, and the associated track is de-associated, and the status of the corresponding radar target and AIS target is changed to pending association, and the target track association check is continued; if the number of unassociated points does not exceed the false association threshold, the associated track is added to the associated target list and track fusion is performed;

[0013] Check the track association conditions of the AIS targets and radar targets in the list of targets to be associated before rough association. When the number of target points in the track exceeds the minimum track points for track association, the track is added to the list of tracks to be associated (radar targets are added to the radar list of tracks to be associated, and AIS targets are added to the AIS list of tracks to be associated). Otherwise, it is not considered as a track to be associated and the target track is added to the single target list for track fusion.

[0014] From the perspective of space, partition processing is performed, and coarse correlation processing is performed in the area. The correlation degree of m radar tracks and n AIS tracks to be correlated in the area is calculated two by two, and an m*n correlation matrix is ​​obtained. In order to reduce the amount of calculation, from the perspective of time, a coarse correlation time window is set, and the correlation degree of the track points within the time window is calculated; then the regional correlated tracks are merged;

[0015] Based on the m*n correlation matrix obtained by rough correlation, the KM algorithm is used to screen out the radar and AIS target combinations with the highest correlation as the correlation tracks. The successfully correlated radar targets and AIS targets are added to the associated list to achieve track fusion. Otherwise, the target track is added to the single target list and track fusion is performed.

[0016] The successfully associated AIS targets and radar targets are added to the associated list to achieve track fusion, and the fused track is sent to the situation or three-party system. Before sending, the AIS target position information is restored to longitude and latitude information, and the radar target position information is restored to distance and azimuth information. When sending, the system track type is sent together; among them, if batch cancellation is required, a batch cancellation message is sent.

[0017] Preferably, the track data reception includes radar detection data reception and AIS reporting data reception; the radar detection data includes target track data and target elimination reporting data; the target track data format includes radar target number, target distance, target azimuth, speed and heading; the target elimination reporting data format includes radar target number and elimination mark; the AIS reporting data format includes AIS target MMSI number, longitude, latitude, ground speed, ground heading and turning rate.

[0018] Preferably, the steps for regularizing the data format of radar track data are as follows: after receiving the radar detection data and the AIS reporting data, adding the current timestamp to the radar detection data and the AIS reporting data as the data reporting time; using the radar target number as the batch number of the radar target track, and the radar detection data as the radar original track data; using the AIS target MMSI number as the batch number of the AIS track, and the AIS reporting data as the AIS original track data.

[0019] Preferably, the batch elimination time window is set as follows:

[0020] The radar debatch time window and the AIS debatch time window are set independently. Taking into account the communication environment and the fault tolerance of data packet loss, the radar debatch window time is set to 3-5 radar average data reporting cycles, and the AIS debatch window is simplified to the association window of coarse association. When the debatch accuracy needs to be improved, the average data reporting cycle of each AIS target is counted and set to the average data reporting cycle of 3-5 AIS targets.

[0021] The steps for timeout batch elimination are as follows:

[0022] Traverse the pre-processed radar original track, determine whether the data time of each point in the track is within the radar batch cancellation time window, delete the points that are not within the batch cancellation time window, and then determine whether the number of points in the track is 0. If the number is 0, the track is determined to be a batch cancellation track, and the track is deleted from the radar original track list. The state of the AIS target associated with the track is changed to pending association, which means that the target of the track has disappeared and batch cancellation is performed; if the number is not 0, the track state remains unchanged;

[0023] Batch cancellation: If the fused track of the radar targets in this batch is a single radar system track, the system track will be deleted directly; if the fused track of the radar targets in this batch is a fusion system track, the type of the system track will be changed to a single AIS system track, and the status of the corresponding AIS target will be changed to pending association; the same applies to the batch cancellation of AIS.

[0024] Preferably, the mis-association sliding time window is set to 1 / 3 of the coarse association window, and the amount of calculation is reduced by judging the mis-association trend. The number of mis-association thresholds is set to 1 / 2 of the number of points after spatiotemporal registration in the sliding window.

[0025] Preferably, the steps of coarse association are as follows:

[0026] In order to reduce the associated calculation amount, the area is partitioned according to the fusion area range, the number of fusion partitions, and the overlap of fusion partitions in the configuration file;

[0027] a. Traverse the targets in each partition, execute steps b to d, and calculate the track correlation degree in each partition;

[0028] b. Rough track association: associate the m radar tracks and n AIS tracks in the state of being associated, find the track points of each track within the rough association time window from the radar registration tracks and AIS registration tracks, calculate the correlation between the two tracks through the KNN algorithm, and form an m*n correlation matrix; the initial value of the correlation matrix is ​​0;

[0029] The KNN algorithm calculates the Euclidean distance of the target at the same time point of two tracks. When the distance is less than the coarse correlation distance threshold, the number of successful correlations is increased by 1. Among N correlations, if there are K successful correlations between the radar and AIS tracks, the two tracks are considered to be correlated. The correlation degree between the two tracks is calculated and set at the corresponding point of the correlation matrix.

[0030] c. When calculating the Euclidean distance of radar and AIS tracks, the position, heading and speed are calculated comprehensively, and different weights are assigned to the position, heading and speed. When the data quality is good, the coarse correlation position weight > coarse correlation speed weight > coarse correlation heading weight. When the heading in the data is unstable, the coarse correlation heading weight can be lowered or set to 0;

[0031] d. The correlation between the two tracks is calculated by the following formula:

[0032] [(association distance threshold * number of associated track points) – sum of Euclidean distances of successfully associated points] / (association distance threshold * number of associated track points);

[0033] e. The results of the rough association of all partitions are merged. If the set partition overlap is greater than 0, the associations of the overlapping parts of adjacent partitions are fused. The fusion method is to retain the one with the larger association.

[0034] Preferably, the steps of track detail association are as follows:

[0035] Based on the fused association matrix, the rows and columns with all weights of 0 are deleted to obtain a new association matrix, which is recorded as the fine association matrix;

[0036] Based on the fine correlation matrix, the radar and AIS with fewer numbers are taken as the left set X of the KM algorithm, and the radar and AIS with more numbers are taken as the right set Y of the KM algorithm;

[0037] Given a weighted bipartite graph Gm,n=(X,Y):X={x1,x2...xm},Y={y1,y2,...yn}, the weight of edge xiyj is wij, which is the weight in the fine incidence matrix, and 0 means unmatchable;

[0038] Use the KM algorithm to solve the maximum weight matching of G;

[0039] The maximum weight match solved, that is, the radar and AIS target combination with the highest correlation, is used as the associated track, and the status of the corresponding track is set to associated.

[0040] Preferably, the track fusion is as follows:

[0041] For radar tracks with less track points than the minimum track points associated with the track, if there is no such system track, it will be batched as a new single radar system track; if there is such a system track, it will not pass the Kalman filter and will be directly used as the latest time data of the single radar system track;

[0042] For AIS tracks with less track points than the minimum track points associated with the track, if there is no such system track, it will be batched as a new single AIS system track; if there is such a system track, it will not pass the Kalman filter and will be directly used as the latest time data of the single AIS system track;

[0043] For radar tracks that have not been successfully associated, i.e., in the state of waiting to be associated, Kalman filtering is performed on the radar track data, and the filtered data is used as the latest time data of the single radar system track;

[0044] For AIS tracks that have not been successfully associated, i.e., in the waiting state, Kalman filtering is performed on the AIS track data, and the filtered data is used as the latest time data of the single AIS system track;

[0045] For the successfully associated tracks, i.e. the associated tracks, Kalman filtering is performed on the corresponding radar targets and AIS targets respectively, and the filtered data is fused as the fused system track. The batch number of the track follows the batch number of the single AIS system track.

[0046] Different weights are assigned to radar tracks and AIS tracks during fusion. The weight setting is based on the ratio of the system errors of the radar track and the AIS track. The one with a smaller system error has a greater weight.

[0047] Since the data time interval of AIS targets is relatively large, ranging from a few seconds to a few minutes, especially when the interval is more than 30 seconds, the credibility of the interpolated data will decrease. Therefore, during fusion, the weight of the AIS target decays as the interpolation time increases. The weight is adjusted according to the confidence in the AIS calibration track. The adjustment method is:

[0048] AIS target weight = AIS target weight * confidence level.

[0049] The present invention discloses a system for real-time track fusion based on radar and AIS, and the system executes the above-mentioned algorithm for real-time track fusion based on radar and AIS.

[0050] The invention discloses a system for real-time track fusion based on radar and AIS.

[0051] The present invention has the following advantages:

[0052] The present invention improves the reliability of input data by cleaning abnormal data during data preprocessing. When the tracks are associated, the track participates in the track association only after the number of points on the track exceeds the set track point number condition. The track that has not been successfully associated at the previous moment is added to the to-be-associated set, and the track that has been successfully associated at the previous moment is subjected to an association check. If the association threshold is not met, the associated track is unbatch and added to the association blacklist. The corresponding radar track and AIS track are added to the to-be-associated set and re-associated. A de-batch strategy is added to de-batch in time when the de-batch condition is met, thereby improving the reliability of the associated track. When the tracks are fused, the weight of the AIS data gradually decays with the increase of the interpolation time, and parameter configuration is added to meet the fusion requirements of different scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The overall architecture diagram of the system provided by the present invention;

[0054] Figure 2 A flow chart provided for the present invention;

[0055] Figure 3 A schematic diagram of batch elimination processing provided by the present invention. DETAILED DESCRIPTION

[0056] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

[0057] like Figure 1-3 As shown, this embodiment provides an algorithm for real-time track fusion based on radar and AIS, based on the monitoring of radar detection data and the monitoring of ship automatic identification system AIS reporting data, track data reception is realized through radar targets and AIS targets, and after receiving the radar track data, the data format is regularized to obtain the radar original track data and the AIS original track data and perform data preprocessing on them.

[0058] Before association, check according to the received radar cancellation report or traverse all target tracks to determine whether there are targets that need to be cancelled, and cancel the cancelled targets, send a cancellation report, notify the situation display to cancel, and eliminate invalid targets in time;

[0059] Check the remaining batches of target tracks for misassociation:

[0060] For the associated target tracks, the KNN algorithm is used to perform association judgment on the associated radar tracks and AIS tracks in the latest time window in the form of a sliding time window; if the number of unassociated points exceeds the false association threshold, it is determined that the associated track has been misassociated, and the associated track is de-associated, and the status of the corresponding radar target and AIS target is changed to pending association, and the target track association check is continued; if the number of unassociated points does not exceed the false association threshold, the associated track is added to the associated target list and track fusion is performed;

[0061] Check the track association conditions of the AIS targets and radar targets in the list of targets to be associated before rough association. When the number of target points in the track exceeds the minimum track points for track association, the track is added to the list of tracks to be associated (radar targets are added to the radar list of tracks to be associated, and AIS targets are added to the AIS list of tracks to be associated). Otherwise, it is not considered as a track to be associated and the target track is added to the single target list for track fusion.

[0062] From the perspective of space, partition processing is performed, and coarse correlation processing is performed in the area. The correlation degree of m radar tracks and n AIS tracks to be correlated in the area is calculated two by two, and an m*n correlation matrix is ​​obtained. In order to reduce the amount of calculation, from the perspective of time, a coarse correlation time window is set, and the correlation degree of the track points within the time window is calculated; then the regional correlated tracks are merged;

[0063] Based on the m*n correlation matrix obtained by rough correlation, the KM algorithm is used to screen out the radar and AIS target combinations with the highest correlation as the correlation tracks. The successfully correlated radar targets and AIS targets are added to the associated list to achieve track fusion. Otherwise, the target track is added to the single target list and track fusion is performed.

[0064] The successfully associated AIS targets and radar targets are added to the associated list to achieve track fusion, and the fused track is sent to the situation or three-party system. Before sending, the AIS target position information is restored to longitude and latitude information, and the radar target position information is restored to distance and azimuth information. When sending, the system track type is sent together; among them, if batch cancellation is required, a batch cancellation message is sent.

[0065] Among them, radar target: ship target detected by radar;

[0066] AIS target: the ship target received by the AIS device;

[0067] Radar original track: a combination of points formed by the same radar target in the order of reception time;

[0068] AIS original track: a combination of point tracks formed by the same AIS target in the order of receiving time; radar registered track: a radar track formed after the radar target is registered in time and space;

[0069] AIS registration track: The AIS track is formed after the AIS target is registered in time and space;

[0070] Single radar system track: For radar targets without associated AIS, the radar track is filtered; Single AIS system track: For AIS targets without associated radar, the AIS track is filtered; Fusion system track: For successfully associated radar and AIS targets, the fusion track is fused;

[0071] The track data reception includes radar detection data reception and AIS reporting data reception; radar detection data includes target track data and target elimination reporting data; the target track data format includes radar target number, target distance, target azimuth, speed and heading; the target elimination reporting data format includes radar target number and elimination mark; the AIS reporting data format includes AIS target MMSI number, longitude, latitude, ground speed, ground heading and turning rate.

[0072] The steps for regularizing the radar track data format are as follows: after receiving the radar detection data and AIS reporting data, add the current timestamp to the radar detection data and AIS reporting data as the data reporting time; use the radar target number as the batch number of the radar target track, and the radar detection data as the radar original track data; use the AIS target MMSI number as the batch number of the AIS track, and the AIS reporting data as the AIS original track data.

[0073] Data preprocessing of radar raw track data and AIS raw track data, including spatial alignment, outlier processing and temporal alignment.

[0074] 1) Spatial registration:

[0075] 1)-1AIS obtains the location information of the target through the built-in satellite positioning (GNSS) receiver or GPS receiver, using the WGS-84 coordinate system, and the location information is reported in longitude and latitude. The spatial registration method for AIS coordinates is to convert longitude and latitude into plane coordinates using Gauss-Kruger projection to calculate the registered coordinates of the AIS target.

[0076] 1)-2 The target location information obtained by the radar is expressed in distance and azimuth, which is a polar coordinate system. Therefore, for the radar target position, the polar coordinates are converted into rectangular coordinates; then the longitude and latitude coordinates of the radar device (radar coordinate position in the configuration file) are converted into plane coordinates through Gauss-Krüger projection; then the coordinates of the radar target position are translated according to the reference coordinates of the radar device to calculate the aligned coordinates of the radar target.

[0077] 2) Outlier processing:

[0078] 2)-1AIS abnormal position check: AIS data with longitude and latitude of 0 in the AIS original track data are discarded as abnormal data. In the actual processing process, abnormal data will not be absolutely 0, but often have floating point values ​​close to zero. As a simplified processing, all data between [-1,1] are discarded.

[0079] 2)-2 Flying point check: For target data of the same batch number, if (current moment position - previous moment position) / (current moment time - previous moment time)>the set effective speed threshold, it is considered that the position of the target point at the current moment has a large error and is a flying point, so the point is discarded.

[0080] 2)-3 Heading correction: For AIS data, when the ground heading = 360, 360 is the default value, and the ground heading at the current moment is set to the ground heading of the target at the previous moment.

[0081] 2)-4 Speed ​​correction: For data with negative speed, the current speed is set to the speed of the target at the previous moment.

[0082] 3) Time registration:

[0083] Since the update cycle of radar targets is 1-2 seconds and the update cycle of AIS targets is as short as 2 seconds and as long as several minutes, they need to be unified at the same time point for association and time alignment.

[0084] The time registration method used in this paper is to sample at a fixed period according to the minimum time period required by the actual business scenario. For example, if the sampling is at a fixed period of 2 seconds, both the radar target and the AIS target are sampled at intervals of 2 seconds. Due to the real-time track fusion, the sampling method adopts the extrapolation interpolation method for sampling (other interpolation methods such as quadratic splines can also be used, adjusted according to performance requirements and accuracy requirements).

[0085] Since the time interval of AIS data is relatively long, the confidence level is increased after the AIS data is time-aligned by extrapolation interpolation. The longer the extrapolation time, the lower the confidence level. (The confidence level ranges from 0 to 1)

[0086] For example, starting from 08:00:00.000 as the reference time, the sampling period is 2 seconds. The following is an example of radar target spatial registration. The same is true for AIS targets.

[0087]

[0088] The batch cancellation time window is set as follows:

[0089] Since the reporting cycle of radar targets (generally 1-2 seconds) and the reporting cycle of AIS targets (generally 2 seconds to 3 minutes, the faster the ship speed, the shorter the reporting cycle) are different, the radar batch cancellation time window and the AIS batch cancellation time window are set independently. Taking into account the communication environment and the fault tolerance of data packet loss, the radar batch cancellation window time is set to 3-5 radar average data reporting cycles, and the AIS batch cancellation window is simplified to a coarse association window. When the batch cancellation accuracy needs to be improved, the average data reporting cycle of each AIS target is counted and set to the average data reporting cycle of 3-5 AIS targets.

[0090] AIS reporting cycle (for reference):

[0091]

[0092] The steps for timeout batch elimination are as follows:

[0093] Traverse the pre-processed radar original track, determine whether the data time of each point in the track is within the radar batch cancellation time window, delete the points that are not within the batch cancellation time window, and then determine whether the number of points in the track is 0. If the number is 0, the track is determined to be a batch cancellation track, and the track is deleted from the radar original track list. The state of the AIS target associated with the track is changed to pending association, which means that the target of the track has disappeared and batch cancellation is performed; if the number is not 0, the track state remains unchanged;

[0094] Batch cancellation: If the fused track of the radar targets in this batch is a single radar system track, the system track will be deleted directly; if the fused track of the radar targets in this batch is a fusion system track, the type of the system track will be changed to a single AIS system track, and the status of the corresponding AIS target will be changed to pending association; the same applies to the batch cancellation of AIS.

[0095] The setting of the false association sliding time window is set to 1 / 3 of the coarse association window. The amount of calculation is reduced by judging the false association trend. The number of false association thresholds is set to 1 / 2 of the number of points after spatiotemporal registration in the sliding window. The threshold is adjusted according to the specific application scenario. If false association is prone to occur in a river scene, the false association threshold number needs to be set larger. If it is in the sea, false association is not likely to occur, so the false association threshold number can be set smaller.

[0096] The steps of coarse association are as follows:

[0097] In order to reduce the associated calculation amount, the area is partitioned according to the fusion area range, the number of fusion partitions, and the overlap of fusion partitions in the configuration file;

[0098] a. Traverse the targets in each partition, execute steps b to d, and calculate the track correlation degree in each partition;

[0099] b. Rough track association: associate the m radar tracks and n AIS tracks in the state of being associated, find the track points of each track within the rough association time window from the radar registration tracks and AIS registration tracks, calculate the correlation between the two tracks through the KNN algorithm, and form an m*n correlation matrix; the initial value of the correlation matrix is ​​0;

[0100] The KNN algorithm calculates the Euclidean distance of the target at the same time point of two tracks. When the distance is less than the coarse correlation distance threshold, the number of successful correlations is increased by 1. Among N correlations, if there are K successful correlations between the radar and AIS tracks, the two tracks are considered to be correlated. The correlation degree between the two tracks is calculated and set at the corresponding point of the correlation matrix.

[0101] c. When calculating the Euclidean distance of radar and AIS tracks, the position, heading and speed are calculated comprehensively, and different weights are assigned to the position, heading and speed. When the data quality is good, the coarse correlation position weight > coarse correlation speed weight > coarse correlation heading weight. When the heading in the data is unstable, the coarse correlation heading weight can be lowered or set to 0;

[0102] d. The correlation between the two tracks is calculated by the following formula:

[0103] [(association distance threshold * number of associated track points) – sum of Euclidean distances of successfully associated points] / (association distance threshold * number of associated track points);

[0104] e. The results of the rough association of all partitions are merged. If the set partition overlap is greater than 0, the associations of the overlapping parts of adjacent partitions are fused. The fusion method is to retain the one with the larger association.

[0105] The steps for track detail association are as follows:

[0106] Based on the fused association matrix, the rows and columns with all weights of 0 are deleted to obtain a new association matrix, which is recorded as the fine association matrix;

[0107] Based on the fine correlation matrix, the radar and AIS with fewer numbers are taken as the left set X of the KM algorithm, and the radar and AIS with more numbers are taken as the right set Y of the KM algorithm;

[0108] Given a weighted bipartite graph Gm,n=(X,Y):X={x1,x2...xm},Y={y1,y2,...yn}, the weight of edge xiyj is wij, which is the weight in the fine incidence matrix, and 0 means unmatchable;

[0109] Use the KM algorithm (Kuhn-Munkras algorithm, i.e., the maximum weight matching algorithm for bipartite graphs) to solve the maximum weight matching of G;

[0110] The maximum weight match solved, that is, the radar and AIS target combination with the highest correlation, is used as the associated track, and the status of the corresponding track is set to associated.

[0111] The track fusion is as follows:

[0112] For radar tracks with less track points than the minimum track points associated with the track, if there is no such system track, it will be batched as a new single radar system track; if there is such a system track, it will not pass the Kalman filter and will be directly used as the latest time data of the single radar system track;

[0113] For AIS tracks with less track points than the minimum track points associated with the track, if there is no such system track, it will be batched as a new single AIS system track; if there is such a system track, it will not pass the Kalman filter and will be directly used as the latest time data of the single AIS system track;

[0114] For radar tracks that have not been successfully associated, i.e., in the state of waiting to be associated, Kalman filtering is performed on the radar track data, and the filtered data is used as the latest time data of the single radar system track;

[0115] For AIS tracks that have not been successfully associated, i.e., in the waiting state, Kalman filtering is performed on the AIS track data, and the filtered data is used as the latest time data of the single AIS system track;

[0116] For the successfully associated tracks, i.e. the associated tracks, Kalman filtering is performed on the corresponding radar targets and AIS targets respectively, and the filtered data is fused as the fused system track. The batch number of the track follows the batch number of the single AIS system track.

[0117] Different weights are assigned to radar tracks and AIS tracks during fusion. The weight setting is based on the ratio of the system errors of the radar track and the AIS track. The one with a smaller system error has a greater weight.

[0118] Since the data time interval of AIS targets is relatively large, ranging from a few seconds to a few minutes, especially when the interval is more than 30 seconds, the credibility of the interpolated data will decrease. Therefore, during fusion, the weight of the AIS target decays as the interpolation time increases. The weight is adjusted according to the confidence in the AIS calibration track. The adjustment method is:

[0119] AIS target weight = AIS target weight * confidence level.

[0120] Through parameter configuration, the integration requirements of different scenarios are met to increase the adaptability of the system. The configuration parameters designed in this embodiment are as follows:

[0121]

[0122] The present invention discloses a system for real-time track fusion based on radar and AIS, and the system executes the above-mentioned algorithm for real-time track fusion based on radar and AIS.

[0123] The invention discloses a system for real-time track fusion based on radar and AIS.

[0124] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made on the basis of the present invention. Therefore, these modifications or improvements made on the basis of not departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. An algorithm based on real-time track fusion of radar and AIS, monitoring based on radar detection data and monitoring based on AIS reporting data of the ship automatic identification system, receiving track data through radar targets and AIS targets, regularizing the data format after receiving the radar track data, obtaining radar original track data and AIS original track data and performing data preprocessing, characterized by: Before association, check according to the received radar cancellation report or traverse all target tracks to determine whether there are targets that need to be cancelled, and cancel the cancelled targets, send a cancellation report, notify the situation display to cancel, and eliminate invalid targets in time; Check the remaining batches of target tracks for misassociation: For the associated target tracks, the KNN algorithm is used to perform association judgment on the associated radar tracks and AIS tracks in the latest time window in the form of a sliding time window; if the number of unassociated points exceeds the false association threshold, it is determined that the associated track has been misassociated, and the associated track is de-associated, and the status of the corresponding radar target and AIS target is changed to pending association, and the target track association check is continued; if the number of unassociated points does not exceed the false association threshold, the associated track is added to the associated target list and track fusion is performed; Check the track association conditions of the AIS targets and radar targets in the list of targets to be associated before rough association. When the number of target points in the track exceeds the minimum track points for track association, the track is added to the list of tracks to be associated. Otherwise, it is not considered as a track to be associated and the target track is added to the single target list for track fusion. From the perspective of space, partition processing is performed, and coarse correlation processing is performed in the area. The correlation degree of m radar tracks and n AIS tracks to be correlated in the area is calculated two by two, and an m*n correlation matrix is ​​obtained. From the perspective of time, a coarse correlation time window is set, and the correlation degree of the track points in the time window is calculated; then the regional correlated tracks are merged; Based on the m*n correlation matrix obtained by rough correlation, the KM algorithm is used to screen out the radar and AIS target combinations with the highest correlation as the correlation tracks. The successfully correlated radar targets and AIS targets are added to the associated list to achieve track fusion. Otherwise, the target track is added to the single target list and track fusion is performed. Add successfully associated AIS targets and radar targets to the associated list to achieve track fusion, and send the fused track to the situation or third-party system; Before sending, the AIS target location information is restored to longitude and latitude information, and the radar target location information is restored to distance and azimuth information; when sending, the system track type is sent together; among them, if debatch processing is required, a debatch message is sent.

2. The algorithm for real-time track fusion based on radar and AIS according to claim 1 is characterized in that: The track data reception includes radar detection data reception and AIS reporting data reception; radar detection data includes target track data and target elimination reporting data; the target track data format includes radar target number, target distance, target azimuth, speed and heading; the target elimination reporting data format includes radar target number and elimination mark; the AIS reporting data format includes AIS target MMSI number, longitude, latitude, ground speed, ground heading and turning rate.

3. The algorithm for real-time track fusion based on radar and AIS according to claim 2 is characterized in that: The steps for regularizing the radar track data format are as follows: after receiving the radar detection data and AIS reporting data, add the current timestamp to the radar detection data and AIS reporting data as the data reporting time; use the radar target number as the batch number of the radar target track, and the radar detection data as the radar original track data; use the AIS target MMSI number as the batch number of the AIS track, and the AIS reporting data as the AIS original track data.

4. The algorithm for real-time track fusion based on radar and AIS according to claim 1 is characterized in that: The batch cancellation time window is set as follows: The radar debatch time window and the AIS debatch time window are set independently. Taking into account the communication environment and the fault tolerance of data packet loss, the radar debatch window time is set to 3-5 radar average data reporting cycles, and the AIS debatch window is simplified to the association window of coarse association. When the debatch accuracy needs to be improved, the average data reporting cycle of each AIS target is counted and set to the average data reporting cycle of 3-5 AIS targets. The steps for timeout batch elimination are as follows: Traverse the pre-processed radar original track, determine whether the data time of each point in the track is within the radar batch cancellation time window, delete the points that are not within the batch cancellation time window, and then determine whether the number of points in the track is 0. If the number is 0, the track is determined to be a batch cancellation track, and the track is deleted from the radar original track list. The state of the AIS target associated with the track is changed to pending association, which means that the target of the track has disappeared and batch cancellation is performed; if the number is not 0, the track state remains unchanged; Batch cancellation: If the fused track of the radar targets in this batch is a single radar system track, the system track will be deleted directly; if the fused track of the radar targets in this batch is a fusion system track, the type of the system track will be changed to a single AIS system track, and the status of the corresponding AIS target will be changed to pending association; the same applies to the batch cancellation of AIS.

5. The algorithm for real-time track fusion based on radar and AIS according to claim 1 is characterized in that: The setting of the sliding time window for misassociation is set to 1 / 3 of the coarse association window. The amount of calculation is reduced by judging the misassociation trend. The number of misassociation thresholds is set to 1 / 2 of the number of points after spatiotemporal registration in the sliding window.

6. The algorithm for real-time track fusion based on radar and AIS according to claim 1 is characterized in that: The steps of coarse association are as follows: In order to reduce the associated calculation amount, the area is partitioned according to the fusion area range, the number of fusion partitions, and the overlap of fusion partitions in the configuration file; a. Traverse the targets in each partition, execute steps b to d, and calculate the track correlation degree in each partition; b. Rough track association: associate the m radar tracks and n AIS tracks in the state of being associated, find the track points of each track within the rough association time window from the radar registration tracks and AIS registration tracks, calculate the correlation between the two tracks through the KNN algorithm, and form an m*n correlation matrix; the initial value of the correlation matrix is ​​0; The KNN algorithm calculates the Euclidean distance of the target at the same time point of two tracks. When the distance is less than the coarse correlation distance threshold, the number of successful correlations is increased by 1. Among N correlations, if there are K successful correlations between the radar and AIS tracks, the two tracks are considered to be correlated. The correlation degree between the two tracks is calculated and set at the corresponding point of the correlation matrix. c. When calculating the Euclidean distance of radar and AIS tracks, the position, heading and speed are calculated comprehensively, and different weights are assigned to the position, heading and speed. When the data quality is good, the coarse-associated position weight > the coarse-associated speed weight > the coarse-associated heading weight. When the heading in the data is unstable, the coarse-associated heading weight can be lowered or set to 0. d. The correlation between the two tracks is calculated by the following formula: [(association distance threshold * number of associated track points) – sum of Euclidean distances of successfully associated points] / (association distance threshold * number of associated track points); e. The results of the rough association of all partitions are merged. If the set partition overlap is greater than 0, the associations of the overlapping parts of adjacent partitions are fused. The fusion method is to retain the one with the larger association.

7. The algorithm for real-time track fusion based on radar and AIS according to claim 1 is characterized in that: The steps for track detail association are as follows: Based on the fused association matrix, the rows and columns with all weights of 0 are deleted to obtain a new association matrix, which is recorded as the fine association matrix; Based on the fine correlation matrix, the radar and AIS with fewer numbers are taken as the left set X of the KM algorithm, and the radar and AIS with more numbers are taken as the right set Y of the KM algorithm; Given a weighted bipartite graph Gm,n=(X,Y):X={x1,x2...xm},Y={y1,y2,...yn}, the weight of edge xiyj is wij, which is the weight in the fine incidence matrix, and 0 means unmatchable; Use the KM algorithm to solve the maximum weight matching of G; The maximum weight match solved, that is, the radar and AIS target combination with the highest correlation, is used as the associated track, and the status of the corresponding track is set to associated.

8. The algorithm for real-time track fusion based on radar and AIS according to claim 1 is characterized in that: The track fusion is as follows: For radar tracks with less track points than the minimum track points associated with the track, if there is no such system track, it will be batched as a new single radar system track; if there is such a system track, it will not pass the Kalman filter and will be directly used as the latest time data of the single radar system track; For AIS tracks with less track points than the minimum track points associated with the track, if there is no such system track, it will be batched as a new single AIS system track; if there is such a system track, it will not pass the Kalman filter and will be directly used as the latest time data of the single AIS system track; For radar tracks that have not been successfully associated, i.e., in the state of waiting to be associated, Kalman filtering is performed on the radar track data, and the filtered data is used as the latest time data of the single radar system track; For AIS tracks that have not been successfully associated, i.e., in the waiting state, Kalman filtering is performed on the AIS track data, and the filtered data is used as the latest time data of the single AIS system track; For the successfully associated tracks, i.e. the associated tracks, Kalman filtering is performed on the corresponding radar targets and AIS targets respectively, and the filtered data is fused as the fused system track. The batch number of the track follows the batch number of the single AIS system track. Different weights are assigned to radar tracks and AIS tracks during fusion. The weight setting is based on the ratio of the system errors of the radar track and the AIS track. The one with a smaller system error has a greater weight. The weight is adjusted according to the confidence level in the AIS calibration track. The adjustment method is: AIS target weight = AIS target weight * confidence level.

9. A system based on real-time track fusion of radar and AIS, characterized by: The system executes the algorithm for real-time track fusion based on radar and AIS as described in any one of claims 1-8.

10. An application, characterized in that: The application is based on the system for real-time track fusion based on radar and AIS as described in claim 9.

Images